ACTUALIZATION

AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE

General Editor
E. F. KONRAD KOERNER
(University of Ottawa)

Series IV - CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.)

John E. Joseph (Edinburgh); Manfred Krifka (Berlin)
Hans-Heinrich Lieb (Berlin); E. Wyn Roberts (Vancouver, B.C.);
Hans-Jürgen Sasse (Köln)

Volume 219

IIenning Andersen (ed.)

Actualization
Linguistic Change in Progress
 ACTUALIZATION
LINGUISTIC CHANGE IN PROGRESS

Edited by

HENNING ANDERSEN
University of California, Los Angeles

JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA
 The paper used in this publication meets the minimum requirements of American
National Standard for Information Sciences — Permanence of Paper for Printed
Library Materials, ANSI Z39.48-1984.

Library of Congress Cataloging-in-Publication Data

Actualization : linguistic change in progress : papers from a workshop held at the 14th International
Conference on Historical Linguistics, Vancouver, B.C., 14 August 1999 / edited by Henning
Andersen.
p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV,
Current issues in linguistic theory, ISSN 0304-0763 ; v. 219)
Includes bibliographical references and index.
Contents: Position paper : markedness and the theory of change / Henning Andersen - Patterns of
restitution of sound change / Kristin Bakken - The role of markedness in the actuation and
actualization of linguistic change / Alexander Bergs and Dieter Stein - On the actualization of the
active-to-ergative shift in pre-Islamic India / Vit Bubenik - Markedness and the use of address
pronouns in early modern English / Ulrich Busse ~ Actualization patterns in grammaticalization :
from clause to locative morphology in Northern Iroquoian / Marianne Mithun - From Latin to
modern French / Lene Schøsler - Issues of explanation in linguistic change / Michael Shapiro - The
role of markedness in morphosyntactic change / John Charles Smith -- Actualization and the
(uni)directionality of change / Henning Andersen.
1. Linguistic change-Congresses. 2. Markedness (Linguistics)-Congresses. I. Andersen,
Henning, 1934-. II. International Conference on Historical Linguistics (14th : 1999 : Vancouver,
B.C.) III. Series.
P142.A28 2001
417'.7--dc21 2001043517
ISBN 90 272 3726 3 (Eur.) / 1 58811 081 8 (US)
© 2001 - John Benjamins B.V.
No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other
means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O.Box 36224 · 1020 ME Amsterdam · The Netherlands
John Benjamins North America · P.O.Box 27519 · Philadelphia PA 19118-0519 · USA
 PREFACE

In his opening address to the Thirteenth International Conference on Historical

Linguistics, held at the Heinrich-Heine University, Düsseldorf in 1997, the
conference director, Dieter Stein, called for greater attention to the actual
progression of language changes through time and encouraged the formation of
what he called a 'chronological linguistics', which would investigate historically
documented changes, past and present, with the aim of establishing the
determinants of their gradual, orderly progression.
As it happened, one of the plenary lectures at the 1997 Düsseldorf
conference provided support for Stein's call for a chronological linguistics by
presenting a theory that makes explicit what appear to be the central notions in
Sapir' s understanding of drift. The paper, which is printed below (21-57), was
well received, and several members of the audience suggested that its claims be
evaluated against data from a variety of languages.
Accordingly a workshop on "Patterns of Actualization in Linguistic Change"
was organized for the Fourteenth ICHL in Vancouver, B.C. in 1999, where
contributions were presented by Kristin Bakken, Alexander Bergs and Dieter
Stein, Vit Bubenik, Ulrich Busse, Susan Herring, Marianne Mithun, Piet van
Reenen and Lene Sch0sler, John Charles Smith, Alan Timberlake, and myself.
Most of these contributions are contained in this volume in revised form. Also
included is a written version of comments made at the time of the Workshop by
Michael Shapiro.
I am grateful to the participants in the workshop for their enthusiastic support
in planning the workshop and preparing this volume, to Lars Heltoft, Brit
Mælum, and Michael Shapiro who served as discussants in the workshop, to the
series editor, Konrad Koerner for his continued collegial support, and to my
assistant Ji-Soo Kim, who helped prepare the volume for publication.

Los Angeles, California, May 2001 Henning Andersen

 CONTENTS

Introduction
Henning Andersen 1
Position paper: Markedness and the theory of change
Henning Andersen 21
Patterns of restitution of sound change
Kristin Bakken 59
The role of markedness
in the actuation and actualization of linguistic change
Alexander Bergs and Dieter Stein 79
On the actualization of the passive-to-ergative shift in Pre-Islamic India
Vit Bubenik 95
The use of address pronouns in Early Modern English
Ulrich Busse 119
Actualization patterns in grammaticalization:
from clause to locative morphology in Northern Iroquoian
Marianne Mithun 143
From Latin to Modern French: actualization and markedness
Lene Sch0sler 169
Markedness, causation, and linguistic change: a semiotic perspective
Michael Shapiro 187
Markedness, functionality, and perseveration
in the actualization of a morphosyntactic change
John Charles Smith 203
Actualization and the (uni)directionality of change
Henning Andersen 225
General Index 249
 INTRODUCTION

HENNING ANDERSEN
University of California, Los Angeles

0. Preamble
The aim of this collection of papers is to consolidate the observation that
the progression of certain kinds of linguistic change typically is grammatically
conditioned—any given innovation being realized in usage, as it is introduced,
accepted, and generalized, over time, in one grammatical environment after
another.
Although it has long been recognized that the progression of change
more often than not is conditioned by social categories (age, gender, class) and
categories of style (from citation forms to casual conversation), medium
(written, spoken) and genre (poetry, prose, artistic, expository, etc.), the fact of
grammatical conditioning has largely been overlooked. Yet, it has been
observed with increasing frequency during the last few decades, as historical
linguists have more and more turned their attention to the actual chronological
attestation of historical changes. But the observation is not only of recent date.
A careful reading of traditional historical grammars and works on dialectology
will undoubtedly reveal many examples like the following.
In the history of Russian, the development of morphological expression
for animacy—which spans the period from the 1000s to the 1700s in our ample
textual record of the language—proceeded earlier and more widely in
pronouns than in nouns, earlier in the productive singular noun declension than
in the unproductive declensions, earlier in singular than in plural nouns, earlier
in masculine than in feminine nouns, earlier in masculines and feminines than
in neuter nouns, earlier in direct objects of verbs than in objects of
prepositions, earlier in an adjacent object than in a postponed object, earlier in
nouns denoting human than nonhuman animates, earlier in reference to a
freeman than to a functionary, servant, or slave, earlier referring to healthy
than to sick individuals, and earlier referring to adults than to adolescents or
2 HENNING ANDERSEN

children (Saxmatov [1911] 1957:220-226, Borkovskij & Kuznecov 1965:

222-226, Klenin 1983, Krys'ko 1994, Klenin 1997).1 See Table 1.

Earlier later
Word class pronouns nouns
Inflectional class productive unproductive
Number singular plural
Gender masculine feminine and neuter
Syntax verbal object prepositional object
adjacent non-adjacent
Lexical meaning human nonhuman
Reference freeman slave
healthy diseased
adult child

Table 1.

The similarity with the phonologically conditioned progression of sound

changes is striking. For example, in the development of velars, palatalization
occurs earlier in stops than in fricatives, earlier before high than before non-
high vowels, earlier before unrounded than before rounded vowels, earlier
directly contiguous to the conditioning vowel than across another segment,
earlier before than after the conditioning vowel, earlier in the narrow domain of
the syllable than across syllable boundaries (that is, in the wider domain of the
word), earlier in stable environments than in alternating environments (cf.
Timberlake 1981, Andersen 2001a:31).

Earlier later
Simultaneous features stop continuant
Contiguous features high nonhigh
unrounded rounded
Phonotactics contiguous noncontiguous
before after
syllable word
Morphotactics stable alternating

Table 2.

1
The most recent major contribution on this topic, Krys'ko (1994), shows that the trickle of
attestation begins somewhat earlier than was previously known. But the relative chronology of
the general flow of attestation in each conditioned environment is not changed by these earlier
dates. They do, however, motivate the use of the phrase "proceeded earlier" (rather than
simply 'is attested earlier').
 INTRODUCTION 3

To anyone familiar with the notion of 'markedness', it will be evident

from these examples that there is some correlation between the environments
in which innovations occur earlier or later and the markedness values of the
phonological, phonotactic, morphophonemic, morphosyntactic, clause-
syntactic, lexical-semantic, referential, and other parameters involved. At the
same time, in each case, there appears to be a distinct ranking of the diverse
features or categories that condition the gradual actualization of the change.
This suggests that the gradualness of such changes may afford us significant
insights not only into the relation between speakers' grammars and speech
performance but also into the cognitive foundations on which speakers'
grammars are formed. Accordingly, we must acknowledge this observed,
orderly gradualness of linguistic change, seek to describe as fully as possible
the progression of documented instances of change, and endeavor to provide
explanations for the observed patterns of actualization.
The papers that have been collected in this volume offer accounts of a
number of changes attested in progress, or whose progression has been
inferred, in various languages and periods. The papers are characterized very
briefly below in Section 1. Several of them explicitly discuss the interpretation
of the changes in terms of markedness. In this way they respond to the theory
put forward in the position paper (pp. 21-57 below), which first stimulated this
collective effort, as mentioned in the Preface. Several of the problems of
interpretation in terms of markedness are taken up for discussion in Section 2.

1. The papers
In the following pages I will briefly characterize the content of the
contributions that follow and draw attention to individual problems of analysis
some of them present. I mention first the position paper that gave the impetus
to this collective effort (Section 1.1) and then the papers on evolutive changes
by Bergs & Stein, Busse, Mithun, Schoesler, and Smith (Sections 1.2-1.6).
Next come the papers on contact change by Bakken and Bubenik (Sections
1.7-1.8), and finally the more general papers by Shapiro and Andersen
(Sections 1.9-1.10).

1.1 Andersen
In the position paper ("Markedness and the theory of linguistic change",
21-57) I begin by setting aside the controversies over the notion of
Markedness, which I believe have arisen due to a misguided insistence on
defining it in observational terms rather than grounding it independently of
observation. I then offer evidence from several areas of culture suggesting that
Markedness is a cognitive principle that underlies the organization of diverse
4 HENNING ANDERSEN

semiotic systems, including all areas of grammar, and proceed to show that
Markedness values not only define synchronic systems, but are intimately
involved in the actualization of change. Both in synchronic language states and
in change, combinations based on Markedness equivalence appear to embody a
Principle of Markedness Agreement. Against the background of these
observations I present a conceptual analysis of Markedness contrasted with the
classic logical relations of contrariety, contradiction, and conversity,
explaining why the asymmetry of Markedness so naturally accompanies these
other, strictly symmetrical logical relations.2 The paper concludes with a
confrontation of Markedness and the various observables (frequency,
complexity, etc.) with which it has often been erroneously identified, and it is
shown, on one hand, why Markedness values cannot always be directly
inferred from observation and, on the other, how Markedness values, once they
are ascribed to the categories of a grammar, may define the direction of its
drift—a topic developed more fully in the Workshop paper on "Actualization
and the (uni)directionality of change" (225-248).

1.2 Bergs and Stein

Alexander T. Bergs and Dieter Stein ("The role of markedness in the
actuation and actualization of linguistic change", 79-93) explore the relation
between Markedness and both the actuation and the actualization of change,
examining two developments, the rise of English wh-relatives {which, who,
whose, whom) and that of periphrastic do. The authors begin by illustrating the
natural occurrence of marked elements in marked environments and then show
how one and the same, hitherto unnoticed semantic category of social rank is
relevant to both sets of data.
Apparently, wh-relatives were first established with highly marked
antecedents, and progressed from reference to God through humans of high
estate to humans in general. Periphrastic do gained currency first with subjects
characterized as 'remarkable' in some respect, to whose narrated actions do-
periphrasis lent special weight; and from there it spread to other contexts in
which it fulfilled a similar function, at least at first, eventually being
generalized. The authors draw attention to cross-linguistic parallels such as the
lexical codification of this parameter of social rank in East Asian languages,
citing examples from Thai and Burmese.

2
Where it is appropriate and feasible I use capitalization to distinguish the explanatory concept
Markedness, which is defined in logical and cognitive terms below (37-47), from the
common-parlance notion of 'markedness'.
 INTRODUCTION 5

Both wh-relatives and do-periphrasis appear to have originated (been

'actuated') as changes from above and their initial actualization in marked
environments then corresponds to the proposed Principle of Markedness
Agreement.

1.3 Busse
Ulrich Busse ('The use of address pronouns in Shakepseare's plays and
sonnets", 119-147) takes up the problem of the use of thou and you in
Shakespeare's works, a particularly interesting body of data since Shakespeare
lived and wrote during "the decisive period" in the gradual replacement of Τ
forms by Y forms, 1550-1620, the period defined by the steepest ascent of the
S-curve of the change. Busse examines the variant accounts of Shakespeare's
usage in the existing literature and shows that the available data from the
corpus argue against a simple markedness-as-frequency account, whether
applied overall or to individual plays. Instead, the data call for a careful sorting
of the works according to date of composition, genre and subgenre (verse vs.
prose, drama vs. poetry, tragedy vs. historical drama vs. comedy, sonnet vs.
other poetic forms) and social distance (public vs. private poems). It is also
useful to compare observations made in this corpus with the pronoun usage in
the same text types produced by other, contemporary writers. But most
significantly for the purposes of this volume, Busse's investigation shows that
an approach that views the just enumerated distinctions in terms of
Markedness values yields a coherent and meaningful account of Shakespeare's
pronoun usage both in terms of a synchronic typology of texts from around
1600 and in the perspective of the chronological progression of the pronoun
change.

1.4 Mithun
Marianne Mithun presents an analysis of the development, from clauses,
of a new category of locative terms in Iroquoian ("Actualization patterns in
grammaticalization: from clause to locative morphology in Northen
Iroquoian", 143-168). The data are provided through internal comparison
within Mohawk with the addition of a few comparative glances at Tuscarora.
The data support an account in terms of reanalysis based on structural
ambiguity and gradual actualization along a gradient of locative categories that
has parallels in unrelated languages.
The gradualness of the actualization is reflected in a number of features
of the modern locatives. To mention a few, one locative still occurs as a main
predicate, the others do not; several still occur with verbal affixes, while the
rest do not; at the other extreme, the unmarked locative has been nominalized
6 HENNING ANDERSEN

to the extent that it combines with nominal allomorphs of incorporated forms

and can even occur without a nominalizer. Thus, generally, the locatives
meaning "beyond" and "in the middle of" show more verbal features than
"beside", "under", "inside", and "place", and the last mentioned, very few
indeed.
Mithun explicitly considers the relation between this gradual
development and the possible Markedness values of the different locatives. It
appears that there is a clear correlation with Markedness, particularly when the
locatives are analysed in terms of more abstract semantic features. Thus, the
locative that is most general in meaning is the most grammaticalized one. It is
also the most frequent one and the only one to show significant allomorphy (cf.
Greenberg 1966). The somewhat different degree of grammaticalization of the
others correlates with their greater semantic or cognitive complexity. In this
connection Mithun draws attention to the widespread grammaticalization
gradient "on > in > under > beside > back > front" established by Svorou
(1994), whose resemblance to the Iroquoian development is very suggestive;
see further Section 2.2.

1.5 Shøsler
Lene Schøsler ("From Latin to Modern French: actualization and
markedness", 169-185) discusses changes in French morphosyntax and syntax
with the explicit intention of evaluating the proposed Principle of Markedness
Agreement (Andersen 1990, 2001a).
Sch0sler's first example, the loss of the Old French two-case system in
the 1000-1400s, apparently was actualized largely in conformity with the
Principle of Markedness Agreement, the case distinction being lost earlier in
marked categories of nouns and pronouns than in corresponding unmarked
categories (in adjectives before substantives, substantives before determiners,
substantives and adjectives before pronouns, feminines before masculines,
plurals before singulars, nonhuman nouns before human nouns); see Section
2.4. One major deviation is the class of personal names, which lose case
marking earlier than appellatives, a particularly striking unconformity
considering that among appellatives, case marking was lost last in nouns
denoting humans; I return to this in Section 2.3. Sch0sler presents the results of
a separate examination of a text sample, which shows more case loss in main
clauses than in subordinate clauses and more in direct discouse than in
narration. On the syntactic and pragmatic level, then, the leading edge of the
change was in unmarked categories; see Section 2.7.
Sch0sler's second example is a complex of changes often associated with
the loss of case in Old French, the change in word order (from the putative
 INTRODUCTION 7

Late Latin OV to Old French VO) and the changes in subject marking (from
verb inflection to obligatory subject pronouns). The development of V2 word
order and the loss of person inflection in the verb proceed earlier in unmarked
environments than in marked and so appear to be conditioned by Markedness
Agreement. But the loss of Pro-drop starts in subordinate clauses and is
attested earlier in the interlocutor persons than in the 3rd person and earlier in
referential than in nonreferential 3rd persons, and so it appears to proceed
earlier in marked environments than in unmarked. This evidently calls for an
explanation; see Section 2.4.
Sch0sler's final example shows how the reorganization of the French
tense system is reflected earlier in direct discourse than in narration and earlier
in prose than in poetry, in conformity with the Principle of Markedness
Agreement.

1.6 Smith
John Charles Smith ("Markedness, functionality, and perseveration in the
actualization of a morphosyntactic change", 203-223) reviews the loss of
object agreement in the participles of Romance compound tenses in the dual
perspectives of markedness and functionality. As for the former, Smith relies
on the understanding of 'markedness' that equates it with observed frequency
(see the critique below 50-51). Smith's functional perspective makes appeal to
perceptual factors and processing strategies. The change examined, which
appears to have followed largely parallel courses in several Romance
languages, has been actualized in a succession of steps that are easily defined
in terms of syntactic environments, which Smith subsumes in four hierarchies
(208; see Section 2.1 below). The question then posed is whether these
hierarchies can be accounted for in terms of markedness-as-frequency. Smith
evaluates each of the four hierarchies, and finds that markedness-as-frequency
alone cannot explain them all. Next he evaluates them in terms of functionality
(as in Smith 1996, 1997) and finds that, on the whole, agreement has been lost
first in contexts where it had least functional value. In other words, it appears
that the slight advantage in decoding that object agreement contributes has
acted as a brake on the actualization of agreement loss (214). Still, in some
French dialects agreement is preserved in one environment where it contributes
no advantage in decoding: agreement with an object relative pronoun is
maintained longer when the relative's antecedent is a pronoun that varies for
gender than when it is a noun with inherent gender. After a brief look at the
influence of pragmatic factors—genre, medium, and style—for which there is
little unequivocal evidence available, Smith concludes that neither
8 HENNING ANDERSEN

functionality nor markedness provides completely satisfactory explanations for

the step-wise actualization of this change.
Smith's critical examination of the potential explanatory value of
markedness-as-frequency is very instructive and can serve to highlight several
of the perennial problems in applying this notion. Accordingly three of the
hierarchies he defines will be discussed in some detail below. It will be
apparent there that if a slightly different approach is taken in the analysis of the
conditioning environments, the gradual actualization of the Romance loss of
object agreement turns out to conform rather well with the Principle of
Markedness Agreement: on the whole, agreement has been lost earlier in
marked than in unmarked environments; see Sections 2.1, 2.4.

1.7 Bakken
Kristin Bakken's "Patterns of restitution of sound change" (59-78) offers
an account of the modern reflexes, in dialects of Norwegian, of two medieval
sound changes whose effects have been partly reversed, /-loss (before
consonants other than d, t) and what she calls 'delateralization' (-ll- > (-dl- >)
-dd-: cf. Sicilian stedda for STELLA 'star'). The questions raised by the modern
data are whether they reflect lexical diffusions that were arrested (and
reversed); what role dialect contact played in the apparent restitutions; whether
they are the result of rule loss or they proceeded lexeme by lexeme; and, if the
latter, what factors determined the progression of the restitutions. The available
data are a set of lexemes in dialect variants from three periods (1698-1821,
1920-1970, and contemporary) and placenames containing the relevant lexical
stems, attested at various times. Bakken's thorough discussion of possible
interpretations concludes that the modern data reflect regular sound changes
and a later replacement of individual lexemes by competing forms introduced
through contact with other dialects, in particular the administrative (standard)
languages.
The gradual progressions of the restitutions are shown to correlate grosso
modo with differences in relative word frequency in spoken usage, suggesting
that some lexemes—those denoting commodities of trade, for instance—may
have been more common in interdialectal communication and hence more
exposed to competition from borrowed variants. As is typical in such cases,
placenames, and especially microtoponyms, preserve reflexes of the medieval
sound changes with the greatest fidelity. (One is reminded of such textbook
examples as the German isoglosses between northern -ss- and southern -ks-
(ses || seks, os \\ oks, wassen || waksen), where deep in the -ks- area placenames
like Vosberg (with vos- for southern fuks 'fox') bear witness to the earlier much
 INTRODUCTION 9

larger distribution of the northern reflex of *-hs-, and which thus indicate how
the old isogloss has been displaced toward the north; cf. Bach 1950:56.)
Some remarks on the application of Markedness considerations to cases
such as this are offered in Sections 2.5-2.6.

1.8 Bubenik
Vit Bubenik ("On the actualization of the passive-to-ergative shift in Pre-
Islamic India", 95-118) presents an account of the drift from accusative to
ergative structure in Indian languages. The account thematicizes the continued
use and, significantly, continued development of earlier stages of the
language—Sanskrit in Early Middle Indic, Sanskrit and regional Prākrits in the
middle period, Sanskrit, Prākrits, and Apabhramsa in Late Middle Indic—in
continuous diglossic and polyglossic contact with local vernacular varieties.
The attested developments progress conditioned by a variety of grammatical,
pragmatic, stylistic, and register categories, in part apparently reflecting the
Principle of Markedness Agreement, but the observed usage appears to be
considerably complicated by the vertical contact relations. The internally
motivated, evolutive changes (from below) are partly countered by the
maintenance of the more archaic, higher diglossic variants, but still gradually
have an impact on them.
Bubenik's account draws on several contemporary conceptions of
syntactic change, but it is mainly couched in terms of the theory that views
innovations as marked and their generalization in usage as the result of a
markedness shift. Section 2.6 below offers some remarks on this question.

1.9 Shapiro
Michael Shapiro's contribution offers "a philosopher's-eye view of
language" ("Markedness, causation, and linguistic change: a semiotic
perspective", 182-201) and addresses questions of the nature of grammatical
competence, of realism and nominalism, of causality, and of markedness. In
the first section of the paper Shapiro argues that the basic units of phonology
(distinctive features) are not signs, he advocates a return to Saussure's (and a
long philosophical tradition's) conception of signifiant and signifié as
inseparable parts of the language sign (as this is traditionally understood), and
he suggests that encoding and decoding both are based on abductive inference.
In Section 2 he discusses the adverse consequences of the tradition of
nominalism for Chomsky's research project and argues that only a realist
linguistics, such as the Peircean-inspired 'neo-structuralism' he advocates, can
pose correctly the problem of the relation between language and the brain,
which, as he argues, does not call for a neuropsychological explanation of
10 HENNING ANDERSEN

language, as in the common view, but rather for a linguistic explanation of (the
evolution of) the higher cortex.
Parts 3 and 4 of the paper are more directly germane to the theme of this
volume. In Section 3 ("Semiosis and linguistic change: efficient and final
causation") Shapiro sketches the conceptual underpinnings of the distinction
between efficient causality and the often misunderstood teleology. In Section 4
("Markedness") he addresses one of the key problems discussed in the
Workshop in Vancouver, a problem that is thematicised in John Charles
Smith's contribution, viz. whether actualization is guided by Markedness
values or by "perceptual factors and processing strategies". In Shapiro's
perspective it is clear that the very formulation of this problem is based on a
misunderstanding. See further Section 3.

1.10 Andersen
Andersen's Workshop paper "Actualization and the (uni)directionality of
change" (225-248) describes the place of actualization—the only observable
aspect of change—in a theory of change; shows how the theory of Markedness
proposed in Andersen 2001a makes it possible to understand change as a
projection of synchronic variation onto the diachronic axis; and tries to clarify
the relation between historical change events, the domain of the language
historian, and the generalized 'change schemas' that the historical linguist can
use to advantage in investigating the origins of types of linguistic change.

2. Discussion
In the following paragraphs I comment on some of the issues of
interpretation that are exemplified in the papers summarized in the preceding
pages.
The comments are organized under the following headings: Clines and
hierarchies (Section 2.1), Specific vs. generic categories (Section 2.2),
Variation in Markedness values (Section 2.3), Alteration vs. loss (Section 2.4),
Markedness, restoration, and restitution (Section 2.5), and Markedness and the
S-curve (Section 2.5).
In several places the relation between innovative and traditional variants
is in focus. To save words, I will occasionally refer to such variant elements
as I-variants and O-variants, with I for "innovative, incoming" and  for
"older, outgoing".

2.1 Clines and hierarchies

It is customary to represent gradations, whether synchronic or diachronic,
in terms as close to the level of observation as possible, and most often such
 INTRODUCTION 11

gradations are presented and referred to as clines or hierarchies without much

attention to the fact that there can be a cline between the values of a feature (or
the opposite terms of a category), but different features (or categories) form a
hierarchy. The distinction between clines and hierarchies is essential for
interpretations in terms of Markedness, for Markedness defines the relation
between the values of individual features (or categories)—whether they form
an actual cline or are contradictories (47-48)—whereas the relation between
different features (or categories) is a matter of their different hierarchical rank.
For the purpose of Markedness interpretations, then, one needs to analyse each
observed gradation into the distinct features (or categories) it involves. Only if
this step is taken can the Markedness values within individual categories be
distinguished from the categories' different rank.
In some instances this is rather unproblematic. Consider Smith's
"Hierarchy 4" (208), which shows the progression of agreement loss in relation
to number and gender: "Masculine Plural > Feminine Plural > Feminine
Singular". What this gradation shows is that number agreement (SingularU vs.
PluralM) is lost earlier in the unmarked than the marked gender (MasculineU vs.
FeminineM), and that number agreement is lost earlier than gender agreement.
The greater tenacity of gender agreement shows that the reference-based
category of Number ranks lower (is less central) than the lexical category of
Gender. This is probably a typological characteristic of the Romance
languages.
Sometimes a little more analysis is called for. This seems to be the case
with Smith's "Hierarchy 3. Person of Clitic Pronoun" (213-214 below). Here
the proposed hierarchy "{First Person, Second Person, Third Person
Reflexive} > 3rd Person Nonreflexive" should surely be separated into two,
one for the category of Person, the other for that of Reflexive. The first of these
would define the progression ExophoricM (first and second person) vs.
Unspecified participant deixisU (exophoric or endophoric third person). Smith
mentions the fact that in some languages, apparently, the Markedness values in
the category of Person are different (211). Perhaps these are languages in
which the primary distinction is defined not in terms of deixis, but in terms of
speech-act role, with an opposition of InterlocutorU > NoninterlocutorM Person.
This possibility illustrates how apparent cross-linguistic variation in
Markedness values may reflect language-particular categories that have similar
purport but are categorized differently. I return to this question below (see
Section 2.3).
The second progression involved in Smith's "Hierarchy 3" is Reflexive >
Nonreflexive, which is independent of the category of Person and hence should
be examined independently in each of the three persons, singular as well as
12 HENNING ANDERSEN

plural. Smith cites word frequency data (from Spanish, French, and Italian) to
show that the reflexive se, respectively si, is less 'marked' (actually, just more
frequent) than the nonreflexive third-person direct object. But frequency
counts of morpheme shapes without regard to their function are probably not
very useful. Leaving aside the minor fact that in Spanish, the apparent
reflexive se serves as an allomorph of nonreflexive le in certain clitic
sequences, consider that in all three languages the distinctions direct vs.
indirect object and singular vs. plural are syncretized in the reflexive se,
respectively si, but not in the nonreflexive third-person pronouns; this alone
would make for a high relative frequency of the se/si forms. And consider the
fact that in all three languages medio-passive verbs are marked with
(historically speaking) 'reflexive pronouns' that cannot by any stretch of the
imagination be considered synchronic direct objects. In view of the difference
in syntactic scope, it seems more reasonable to posit the Markedness values
ReflexiveM vs. NonreflexiveU. Whether these values correlate directly with text
frequency can only be established through a word count that examines
reflexive and nonreflexive pronoun forms with a view to their diverse
functions. Such a count should naturally include first and second person
pronouns, in which reflexive and nonreflexive functions happen to be
syncretized. It seems possible that the outcome will correlate with the
progressions ExophoricM (first and second person) > Unspecified participant
deixisU (exophoric or endophoric third person) and Reflexive M >
NonreflexiveU Clitic Pronoun.

2.2 Specific vs. generic categories

In some instances it may be useful to recognize the generic category to
which specific conditioning categories belong.
This is exemplified by Smith's "Hierarchy 2. Identity of Preceding Direct
Object", which captures the progression "{Topics, Interrogatives,
Exclamatives} > Relatives > Clitic Pronouns" (208). It is perhaps not
immediately obvious why topics, interrogatives and exclamatives would form
a natural class, or why they would be more or less marked than relative
pronouns. However, from the point of view of information structure—which
seems to be the generic category conditioning this progression—interrogatives
are the topics of interrogative clauses, and exclamatives are the topics of
exclamatory clauses. Together with the simple declaratives, these are the basic
categories of main clause. Relative pronouns, on the other hand, are the
(functional equivalent of) topics in relative clauses. This part of the
progression, then, can be understood as a manifestation of the common
distinction between nouns M and pronoun U topics. The second part of the
 INTRODUCTION 13

progression, "relatives vs. clitic pronouns", represents another distinction in

information structure—the distinction between topical M and nontopical U
presupposed participants. The former are restricted to the relative clauses, the
latter, occur freely in all clauses—hence their presumptive markedness values.
Thus the progression can be restated as Topic object nouns M > Topic object
pronounsU and Topic pronouns M > Nontopic pronounsU.
In other cases it may be necessary to analyse observed linguistic
categories into more abstract, semantic categories that are not necessarily
directly expressed. A possible example is the progression identified in
Mithun's paper, "place" > "inside" > "under" > "beside" > "in the middle of" >
"beyond". Mithun's comparison of this with the widespread
grammaticalization gradient "on > in > under > beside > back > front"
established by Svorou (1994) (165) is very suggestive indeed. But perhaps a
more precise understanding of the progression of the Iroquoian development
can be attained if the locatives are resolved in terms of the cognitive
dimensions involved. One tentative analysis: adessiveU ("place") vs. inessiveM
("in"), contiguousU ("under", "beside") vs. noncontiguousM ("near"), singleU
point of reference (the preceeding locatives) vs. multipleM points of reference
(the following), relation to several referentsU ("between, in the middle of") vs.
to referent(s) and speakerM ("beyond").

2.3 Variation in Markedness values

Since the concept of Markedness was first introduced into linguistics,
linguists have envisaged the possibility that one and the same category may be
ascribed different values in different systems. In his reply to the letter in which
Trubetzkoy first adumbrated the notion of phonological markedness, Jakobson
pointed to the importance this concept would have also outside linguistics—for
anthropology and cultural history, for instance—and illustrated this with
several examples, one of them being the different values that may be ascribed
to life and death (1985:162).
More often, perhaps, an appearance of such system-specific Markedness
values arises when similar areas of purport (such as the three persons) are
defined by different categories in different languages (in casu, deixis or
participant role), as suggested in the discussion above, Section 2.1).
Sch0sler's contribution points to a detail in the history of French that
seems hard to reconcile with Markedness theory and may call for such
consideration. In the loss of case distinctions in Old French, human nouns
retain case longer than nonhuman nouns, but case loss proceeds earlier in
personal names than in appellatives. This is surprising if one assumes that
personal names are a subcategory of human nouns, as is usually done. But
14 HENNING ANDERSEN

perhaps this is not a necessary assumption. In fact, quite unlike common

nouns, names have no descriptive content (leaving aside occasional
connotative content); they function as individualized deictics (Zeno means "the
one named Zeno"). This suggests the possibility that personal names may be
categorized (in some languages or universally) as a subclass of pronouns—in
which case their role in the hierarchy of categories that condition the
progression of case loss in Old French is in accord with the Principle of
Markedness Agreement.

2.4 Alteration vs. loss

In Andersen (2001a:33-35) it is acknowledged that changes progress
differently depending on their character. As mentioned there, it may be the
case that change from above typically progresses from marked to unmarked
environments, change from below, the other way around. Clarification changes
similarly may typically spread from marked to unmarked environments,
whereas obscuration changes may typically be actualized earlier in unmarked
than in marked environments. The essential point in understanding these
differences is that innovative variants may be valuated differently—as marked
or unmarked in relation to established variants—in different kinds of change
and hence will be compatible with different environments, according to the
Principle of Markedness Agreement. See further Section 2.6.
In some instances an observed change can be interpreted in more than
one way. The loss of morphological case in Old French, for example, poses
such a dilemma. The development can be viewed as the introduction of the
unmarked case form into one environment after another, as Sch0sler suggests
(174). But then the progression does not correlate well with the Principle of
Markedness Agreement. Alternatively, the development can be viewed as the
loss of a grammatical distinction. If this is the essence of the change, one
would expect it to be actualized earlier in marked than in unmarked
environments, for it is commonly the case that distinctions combine more
readily with the unmarked than with the marked term of the same category;
this is the Principle of Compensation, first formulated by Br0ndal (1943:105;
cf. Andersen 1974, 1979; Battistella 1996:24 with further references). If this
consideration is valid, it would not be surprising to see the case distinction
linger longer in unmarked than in marked environments. This is the
interpretation I have put forward in the summary above (Section 1.5).
Similar considerations seem to be called for in interpreting the loss of
direct object agreement in Romance discussed by Smith (see Sections 1.7, 2.1).
Agreement is abandoned earlier when the controller follows than precedes,
earlier when it is a noun than a pronoun, earlier when it is a topic than when it
 INTRODUCTION 15

is presupposed, earlier when it is an exophoric person than third person, earlier

when it is reflexive than when not. In Gascony some dialects loose agreement
with relatives earlier when the antecedent is a noun than when it is a pronoun.
The one notable deviation from this marked-before-unmarked pattern is that
number agreement is lost earlier in the masculine than the feminine (Section
2.1).
Sch0sler describes the development of obligatory subject-person marking
in Old French, which progresses earlier in marked than in unmarked
environments (subordinate before main clauses, exophoric persons before
third, referential before nonreferential third person). It is tempting to view this
as another instance of loss—"the loss of Pro-drop" Sch0sler calls it. But
perhaps it should be understood as a clarification change.
It is interesting that both in the loss of case in Old French and in the loss
of object agreement in Romance, apparently, the changes progressed earlier in
unmarked styles and genres than in marked. This is consistent with their being
internally motivated changes.

2.5 Markedness in restoration and restitution

Bakken's paper (59-78) makes no explicit reference to Markedness, but
its discussion of the reintroduction of long-lost phonotactic combinations
naturally raises questions of motivation and, hence, Markedness. First of all,
the Norwegian data suggest that a distinction should be drawn between
'restoration' and 'restitution'.
There are restorations, in which the loss of a constraint (say, through
phonological reanalysis) allows underlying representations to resurface.
Restorations are typically grammatically conditioned in that 'original'
morpheme shapes are restored only in environments where they were subject
to alternation. In such internally motivated changes, where the (typically
gradual) curtailment of a rule allows formerly underlying (or intermediate)
representations to surface, the innovative variants are unmarked, and the
change results in simplification—first in the relation between underlying and
surface forms, eventually, when the rule has been curtailed and lexicalized out
of existence, of speakers' grammars.
Distinct from such changes are restitutions, such as those exemplified in
Bakken's paper, which ensue from contact with a closely related language
variety (dialect or sociolect) with pronunciation norms that happen to be
phonologically more conservative in some respect. It is natural for the linguist
who is familiar with the historical phonology of a dialect to focus attention on
the reintroduction of 'original' phonological shapes. But in reality such
restitutions owe their appearance of phonological unity largely or entirely to
16 HENNING ANDERSEN

the linguist's expert construal of the data and do not differ from other phoneme
substitutions in individual lexemes that may occur through dialect contact (cf.
Andersen 1988:40-44). Such a set of restitutions or substitutions is not a
phonological change—or even a single change in the sense of a bounded,
internally coherent historical event in the given community's tradition of
speaking. It is, properly speaking, just a subset of a series of individual
replacements of local word shapes with borrowed ones, part of a smaller or
larger relexification, motivated by the individual word shapes' greater utility in
interdialectal communication and hence defined in pragmatic and semantic
terms. The progression of such a relexification begins as an elaboration of
speakers' grammars, as elements of a local tradition of speaking are matched
with marked covariants appropriate for specified pragmatic purposes. It runs to
completion lexeme by lexeme, as the traditional elements one by one fall into
disuse, superseded by the borrowed, more widely used, more viable
alternatives. (See further Section 2.6.)
Bakken's data, and especially those obtained in recent informant
interviews, are a nice illustration of such a course of events.

2.6 Markedness and the S-curve

It is a well-known fact that when a change is in progress in a speech
community, the innovated element typically increases in frequency of
occurrence in a way that can be diagrammed with an S-shaped curve. For a
number of reasons the slow-fast-slow progressions captured by S-curves are
usually not observed in the textual record, the most obvious reasons being that
the written attestation does not conform to any consistent sampling method,
written usage is often a compromise between writing norms and spoken usage,
written texts may reflect a haphazard assortment of such compromises, and
some texts may be directly copied from or otherwise based on material written
by earlier writers and hence contain reflections of earlier norms of writing or
spoken usage. The historical linguist consequently has to make do with
whatever evidence can be drawn from the written attestation, but can
reasonably assume behind this attestation a regular S-curve progression in
spoken usage, as for instance Bubenik does (96).
It is often assumed that any innovative element (I-variant) is marked in
relation to the element it is called on to replace (O-variant). This assumption is
surely correct for some innovations (though not for all; see below). When a
marked I-variant is widely adopted and gains currency in a speech community,
it may, at some point, be judged to be unmarked—through a revaluation by
some speakers or a reanalysis by new cohorts of learners or a combination of
the two. If the I-variant was at first limited to specific environments, its
 INTRODUCTION 17

reinterpretation as unmarked allows it to spread to unmarked environments and

thus to gain ascendancy in community usage and eventually supersede the O-
variant.
The change in the variants' Markedness value is sometimes referred to as
a 'markedness shift'. Bubenik (96) credits Simon Dik with this term, but note
that in Dik's view markedness is identified with frequency; hence the
'markedness shift', in this conception, is merely the observable (or assumed)
change in the ratio of I-variants to O-variants, in the middle of the S-curve,
from 49% : 51% to 51% : 49% of usage tokens. In the restatement offered in
the preceding paragraph, by contrast, the Markedness shift is described as
occurring in speakers' grammars (revaluation) and between speakers'
grammars (reanalysis). This makes it possible to understand the observable
changes in relative frequency as actualizations (cf. Andersen 2001a:51; see
further p. 238).
In speaking of Markedness, however, it is essential to define the
categories with respect to which a given pair of variants is ascribed value. In
the case of such complex grammatical systems as the diglossic or polyglossic
grammars discussed in Bubenik's paper, evidently, distinct sets of
appropriateness norms exist for the diverse, complementary functions the
conjoined (partial) grammars conventionally serve. Accordingly, elements that
are O-variants (marked as obsolescent or obsolete) in the norms of a lower
diglossic grammar may be neutral and stable (unmarked) in a higher grammar.
And well-established, neutral elements of a lower grammar may be I-variants,
proscribed or marked as novel in the higher grammar.
It is very useful to see the historical development of such complex
linguistic traditions subjected to historical analysis, as in Bubenik's paper, not
least because they draw attention to the prominent role of convention in the
usage norms of natural languages. In this regard studies of diglossia offer an
instructive perspective on the often facile appeal to 'processing ease' as a
determinant of change, which is a recurrent theme in the history of the
discipline.

3. Conclusion
All the papers in this volume are contributions to the long-standing
dialog on the determinants of linguistic change.
Most of the papers exemplify the grammatical conditioning of attested
changes and evaluate the extent to which their progression can be interpreted
in terms of Markedness and the Principle of Markedness Agreement.
The focus on Markedness in the position paper to which most of the
Workshop papers responded naturally raised the question whether change is
18 HENNING ANDERSEN

better understood in a Markedness perspective, which views a grammar as a

kind of value system and change as a manifestation of the inherent
asymmetries of such systems, or from a speech-processing point of view,
which endeavors to explain change and the gradualness of change as resultants
of the relative strengths of encoding and decoding factors.
It is interesting to note that this disjunction, which has been debated off
and on for several generations now, is put in a new light in the philosophical
piece by Shapiro (196-200). Shapiro argues that language change belongs to
the category of 'finious processes' described by Charles S. Peirce, that is, the
nonmechanistic processes that "act in one determinate direction and tend
asymptotically toward bringing about an ultimate state of things" (198). In this
view, any finious process is the result of fortuitous variation plus a principle of
selection.
If this is a correct understanding of language change, then the disjunction
between speech-processing factors and Markedness relations as determinants
of change is resolved with a 'both-and'. Encoding and decoding conditions
produce the fortuitous fluctuation in speech production and perception which is
not change, but without which no change could occur. The usage categories
against which speakers evaluate this fluctuation allow it to be reanalysed as
variation. And the inherent asymmetry of these categories, their Markedness
values, provides the persistent skewing—the slant, as Sapir called it—that
transforms the variation into change as it is transmitted through successive
cohorts of speakers (cf. Andersen 1989, 1990, and 2001b:236-246).

REFERENCES
Andersen, Henning. 1974. "Towards a typology of change: bifurcating changes
and binary relations". Historical Linguistics. Proceedings of the First
International Conference on Historical Linguistics, I—II ed. by John M.
Anderson & Charles Jones, vol. II, 17-60. Amsterdam: North-Holland.
Andersen, Henning. 1979. "Phonology as semiotic", A Semiotic Landscape.
Proceedings of the First Congress of the International Association for
Semiotic Studies ed. by Seymour Chatman, 377-381. The Hague: Mouton.
Andersen, Henning. 1988. "Center and periphery: adoption, diffusion, and
spread", Historical Dialectology. Regional and Social ed. by Jacek Fisiak,
39-84. (Trends in Linguistics. Studies and Monographs, 37.) Berlin & New
York & Amsterdam: Mouton de Gruyter.
Andersen, Henning. 1989. "Understanding linguistic innovations". Language
Change. Contributions to the Study of Its Causes ed. by Leiv Egil Breivik
& Ernst Håkon Jahr, 5-28. Berlin & New York: Mouton de Gruyter.
Andersen, Henning. 1990. "The structure of drift". Historical Linguistics 1987.
Papers from the 8th International Conference on Historical Linguistics ed.
 INTRODUCTION 19

by Henning Andersen & Konrad Koerner, 1-20. (Current Issues in

Linguistic Theory, 66.) Amsterdam & Philadelphia: John Benjamins.
Andersen Henning. 1998. "Dialektnaja differenciacija obščeslavjanskogo
jazyka. Paradoks obščix tendencij razvitija s različnymi lokal'nymi
rezul'tatami". [The dialect differentiation of Common Slavic. The paradox
of shared tendencies of development having distinct local outcomes.]
American Contributions to the Twelfth International Congress of Slavists.
Cracow, Aug-Sept. 1998. Literature. Linguistics. Poetics ed. by Robert A.
Maguire & Alan Timberlake, 565-600. Bloomington, Ind.: Slavica
Publishers.
Andersen, Henning. 2001a. "Markedness and the theory of change". In this
volume, 21-57.
Andersen, Henning. 2001b. "Actualization and the (uni)directionality of
change". In this volume, 225-248.
Bach, Adolf. 1950. Deutsche Mundartforschung. Ihre Wege, Ergebnisse und
Aufgaben. (Zweite Auflage.) Heidelberg: Carl Winter Universitätsverlag..
Battistella, Edwin L. 1996. The Logic of Markedness. Oxford: Oxford
University Press.
Borkovskij, Viktor I. & Petr S. Kuznecov. 1965. Istoriceskaja grammatika
russkogo jazyka. [A Historical Grammar of Russian.] (Second edition.)
Moskva: Izdatel'stvo Akademii Nauk.
Br0ndal, Viggo. 1943. Essais de linguistique générale. Copenhagen:
Munksgaard.
Dik, Simon . 1989. The Theory of Functional Grammar. Dordrecht: Foris.
Greenberg, Joseph H. 1966. Universals of Language. With special attention to
feature hierarchies. (Janua Linguarum, Series Minor, 59.) The Hague &
Paris: Mouton.
Jakobson, Roman (ed.). 1985. N. S. Trubetzkoy's Letters and Notes (Janua
Linguarum, Series Major, 47.) Berlin & New York & Amsterdam: Mouton.
Klenin, Emily. 1983. Animacy in Russian. Columbus, Ohio: Slavica
Publishers.
Klenin, Emily. 1997. Review of Krys'ko 1994. Russian Linguistics
21.108-120.
Krys'ko, Vadim B. 1994. Razvitie kategorii odusevlennosti ν istorii russkogo
jazyka. Moscow: Liceum.
Saxmatov, Aleksandr A. [1911] 1957. Istoriceskaja morfologija russkogo
jazyka. Moscow: Gosudarstvennoe učebno-pedagogičeskoe izdatel'stvo
Ministerstva prosveščenija RSFSR.
Smith, John Charles. 1996. "Surfonctionnalité et hyperanalyse: l'accord du
participe passé dans les langues romanes à la lumière de deux théories
récentes". Faits de langues 8.113-120.
Smith, John Charles. 1997. "Types and tokens in language change: some
evidence from Romance". Language History and Linguistic Modelling: a
Festschrift for Jacek Fisiak on his 60th birthday ed. by Raymond Hickey
& Stanisław Puppel, 1099-1111. (Trends in Linguistics—Studies and
Monographs, 101.) Berlin & New York: Mouton de Gruyter.
20 HENNING ANDERSEN

Svorou, Soteria. 1994. The Grammar of Space. Amsterdam & Philadelphia:

John Benjamins.
Timberlake, Alan. 1981. "Dual reflexes of *dj and a morphological constraint
on sound change". International Journal of Slavic Linguistics and Poetics
23.25-54.
 MARKEDNESS AND THE THEORY OF LINGUISTIC CHANGE

HENNING ANDERSEN
University of California, Los Angeles

L'opposition dans les faits linguistiques n'est pas un

schéma que la science introduit pour maîtriser les faits, et
qui resterait extérieur à ceux-ci. Son importance dépasse
l'ordre épistémologique: quand la pensée linguistique range
les faits d'après les principes d'opposition et de système,
elle rencontre une pensée qui crée ces faits mêmes. (Pos
1938:246)

0. Introduction
0.1 Preamble
To our structuralist predecessors our generation of linguists owes a rich
heritage of technical vocabulary we use every day.1 Much of this vocabulary
was created for specific purposes, in specific structuralist theories, and was
provided with more or less precise definitions, some terms even being defined,
in true structuralist fashion, in relation to other terms. But in contemporary
linguistics many of these lexemes have lost their status as terms and are used as
common-parlance words. Among these is the term markedness, and the
correlative terms marked and unmarked, which were coined (first in Russian and
German; see below) and defined by Trubetzkoy and Jakobson in 1930 (cf.
Jakobson 1985:162). That these lexemes are now common-parlance words is
shown by the fact that they are used entirely in accordance with the principle of
cooperation—you can use the word markedness freely without anyone
demanding that you define what you mean by it. And if asked, most linguists are
quite content with an informal characterization of, say, unmarked that equates it
with approximate synonyms such as simple, common, basic, default and easily
agree on a shared understanding of markedness as 'relative complexity or
frequency' or, on a more abstract level, 'a sort of asymmetrical relation'.

l
This work was supported by research grants from the John Simon Guggenheim Memorial
Foundation and the President of the University of California.
22 HENNING ANDERSEN

0.1.1. An early indication that markedness had changed status from technical
term to everyday word is, perhaps, Joseph Greenberg's monograph Universals
of Language (1966). Here markedness is treated as a 'found object': it is
described as a "Protean notion" that is acknowledged as being familiar to
everyone, but thought to be badly in need of a definition. In other words, the
monograph starts not from the understanding that markedness is a formal
principle that deductively explains a variety of observed phenomena in
language—which was the way Trubetzkoy and Jakobson first grasped it—but
from the idea that the word markedness presumably refers to some characteristic,
present in all the different observables that linguists intuitively recognize as
instantiations of markedness—a single, unifying criterion that would serve as a
guide in determining the markedness attributes of any observed phenomena, and
which one might discover through an inductive search through the different
instantiations of markedness.
As you may recall, Greenberg's search for such a single criterion was not
successful, and as a consequence his monograph failed to resolve the conflict
between the two predominant attitudes to markedness—in the sixties, when he
wrote, as well as now, thirty-odd years later. One of these is to consider
markedness simply a handy label for a large number of disparate observables;
the other is to regard it as a hazy (non)concept that contributes nothing to
linguistics. Greenberg, in the end, concluded that markedness attributes can at
best be identified on the basis of a number of indications such as these: (a)
unmarked terms often occur in positions of neutralization; (b) unmarked terms
usually have greater relative text frequency; (c) unmarked terms show more
allophonic or allomorphic variability; (d) unpaired phonemes are common in
marked phoneme classes, and syncretism, in marked categories; (e) unmarked
terms are often indicated by the features of basic allophones in phonology and by
agreement a potiori in morphosyntax (1966:58-59). For a detailed analysis and
critique of Greenberg's contribution, see Andersen (1989a).

0.1.2. More recent literature on markedness only serves as further illustration

that the word has lost its terminological status. For example—to take a linguist
who makes extensive use of the word—in Givon's writings, say, the two-
volume Syntax (1984, 1990), markedness is not given a precise definition, but
serves as a cover term for the range of substantive phenomena in which it is
manifested, mainly complexity of expression, relative frequency, cognitive
complexity (1990:945-966). Battistella's theoretically oriented monograph, The
Logic of Markedness (1996) reports on numerous definitions and
characterisations the word has been given since 1930. In this way, this work is a
useful contribution to a history of the lexeme markedness. But despite its title,
 MARKEDNESS AND THE THEORY OF CHANGE 23

the monograph does not offer anything like a logical analysis of what
markedness is. It does not even analyse the differences between structuralist
usage and poststructuralist usage.

0.2 Issues in 'markedness'

It seems to me that the existing literature on markedness from Greenberg
(1966) to the present conflates a number of issues concerning this notion which
can only be clarified if the issues are distinguished and approached one by one.

0.2.1. One issue is the synchronic one of what the word markedness means.
This issue can be resolved only through an essentially lexicographic expedition
through the current literature that would register the actual use to which the
words marked, unmarked, and markedness are put and classify the words'
referents. The outcome of such an undertaking would be useful as a purely
descriptive stock-taking. It might perhaps be supplemented with judgements by a
usage panel, in the style of the American Heritage Dictionary, which could
establish what are customary (or appropriate), and what are unusual (or
inappropriate) uses of these words according to the understanding of a
representative sample of practicing linguists. Much of what is in Battistella
(1996) can be appreciated as a first step towards such a study.

0.2.2. Another issue is a diachronic one, which calls for an investigation of

several strands of development that would trace (i) the history of the words and
terms for markedness and (ii) the history of the notion.
The former effort would recognize the varying terminology of different
schools (e.g., the Copenhagen School's intensive vs. extensive distinction) and
periods (e.g., Gm. merkmalhaft, Russ. priznakovyj (1930s) > Gm. markiert,
Russ. markirovannyj (1950s and later), both, "marked"; cf. Jakobson [1971]
1971b). It would also pay attention to ways of speaking of markedness in the
pre-terminological period, say, in nineteenth-century European linguistics or in
medieval Arabic linguistics (see Owens 1988:199-220).
The history of the notion of markedness, by contrast, would trace different
conceptions back in time beyond the explicitly named stages in the recent history
of grammatical scholarship. It would pay attention, for instance, to such implicit
recognitions of markedness as the organization of morphological paradigms in
grammatical texts. Recall the traditional presentation of verb morphology in our
Latin grammars, where the order of forms in paradigms mirrors the markedness
relations of the respective categories in that singular (u) forms precede plural ()
forms, the forms of the present (u) tense precede those of the preterite (), the
paradigms for these two historical (u) tenses precede those for the future (), the
24 HENNING ANDERSEN

tenses of the infective (U) aspect are presented before those of the perfective (),
the indicative mood (u) before the subjunctive (), and so on; cf. Table 1.
Comparable principles of organization can be discerned in bilingual
Sumerian-Akkadian grammatical texts from Babylon (1900-1600s B.C.); see
Jakobsen (1974). Grammatical texts in other ancient grammatical traditions await
exploration.

Unmarked Marked
Singular Plural number
present preterite tense
historical tenses: present, preterite future tense
infective perfective aspect
indicative subjunctive mood
descriptive moods: indicative, subjunctive directive mood: imperative
finite forms nominal and adverbial forms: infinitive;
participles, gerundive; supines, gerund
Table 1.

0.2.3. Finally, there is the analytic issue, which calls for an examination of the
logical nature of markedness. On this issue, it seems we have been beating about
the bush for most of this century. Looking back, one can see that among the
structuralists, Hjelmslev alone had a precise understanding of markedness
(1939:87; see Section 3.4.4 below), but he had no impact on the mainstream of
linguistics at the time. Jakobson, on the other hand, who throughout his
scholarly career served as an authority on markedness, consistently defined
markedness in logically incoherent terms (cf. Section 3.2). When you compare
Givon's characterization of markedness with Greenberg's, and Battistella's with
Jakobson's, you have to conclude that the poststructuralist period has produced
no advance in the clarity of this notion.

0.2.4. I want to return to the analytic issue towards the end of this presentation
(Chapter 3), but I think it will be useful to give priority to some examples of the
manifestations of markedness in synchrony (Chapter 1) and in diachrony
(Chapter 2).

1. Markedness in synchrony
Much skepticism has been expressed in this century about the utility of the
notion of markedness.
It is my impression that this skepticism has been characteristic, first of all,
of linguists who by virtue of their personal cognitive style or academic training
 MARKEDNESS AND THE THEORY OF CHANGE 25

(or both) are skeptical of anything that cannot be directly observed and tend to
adopt what you might call a nominalist attitude to language and language
description. To anyone who sees linguistic description as essentially a way for
the linguist to organize his data—rather than as a hypothesis about the
competence of the speakers of the language—markedness can easily seem
readily dispensable. But secondly, many linguists who have understood
markedness as primarily the difference in relative text frequency of opposites, or
as a concept covering this and a number of other observable phenomena, have
felt justified in considering the notion (and the word) redundant. A particularly
eloquent discussion of markedness from these two points of view is contained in
Roger Lass's On Explaining Language Change (1980).
I would like to shift attention from the question of the utility of markedness
as a theoretical concept in linguistics to the reality of markedness as a principle of
cognitive organization that is reflected in human behavior and apparently
fundamental to it. It is only in this sense of markedness that we can appreciate
the analytic thinking of our Sumerian and Akkadian-speaking colleagues
working in applied linguistics almost 4000 years ago. I hope the examples I
present here will be understood in this spirit.

1.1 Ritual
I begin with an example of the manifestation of dual symbolic
classification in ritual.
On the island of Roti (an island southwest of Timor in Indonesia; see Fox
1973), the usual course of events in a funeral can be summed up briefly as
follows (bracketed numerals refer to the terms in Table 2. First the coffin is
brought to the house of the deceased amidst great uproar [1] and is put down
outside the house, parallel to it at its west end. The coffin is then raised, carried
under the roof [3], through the forecourt, and up the ladder into the house [5],
which stands on piles. The corpse is laid out in the men's [7] (the eastern [9])
half of the house, its head [11] to the east [9] and feet to the west [13]. The
corpse is then placed in the coffin with the same orientation, and the mourners
are admitted to the house [1]. Subsequently the coffin is brought down into the
forecourt, where it may be rested on the east side [9], still with the same
orientation [13]. The deceased is then carried out of the house, feet first [12], and
in this way the body is conducted in a noisy stampede [1] to the grave. At the
side of the grave, which has been dug running east to west [13], the coffin is
turned so that the corpse is headed [11] westward [10], and in this position the
coffin is lowered into the grave for the decedent's journey to the land of the dead
in the west [10].
26 HENNING ANDERSEN

A different rite is used when someone has died a bad death—by drowning,
falling from a tree, being stabbed, gored by a buffalo, or mauled by a crocodile,
or in childbirth. The deceased who has died such an inauspicious death [20] is
not brought into the house proper, but either placed outside [4] or in the
forecourt [6], but in this case on the west side [10]. A woman [8] who has died
in childbirth remains in the women's half of the house [10], but the body is laid
out north to south [14]. No mourners are admitted to the house [2].
Subsequently the coffin with her corpse is carried down from the house and
rested on the west side of the house [10]. The coffin of the inauspicious decedent
is then carried out and to the grave, head first [11]. The graves of the "bad dead"
are dug running north to south [14], and the coffin is lowered into the grave
headed north [16].

Unmarked Marked
Ceremony [1] No ceremony [2]
inside [3] outside [4]
house [5] forecourt [6]
man [7] woman [8]
east [9] west [10]
head [11] feet, tail [12]
east—west [13] north—south [14]
south [15] north [16]
right [17] left [18]
auspicious [19] ominous [20]
Table 2.

I have omitted several interesting details and elements of interpretation, but

it is clear enough that the two alternative rites have one thing in common: each is
composed almost exclusively of terms that are equivalent in markedness,
unmarked in the case of the good death, marked for the bad death. The few
exceptions to this homogeneity in markedness have substantive motivation. This
is the case with the orientation of the body when it is carried to the grave: we
usually enter this world head first, and so it is 'natural' that we should leave it
feet first [12]. It is in an inversion of this 'natural' order that the 'bad dead' are
carned to the grave head first [11].
The funeral rites on Roti are one of many examples of ritual behavior in
which a series of symbolic elements are concatenated predominantly or wholly
on the basis of their markedness values. Similar patterns of rule-governed
behavior have been observed and described all over the world since the
pioneering work of Herz (1909, cf. R. Needham 1973, 1979; a useful collection
of references to the use of dual symbolic terms in ritual and on the typology of
 MARKEDNESS AND THE THEORY OF CHANGE 27

systems of binary symbolic classification can be found in Ivanov & Toporov

1974:259-305; see also Andersen 1991:94-97). Such rule-governed behavior
everywhere manifests systems of opposites, correlated as in Table 2 according to
their positive or negative values. In ritual behavior the opposite terms of the
given symbolic categories are concatenated in such a way as to maximize
syntagmatic combinations that are homogeneous in markedness. In the following
I will speak of this homogeneity as 'markedness agreement'.

1.2 Texts
Readers who have little acquaintance with ritual behavior in exotic cultures
may be more familiar with the complementary categories of yang and yin that
used to regulate formal behavior in traditional China, and which are fundamental
to Chinese science and philosophy. Here yang (U) is correlated with light,
warmth, male, day, sun, heaven, east, south, hard, light, strong, before, above,
left, life, noble, joy, wealth, honor, celebrity, love, and profit whereas yin () is
linked with darkness, cold, female, night, moon, earth, west, north, soft, heavy,
weak, behind, below, right, death, common, sorrow, poverty, misery, bitterness,
ignominy, rejection, loss (J. Needham 1954:293-304). The similar role played
by pairs of polar opposites in ancient Greek science is perhaps less well known.
The earliest attributed record is probably the Table of Opposites of the
Pythagoreans—it links limit (u) with odd, one, right, male, at rest, straight, light,
good, square, and limitless () with even, many, left, female, moving, curved,
darkness, evil, oblong. It is remarkable to observe, in the early development of
Greek scientific thinking, how much intellectual effort was expended in attempts
to reconcile observations in medicine and the natural sciences with the
correlations of values that were part of this traditional, implicit understanding of
the order of things, and how slow was the process of emancipating observation
from these preconceived correlations (cf. Lloyd 1966).
In our own time and culture, the extent to which our everyday behavior
conforms to such schemes of binary symbolic values may escape our awareness,
but it is no secret to the observant anthropologist (Sahlins 1978). But perhaps
these schemes are nowhere more pervasively documented than in our literature,
in which semiotic space is organized by such oppositions as up vs. down, above
vs. below, distant vs. near, spacious vs. confined, movement vs. immobility,
freedom vs. slavery, culture vs. nature, creativity vs. fossilized forms, harmony
vs. disharmony (cf. Lotman 1970:275). The deployment of these paradigms of
values moulds the composition of the plot, the interaction and development of
characters, the selection of settings for the action, and the sequence of themes in
the narrative structure (cf. Jakobson & Pomorska 1983:107). The coherence and
verisimilitude of the world in which the fictional action occurs is largely—in
28 HENNING ANDERSEN

trivial literature, entirely—a product of the orderly concatenation of such

correlative values.
On the more concrete, linguistic levels of textual organization of literary
texts, written as well as oral, we have the poetic constraints of lexical and
grammatical parallelism on one hand and, on the other, the categories of
phonological parallelism such as meter, rhyme, alliteration, and assonance. These
are constraints that produce homogeneous syntagms of lexical or grammatical
semantic features or of prosodic or segmental features. In many works, genres,
or periods the constraints governing the prosodic and segmental features operate
independently of any constraints governing grammatical or lexical ones, and vice
versa. But if we tentatively adopt the hypothesis that text organization, like ritual,
is governed, to some extent, by a Principle of Markedness Agreement, all we
need do is recognize that literature, oral as well as written, poetry as well as
prose, conforms to this principle on certain levels of structure in accordance with
conventions that may be more or less culture or period specific.

1.3 Discourse
Turning from 'ritualized speech' to normal (narrative) discourse, we recall
that clauses are of varying degrees of transitivity, as Hopper & Thomson have
demonstrated (1980).

High transitivity (u) Low transitivity (M)

Two or more participants One participant
action no action
telic atelic
punctual nonpunctual
volitional nonvolitional
affirmative negative
realis irrealis
agent high in potency agent low in potency
object totally affected object unaffected
object individuated object nonindividuated
Table 3.

As is well known, Hopper & Thomson drew up a list of semantic

categories or features (see Table 3) and pointed out that whenever a language has
a constraint on the combination of these features in the form of an obligatory
pairing of two transitivity features, "the paired features are always on the same
side of the high—low transitivity scale" (1980:254). For example, the perfective
aspect of action verbs (telic, u) may correlate with the definite object (u), but not
 MARKEDNESS AND THE THEORY OF CHANGE 29

the imperfective (atelic, ) with the definite object (u), nor the perfective (telic, u)
with the indefinite object ().
Interestingly, Hopper & Thomson's wider findings transcend grammar
rules: their study of running text shows that there is an overwhelming
predominance of high transitivity features in foregrounded text portions and of
low transitivity features in backgrounded material. As they put it, "grounding
itself reflects a deeper set of principles—relating to decisions speakers make, on
the basis of their assessment of the hearer's situation, about how to present what
they have to say" (295).
What Hopper & Thomson's findings suggest is that in the casting of
conceptual representations—prior to the action of any linguistic formation rules
(in the sense of Chafe 1970)—humans select and combine conceptual categories
(admittedly with a fair degree of freedom of choice and in accordance with their
communicative intentions) by and large in an orderly fashion, so that the
resulting linguistic representation—by its clustered distribution of unmarked and
marked categories—diagrams the distinction between backgrounded and
foregrounded material in the speaker's conceptual representation. Backgrounded
and foregrounded portions of a text are what they are, and are cognized as such,
because they are comprised of largely homogeneous syntagms of features of
transitivity. From the encoding point of view, they are formed the way they are,
presumably, because wherever the speakers' communicative intentions leave any
category unspecified, the values of that category are assigned by default in
accordance with the Principle of Markedness Agreement.

1.4 Agreement
Greenberg mentions, as one of the manifestations of markedness,
'agreement a potiori', the special cases in which agreement conflict is resolved in
favor of an unmarked category, as in Sp. cuello i camisa blancos (1966:60).
Normally linguists describe gender agreement entirely in substantive terms—in
terms of specific genders—masculine agreeing with masculine, feminine with
feminine, etc. But if the special case of agreement a potiori is to be understood in
terms of markedness, then we should recognize that all agreement patterns can be
so described, and that in fact it is simpler to describe all agreement patterns in the
same terms. If they are so described, it is clear that in normal agreement in case,
number or gender, the rules produce syntagms that are homogeneous in
markedness, that is, conform to the Principle of Markedness Agreement.

1.5 Allomorphy
Similar homogeneous syntagms are generated by rules of allomorphy. An
alternation can be thought of as a paradigm of allomorphs comprising one or
30 HENNING ANDERSEN

more derived () allomorphs and one basic (u) allomorph. The contexts across
which the allomorphs are distributed form another, correlated paradigm
comprising one or more specified () environments, defined in phonological,
morphosyntactic, or lexical terms, and an elsewhere (u) environment. This being
so, the effect of rules of allomorphy is to assign marked allomorphs to marked
contexts and unmarked (basic) allomorphs to unmarked (elsewhere) contexts.
For examples, see Andersen (1980).

1.6 Allophony
It is obvious that rules that assign allophonic features work exactly the
same way. To take the most pedestrian of examples, in (some varieties of)
American English, for instance, vowels are specified as [+nasal] () before
[+nasal] (M) consonants and [-nasal] (u) elsewhere (u). And velar plosives are
assigned different degrees of the [front] feature ( for velars) before [front]
vowels, but none (u) elsewhere (u). Similarly, when a distinctive feature is
neutralized, and its opposite values are assigned in complementary distribution:
in Russian, for example, the voicing distinction is neutralized in any obstruent
followed by another obstruent or by a phrase boundary, a word boundary, or an
enclitic boundary. In these environments obstruents are specified as [+voice] ()
when the next following segment is [+voice] (), but [-voice] (u) otherwise (u),
that is, if the next segment is a [-voice] obstruent or a sonorant or a vowel and
before pause.

1.7 Conclusion
We are led to conclude that in ritual, in the thematic and plot structure of
texts, in lexical, grammatical, and phonological parallelism, in the grounding
structure of narrative discourse, and in the regularities of morphosyntax,
morphophonemics, and phonology, syntagmatic structures are commonly
formed in accordance with one and the same Principle of Markedness
Agreement.
The manifestation of this principle in allophonic rules was observed by
Schachter in (1969). I myself drew attention to the phenomenon in Andersen
(1968) and called it "markedness assimilation". But it seems it was first descried
by František Mareš, who proposed the generalization that in all allophonic
change, phonemes develop marked allophones in marked environments (1952).

2. Markedness in diachrony
Mareš's generalization, whether one calls it markedness assimilation or
not, is evidently the dynamic counterpart to the synchronic markedness
agreement observed in established rules of allophony. It is natural to ask if the
 MARKEDNESS AND THE THEORY OF CHANGE 31

actualization of other kinds of linguistic change can be understood as similarly

governed by the Principle of Markedness Agreement. If so, we should be able to
observe, in the progression of such changes, that as a linguistic innovation gains
currency and is generalized in a language, the process of actualization conforms
to the Principle of Markedness Agreement in that the innovated element is
favored first of all in marked environments, if the innovated element is marked,
but in unmarked environments if it is unmarked.
In fact, it has been known for some time that many kinds of linguistic
innovation are actualized, if not in precisely this manner, at least in part along
these lines.

2.1 Phonology
In phonology, for instance, allophonic fortition occurs earlier in stops (u)
than in fricatives (), earlier in coronals (u) than in back consonants ()
(Andersen 1972:17; cf. Zabrocki 1951, Back 1989). In velars, palatalization
occurs earlier in stops (u) than in fricatives (), earlier before high (u) than
before non-high () vowels, earlier before unrounded (u) than before rounded
(M) vowels, earlier directly contiguous to the conditioning vowel (u) than across
another segment (), earlier before (u) than after () the conditioning vowel,
earlier in the narrow domain of the syllable (u) than across syllable boundaries
(that is, in the wider domain of the word) (), earlier in stable environments (u)
than in alternating environments (M) (cf. Timberlake 1981, Andersen 1998).

2.2 Case marking

In morphosyntax we find similar examples. Timberlake (1977)
investigated the Russian change in case marking of direct objects in negative
sentences. He found that the older use of the genitive is giving way to the use of
the accusative (which is regular in affirmative sentences) in an ordered
progression such that the accusative occurs earlier and more widely in proper

Unmarked Marked
(a) Proper Common
(b) human non-human
(0 animate inanimate
(d) concrete abstract
(e) singular plural
(f) definite indefinite

Table 4.
 32 HENNING ANDERSEN

nouns than in common nouns ((a) in Table 4), earlier in nouns denoting humans
than in other nouns (b), earlier in animates than in inanimates (c), earlier in
concrete nouns than in abstract nouns (d), earlier in singulars than in plurals (e),
earlier in definite than in indefinite noun phrases (f). Timberlake was able to
subsume these different categories under the abstract semantic label of
individuation. But in addition to this substantive characterization, he identified
the features favoring innovation as unmarked (162), as in Table 4.

2.3 Morphosyntax
In my own study of the development of the Polish enclitic auxiliary
paradigm into bound person-and-number markers (1987, 1990), I observed that
agglutination of these markers to verb stems occurred earlier in the present tense
(of być "be") than in the preterite ((a) in Table 5), earlier in the (present or
preterite) indicative than in the conditional mood (b), earlier in the first persons
than in the second persons (c), and earlier in singular than in plural forms (d); the
initial displacement of the clitics from Wackernagel's position, as they
(statistically speaking) drifted rightward in sentences, occurred earlier in main
clauses than in subordinate clauses (e), earlier in asyndetic clauses than in
clauses with a conjunction (f), and earlier when the initial constituent was a
lexical NP than when it was a pronoun (g); the concatenation of the earlier
enclitics with the former participles, which now are past-tense stems, occurred
earlier in main clauses than in subordinate clauses (e), earlier in prose than in

Unmarked Marked
(a) Present Preterite
(b) indicative subjunctive
(c) 1st person 2nd person
(d) singular plural
(e) main clause subordinate clause
(f) asyndetic syndetic
(g) lexical NP pronoun
(h) prose poetry
(i) expository artistic
(j)) speech writing
(k) casual formal

Table 5.

poetry (h), and earlier in expository than in artistic prose (i), and is still more
frequent in speech than in writing (j) and more common in casual than in formal
style (k).
 MARKEDNESS AND THE THEORY OF CHANGE 33

In Timberlake's study, the data presented a clear choice, in explaining the

orderly progression of the change, between a substantive feature, the degree of
individuation, and the formal principle of markedness. In the Polish development
summed up here, by contrast, which is documented in much more detail, some of
the substantive categories are mophological—(a) to (d), some involve features of
information structure related to grounding distinctions—(e) to (g), some are
genre categories—(h) and (i), one is a distinction between media—(j), and one,
between styles—(k). The only generalization that these data will support is
evidently that throughout this long drawn-out development, which started
perhaps eight hundred years ago, and which is far from completed, unmarked
environments have been hospitable to the three kinds of innovation mentioned
here earlier than the corresponding marked environments.

2.4 Different pragmatic motivation

It is important, however, to acknowledge types of change with different
actualization patterns. In phonology, it has been known for some time that
lenition and other obscuration innovations are favored by unmarked
environments, whereas clarification innovations are favored by marked
environments. Vowel reduction and syncope, for example, arise and gain
acceptance first in casual styles, whereas diphthongization and epenthesis arise
first in maximally explicit styles (see Dressier & Drachmann 1977, Dressier
1980)
Also the established distinction between internally and externally motivated
changes is important. Evolutive changes and contact changes appear to follow
opposite paths or at least partly different paths of actualization. Romaine, for
example, has shown that when wh-relativization was adopted into Scots English,
it was established first in the most complex () styles and in the least frequently
relativized syntactic positions () in the case hierarchy (1982). Similarly,
Fischer, in her study of the accusative-cum-infinitive construction in English, has
shown that this, too, was manifested in the most salient environments () first
(1992).
These and similar differences in actualization can be understood in terms of
the traditional distinction between grammatical system and usage rules—
Coseriu's system and norms (1952, 1965). If we assume there is a distinction in
any speaker's grammar between an internally coherent structure of productive
rules and an additive system of usage rules, then the observed differences in
actualization can be understood in terms of the source or motivation of different
changes. In the internally motivated, evolutive change, perhaps, the usage rules
are gradually adjusted to incorporate an innovation that is unmarked in relation to
the productive rules of the core grammar, and which is first admitted to
34 HENNING ANDERSEN

unmarked environments; only as the innovation loses its novelty does it spread
from unmarked contexts to marked contexts. In the externally motivated change,
by contrast, usage rules are presumably directly modified to conform to the
external model; the innovation is pragmatically motivated and occurs first in the
most salient, most monitored, marked environments, from which it may spread,
as it loses its novelty, to less salient, unmarked environments.

2.5 Open questions

There is much more that we need to learn about actualization, as the
following example of lexical borrowing illustrates.
The often-cited English borrowings from Norman French, beef, veal, pork,
etc., enter into obvious markedness relations with the native lexemes: the
borrowed words for kinds of meat are marked in relation to the unmarked native
words, in terms of which they are most naturally explicated; cf. Table 6.

Unmarked Marked
Ox Beef
calf veal
pig pork
sheep mutton
deer venison
swan cygnet

Table 6.

In a comparable instance of borrowing, in the Spanish creole of

Zamboanga in the Philipines, "where a Filipine and a Spanish-derived form
participate in a marked vs. unmarked relation in the same contrast set, the Filipine
form designates the marked category: it ... signifies lesser magnitude, shorter
distance, worse evaluation, female sex, junior generation, or polarity" (Frake
1971). See Table 7.
Both in Table 6 and in Table 7 we are dealing with what must at first have
been stylistic lexical variation. One can understand why in English the Norman
borrowings would have been codified, at first, as upper-class culinary terms.
They are an example of 'change from above', but their semantic specialization, it
would seem, fully explains the modern-day markedness relations. In the case of
Zamboangueño, one can correspondingly suppose that the codification of the
Spanish borrowings for unmarked categories reflects the fossilized, positive,
stylistic valuation of Spanish vocabulary. But why this positive valuation was
associated first or only with the unmarked members of the diverse categories
 MARKEDNESS AND THE THEORY OF CHANGE 35

mentioned in Table 7—Was this perhaps a change from below?—must remain

an open question, at least for the time being.

2.6 Conclusion
In any case, it is clear enough from instances where there is adequate
historical evidence, such as the Russian and Polish examples presented in
Sections 2.2-2.3 above, that when an innovation is generalized across the

a. Zamboangueño adjectives of polarity, potency, and evaluation

grande "large" dyútay "small"
?dlto "tall" pandak "short"
lihéro "fast" mahinay "slow"
kórre "fast" páta? "slow"
mapwérso "strong" malúya "weak"
?drde "bright" ?amamalun "dim"
?apretdo "tight" haluga? "loose"
?agúdu "sharp" mapurul "dull"
lisu "smooth" makasap "rough"
sabróso "tasty" mata?ban "tasteless"
dúlse "sweet" mapa?it "bitter"
madúru "ripe" mihilau "raw"
mánso "tame" ma?ilap "wild"
buníto "pretty" ?umálin "ugly"
limpyo "clean" bulin "dirty"
kláro "clear" lubug "turbid"
deréčo "straight" tiku? "bent"
balyénte "bold" mahuya? "shy"
?umilde "modest" hambuk "vain"
byého "old" báta? "young"
nwébo "new" da?an "old"
b. Zamboangueño nouns contrasting in generation, age, or sex
tdta/nána "father/mother" ?anak "son/daughter"
lólo/lóla "grandfather/-mother" ?apu "grandchild"
soltéro "bachelor" dalága "unmarried girl"
plóres "blossom" putut "bud"
?ohas "mature leaf" talbus "young leaf"
 Zamboangueño neutral pronouns, singular and plural
yo "I" kanú : kitá "we (EXCL) : (INCL)"
?éle "he" sliá "they"
d. Zamboangueño second person pronouns, degrees of respect
?usté, ?ustédes "2SG, 2PL; polite" ?ebós, kamó "2SG, 2PL; familiar"
tu, bosótros "2SG, 2PL; neutral"
Table 7.
36 HENNING ANDERSEN

grammatical, lexical, stylistic, and social categories of a language, this may occur
as an orderly progression in which the only common denominator for the diverse
categories involved is markedness.
There is hardly any way to understand this without assuming, first, that in
the case of any successful innovation, the speakers—or a majority of
speakers—are in tacit agreement regarding the value they ascribe to the
innovative variant vis-à-vis its traditional alternative (cf. Andersen 1989b:23,
25). If they are not, the innovation will not gain currency.
Secondly, one must assume that in speakers' grammars all the different
categories that are relevant to the generalization of an innovation are associated
with one another in terms of markedness values.
Thirdly, one must assume that it is something like the proposed Principle
of Markedness Agreement that allows the innovation to occur earliest in
environments with equivalent markedness value and to subsequently gain
ascendancy first in such contexts and then, as it loses its novelty, in the
complementary contexts with opposite markedness value.
The big question in historical linguistics is how the individual speakers
who acquire a community language can know or infer all the multifarious
parameters of variation that they need to master in order to function as full-
fledged members of the community.
It seems that the orderly progression of such well-documented changes as
the Polish one I mentioned in Section 2.3 holds the answer to this question. The
progression can be modeled as a series of step-by-step modifications of variable
rules, and hence it presupposes the formation and existence—in each speaker's
competence, at any time during the progression of the change—of a
comprehensive network of association that readily relates unmarked terms with
unmarked, and marked with marked terms across categories, in part without
regard to the substantive character of the categories, in part, apparently,
constrained by reference to the substantive content of some categories.
In supposing that such a network of association is part of every speaker's
competence, let us acknowledge that we are not going beyond what has
traditionally been assumed. For this has been the standard assumption of
grammarians and linguists since antiquity. This assumption is implicit in the
ancients' understanding of proportional analogy; it is explicitly described in the
1800s, for instance, by the great neogrammarian theoretician Hermann Paul (cf.
[1881] 1970:26-27, 106-109); it is explicated in Ferdinand de Saussure's
multidimensional "rapports associatifs" (1916:252-263); and it was
rediscovered—and restated in semiotic terms—by Roman Jakobson, who spoke
of the '"system of diagrammatization', patent and compulsory in the entire
syntactic and morphological pattern of language" ([1966] 1971:357).
 MARKEDNESS AND THE THEORY OF CHANGE 37

But by integrating this understanding with the proposed Principle of

Markedness Agreement, I think we can take up the challenge Sapir posed in
Language, when he insisted on the need to study the intuitional bases of speech
and asked the rhetorical question, "How can we understand the nature of the drift
that frays and reforms ... [grammatical] patterns" unless we study "patterning as
such and the 'weights' and psychic [= cognitive] relations of the single elements
... in these patterns?" (1921:183; Shapiro 1991:48ff.).
Sapir's rhetorical question implies another question: How can we study
"patterning as such" unless we have a theory of the "'weights' and psychic [=
cognitive] relations of the single elements" in the patterns we observe? The
hypothetical Principle of Markedness Agreement implies precisely such a theory,
and it is my hope that in testing this principle we can follow Sapir's lead and gain
a better understanding both of the relations among linguistic elements in
grammatical space and of the role markedness values play in their selection,
combination and concatenation—in synchrony as well as diachrony.

3. An analytic account of markedness

As I mentioned above (Section 0.2.3.), there is a need for a conceptual
analysis of markedness which has not been filled by previous discussions of the
notion. What I have in mind is not the sort of semantic analysis that will naturally
develop out of the synchronic investigation of current (common-parlance) usage
of the markedness words, which I advocated in Section 0.2.1. This will merely
chart the wide range of understandings different linguists have of what
markedness is, from the frankly vague notions of marked as "linguistically
undesirable" or "descriptively costly" to such a precise, but arbitrary suggestion
as Kean's "occurring in less than 10% of the languages of the world" (1980).
What is needed, rather, is an analysis that accounts for the intuitive
understanding of markedness that is reflected in native speakers' language use
(cf. Sections 1.3-1.7) and in linguistic change (Sections 2.1-2.6), and which
accounts for the implicit grasp of markedness that is revealed in the ordering of
morphological paradigms in various grammatical traditions and for the explicit
descriptions of markedness by, for example, Arabic and Western grammarians
(Section 0.3.2)—an analysis that accounts for these and other observed
manifestations of the intuitive and reflective, implicit and explicit, object-
linguistic and metalinguistic recognition of markedness by explaining how
markedness fits into the relations among linguistic units and by clarifying the
apparent, equal compatibility of the most diverse logical relations with the
"Protean", asymmetrical relation that markedness is.
38 HENNING ANDERSEN

3.1 The problem

It is rather remarkable that so much of the standard literature on the notion
of markedness has sidestepped the logical problem the notion poses. I do not
wish to belabor this point, but it is hardly possible to appreciate this unless it is
illustrated with some concrete examples and some mention of the ways this issue
has been ignored in the literature.
If we examine, for instance, the logico-semantic relations exemplified in
the Zamboangueño lexeme pairs in Table 7 we find that they are of several kinds.
Some of the lexical pairs are strictly speaking logical contradictories; such are
"clean : dirty", "clear : turbid", "straight : bent", as well as the opposition in
biological sex in "bachelor : unmarried girl" (an unmarried person is either a
male or a female—tertium non datur). Other pairs are logical contraries; among
these are "large : small", "tall : short", "fast : slow", "strong : weak", "tight :
loose", "sharp : dull", "smooth : rough" (the predicate "not large" does not entail
"small"—there is a third possibility, viz. "neither large nor small"); perhaps the
language-specific pair "sweet : bitter" and the stages of maturation in "blossom :
bud", "mature leaf : young leaf' belong here too. Still other lexical pairs are
converse opposites: "father/mother : son/daughter", "grandfather/grandmother :
grandchild". Contradictory, contrary, and converse relations are symmetrical in
the sense that the assertion of either term entails the denial of its opposite; see the
examples in Table 8.a, and contrast the asymmetrical, inclusive relation of
hyponymy in Table 8.b.
And yet, in the formation of Zamboangueño, despite their logical
symmetry, all these modes of opposition have been treated as asymmetrical.
In the categories investigated by Hopper & Thompson (Section 1.3),
similarly, some binary oppositions are contradictory (e.g., telic vs. atelic, realis
vs. irrealis), while others are contrary (e.g., agent high in potency vs. agent low
in potency); still other categories are viewed as scalar (e.g., individuation
—though this can be analysed into a hierarchy of binary oppositions, as shown
by Timberlake (1977) (cf. Section 2.2).
A similar mixture of contradictory and contrary binary and scalar
categories come into play in the Polish example (Section 2.3). All of them are
intrinsically symmetrical, but all of thern are engaged in the attested diachronic
processes on the basis of their asymmetrical markedness values.

3.2 Roman Jakobson

Despite the obvious logical diversity of lexical and grammatical relations,
and despite the acknowledged invariable asymmetry of markedness, theoreticians
who have attempted to explicate the notion of markedness have fairly
 MARKEDNESS AND THE THEORY OF CHANGE 39

consistently tried to present it as a sui generis contradictory relation or to resolve

it in terms of contradictory, that is, symmetrical oppositions.
This is true of all of Jakobson's published characterizations of markedness
(cf. Andersen 1989a:23-24; Battistella 1996:19-34). In the earliest and most
explicit of these Jakobson acknowledges the existence of contrary and

a. Exclusive relations
i. Contradictory opposites (e.g., A = "male",  ="female")
χ is A χ is  Τ—F χ is Β χ is A T—F
χ is A χ is not  Τ—Τ χ is not  χ is A Τ—Τ
χ is not A χ is  Τ—Τ χ is  χ is not A T—T
χ is not A χ is not  Τ—F χ is not  χ is not A T—F
ii. Converse opposites (e.g., A = "parent of y", B= "child of x")
χ is A  is  Τ—Τ  is  χ is A Τ—Τ
χ is A  is not  Τ—F  is  χ is not A T—F
χ is not A  is  Τ—F  is not  χ is A Τ—F
χ is not A  is not  Τ—Τ  is not  χ is not A Τ—Τ
iii. Contrary opposites (e.g., A = "wide",  = "narrow")
χ is A χ is  Τ—F χ is  χ is A Τ—F
χ is A χ is not  Τ—Τ χ is not  χ is A T—T
χ is not A χ is  Τ—T/F χ is  χ is not A Τ—T/F
χ is not A χ is not  Τ—T/F χ is not  χ is not A Τ—T/F
b. Inclusive opposites (e.g., A = "flower",  = "rose")
χ is A χ is  Τ—T/F χ is  χ is A Τ—Τ
χ is A χ is not  Τ—T/F χ is  χ is not A Τ—F
χ is not A χ is  Τ—F χ is not  χ is A Τ—T/F
χ is not A χ is not  Τ—Τ χ is not  χ is not A Τ—T/F

Table 8.

contradictory oppositions in meaning, but tries to explicate the asymmetry of

markedness as a (contradictory) opposition, not in meaning, but in what the
marked and the unmarked members of a grammatical opposition can be used to
assert (in his words, bezeichnen "denote", besagen "signify", ankündigen
"indicate", signalisieren "signal"):
As he examines two opposed morphological categories, the linguist often assumes
that these categories are equals, and that each of them has its own positive meaning:
category I denotes A, category II denotes B; or, at least, I denotes A and II denotes the
absence or negation of A. In reality the general meanings of correlative categories
stand in a different relation to each other: if category I indicates the presence of A,
category II does not indicate the presence of A, that is, it does not signify whether A
is present or not. In comparison with the marked category I, the general meaning of
the unmarked category II is limited to the absence of A-signaling [my translation;
HA]" ([1932] 1971b:3; see also [1936] 1971b:29-30).
40 HENNING ANDERSEN

But then, when he describes the extensive "use of unmarked forms at the
expense of the marked ones (e.g., infinitive for finite forms, present tense for
preterite, second person for first ...)", from which one may infer that "the
unmarked form represents both of the terms of the opposition in linguistic
consciousness" ([1932] 1971b: 14), Jakobson tries to account for this by adding
to the "signalization vs. non-signalization of A" a doubly contradictory relation
allegedly implicit in all such oppositions. Thus, on one hand, in a pair such as
lioness and lion (I choose these in lieu of Jakobson's examples, Russ. oslica f.,
osel m. "ass", which do not translate well), there is an opposition "signalization
of 'female'" vs. "non-signalization of 'female'"; this explains why both lioness
and lion can be used to refer to a female lion. But on the other hand, he says,
there is an opposition "non-signalization of 'female'" vs. "signalization of 'non-
female'", which explains why lion can be used both as a generic term and
specifically in reference to a male lion. At this point in Jakobson's explication
both the apparent simplicity of the account in the quotation above and the
meaning of the ad-hoc term 'signalization' have been severely compromised.
And what is worse, the multiple contradictories that have been posited shed
absolutely no light on the markedness relations in such simple pairs as countess
and count or parent and child, in which it is not true that both terms (say,
countess and count) can be used to refer to the designatum of the marked term (a
countess), nor that the unmarked term {count) can be used both to refer to a male
count and as a generic term to refer to any count or countess without
specification of biological sex.
Givón achieves as much precision, and with fewer words, when he credits
the Prague school linguists with the discovery that "binary distinctions in
phonology and grammar were systematically skewed or asymmetrical" and
explains that "one member of the contrasting pair acted as the presence of a
property, the other as its absence" [his italics; HA] (1990:946). Since Givon
does not worry much about precise definitions, he can overlook the fact that the
logical relation between presence and absence is contradictory, that is,
symmetrical, and hence he does not have to face the logical problem of how a
symmetrical relation can be "systematically skewed or asymmetrical".

3.3 John Lyons

Lyons (1977) is one of the few in recent times who have appreciated that
most lexical "sense-relations", as he calls them, are logically problematic. In
recognition of this he carefully separates his presentation of the logical modes of
opposition (271-273) from the lexical relations and even introduces separate
terminology for lexical relations in order to avoid using for them the standard
terms of logic (279).
 MARKEDNESS AND THE THEORY OF CHANGE 41

Thus he decides to reserve the term antonym for the contrary opposites in
the lexicon and to call the contradictory opposites complementarles, because
lexical opposites of both kinds differ from their logical counterparts by
"manifest[ing] the property of polarity" (279), that is, they combine their logical
symmetry with a kind of asymmetry, one term of each such opposition being
"positive", the other, "negative". (Lyons does not notice that the terms positive
and negative, too, form a symmetrical opposition.) He notes that the asymmetry
of such binary oppositions is manifested in a number of ways. One of these is
the irreversible binomials, first described by Malkiel (1959), in which the
positive opposite usually precedes its negative counterpart when the opposites
are conjoined, as in men and women, ducks and drakes (contradictories), good
and bad, high and low (contraries), buy and sell, parents and children, east and
west (converses), up and down, right and left (directionals) (Lyons 1977:276).
Another manifestation is the common development (or derivation) of generic
terms from positive terms or vice versa. For instance duck is both the
contradictory opposite of drake and the hypernym of duck and drake, just as lion
is both the contradictory opposite of lioness and the hypernym of lion and
lioness', similarly, wide serves both as the contrary opposite of narrow and as the
generic adjective for the dimension of width, long is the contrary opposite of
short as well as the generic adjective for the dimension of length, and so on and
so forth.
But Lyons's move to establish new, specialized terminology to distinguish
the lexical "sense-relations" from the logical modes of opposition is either just a
cosmetic cover-up for a reluctance to analyse the "sense-relations" into their
constituent logical relations, or it reflects a failure to recognize the need for such
an analysis. Interestingly, his discussion of the "polarity" of contradictories and
contraries (275) is divorced from his presentation of hyponymy {cow : animal,
rose :flower, buy : get, crimson : red, 291-301) as well as from his discussion
of markedness ("semantic marking", 307-311). If these matters had not been
separated in the exposition, but their presentation integrated, it might have been
clearer that the defining feature of Lyons's "property of polarity" is that exclusive
oppositions (of contradiction, contrariness, converseness, and direction) are
accompanied by (simultaneously combined with) the asymmetry that is
characteristic of the inclusive relation of hyponymy.

3.4 Analysis
This is unquestionably a topic that calls for the clear distinction between
logical and linguistic relations Lyons drew. But rather than Lyons's separation of
the two, the topic calls for the use of the language-independent concepts of logic
as tools in the analysis of semantic and other relations in grammar. In the
42 HENNING ANDERSEN

following I will show that the seemingly paralogical character of linguistic

relations—the peculiar fact that the linguistic counterparts of the symmetrical
modes of opposition are asymmetrical—reflects a hierarchical order among the
modes of opposition of normative logic that may be rooted in a universal strategy
applied in the cognition of all such relations.

3.4.1. Consider first the difference between contradictories and contraries. From
the point of view of normative logic, there is an absolute divide between the two,
defined by the entailment of negation (cf. Table 8.a.i-ii). In reality, however,
many relations that are in principle contradictory are practically contrary. For
example, strictly speaking every person is either married or unmarried
(normatively, tertium non datur). But real life presents us with intermediate states
and borderline cases—the union that has not been consecrated, the marriage that
has not been consummated (or, in some cultures, which has produced no issue),
the spouse that has been widowed or abandoned (and may or may not be at
liberty to marry again)—which persuade us to recognize a looser, essentially
contrary understanding of the distinction "married" vs. "unmarried", that is, to
acknowledge that there are people who are not 'really' one nor t'other, but
somewhere in between. The contradictory opposition "married" vs. "unmarried"
does not thereby disappear. It remains as a stricter, more principled, or technical
sense of the distinction "married" vs. "unmarried"; cf. Figure 1 (a) and (b). If we
wish, we can analyse the contrary opposition into a bundle of contradictories,
each defined by a separate criterion ("consecrated" vs. "non-consecrated",
"consummated" vs. "non-consummated", etc.).

Figure 1. Contradictory opposites construed (a) as contradictory and (b) as contrary;

(c) a contrary view of a strict contradiction.

Examples such as this, which are common, show that for a practical logic,
the contradictory and contrary modes of opposition are not miles apart, but
 MARKEDNESS AND THE THEORY OF CHANGE 43

closely related. Indeed, from a practical point of view, true contradictories appear
as a species of contraries—they are just those exclusive distinctions in which no
borderline case or intermediate state, no 'tertium' is conceivable; cf. Figure 1 (c).
This special class probably includes as a subspecies all true converse oppositions
(such as the divalent "parent of' vs. "child of') (Lyons 1977:279-280).
Directional opposites (e.g., "up" vs. "down", "in" vs. "out") (Lyons
1977:281-282) are another subspecies. They may be genuinely contradictory
(witness the directions of a moving elevator), but are generally contrary
(consider "look up" vs. "look down").

3.4.2. Similar considerations may clarify the relationship between exclusive

modes of opposition (Table 8 (a)) and inclusive ones (Table 8 (b)). In normative
logic, these are entirely distinct modes of opposition. The exclusive opposites are
intrinsically symmetrical, whereas the inclusive ones are intrinsically directed,
asymmetrical.
But recall the lexical examples above (Section 4.2) in which the "positive"
term (e.g., duck, wide) functions both as the contradictory or contrary opposite of
the "negative" term {drake, narrow) and as the hypernym subsuming both
opposites. The language historian knows that in some instances of this kind, a
generic term (e.g., duck, dog) has been extended to serve as "its own hyponym"
("female duck", "male dog"), as Lyons puts it in his discussion of dog and bitch
(1977:308), while in other instances a specific term has been extended to serve
as cover term for itself and its opposite; this is certainly the way Sp. padres
"parents" is related to Sp. padre "father" vs. madre "mother" and hermanos
"siblings", to hermano "brother" vs. hermana "sister".
Characteristic of all such examples is that one of the terms of an opposition
is construed both as superordinate and subsumed, inclusive and included, cf.
Figure 2. But in this inclusive construal the contrary or contradictory opposition
does not disappear. It remains easily accessible to analysis in terms of the
concepts of normative logic.

Figure 2. (a) A contradictory opposition and

(b) a contrary opposition, both construed as inclusive relations.
44 HENNING ANDERSEN

The normatively contradictory oppositions of grammatical categories work

precisely this way, as Jakobson endeavored to clarify (cf. the quotation in
Section 3.2 above). Typically, for example, in the oppositions of present tense
vs. preterite, the present tense may be used to refer not only to events
contemporary with the speech act, but also to events prior to it (the 'present
historical'), that is, its reference potential includes both the actual present and its
opposite, the actual past, or, to put it differently, it serves as a generic historical
tense.
Phonology and morphophonemics abound in closely parallel examples.
Any simple alternation between two allomorphs, for instance, is based on a pair
of alternants in complementary distribution, that is, with logically contradictory
privileges of occurrence—at least in the static view; e.g., Eng. /rayd-/ ~ /rowd-/
"ride", /nayf-/ ~ /nayv-/ "knife". But when there is a synchronic variation
between two such alternants, and we view them in dynamic terms, one of them
in effect encroaches on the other's privileges of occurrence; cf. the variation
/strayv-d/ ~ /strowv-0/ "strive; past" or /skarf-s/ ~ /skarv-z/ "scarf; pl.". That is
to say, the privileges of occurrence of one alternant include those of its opposite,
and in the course of time, this 'generic' alternant may entirely supplant its
covariant.

3.4.3. With this last example, perhaps, we come close to the source of the
asymmetry of markedness. In terms of normative logic, there is nothing
asymmetrical about the two complementary sets of environments to which two
covariant allomorphs are assigned. But if they are initially construed as an
inclusive opposition, then we can understand why one of the allomorphs would
be allowed to substitute for the other and might in time completely replace it.
Similarly with the contradictory and contrary oppositions of grammatical
and lexical semantic categories. In and of themselves these form perfectly
symmetrical oppositions. But it appears that they are initially construed as
inclusive relations. Hence one term in every such opposition is cast as the
superordinate, potentially generic, representative of both the terms.
This is true even of those normatively symmetrical oppositions in which it
is practically inconceivable (to the speakers of a given language) that the
reference potential of one term could include that of the other (e.g., "count" vs.
"countess", "parent of' vs. "child of', "east" vs. "west"). Such oppositions
appear to be construed first as inclusive, which makes the relation between the
two opposites asymmetrical and casts one of the opposites as superordinate. The
superordinate term remains a merely virtual hypernym, for there is no use for its
generic reference potential. In Hjelmslev's words, such oppositions form
inclusive relations in which part of the reference potential is blank (1939:87; see
 MARKEDNESS AND THE THEORY OF CHANGE 45

further Section 3.4.4 below). But the inclusive term retains its status as the
superordinate, unmarked term, as witnessed, among other things, by its place in
irreversible binomials and perhaps by its greater derivational potential. And its
virtual status of hypernym may in time be realized through linguistic innovation,
as in the case of Sp. padres "fathers" > "parents" and Eng. fathers "fathers" in
fore-fathers "ancestors"; cf. Figure 3.

Figure 3. True contradictory opposites construed as inclusive relations, (a) a simple

contradictory with the generic reference potential realized, (b) a conversity with the
generic reference potential latent.

3.4.4. In the past, several attempts have been made to explicate markedness in
terms of cognitive psychology.
Trubetzkoy, who originated the notion of markedness in phonology
(Jakobson 1985:162), was the first linguist to draw the parallel between the
distinction marked vs. unmarked and the figure-ground relation of Gestalt
psychology (1936). Since then, especially in more recent years, the
figure-ground concept has been invoked many times, either directly (e.g.,
Greenberg 1966:60, Wallace 1982, Givón 1990:947) or through the kindred
notions of prototype theory (cf. Lakoff 1987:59-61). This cognitive perspective
has been very useful in clarifying the source of the asymmetrical relation
between marked and unmarked opposites. But it has done nothing to clear up the
mystery of how this asymmetry is imposed on the logically symmetrical modes
of opposition, not only the privative (contradictory) ones Trubetzkoy
acknowledged, but also other binary distinctions, contrariety, converseness, and
direction.
Here I have sketched an account that clears up this mystery, so to say, in
three moves.
First, instead of thinking of markedness as an asymmetry that is imposed
on other modes of opposition, with which it is essentially incompatible, I
hypothesize that markedness arises in the initial cognition of any and all
distinctions thanks to the inherently asymmetrical, inclusive relation that obtains
between any concept that is formed () and the conceptual space that surrounds
it(U).
46 HENNING ANDERSEN

Secondly, from the cognitive priority of inclusion follows its logical

primacy. It is only in a subsequent step of analysis that a concept is cognized
either as simply included (in lexical terms, as a hyponym)—in which case the
conceptual space that surrounds it is formed as a correlative inclusive,
superordinate concept, e.g., ¡flower [rose]]—or it is cognized as one of the terms
of an included exclusive opposition. In the latter case, some of the surrounding
conceptual space may be formed as a hypernym—e.g., [duck [duck : drake]],
[wide [wide : narrow]]—or, if it does not correspond to any experience, it may
remain blank—[0 [east : west]].
Thirdly, I hypothesize that contraries are cognitively prior and logically
primary in relation to contradictions. This means that true contradictions are
cognized as a species of contraries in which the intermediate area between the
two opposites is not supported by experience and hence is not conceptually
formed, but remains virtual.
The first of these moves has roots in the thesis of Lévy-Bruhl (1910,
1922), the French anthropologist who was the first to discover the peculiar fact
that in ritual, occasionally, one of two symbolic opposites may substitute for its
counterpart, a phenomenon for which he coined the term "participation". The
illogical character of this phenomenon, which is in open defiance of the law of
contradiction, led Lévy-Bruhl to suppose that this observed peculiarity of
primitive cultures reflected a prelogical stage of cultural development, and he
formulated a theory of the primitive mind that seemed attractive and stimulated
discussion for some time, but was soon abandoned (also by Lévy-Bruhl) in the
face of evidence that the human capacity for logical thinking is the same in all
cultures (Lloyd 1966:3-6). Instead of Lévy-Bruhl's cultural-historical
interpretation of "participation" I have here suggested a universal of concept
formation, which is much more in agreement with the evidently universal
presence of markedness. If markedness is "prelogical", it is so in the sense of
being 'preanalytic'.
My second move exploits Hjelmslev's insight that "l'exclusion ne constitue
qu'un cas spécial de la participation, et consiste en ceci que certaines cases du
terme extensif ne sont pas remplies", that is, "exclusion is merely a special case
of inclusion, in which certain of the unmarked term's areas of manifestation
remain blank" (1939:87). This is what is illustrated in Figure 3. Note that this
move logically presupposes the primacy of inclusion. Nothing in experience
would motivate an analytic progression from a symmetrical, exclusive relation to
an asymmetrical, exclusive one. But the inclusive construal of exclusive relations
is easily understood as the result of an initial cognitive "overshoot".
The third and final move takes Hjelmslev's understanding of the
relationship between inclusion and exclusion one step further and reveals a
 MARKEDNESS AND THE THEORY OF CHANGE 47

similar relationship between the two chief modes of exclusive opposition,

contrariety and contradiction. A contradictory relation can be cognized inside a
contrary one, but not vice versa.
These three moves have a number of presuppositions, corollaries, and
consequences that cannot be developed here (a few are mentioned in Andersen
1989a: 38-40), but which I hope to return to elsewhere. But to conclude this
sketch, let me draw attention to the internal consistency of this account of the
modes of opposition, in which a traditional taxonomy of these such as Table 9.a
is superseded by the consistently inclusive taxonomy in Table 9.b. Note in Table
9.b how the analytic progression from I. to i. reflects the primacy of inclusion at
every single step. In other words, on this metalevel, inclusion is the unmarked
mode of opposition, exclusion, the marked, contrariety is unmarked in relation to
contradiction, and contradiction, in relation to conversity.

a. b.
I. inclusion I. inclusion
II. exclusive oppositions A. exclusion
A. contrariety 1. contrariety
B. contradiction a. contradiction
C. conversity i. conversity
Table 9.

3.5 'Markedness' and Markedness

Before we leave this topic it will be useful to confront the theory of
Markedness that has been developed here with some of the most widely accepted
characterizations of the notion. I will consider just a handful of these: (i) the
traditional characterization that comes closest to the understanding that has been
presented here (Section 3.5.1); three manifestations of Markedness that have
often been identified as 'markedness', (ii) the (iconic) reflections of Markedness
in expressions (Section 3.5.2), (iii) the syntagmatic manifestations of
Markedness (Section 3.5.3), and (iv) differences in relative text frequency
(Section 3.5.4); and, finally, (v) the frequency of grammatical phenomena in
cross-linguistic comparison (Section 3.5.5).

3.5.1 Markedness as semantic complexity. Markedness has been analysed here

as the distinction in semantic depth that is proper to inclusive relations such as
hyponymy. In hyponymy, the hyponym (e.g., "rose" ()) is semantically more
complex than its hypernym (e.g., "flower" (u)), the hyponym has more semantic
features (i.e. greater intension or semantic depth) than its hypernym, and
correspondingly the hyponym has lesser reference potential (lesser extension or
48 HENNING ANDERSEN

semantic breadth) than its hypemym. This is the relation that is commonly (a
priori) extended both to contrary opposites ("wide" vs. "narrow") and to
contradictory opposites ("duck" vs. "drake"), whereby the unmarked term of
such logically symmetrical relations, too, may have or may acquire greater
reference potential (extension) than its marked opposite (see Section 3.4.3).
Although this understanding of Markedness is based on an analysis of
(binary) oppositions, it can easily be extended to clines, the scalar differences
that are asymmetrical. Since clines are intrinsically asymmetrical, any two values
on a cline are in a relation of inclusion, and consequently points on a cline evince
different degrees of Markedness. Although scales whose polar values are
contraries are based on exclusive, symmetrical oppositions with equipollent
terms, the imputation of Markedness values to their opposites enables us to view
them too as clines.
Semantic complexity and similar notions have been recognized as the
defining criterium for markedness by many investigators since Trubetzkoy and
Jakobson, e.g., Lyons (1970:307), Shapiro (1983:79), Chvany (1985:248),
Givon (1990:946), Battistella (1996:56). But note that semantic complexity is
logically characteristic only of the relation of true inclusion, such as hyponymy.
The terms of exclusive relations are logically equally complex. It is only thanks
to the fact that they are cast (a priori) in terms of inclusion that they appear not to
be equipollent.

3.5.2 Markedness reflected in expressions. The first characteristic of

'markedness' mentioned in Givon's account (1990:946) is what he calls
"structural complexity": "the marked structure tends to be more complex—or
larger—than the corresponding unmarked one". Good examples might be Eng.
female vs. male or woman vs. man, where the complexity in expression mirrors
that in content. It is not clear whether Givón's statement distinguishes between
meaning (content) and form (expression) or conflates the two. But to the extent
that it refers to form (expression), Givón's statement obviously calls for
significant hedging. A "larger structure" reflects the marked term of an
opposition only provided there is an iconic relation between content and
expression—which is not always the case—and, furthermore, only if such an
iconic relation does not reflect some other difference between the terms of the
opposition. Consider, for example, the Russian polarity adjectives n'íz-k-ij "low"
(M) VS. vis-ok-ij "high" (u), bl'íz-k-ij "near" () vs. dal'-ok-ij "distant" (υ), úz-
k-ij "narrow" (M) VS. sir-ok-ij "wide" (u), m'él-k-ij "shallow" () vs. glub-ok-ij
"deep" (u), where the longer, dissyllabic stem in each pair correlates not with the
marked term, but with the term that denotes the unrestricted, greater extension in
space.
 MARKEDNESS AND THE THEORY OF CHANGE 49

Lyons speaks about the complexity of expressions as "formal marking"

(1977:305), and this notion figures prominently as "constructional iconicity" in
morphological naturalness theory (Mayerthaler 1981) and is dubbed "structural
markedness" by Croft (1990); cf. also Jakobson (1939, 1965), Greenberg
(1969), Haiman (1980). There is undoubtedly a universal tendency for
(semantic) markedness values to be(come) reflected in differences in the size of
expressions. But evidently Markedness is only one of several dimensions of
content that can be represented iconically by the relative size (or complexity) of
expressions. If we wish to clarify the possible iconic relations between meaning
and sound, we had better keep the notions of Markedness and size or complexity
of expression distinct.

3.5.3 Markedness manifested in syntagms. Just as Markedness may be reflected

iconically in simple and complex expressions, so it may be reflected in syntactic
properties. Typically, the marked term of an opposition has narrower privileges
of occurrence than its unmarked counterpart. One obvious consequence of this is
a difference in their relative text frequency, as we will see in the next section.
Here I will mention two principles that govern the regular manifestation of
Markedness in sequences. The first of these applies specifically when members
of one and the same opposition are concatenated; it sequences them, with some
regularity, in the order unmarked-marked. This was first observed by Malkiel in
his study of irreversible binomials (1959), and it was acknowledged as a
significant manifestation of markedness by Lyons (1977:276; cf. Section 3.3
above). There appears to be an analogous sequencing regularity in phonology: in
diphthongization (more precisely in primary diphthongization), the opposite
values of a distinctive feature come to be juxtaposed in the order
unmarked-marked. In earlier work I ascribed this to a 'Principle of Intra-
Segmental Variation', which I referred to informally as the 'principle of
unmarked beginnings' (Andersen 1972:23, 43, and passim).
The other prominent syntagmatic manifestation of Markedness is the
Principle of Markedness Agreement, the favoring of combinations or
concatenations of different features that are homogeneous in Markedness value.
This was illustrated above with examples from diverse levels of grammar,
ranging from text structure through grammatical agreement to allophonic
variation (Sections 1.2-1.7).
In Chapter 1 I characterized the Principle of Markedness Agreement in
purely descriptive terms, noting that it produces maximally orderly sequences
(see also Andersen 1991). One can imagine that this orderly distribution of
default category values to some extent facilitates speech processing, allowing
attention to be focused on the meaningful variables.
50 HENNING ANDERSEN

In addition to these, there are several other sequencing principles that may
be stated in terms of Markedness, among them Behagel's laws and the principle
that sequences modifiers according to their "essentiality to the head" (Nida
[1943] 1964:59) and affixes according to their category affinity (Jakobson
[1957] 1971b: 146) or "semantic relevance to the meaning of the stem" (Bybee
1985).

3.5.4 Markedness manifested in text frequency. In previous discussions of the

relation between markedness and frequency, it has often been unclear whether
frequency was considered a symptom of markedness or the source or cause of
markedness, and some linguists have simply identified frequency with
markedness (cf. Section 1.0). In Greenberg's 1966 study, relative text frequency
was identified as one of the most widely attested and reliable criteria of
markedness. Still, Greenberg understood that observed differences in frequency
often are "merely resultants" of diachronic tendencies, and in the end he
conceded that "frequency is itself but a symptom, and the consistent relative
frequency relations which appear to hold for lexical items and grammatical
categories are themselves in need of explanation" (1966:70).
Frequency is one of the three characteristics of markedness mentioned by
Givón: "The marked category (figure) tends to be less frequent, thus cognitively
more salient, than the corresponding unmarked one (ground)" (1990:947). Note
how Givón here establishes the link between frequency and saliency, but sees no
need to clarify the relation between markedness and frequency. Frequency is
simply part of his 'cluster-definition' of markedness, just as it is of Croft's
(1990).
The theory of Markedness that has been developed in this chapter views
Markedness as in principle independent of text frequency, but explains why
some differences in relative frequency arise as a natural consequence of
differences in Markedness: since marked terms have lesser reference potential
(or more narrowly defined privileges of occurrence) than their unmarked
counterparts, their frequency is lower. This is true not only of the terms of
logically inclusive oppositions, but also of the terms of exclusive oppositions, to
which the difference in value proper to inclusive relations has been imputed a
priori. Where there is a genuine correlation between Markedness and text
frequency, the latter is a manifestation of the former—that is, relative text
frequency is an epiphenomenon, as Greenberg concluded.
This does not mean that learners of a language may not infer Markedness
values from observed differences in frequency, or that a linguist should not form
hypotheses about Markedness values on the basis of frequency counts. But it
should be recognized that relative text frequency is determined by a number of
 MARKEDNESS AND THE THEORY OF CHANGE 51

factors, not least the practical, conventionalized communicative needs of

speakers. Hence, whatever the Markedness relation between, say, Aorist and
Perfect, in some language, or Imperative and Optative, there may not be any
clear-cut difference in relative frequency between them, or they may occur with
different frequencies in different text types or in different styles. The speakers of
a language may have more reliable criteria for Markedness values than the more
or less obvious differences in frequency—not least, presumably, those innate
assumptions about the internal coherence of grammatical systems which we
linguists hope eventually to discover. Furthermore, in considering any putative
correlation between Markedness values and relative frequency the historical
linguist should envisage the possibility that reanalysis may entail a shift in
Markedness values, and that such a shift cannot be reflected in usage
immediately, but can only emerge gradually, as the usage rules of the language
are adjusted to the underlying system.
If one wishes to understand synchronic variation or diachronic changes in
the correlation between Markedness values and relative text frequency, then, it is
essential to keep the two distinct, and it is totally counterproductive to define one
in terms of the other.

3.5.5 Markedness in cross-linguistic comparison. Since Roman Jakobson

observed that unmarked categories are more widely distributed in the world of
languages than the corresponding marked ones (1941), cross-linguistic
comparison has been used as a source of indications about the markedness of
language particular categories. But the words unmarked and marked have also
been used to characterize language categories as more or less wide-spread in the
currently known sample of natural languages.
As the concept of Markedness has been defined in this paper, it is
understood as an intrinsic characteristic of linguistic oppositions, as values
speakers impute to the terms of any and all oppositions in the process of
grammar formation. Surely no great harm can come from using the word
markedness about the linguist's evaluation of the relative unusualness of
structural features among the languages we know about. But there may be some
advantage, at the very least some pedagogical value, in reserving the words
markedness, marked, and unmarked for terminological use. They are certainly
not necessary when one describes the cross-linguistic incidence of linguistic
phenomena. These can be characterized perfectly well with such honestly
impressionistic words as infrequent, uncommon, unusual, or rare, or better
even—whenever this is possible—numerically with reference to a precisely
defined language sample.
52 HENNING ANDERSEN

4. Conclusion
I aimed to make two points in this paper. The first was to show that
Markedness, although it is mostly thought of as a synchronic property of
speakers' grammars, is a significant conditioning element in the progression of
linguistic change. The second was to clarify to some extent how Markedness is
related to the diverse modes of opposition on which it appears to be imposed,
and to ground the phenomenon independently of the level of observation.
As for the second of these points, it may not seem so important in and of
itself. I personally consider it essential, as Chapter 3 of this paper demonstrates.
But Markedness is such an easy concept to grasp that its proper logical analysis
may perhaps, by some, be considered an academic issue.
In relation to my first point, however, the theory of Markedness takes on
some importance. It is simple enough to make observations in the attested
progression of linguistic changes, but without a theory of Markedness such
observations cannot be conducted in a systematic way. The Principle of
Markedness Agreement which has been proposed here provides a basis for
making systematic observations of details in the actualization of linguistic
changes of all kinds. More than that, if the cognitive underpinnings I have
hypothesized for the principle are valid, such investigations will help us proceed
to the next step: understanding how the category values of a synchronic language
system both define its possible future changes and determine their gradual
actualization, or—paraphrasing Sapir (cf. Section 2.6 above)—how the cognitive
relations and the 'weights' of the individual elements of language patterns guide
and shape the drift that frays and reforms those patterns over time.

REFERENCES
The American Heritage Dictionary of the English Language ed. by Anne H.
Soukharov et al. Third edition. Boston & New York & London: Houghton
Mifflin Co., 1992.
Andersen, Henning. 1968. "IE *s after i, , r,  in Baltic and Slavic". Acta
Linguistica Hafniensia 11.171-190.
Andersen, Henning. 1972. "Diphthongization". Language 48.11-50.
Andersen, Henning. 1980. "Morphological change: towards a typology".
Historical Morphology ed. by Jacek Fisiak, 1-50. (Trends in Linguistics.
Studies and Monographs, 17.) The Hague: Mouton.
Andersen, Henning. 1987. "From auxiliary to desinence". Historical
Development of Auxiliaries ed. by Martin Harris & Paolo Ramat, 21-52.
(Trends in Linguistics. Studies and Monographs, 35.) Berlin & New York:
Mouton de Gruyter.
Andersen, Henning. 1989a. "Markedness theory: the first 150 years".
Markedness in Synchrony and Diachrony ed. by Olga Mišeska Tomic,
 MARKEDNESS AND THE THEORY OF CHANGE 53

11-46. (Trends in Linguistics. Studies and Monographs, 39.) Berlin & New
York: Mouton de Gruyter.
Andersen, Henning. 1989b. "Understanding linguistic innovations". Language
Change. Contributions to the Study of its Causes ed. by Leiv Egil Breivik &
Ernst Håkon Jahr, 5-28. (Trends in Linguistics. Studies and Monographs,
43.) Berlin & New York : Mouton de Gruyter.
Andersen, Henning. 1990. "The structure of drift". Historical Linguistics 1987.
Papers from the 8th International Conference on Historical Linguistics ed.
by Henning Andersen & Konrad Koerner, 1-20. (Current Issues in
Linguistic Theory, 66.) Amsterdam & Philadelphia: John Benjamins.
Andersen, Henning. 1991. "On the projection of equivalence relations into
syntagms". New Vistas in Grammar: Invariance and Variation ed. by Linda
R. Waugh & Stephen Rudy, 287-311. (Current Issues in Linguistic Theory,
49.) Amsterdam & Philadelphia: John Benjamins.
Andersen Henning. 1998. "Dialektnaja differenciacija obščeslavjanskogo jazyka.
Paradoks obščix tendency razvitija s različnymi lokal'nymi rezul'tatami" [The
dialect differentiation of Common Slavic. The paradox of shared tendencies
of development having distinct local outcomes]. American Contributions to
the Twelfth International Congress of Slavists. Cracow, Aug-Sept. 1998.
Literature. Linguistics. Poetics ed. by Robert A. Maguire & Alan
Timberlake, 565-600. Bloomington, Ind.: Slavica Publishers.
Back, Michael. 1989. Die synchrone Prozessbasis des natürlichen Lautwandels.
(Zeitschrift für Dialektologie und Linguistik. Beihefte, 710.) Stuttgart: F.
Steiner.
Battistella, Edwin L. 1996. The Logic of Markedness. Oxford: University Press.
Bybee, Joan L. 1985. "Diagrammatic iconicity in stem-inflection relations".
Iconicity in Syntax ed. by John Haiman, 11-48. (Typological Studies in
Language, 6.) Amsterdam: John Benjamins.
Chafe, Wallace. 1970. Meaning and the Structure of Language. Chicago:
University of Chicago Press.
Chvany, Catherine V. 1985. "Backgrounded perfective and plot line
imperfécrives: toward a theory of grounding in text". The Scope of Slavic
Aspect ed. by Michael S. Flier & Alan Timberlake, 247-273. Columbus,
Ohio: Slavica
Coseriu, Eugenio. [1952] 1962. "Sistema, norma y habla". Revista de la
Facultad de Humanidades y Ciencias (Montenegro) 9.113-177. Reprinted
in his Teoría del lenguaje y lingüistica general. Cinco estudios, 11-113.
Madrid: Editorial Gredos.
Coseriu, Eugenio. [1965] 1968. "Sincronía, diacronía y tipología". Actos del XI
Congreso Internacional de Lingüistica i Filología Románicas, vol. 1,
269-283. Madrid, 1965. Reprinted as "Synchronie, Diachronie und
Typologie" in his Sprache. Strukturen und Funktionen. XII Aufsätze zur
allgemeinen und romanischen Sprachwissenschaft, 91-108. (Tübinger
Beitrage zur Linguistik, 2.) Tübingen: Gunter Narr.
Croft, William. 1990. Typology and Universals. Cambridge: Cambridge
University Press.
54 HENNING ANDERSEN

Dressier, Wolfgang U. 1980. "A semiotic model of diachronic process

phonology". Wiener Linguïstische Gazette 22/23.31-94.
Dressier, Wolfgang U. & Gaberell Drachman. 1977. "Externe Evidenz für eine
Typologie der Vokalprozesse". Salzburger Beitráge zur Linguistik
3.285-297.
Fischer, Olga. 1992. "Syntactic change and borrowing: The case of the
accusative-and-infinitive construction in English". Internal and External
Factors in Syntactic Change ed. by Marinel Gerritsen & Dieter Stein, 17-88.
(Trends in Linguistics. Studies and Monographs, 62.) Berlin & New York:
Mouton de Gruyter.
Fox, James. 1973. "On bad death and the left hand". Needham 1973:342-368.
Frake, Charles O. 1971. "Lexical origin and semantic structure in Philippine
Creole Spanish". Pidginization and Creolization ed. by Dell Hymes,
223-242. Cambridge: University Press.
Givon, Talmy. 1984-1990. Syntax. A Functional-Typological Introduction, I—II.
Amsterdam & Philadelphia: John Benjamins.
Greenberg, Joseph H. 1963. "Some universals of grammar with particular
reference to the order of meaningful elements". Universals of Language ed.
by Joseph H. Greenberg, 73-113. Cambridge, Mass.: The MIT Press.
Greenberg, Joseph H. 1966. Universals of Language. With Special Attention to
Feature Hierarchies. (Janua Linguarum, Series Minor, 59.) The Hague &
Paris: Mouton.
Greenberg, Joseph H. 1969. "Some methods of dynamic comparison in
linguistics". Substance and Structure of Language ed. by Jan Puhvel,
147-203. Berkeley & Los Angeles: University of California Press.
Haiman, John. 1983. "Iconic and economic motivation". Language 59.781-819.
Hertz, R. 1909. "La préeminence de la main droite: étude sur la polarité
religieuse". Revue philosophique 68.553-580. Translated by Rodney &
Claudia Needham in Death and the Right Hand ed. by Rodney Needham,
89-113, 155-160. Aberdeen: Cohen & West, 1960.
Hjelmslev, Louis. [1939] 1970. "Notes sur les oppositions supprimables".
Travaux du Cercle linguistique de Copenhague 8.51-57. Reprinted in his
Essais linguistiques, 82-88. (Travaux du Cercle linguistique de Copenhague
12.) Second edition. Copenhagen: Nordisk Sprog- og Kulturforlag.
Hopper, Paul J. & Sandra A. Thompson. 1980. "Transitivity in grammar and
discourse". Language 56.251-299.
Ivanov, Vjačeslav V. & Vladimir N. Toporov. 1974. Issledovanija ν oblasti
slavjanskix drevnostej. Leksiceskie i frazeologiceskie voprosy rekonstrukcii
tekstov [Studies in Slavic antiquities. Lexical and phraseological issues in the
reconstruction of texts]. Moscow: Nauka.
Jakobsen, Thorkild. 1974. "Very ancient texts: Babylonian grammatical texts".
Studies in the History of Linguistics. Traditions and paradigms ed. by Dell
Hymes, 41-62. Bloomington & London: Indiana University Press.
Jakobson, Roman. [1932] 1971b. "Zur Struktur des russischen Verbums".
Charisteria Gvilelmo Mathesius quinquagenario a discipulis et Circuli
 MARKEDNESS AND THE THEORY OF CHANGE 55

Linguistici Pragensis sodalibus oblata, 74-84. Prague: Cercle Linguistique

de Prague. Reprinted in Jakobson 1971b:3—15.
Jakobson, Roman. [1936] 1971b. "Beitrag zur allgemeinen Kasuslehre
(Gesamtbedeutungen der russischen Kasus)". Travaux du Cercle
Linguistique de Prague 6.240-288. Reprinted in Jakobson 1971b:23-71.
Jakobson, Roman. [1939] 1971b. "Signe zéro". Mélanges de linguistique, offerts
à Charles Bally, 143-152. Geneva: Librairie de l'Université. Reprinted in
Jakobson 1971b:211-219.
Jakobson, Roman. [1941] 1971a. Kindersprache, Aphasie und allgemeine
Lautgesetze. Uppsala: Almqvist & Wiksell. Reprinted in Jakobson
1971a:328-401.
Jakobson, Roman. [1966] 1971b. "Quest for the essence of language". Diogenes
51.21-37. Reprinted in Jakobson 1971b:345-359.
Jakobson, Roman. [1957] 1971b. "Shifters, verbal categories, and the Russian
verb". Cambridge, Mass.: Department of Slavic Languages, Harvard
University. Reprinted in Jakobson 1971b: 130-147.
Jakobson, Roman. [1971] 1971b. "Krugovorot lingvistiöeskix terminov"
[Terminological shifts in linguistics]. Fonetika. Fonologija. Grammatika, 
semidesjatiletiju A. A. Reformatskogo, 384-387. Moscow: Nauka. Reprinted
in Jakobson 1971a:734-737.
Jakobson, Roman (ed.). 1971a. Selected Writings, vol. I. Phonological Studies.
Second edition. Berlin & Amsterdam & New York: Mouton.
Jakobson, Roman (ed.). 1971b. Selected Writings, vol. II. Word and Language.
Berlin & Amsterdam & New York: Mouton.
Jakobson, Roman (ed.). 1985. N. S. Trubetzkoy's Letters and Notes. (Janua
Linguarum, Series Major, 47.). Berlin & New York & Amsterdam: Mouton.
Jakobson, Roman & Krystyna Pomorska. 1983. Dialogues. Translated by
Christian Hubert. Cambridge, Mass.: The MIT Press.
Kean, Mary-Louise. 1980. The Theory of Markedness in Generative Grammar.
Bloomington, Ind.: Indiana Linguistics Club.
Lakoff, George. 1987. Women, Fire, and Dangerous Things. What Categories
Reveal about the Mind. Chicago & London: University of Chicago Press.
Lass, Roger. 1980. On Explaining Language Change. Cambridge: University
Press.
Lévy-Bruhl, Lucien. [1910] 1922. Les fonctions mentales dans les sociétés
inférieures. Paris: Felix Alcan. Translated by L. A. Clare as How Natives
Think. London: George Allen & Unwin, 1922.
Lévy-Bruhl, Lucien. 1922. La mentalité primitive. Paris: Félix Alcan. Translated
by L. A. Clare as Primitive Mentality. London: George Allen & Unwin and
New York: Macmillan, 1923.
Lloyd, G. E. R. 1966. Polarity and Analogy. Two Types of Argumentation in
Early Greek Thought. Cambridge: University Press.
Lotman, Jurij M. 1970. Struktura xudozestvennogo teksta [The structure of the
literary text]. Moscow: Izdatel'stvo Iskusstvo. Translated as The Structure of
the Artistic Text. (Michigan Slavic Contributions, 7.) Ann Arbor, Mich.:
Department of Slavic Languages and Literatures.
56 HENNING ANDERSEN

Lyons, John. 1977. Semantics, vol. I. Cambridge: University Press.

Malkiel, Yakov. 1959. "Studies in irreversible binomials". Lingua 8.113-160.
Reprinted in his Essays on Linguistic Themes, 311-355. Oxford: Blackwell,
1968.
Mareš, František V. [1952] 1969. Diachronische Phonologie des Ur- und
Frühslavischen. (Slavistische Beiträge, 40.) Munich: Otto Sagner.
Mayerthaler, Willi. 1981. Morphologische Natürlichkeit. Wiesbaden: Athenaion.
Needham, Joseph. 1954. Science and Civilization in China, vol. I. Cambridge:
University Press.
Needham, Rodney. 1973. Right and Left. Essays on Dual Symbolic
Classification. Chicago & London: University of Chicago Press.
Needham, Rodney. 1979. Symbolic Classification. Santa Monica, Calif.:
Goodyear Publishing.
Nida, Eugene. [1959] 1964. A Synopsis of English Syntax. (Publications of the
Summer Institute of Linguistics, 4.) Norman, Okla.: University of
Oklahoma. Forth printing, Mexico, D.F.: Instituto Lingüístico de Verano.
Owens, Jonathan. 1988. The Foundations of Grammar. An Introduction to
Medieval Arabic Grammatical Theory. (Studies in the History of the
Language Sciences.) Amsterdam & Philadelphia: John Benjamins.
Paul, Hermann. [1881] 1970. Prinzipien der Sprachgeschichte. (Reprint of the
fifth edition, 1920.) Tübingen: Max Niemeyer.
Pos, Hendrik. J. 1938. "La notion d'opposition en linguistique". Onzième
Congrès International de Psychologic Rapports et comptes rendues,
246-247. Paris: Félix Alcan.
Romaine, Suzanne. 1982. Socio-Historical Linguistics, its Status and
Methodology. Cambridge: Cambridge University Press.
Sahlins, M. D. 1978. Culture and Practical Reason. Chicago: University of
Chicago Press.
Sapir, Edward. [1921] 1949. Language. An Introduction to the Study of Speech.
New York: Harcourt, Brace & World.
Saussure, Ferdinand de. [1915] 1973. Cours de linguistique générale. Edition
critique préparée par Tullio Mauro. Paris: Payot.
Schachter, Paul. 1969. "Natural assimilation rules in Akan". International
Journal of American Linguistics 35.342-355.
Shapiro, Michael. 1983. The Sense of Grammar. Language as Semeiotic.
(Advances in Semiotics.) Bloomington & Indianapolis: Indiana University
Press.
Shapiro, Michael. 1991. The Sense of Change. Language as History. (Advances
in Semiotics.) Bloomington & Indianapolis: Indiana University Press.
Timberlake, Alan. 1977. "Reanalysis and actualization in syntactic change".
Mechanisms of Syntactic Change ed. by Charles N. Li, 141-180. Austin:
University of Texas Press.
Timberlake, Alan. 1981. "Dual reflexes of *dj and a morphological constraint on
sound change". International Journal of Slavic Linguistics and Poetics
23.25-54.
 MARKEDNESS AND THE THEORY OF CHANGE 57

Troubetzkoy, Nicolas S. 1936. "Essai d'une théorie des oppositions

phonologiques". Journal de psychologie 33.5-18.
Wallace, Stephen. 1982. "Figure and gound: the interrelationships of linguistic
categories". Tense-Aspect: Between Semantics and Pragmatics. Containing
the Contributions to a Symposium on Tense and Aspect, UCLA, May 1979
ed. by Paul Hopper, 201-226. Amsterdam & Philadelphia: John Benjamins.
Zabrocki, Ludwik. 1951. Usilnienie i lenicja w jezykach indoeuropejskich i w
ugrofinskim [Fortition and lenition in Indo-European and Finno-Ugric].
(Poznanskie Towarzystwo Przyjaciól Nauk, Wydzial filologiczno-
filozoficzny. Prace komisji filologicznej, torn XIII, zeszyt 3.) Poznan: Naklad
Poznanskiego Towarzystwa Przyjaciól Nauk.
 PATTERNS OF RESTITUTION OF SOUND CHANGE

KRISTIN BAKKEN
University of Oslo

1. Introduction
The data that will be discussed in this article are from the dialect area
Western Telemark in southern Norway. They exemplify the phonotactic
requirements of two historical phonological processes that characterize the
dialect. I will refer to these processes by the terms 'l-loss' and 'delateralization'.
Although my focus here will be the reversal of these processes, or the undoing
of their effects, I will first characterize the phonological processes as such.
In /-loss a postvocalic /1/ is lost when it occurs before any consonant other
than /d/ or /t/.1 In the opinion of most scholars, the loss of / followed a
lengthening and a subsequent qualitative change of the preceding vowel (cf.
Larsen 1976:144-148, Indreb0 1951:132, Seip 1955:109f.), although this view
has been challenged recently by Sand0y (1994, 1997, 1999). Sand0y regards the
lenghtening as compensatory for the /-loss and the qualitative changes as
primary. Examples of this process are ON holmi "island" > dial, Phu:re/, ON
kalfr "calf > dial. /1kc:ve/. There is evidence of /-loss in the oldest written
records from Western Telemark, i.e. from the fourteenth century on. Examples
can also be found in earlier runic inscriptions, indicating that the process had
started before 1200 (Skulerud 1918:37f., Seip 1955:109f., NIYR 6;2:197). The
end of the productive period of the process is difficult to delimit.
The other restitution I will examine annuls a somewhat younger
phonological process that belongs to the same area. It too affects the lateral. By

1
During the Old Norse (ON) period the fricative /Ö/ and the plosive /d/ merge after /1/ and
leave/d/ (cf. Noreen 1923:175), but this presumably happens later after Proto-Nordic light
stems, and it is therefore possible that /-loss in an early phase took place before a Proto-Nordic
/ð/ retained after such Proto-Nordic light stems. Local examples that support a hypothesis of /-
loss after etymological /ð/ are ON fiolÐ > /fjcd:/ "crowd" and ON hulð > hud(de) "supernatural
human-like creature". It seems as if that the merger of /d/ and /ð/ would rule out such
examples, but it is also possible that two /-allophones had developed before /d/ and /ð/, and
that these were phonemicized as a consequence of the merger. Icelandic data from the same
period support positing two /-phonemes (cf. Jakob Benediktsson 1960, Sand0y 1999:71-73).
60 KRISTIN BAKKEN

'delateralization' I will refer to the fact that Old Norse postvocalic or intervocalic
ll changed into /d:/. Examples are ON allir "all" > dial. /2adi9/ and ON fjall
"mountain" > dial. /fj0d:/. The process is generally assumed to be the result of a
"segmentation", :/ > /dl/ and an assimilation, /dl/ > /d:/ (Christiansen 1976:172).
When ON ll changed into /d:/ the phonological system lost its opposition
between long and short /. When ON ld was assimilated to /1:/ later on, the system
was replenished on this point.
The (¿/-segmentation characterizes a large area west of Telemark and
Setesdal today, and there is evidence of this segmentation process in medieval
documents from this area (Hægstad 1916:142f.). The oldest written examples
with dd, however, are younger. Two uncertain placename examples are from ca.
1500, whereas the first solid evidence dates from 1585 and onwards.
The sources are written in Danish at this point, however, and even if
names occasionally can pass through the net of normalization, usually they are
danified (cf. Bakken 1999). Even though the written evidence is scant, the
change may therefore still have occurred by 1500 as indicated by the tentatively
dated Rauland example Haddeland (DN XI 275).2
To illuminate the productivity period of delateralization, the first local
dialect glossary which was compiled in 1698, is useful (Bloch 1698). At this
point in time we find ample evidence of delateralization. Interestingly enough the
dd-forms are often marked as alternating with the //-forms, cf. "Fulle sive
Fudde" "full" and "Kallen sive Kadden" "the man". Bloch writes words
containing ON ld with ll. Since secondary ll of this origin is never turned into
/d:/, one has to suppose that the delateralization of original ll had ceased to be
productive by the time ld assimilated to /1:/. On the basis of Bloch 1698, one
must infer that this happened before 1700.
Placename data indicate that both the /-loss and delateralization processes
have greatly reduced their geographical domain over the years. Names
instantiating both processes are found outside the area that traditionally has been
associated with them, and the vocabulary at large shows increasingly more
instantiations of the processes as one moves westward. It seems reasonable to

2
The example Hallingdal has traditionally been related to the discussion of the domain and
productivity period of the delateralization process (cf. Larsen 1917:45). The name of this
valley east of Telemark was written with dd in the oldest medieval records. In 1443 it was
written with ll for the first time (DN X 137). Larsen interpreted this as a hypercorrection
created against the background of a productive delateralization process in Telemark. Myhren
(1994:14f.) has pointed to an alternative explanation of the nameform Hallingdal that does not
need to implicate the Telemark delateralization process. If hypercorrection is to be
hypothesized for the nameform Hallingdal in 1443, this would date the delateralization
process back 50-100 years before the first written instance of delateralization in Telemark. This
seems unlikely and in my opinion renders chronological support for Myhren's alternative
interpretation.
 PATTERNS OF RESTITUTION OF SOUND CHANGE 61

infer that the effects of /-loss and delateralization have for quite some time been
restored, consequently leading to the restitution of etymological wordforms. It is
the nature of these two restitution processes that will be examined in this article. I
will address four questions:

• Does the restitution of /-loss and delateralization imply that these

processes initially were lexically diffused?
• What role does dialect contact play in the restitution?
• Does the restitution proceed lexeme by lexeme, or by rule loss on a
general basis?
• If the restitution proceeds lexeme by lexeme, in what manner does it
proceed?

2. Data
Before I go on to answer these questions, I will briefly account for the data
that have been analysed, and which constitute the empirical basis for my
discussion.
Investigating diachronic phonology always raises the methodological
problem of studying oral phenomena through imperfect written sources. I have
shown elsewhere (Bakken 1987, 1999) how dialect wordforms as a rule are kept
out of the written norm in the Middle Norwegian period (1350-1550) as well as
in the later Danish period (1550-1900). Exceptions in both periods are local
nameforms which may be spelled as they were pronounced within both the
Middle Norwegian and the Danish written norm. The shortcomings of ordinary
written sources are particularly deeply felt when the object of study is the lexical
actualization of change, as is the case here.
Although the processes under scrutiny were productive in the period
before 1700, the restitution of their lexical effects are gradually carried through
after 1700. As a consequence I have based my discussion on data drawn from
the following four kinds of sources. The backbone of my data is a list of forty-
seven words drawn from the synopsis list in the dialect archives at the
University of Oslo. The synopsis consists of spoken data collected from most
Norwegian counties in the years between 1920 and 1970.I have checked this list
of forty-seven words at the thirteen locations in Western Telemark that are
included in the synopsis data (see Figure 1 below). The counties represented are
Seljord, Kviteseid, Tokke, and Vinje.
Secondly I have systematically compared these thirteen lists of words from
the synopsis with responses to the same forty-seven words from young
informants. These interviews were conducted by telephone, and as a rule I only
spoke to one informant at each location. However, as well as inquiring into
62 KRISTIN BAKKEN

personal pronunciation habits, I also asked the informants how their own
pronunciations relate to the linguistic norms of their community.
Looking further back in time, my main source has been the four local
dialect glossaries collected by amateur linguists in the period 1698-1821.3 These
glossaries are a less systematic point of comparison, but they are invaluable
sources of dialect phonology from a time when written sources of this kind
generally are scarce.

Figure 1. Map of the Telemark region in southern Norway showing

the thirteen synopsis locations.

My fourth source is placename data. There are extensive records of

placename pronunciations in The Name Archives at the University of Oslo.
3
The glossaries are: Bloch (1698) from Fyresdal, a western Telemark county adjacent to Tokke
and Kviteseid; Paus (1743), not restricted to any one western Telemark county, but its author
was born in Kviteseid; Wille (1786), which reflects the spoken dialect of Seljord; and
Svenungsen (1821), which basically reflects the dialect of Vinje and Tokke.
 PATTERNS OF RESTITUTION OF SOUND CHANGE 63

These date from the 1880s to the present, but they have a bias towards the
archaic. As indicated above, names in medieval as well as Danish sources are
often written in a surprisingly phonetic form, and they are consequently also
included in my data.

3. Results
I will present a more detailed analysis of my data elsewhere. Here I can
only give a general outline of my results. Nevertheless it is important to bear in
mind that each individual wordform needs to be interpreted in light of the source
in which it is recorded, and that the formulation of geographical and
chronological patterns over these forty-seven words demand such an
interpretative, atomistic approach. The data are complex; the degree of
commensurability is often variable, sources need interpretation, and deficiencies
in representativity and continuity need to be taken into consideration.
Nevertheless, diachronic, lexical, and geographical patterns do emerge
from my data, and I will give an outline of these patterns here. First, however, I
will present the forty-seven words in their order of frequency according to
Vestb0stad (1989); see Table 1.4
If modern data were all we had to go by, both delateralization and /-loss
would seem to be highly irregular processes. At some locations regular and
irregular dialect forms seem to balance each other off, at other locations there are
only stray dialect forms in a very limited set of words. When one adopts a
diachronic perspective, however, this impression of irregularity is altered.
Synchronic exceptions to the rules are often found to have been regular in the
past; in Kviteseid main parish kjelke "sled" and belg "pod" are pronounced
[2ç0lçs] and [belg], but the synopsis data give [2çoçs] and [bae:g]. From Seljord
the synopsis has [kalv] "calf, [halv] "half and [fj0l:] "mountain" but Wille
1786 lists the compound Haavhœmpe (first element halv, the compound is a
pejorative for "woman"), Kaav and Fi0ddi from Seljord. No-one in this century
pronounces elg "moose" without an /1/, and the only form recorded in the
glossaries is written with an / as well (Bloch 1698), but Ægshov, Æjestigane,
and Ægstj0nn are placenames in Vinje. Synchronic exceptions can as a rule be
accounted for in ways that seem to undermine a lexical diffusion hypothesis.
First, many of them turn out to be comparatively late restitutions, such as the first
examples above; secondly, some exceptions turn out to be late loans; thirdly,
some are words that have gone out of use and been reintroduced at a

4
Vestb0stad (1989) is the New Norwegian frequency dictionary. The figures represent numbers
of tokens in a corpus of some one million words of written text.
64 KRISTIN BAKKEN

/-loss Delateralization
sj0lv, pron. "(-)self 1579 alle, pron. "all" 3619
folk, n. "folk, people" 1523 full, adj. "full" 494
fylgje, v. "follow" 431 kalle, v. "call" 317
helg, n, "holiday, weekend" 237 ƒalle, v. "fall" 215
halv, n. "half" 207 fjell, n. "mountain" 184
mj0lk, n. "milk" 76 fylle, v. "fill" 134
golv, n. "floor" 67 stelle, v. "care for, serve" 73
hals, η. "neck, throat" 52 rulle, v. "roll" 39
elg, n. "moose, elk" 43 imillom, adv. "in between" 35
tolv, num. "twelve" 36 smelle, v. "bang, slam" 27
kalv, n. "calf 31 voll, n. "meadow" 22
ulv, n. " w o l f 18 ull, n. "wool" 16
stolpe, n. "post, pole" 17 elleve, num. "eleven" 14
sylv, n. "silver" 14 skilling, n. "shilling" 14
halvt anna, adj. "one and a half" 13 troll, n. "troll, giant" 14
kvalp, n. "puppy" 6 skalle, n. "head, skull" 13
kvelve, v. "turn over" 5 bolle, n. "bowl" 5
pulse, n. "sausage" 5 heller, n. "cave" 5
belg, n. "pod, body, stomach" 1 kolle, n. "hill" 4
kjelke, n. "sled" 2 gylle, v. "gild" 3
stylk, n. "stem, stalk" 1 myllar, n. "miller" 1
skalk, n. "heel of a loaf of bread" 0 eismall, adv. "alone" 0
talg, η. "tallow" 0 mj0llaus, adj. "without flour" 0
halm, n. "straw" 0

Table 1. The test words in descending order of text frequency.

later stage. Examples of the last type are elg and also ulv "wolf. The
reintroduction of the word elg must be understood against the background of the
several hundred years when the moose was virtually extinct in the area. The data
indicate that the taboo term skrubb replaced the ON ulfr "wolf at some point,
and that word was only reintroduced after the l-oss had ceased to be productive.
(In old personal names the form -uv is well documented.) The notion of
productivity gives coherence to several apparent exceptions in this way, and my
first conclusion is therefore that delateralization and /-loss most likely were
regular phonetic processes at the outset and, consequently, that it is the restitution
that was lexically diffused.
A second conclusion that can be drawn from my data is that not only were
the restitution processes in general lexically diffused but, more specifically,
 PATTERNS OF RESTITUTION OF SOUND CHANGE 65

the test words can be grouped together according to the manner in which they
relate to the restitution process. One conspicuous group is marked by very
strong resistance to restitution. Among the synopsis words, these are all
examples of /-loss. The wordforms hœg (< ON helgr "holiday"), fyje (ON fylgja
"follow"), and sjav (ON sjalfr "self") are used uniformly all over Western
Telemark, and even partly outside the area. Other resistant dialect forms not
included in the synopsis list are syje (ON sylgja "silver ornament"), kvœv (ON
*kvelfr "shallow valley"), home (ON holmi with the secondary meaning "grassy
elevation in an otherwise flat field"), and fjåg (ON fjalgr "merry"). The noun
hadd and the verb hadde (ON hallr "slope", halla "to slope") are examples of
resistant dd-words. These wordforms seem to have a special status within the
dialect vocabulary; they are somehow perceived as independent words without
links to cognates in neighboring dialects. This is illustrated by the fact that fjåg
has been included in the New Norwegian written standard.
Another group of words is marked off by the fact that they never occur in
the expected dialectal form. As already indicated, most of these words are very
early restitutions (golv "floor", hals "throat"), late loans (rulle "roll", stelle "care
for"), or reintroductions into the dialect (falle "fall", elg, and ulv). Two
candidates for /-loss, sylv "silver" and stylk "stalk", never show l-loss, and since
both words had ON i in the nucleus, this may be phonetically motivated. At any
rate the Mid-Scandinavian flap (/r/)never develops after long and short i either
(Sand0y 1999:73), and this is relevant since the /-loss and the development into
/r/ are parallel and geographically complementary processes.
The most interesting group of words are those which are basically variable.
Synchronically one can detect a geographical pattern of variation in that Vinje
seems to be the core area for these dialect forms, and their numbers decrease as
one moves eastward. It is interesting that different communities behave similarly
as to which words are most resistant to restitution. Again the shift from a
synchronic to a diachronic perspective is illuminating. Lexical patterns that at one
point in time characterize one location are transposed geographically when one
moves back in time. This indicates that the restitution processes move through
the lexicon in the same way, although the vocabularies of different locations are
affected by these processes at different times. When the geographical patterns are
viewed diachronically, it is clear that both of the restitution processes originate in
the east and move westward as time passes. Generally one can note that among
the last dialectal wordforms to be restored are fjødd (fjell "mountain"), fokk (folk
"people"), kjàkkje (kjelke "sled"), mjåkk (mj0lk "milk"), kvœve (kvelve "turn
over"), and hedder (heller "cave"). This is true whether one peruses the old
glossaries in the case of the eastern locations, or one examines the synopsis data
in the case of the western locations. In the recent telephone interviews my
66 KRISTIN BAKKEN

informants were familiar with these dialect forms even if they did not actively
use them themselves.
One additional point that can be made is the fact that delateralization
shows a much more tidy lexical distribution than /-loss. There are both more
dialectal relics and more exceptions to the rule among the /-loss words.
Finally, my analysis has demonstrated that words with certain semantic
characteristics lag behind in the restitution process. Placenames instantiating both
processes are found in the easternmost counties, where only lexicalized relics
like hœg are left in the vocabulary at large, and such nameforms are also found
outside Western Telemark proper. Personal names to some degree align with the
placenames in this respect; most notable are the names containing the lexeme ulv,
which until modern times have been pronounced without the /, although the
common appellative is ulv. Some of these names, such as Τον and Bjug, have
been used exclusively in their dialectal form (cf. Bakken 1999). Today the
tendency is for the dialectal personal names to be going out of use. The analysis
has also demonstrated that dialectal wordforms tend not to be restituted when
they appear as elements in lexicalized compounds. Examples are hals "neck" but
håsklute "necktie" and håslene "neck + linen" (Ross 1906:12), halve "half but
håvravle "halfwit" (synopsis, Grungedal 1959), ulv "wolf generally in Western
Telemark, but Soluve "sun + wolf = "halo around the sun" (Ross 1895:995)
and heluve "hell + wolf = "godless person". The data also give evidence of
lexical splits that arise when only one meaning of a polysemous word is
restituted phonologically. I will return to these examples in the general
discussion below.
In cases where lexical restrictions on change can be identified, one
explanatory factor that is often invoked, is frequency. In the following Section I
include a discussion of frequency in this respect, but for now only point to the
fact that the three relic wordforms, sjav "(-)self", hœg "holiday", and fyje
"follow" are the first, third and fourth most frequent among the /-loss words (cf.
Table 1 above). Folk "people" and mj0lk "milk" are among the last to be restored,
and they are second and sixth on the frequency list. Whereas these two words
differ substantially in frequency, no corresponding difference can be detected in
their restitution. The single most resistant dd-wordform is fjødd "mountain", but
it only appears as the fifth most frequent word on the dd-list. As can be
expected, the pronoun alle "all" is by far the most frequent word among the
synopsis words, and although it is among the most resistant dd-forms in Vinje
today, it does not stand out compared to words like fjøll and elleve "eleven" that
are much less frequent in Vestb0stad (1989). Moreover, words that in
Vestb0stad (1989) appear to be very infrequent, such as heller "cave", eismall
"alone", kjelke "sled", and talg "tallow" are among the last to be restored. Since
 PATTERNS OF RESTITUTION OF SOUND CHANGE 67

Vestb0stad 1989 cannot be relied upon in an automatic fashion in this context, I

will address the frequency issue in a more nuanced manner in the final
discussion.

4. Discussion
The starting point of this paper was four questions of general interest.
Against the background of the analysis just presented, I will proceed to discuss
possible answers to them.
The first question concerns the relation between the nature of the original
change and the restitution: Does the restitution imply that the change initially was
lexically diffused?
In speaking of lexical diffusion I will follow Wang (1969) and many
others and define it principally as a sound change that is phonetically abrupt, but
lexically gradual. Although it is questionable whether these two characteristics of
a process really are interdependent or complementary (cf. Janson 1983), I will
not discuss the phonetic characteristics of the processes under scrutiny here. The
related question of phonetic regularity is central, however. A phonetically regular
process is characterized by general productivity, and such productivity seems to
be incompatible with lexical exceptions, i.e. lexical diffusion. Lexical diffusion
and phonetic regularity are therefore truly mutually exclusive. The notion of
phonetic regularity, however, needs some comments.
In the literature on lexical diffusion and variation it is often unclear whether
it is the individual or the community that is characterized by linguistic variability.
When Weinreich, Labov & Herzog (1968) express the goal of describing the
structured heterogeneity of language, this heterogeneity is expected to be found
in the usage of both the community and the individual. Social subgroups exist
within a dialect community, and moreover, the individual is not statically defined
according to any one such subgroup. He masters several registers, at least
passively, and can choose to exploit more than one. In practice speakers do. The
consequence would seem to be that there is no such thing as a homogeneous
language system. What status can be given to the notion of regularity or phonetic
productivity within such a theoretical framework?
It seems to me that a regularly productive phonetic rule eventually must be
located within the individual. And given the notion of individual language
variability, the regular phonetic rule must be the most natural, the least conscious
and consequently the dominant among possibly alternating phonetic patterns that
the individual more or less actively employs. Within the framework of Cognitive
Grammar (Langacker 1987) a regular rule can be interpreted as a pattern or
schema that is very general, i.e. has a low degree of specificity, and which is
deeply entrenched. This underspecification is for instance a precondition for the
68 KRISTIN BAKKEN

schema's ability to attract new words on a general basis. This line of thought
does not absolutely rule out "irregular" exceptions, but generally one would
presume the exceptions to be sanctioned by a competing general schema.
Cognitive Grammar would seem to predict that the units that do not align with
such a general, deeply entrenched schema need the support of another, equally
general and equally entrenched schema to come about. Wang (1969) attributes
evidence of change by lexical diffusion to the existence of competing phonetic
rules. However, such competing phonetic rules are chronologically or
geographically distinct or at least individually distinct. Following Wang,
competing phonetic rules would by definition not be "regular", and their
interaction would cause lexical diffusion.
Hudson (1997) and Kemmer & Israel (1994:173), however, show how
morphologically general schemas can compete with phonetic schemas. The result
is exceptions to the phonetic rule, but these exceptions are in my opinion
compatible with the notion "phonetic regularity", since they are sanctioned
morphologically, not phonetically. Chen & Wang (1975:259) define sound
change by lexical diffusion in such a way that exceptions caused by paradigmatic
pressure or dialect loans do not count as evidence of such change. In the
following, the expression 'phonetic regularity' will accordingly imply "having no
competing phonetic schema undermining it".
Against this background I will now discuss my first question. When I
started out investigating the /-loss and delateralization processes, their
inconsistency immediately struck the eye. The processes seemed to be clear
instances of lexically conditioned change. What this analysis has demonstrated,
however, is that the unsystematic picture that can be drawn on the basis of
contemporary data is considerably altered when viewed in a diachronic
perspective. Lexical exceptions are found to have had regular phonetic forms in
the past, indicating that the contemporary inconsistency is a historical
coincidence rather than an aspect of the original processes. Moreover, many of
the lexical exceptions that can be identified are most coherently explained by
referring to the notion of productivity. Words that are either late loans or recent
reintroductions into the local vocabularies are seen to be outside the scope of the
l-loss and delateralization processes.
From an empirical point of view, then, it seems most likely that /-loss and
delateralization at some point were regular processes. In other words, the /-loss
and delateralization schemas at one point had no competition and were basic to
the structuring of dialect phonotactics. One possible exception must be
mentioned, however. I have not been able to attest /-loss in the preterite of verbs
like fortelje "tell" or velje "choose". The preterite formative in such verbs was -ð-
, at least in an early phase of Old Norse. Although the status of -d- vs. -ð- and
 PATTERNS OF RESTITUTION OF SOUND CHANGE 69

their respective relation to /-loss in Old Norse is controversial (cf. Footnote 1),
the vowel preceding this -ð-/-d-is lengthened and rounded in fortálde, which
indicates that the word at least has undergone one phase of the /-loss process. In
this light, words with retained / may be regarded as exceptions to the more
general rule. These exceptions are clearly morphologically motivated, however; a
general morphological schema for the preterite forms is here seen to override the
phonetic schema. Verbs with stems ending in -/ are rather few, and for these
there were at one point three preterite formatives available. Since the / is
preserved in front of two of these, d and t, there may have been strong analogical
pressure to annul the phonetic /-loss rule here. This mirrors exactly the
interpretation of the English t/d-deletion given in Kemmer & Israel (1994:173).
So it would seem that both /-loss and delateralization at one point were
phonetically regular. How then did the restitution come about? Both Chen &
Wang (1975) and Janson (1977) describe restitution as lexical diffusion in
reverse. The existence of competing phonetic schemas, such as would be the
situation for a lexically diffused change, would seemingly ease the restitution
process. The existence of an alternative phonotactic pattern would imply that the
restitution track was kept open, and the processes would easily be reversible. But
if all //-words had developed dd, and / was not admitted before consonants, it
would seem not only unmotivated, but phonotactically impossible to reverse the
processes.
Three internal factors are important in this connection. First it is a fact that
after ld had assimilated to /1:/, the system once again had the //-option, and
presumably a phonotactic //-schema. Secondly, as a result of the merger of d and
ð, and due to the absence of /-loss in preterite verb forms, /-loss would have
appeared optional before /d/. Similarily, I have suggested that /-loss never
happened after the front, high /i/, cf. examples like ON silfr "silver" and ON
stilkr "stem". The restituted pronunciations would therefore have some, if
marginal phonotactic support in the dialect.
Then there is the question of inherent variability. Even if both these
processes were regular within the individual in the way discussed above, and
even if all individuals in the community shared the same norms in this respect,
people would in most cases be aware of the pronunciation of the //-words and
/C-words in other dialects or registers. This passive knowledge, could in certain
contexts be activated, for example, when a peasant talked to the Danish minister,
the travelling horsetrader, the Dutch falconer, or the nonlocal tax collector. This
kind of variability in addition to more massive dialect contact in the border areas
would certainly be necessary to trigger restitution processes such as the two that
are described here. Thus, restitution does not seem to presuppose that the
reversed processes initially were lexically diffused.
70 KRISTIN BAKKEN

I have so far only barely touched upon the importance of dialect contact,
but it is obvious that this factor is vital when discussing the restitution processes.
My main argument in this respect can be drawn from the geographical patterns
that emerge from the diachronic data. Both the restitution of delateralization and
of /-loss are processes that start in the southeast and gradually move westward.
To the north and west Telemark is bounded by vast mountaineous areas, and the
region faces southeast in most respects. It is in my opinion not coincidental that
the restitutions originated in these dialect border areas, and that the counties
Kviteseid and Seljord now only have a few lexical relics of the sound changes
left.
Even without this empirical support, dialect border areas would seem the
most likely environment for restitution processes to start. It is in such areas that
speakers are most familiar with alternative linguistic forms. Trade, intermarriage,
permanent moves, and general mobility favor dialect contact and thus greater
familiarity with other dialects. The distinction between passive familiarity with
alternative forms and active use of these forms could easily be blurred, and
individual variability would be the result.
Lisse (1964) discusses the role dialect contact has played in some local
Danish restitution processes. He raises the question whether these restitutions
happened under the relatively recent influence of the national standard, or they
were caused by "an old restitution tendency" (Lisse 1964:180). The latter he
views as triggered by dialect contact. Although many examples seemingly are
recent introductions from the national standard, Lisse shows convincingly that
they are part of a much older restitution process that occurred under the influence
of neighboring dialects (Lisse 1964:185). Hyperrestitutions serve as an argument
in his article. Larsen (1917) also explains hyperrestitutions as a result of dialect
contact.
My analysis supports a conclusion like Lisse's. First the geographical
distribution of restituted forms cannot be given a reasonable interpretation under
the national-standard hypothesis. Secondly, the restitution processes go back in
time so far that the concept of national standard loses much of its content. There
was no spoken Norwegian standard till recently (some still contest that such a
standard exists in Norway), and if there had been, there were no channels to
transmit it. Some might argue that the official written language, i.e. first Danish
and then modern Norwegian, triggered restitution. This might hold for sporadic
examples, e.g., shilling "shilling", but the fact that many of the restituted words
are primarily local, does not favor this hypothesis. Eismadd "alone" could not
become eismall under the influence of any standard. Finally, it is worth noting
that even if the restituted words belong to the standard lexicon, they do not get
standard pronunciations after the restitution; fj0ll "mountain" has no support in
 PATTERNS OF RESTITUTION OF SOUND CHANGE 71

the standards, nor has kj0lkje "sled".5 Linguistic variability primarily caused by
dialect contact seems to be the main reason for the two restitution processes
discussed here.
I now address the third question: Did the restitution proceed lexeme by
lexeme, or by rule loss on a general basis? These alternatives may be too
simplistically phrased, because on the basis of my analysis I would answer both
affirmatively. This calls for comment. When /-loss and delateralization ceased to
be productive, it would seem that the rules as such were lost. Loss of phonetic
productivity would then change the status of the words that actualized the
phonetic processes. Their phonotactic form would be lexicalized, and if the
connection between them was still perceived, this would be in the form of a
loose lexical network (cf. Bybee 1988:125). Even when the phonetic processes
had ceased to be productive, however, their effects characterized parts of the
lexicon. The undoing of these effects constitutes the restitution process, and as I
have demonstrated in my analysis, this undoing proceeded lexeme by lexeme.
Janson (1977) describes a similar process, which he perceives as "reversed
lexical diffusion". His interpretation of the restitution phase differs from mine,
however, in that he presupposes a lexically diffused rule that comes to a halt
before its complete actualization. The restitution is consequently viewed as a
process whereby a rule loses ground.
Janson's interpretation raises the question of whether the restitution of
single words happens by rule reversal or by substitution of one wordform with
another. There are arguments to support both hypotheses. Larsen (1917) and
Lisse (1964) point to sporadic examples of hyperrestitution, and this would
imply restitution by analysis and rule reversal. Some of the examples are
dubious, however, cf. the interpretation of Hallingdal (Larsen 1917:45, Myhren
1994:14f.). Moreover, I have not been able to attest any instances of
hyperrestitution in my data.6 It is nevertheless intuitively likely that speakers
would observe the systematic correspondence between //-forms and dd-forms,
and that they could occasionally make an 'incorrect' restitution. To attribute these
odd examples to a more general restitution by rule reversal seems unnecessary,
though. The lexical substitution hypothesis does not rule out the possibility of
substitution by word-based, irregular analogy. Or to rephrase this in terms of
Cognitive Grammar: the prominence given to lexical units within this theory
5
The standard hypothesis might allow for influence by the way of intermediate forms, but
since these wordforms actually are the ones used in the neighboring dialects, it does not seem
reasonable to interpret them as 'intermediate'.
6
Hannaas (1911:8) comments on five examples of hyperrestitution of etymological dd into 11
given in a manuscript from the early part of the seventeenth century from Vest-Agder. He
confirms that one of the examples is in contemporary use and adds that such hyperforms can
be encountered in the border areas between 11-dialects and dd-dialects.
72 KRISTIN BAKKEN

does not rule out the possibility of perceiving phonological correspondences

between words and subsequently producing analogical phonological forms on
this basis. However, such 'incorrect' analogy would be constrained by the lack
of support the analogical form would have from cognates in neighboring
dialects.
To replace one wordform with another existing in a neighboring dialect,
one has to establish correspondences between dialectal cognates. Although
phonological form is relevant here, such correspondences are probably mainly
drawn on the basis of lexical meaning, and once a semantic correspondence was
established, restitution by substitution would not necessarily have to activate any
phonological redundancy rule. The very fact that restitution proceeds lexeme by
lexeme gives support to the idea of restitution by lexical substitution. When
semantically opaque placenames, unanalysable compounds and local words with
nonstandard meanings tend not to be restituted, this can easily be accounted for
under the lexical substitution hypothesis, for these are the exact cases where it
would be difficult to establish semantic counterparts. If restitution were
actualized by phonological analysis and rule reversal, such semantic
considerations would be irrelevant. My data therefore seem to be compatible
with the hypothesis of restitution by lexical substitution. Thus, in light of Wang
(1969) and Chen & Wang (1975), the restitution processes described here would
not meet the classic definition of lexically diffused sound change. These
restitutions seem to be not so much phonetic as lexical processes. I will return to
the relevance of phonetics and semantics when I return to the conditions relevant
to the lexical reversal of the effects of l-loss and delateralization.
My analysis has indicated that restitution proceeded lexically, i.e. some
words were restituted earlier, some later, and some not at all. This lexical
variation was systematic in that specific lexemes tended to relate to the restitution
process in the same way independently of geographical, individual, or diachronic
variation. It would therefore seem that the restitution process was conditioned by
inherent lexical or by social factors. I will briefly discuss some of these factors.
One possibility is that words sharing certain phonetic traits relate to the
restitution process in the same way. It is for example a possibility that the quality
of the vowel preceding the lost /, or the quality of the consonant following it, is
relevant for the restitution of the /. However, apart from the possibility of a
preceding /i/ preventing /-loss in the first place, I have not been able to identify
any such tendency. The three most resistant words, fyje for fylgje, hœg for helg,
and sjav for sj0l(v) have different vowels and in part different consonants. Even
if the number of examined words provides too small a basis for phonetic
generalizations of this kind, one would not expect to detect any such phonetic
patterns. If there had been such phonetic conditioning of the restitution process it
 PATTERNS OF RESTITUTION OF SOUND CHANGE 73

would indicate that the process itself was a rule-govemed phonetic process. But
as I have demonstrated above, restitution by lexical substitution is the most
realistic interpretation, and this would marginalize phonetic conditioning.
If restitution really proceeded by lexical substitution, speakers had to
identify one wordform with another. As I have already suggested, such
identification depends on lexical content and phonetic form. From this
perspective restitution might depend on a factor that I will label 'phonological
distance'. If the two cognate wordforms are very far apart phonologically, they
might not be identified so easily, and restitution would be hindered.
The examples of delateralization are equal as to phonological distance, all
instances of /1:/ are changed into /d:/ without any further, secondary change. In
the case of /-loss, however, this loss was often accompanied by secondary
qualitative and quantitative vowel changes of different degrees. Accordingly, the
phonological distance between dialectal cognates like fylgje and /2fy:jә/, kjelke
and /2çɔç:ә/, helgi (def.) and /2has:ji/ would seem comparatively large. It is
difficult to assess this factor, and one would really need to formalize the concept
'phonological distance' to put it to a test. However, there is one characteristic of
my data that would seem to support the relevance of such a factor: I have shown
that delateralization effects are much more uniformly restored than /-loss. Even if
this fact may be accounted for in a number of ways, for instance by the
difference between the productivity periods of the changes, phonological
distance could also be a possible explanatory factor. Whereas this factor is
constant where delateralization is concerned, it is variable in the cases of /-loss.
Although phonetic form may be relevant in etablishing a connection
between two cognate wordforms, restitution by substitution is primarily
dependent on the perception of semantic correspondences. It is therefore to be
expected that lexically diffused restitution is semantically conditioned. In my
analysis I have demonstrated that names are not restituted in the same way as
other words. This must be due to their semantic opacity. Speakers do not
necessarily recognize the semantic correspondences between lexical stems in
names and their appellative counterparts. This accounts for a nameform like
Fjøddet in an area where fjøll "mountain" is the appellative wordform in
contemporary use. Names often contain lexical relics that have gone out of use,
for example the word poll "small bay, inlet" in Bakkepodden; and these words
are not restituted either. I have also found several examples in the written sources
where lexicalized compounds contain stems that are not restituted, whereas the
simplexes are. Examples are hals "neck", but håsklute "necktie" (Ross 1906:12)
and halve "half, but håvravle "halfwit" (Grungedal, synopsis 1959). In Bakken
(1997) I argue that such compounds and names have been lexicalized as
simplexes because they are unanalysable.
 74 KRISTIN BAKKEN

Meaning must also be a relevant factor when interpreting the several

instances of lexical split in the data. Some of these must be explained by
referring to semantic idiosyncracies that tend to obscure identification. When
home is used for an elevation in farmland, whereas holme is used for an island, it
appears that the word with primary meaning has been restituted, but that the
secondary local meaning has prevented restitution here. Other lexical splits occur
in my data as well. Wille 1786 reports that the form skadde was used especially
for the crown of the head, whereas skalle was used for the head in general. The
informant from Vinje who reported a difference in the pronunciation of full
"drunk" and fudd "full" deviated from this pattern in that here it is the secondary
meaning that appears in the restituted form. This is best accounted for by
regarding the secondary meaning as a loanword in the dialect. (The local word
for "intoxicated" was drukkjen.) It is nevertheless interesting that the two
meanings are kept apart phonologically. Meaning is obviously a relevant factor
that can account for some of these lexical discrepancies.
This is not the whole story, however, for I have pointed to several
instances where neither semantics nor phonological distance suffices to explain
lexical differences in the restitution process. Since word frequency generally is
thought to be a decisive factor in lexically diffused change (cf., for instance,
Bybee 1994:296, Phillips 1984), I have also tried to assess the importance of
frequency in regard to these two restitution processes. The task is made difficult
on methodological grounds, however. One would really need frequency ratings
for different chronological stages in the development of the dialects in addition to
contrastive frequency ratings for neighboring dialects and the national standard.
As it happens, only the usage in the written national standard is reflected in the
one existing frequency dictionary. I will nevertheless list some possible
scenarios here.7
Scenario 1. The word is frequent both locally and in the contact dialects.
On one hand this would give strong support to the local wordform, but on the
other hand the pressure from competing alternative wordforms would also be
strong. The words fylgje, helg, sj0l(v), fjell, and folk are all of this type. The first
three remain as relics in unrestituted form, whereas the last two are restituted. It
is noteworthy, however, that these two words generally are among the last to be
restored. Since these five words are all among the most frequent words on the
synopsis list, and since local usage is not likely to differ from standard usage in
these cases, it would seem that frequency is a factor that makes wordforms
resistant to restitution. However, the difference between helg and folk cannot be
accounted for in terms of frequency.

7
See the word frequencies in Table 1.
 PATTERNS OF RESTITUTION OF SOUND CHANGE 75

Scenario 2. The word is frequent locally, but rare in the contact dialects.
This situation would probably prevent restitution, because the pressure from
competing cognate wordforms would be weak. There are nevertheless few
examples that exemplify this situation, the main reason being that the lexical
isoglosses are comparatively wide. However, very local personal names, like
Τον and jug are relevant examples (cf. Bakken 1999).
Scenario 3. The word is frequent in the contact dialects, but not locally.
This would favor early restitution. Many of the examples I have described as
reintroductions to the dialect may have had a low text frequency locally for one
reason or another, for example, the existence of a more common synonym. The
local wordform would then early fall prey to massive external influence.
Examples are elg and falle.
Scenario 4. Finally, the text frequency of words may change. It is obvious
that word frequency varies according to differing cultural domains. Cultural
change may therefore influence word frequency. When society is modernized,
traditional concepts may lose their culturally central position, and the words for
them be marginalized. As long as bœg and tåg were culturally central concepts,
and the words therefore frequent, their specialized local meanings would prevent
them from being restituted. The wordforms bœg and tåg were in fact for a long
time resistant to change as witnessed by the synopsis. Among the young people,
however, these wordforms have suddenly disappeared along with the way of life
they were integral to. The introduction of belg and talg for somewhat different
concepts, must be regarded as independent loans and not as restitutions proper.
Frequency is obviously a relevant factor, but as my analysis has
demonstrated, not one that can be appealed to in an automatic fashion.
Many other conditioning factors could be probed, but on the basis of my
data they are difficult to assess. It would seem that the culturally central position
of concepts such as "mountain" and "milk" would favor the traditional
wordforms, and indeed these words are among the last to be restituted. But it is
difficult to see how they differ from kalv "calf or ull "wool", except for the fact
that the latter were objects of trade.
I have shown that the restitution of the effects of /-loss and delateralization
in Western Telemark most likely was a lexically diffused substitution process
triggered by dialect contact and linguistic variability. As such the restitution
implies a more or less conscious choice between lexical alternants. Such choices
are obviously bound to be sociologically and culturally determined. Diachronic
data have severe limitations as evidence of these kinds of determinants (cf.
Milroy 1992:45-47). This holds for the data examined here as well, and I will
refrain from invoking such socio-cultural explanatory factors. It is nevertheless
very likely that some of the apparent discrepancies in the recorded distributions
76 KRISTIN BAKKEN

would appear in a different light if we could relate them to the cultural and social
context that at one time conditioned them.

REFERENCES
Bakken, Kristin. 1987. "Rettsarkiva som kilder til stedsnavn". Institutt for
namnegransking. Årsmelding 1986, 27-42. Oslo: Institutt for namne-
gransking, Universitetet i Oslo.
Bakken, Kristin. 1997. "Form and meaning. The basis for name-specific
phonological development". You name it. Perspectives on onomastic
research ed. by Ritva Liisa Pitkanen & Kaija Mallat, 21-30. (Studia Fennica
Linguistica, 7.) Helsinki: Finnish Literature Society.
Bakken, Kristin. 1999. "Personnavn i mellomnorske diplomer. Dialektform eller
skriftnorm?". Studia Anthroponymica Scandinavica 17.27-49.
Benediktsson, Jakob. 1960: "Urn tvenns konar framburð á ld í íslenzku".
Íslenzk tunga. Tímarit um íslenzka og almenna málfrceoi 2.32-50.
Bloch, Thomas. [1698] 1956. Glossemata Tellemarchica. Fyresdal 1698.
(Skrifter fraa Norsk Maalf0rearkiv ed. by Sigurd Kolsrud.) Oslo: Norsk
Maalf0rearkiv.
Bybee, Joan L. 1988. "Morphology as lexical organization". Theoretical
Morphology. Approaches in Modern Linguistics ed. by Michael Hammond
& Michael Noonan, 119-141. San Diego: Academic Press Inc.
Bybee, Joan L. 1994. "A view of phonology from a cognitive and functional
perspective". Cognitive Linguistics 5.285-306.
Chen, Matthew Y. & William S.-Y. Wang. 1975. "Sound change: actuation and
implementation". Language 51.255-281.
Christiansen, Hallfrid. 1976. (1946-48). Norske dialekter. Oslo: Tanum-Norli.
DN = Lange et al. 1849-1995.
Gjermundsen, Arne Johan. 1980. "Ei ordsamling frå 1700-talet". Årbok for
Telemark 26.9-35.
Hagland, Jan Ragnar. 1975. Hans Jacob Wille Norsk Ordbog som isœr
indeholder en Samling af Norske Ord, som iscer bruges i Sillejord og fleere
Stœder i Norge. (Kongelig norske Videnskabers Selskabs Skrifter, 3.)
Trondheim: Universitetsforlaget.
Hannaas, Torleiv (ed.). 1911. Maallœre g ordt0ke fraa Vest-Agder fraa fyrste
helvti av 1600-talet. Kristiania: Den norske kildeskriftkommission.
Hudson, Richard. 1997. "Inherent variability and linguistic theory". Cognitive
Linguistics 8.73-108.
Hægstad, Marius. 1916, Vestnorske maalf0re fyre 1350, bind II. Sudvestlandsk.
(Videnskapsselskapets skrifter, bind II. Hist.-Filos. Klasse, 1915, No. 3.)
Kristiania: I kommission hos Jacob Dybwad.
Indreb0, Gustav. 1951. Norsk målsoga. Bergen: A.S. John Griegs boktrykkeri.
Janson, Tore. 1977. "Reversed lexical diffusion and lexical split: Loss of -d in
Stockholm". The Lexicon in Phonological Change ed. by William S.-Y.
Wang, 252-265. (Monographs on linguistic analysis, 5.) The Hague:
Mouton.
 PATTERNS OF RESTITUTION OF SOUND CHANGE 77

Janson, Tore. 1983. "Sound change in perception and production". Language

59.18-34.
Kemmer, Suzanne & Michael Israel. 1994. "Variation and the Usage-Based
Model". Chicago Linguistic Society Regional Meeting 1994, vol. 2: The
Parasession on Variation in Linguistic Theory ed. by Katharine Beals et al.,
165-179. Chicago: The Chicago Linguistic Society.
Knirk, James (ed.). 1990. Norges innskrifter med de yngre runer, vol. VI, 2.
Bryggen i Bergen, bind I ed. by Ingrid Sanness Johnsen. Oslo: Norsk
historisk kjeldeskriftinstitutt.
Langacker, Ronald W. 1987. Foundations for Cognitive Grammar, vol. I.
Theoretical Prerequisites. Stanford: Stanford University Press.
Lange, Chr. . ., . R. Unger, H. J. Huitfeldt-Kaas, G. Storm, A. Bugge, Chr.
Brinchmann, O. Kolsrud, H. Mager0y & Fr. Næshagen (eds.). 1849-1995.
Diplomatarium Norvegicum, I-XXII. Oslo: Kommisjonen for Diplomata-
rium Norvegicum.
Larsen, Amund B. 1976. "Indberetning fra adjunkt Amund B. Larsen om reiser
foretagne med offentlig stipendium i arene 1888-90 for at studere
bygdemalene i Kristiansands stift". Amund . Larsen. Skrifter ed. by Magne
Myhren, 131-205. Oslo: Universitetsforlaget. (Originally printed in
Universitetets- og skole-annaler. Ny rcekke. 5th report 1890, 267-289, 6th
report 1891, 209-260.)
Larsen, Amund B. 1917. "Naboopposition - knot". Maal og Minne 1917.34-46.
Lisse, Christian. 1964. "Nogle lydlige restitutioner i 0mâlene". Dialektstudier
udgivne af Institut for dansk dialektforskning, 171-208. K0benhavn: J. H.
Schultz.
Milroy, James. 1992. Linguistic Variation and Change (Language in Society,
19.) Oxford, England & Cambridge, Mass.: Blackwell.
Myhren, Magne. 1994. "Til tvo stadnamn i Buskerud". Namn g Nemne
11.11-16.
NIYR 6:2 = Knirk 1990.
Noreen, Adolf. 1923. Altisländische und altnorwegische grammatik (laut- und
flexionslehre) unter berücksichtigung des urnordischen. (Altnordische
grammatik, 1. 4. aufl.) Halle (Saale): Max Niemeyer Verlag.
Paus 1743 = Gjermundsen 1980.
Phillips, Betty S. 1984. "Word frequency and the actuation of sound change".
Language 60.320-342.
Ross, Hans. 1906. Norske bygdemaal, II. Vesttelemaal. (Videnskabs-Selskabets
Skrifter, II. Hist.-Filos. Klasse, 1905, No. 3.) Christiania: I kommission hos
Jacob Dybwad.
Ross, Hans. 1895. Norsk Ordbog. Tillœg til "Norsk ordbog" af Ivar Aasen.
Christiania: Cammermeyer.
Sand0y, Helge. 1994. "The nature of 'overlong syllables' in the Scandinavian
languages". Phonologica. Proceedings of the 7th International Phonology
Meeting 1992 ed. by Wolfgang U. Dressier et al., 233-242. Torino:
Rosenberg & Sellier.
78 KRISTIN BAKKEN

Sand0y, Helge. 1997. "Overlange stavingar i nordisk". Nordiska dialektstudier

ed. by Maj Reinhammar, 273-283. (Skrifter utgivna av Sprak- och
folkminnesinstitutet genom dialektenheten i Uppsala, Ser. A 27.) Uppsala:
Sprâk- og folkminnesinstitutet.
Sand0y, Helge. 1999. "Breyting á hljóðlengd eða hljóðgildi? Tilhraun til
endurskoðunar á hljóðbreytingum í vesturnorrœnni málsogu". Íslenskt mál
19/20.45-83.
Seip, Didrik Arup. 1955 (1931). Norsk sprákhistorie til omkring 1370. Oslo:
Aschehoug.
Skulerud, Olai. 1918. Telemaalet i umriss. Kristiania: Olaf Norlis forlag.
Svenungsen 1821 = Svenungsen 1985.
Svenungsen, Niels. 1985. Det Norske Fjeldsprog. (Norske studiar, III.
Published by Norsk Malf0rearkiv, University of Oslo.) Oslo: Novus.
Vestb0stad, Per (ed.). 1989. Nynorskfrekvensordbok. Bergen: Alma Mater.
Wang, William S.-Y. 1969. "Competing sound changes as a cause of residue".
Language 45.9-25.
Weinreich, Uriel, William Labov & Marvin Herzog. 1968. "Empirical
Foundations for a Theory of Language Change". Directions for Historical
Linguistics ed. by Winfred Lehmann & Yakov Malkiel, 95-188. Austin:
University of Texas Press.
 THE ROLE OF MARKEDNESS IN THE
ACTUATION AND ACTUALIZATION OF LINGUISTIC CHANGE

ALEXANDER T. BERGS AND DIETER STEIN

Heinrich-Heine-Universität Düsseldorf

0. Introduction
Although the notion of 'actuation' has been recognized as a problem in
historical linguistics since the first use of this term in Weinreich, Labov &
Herzog's seminal article (1968), only few attempts have been made to come to
grips with it. The same holds true for the notion of markedness in a way—since
many linguists make use of the term, but only very few ever explore the depths
of this Prague school concept (for an overview, see Andersen 1989, 2001; for a
notable exception, see Battistella 1996). The present paper will be on the daring
side, as it is an attempt to combine the two concepts into a framework for the
understanding of language change in such a way that markedness as a cognitive
dimension is related to both the actuation and the actualization of new linguistic
forms. It will be argued that markedness on a matrix level is applicable to both
the signifiant and the signifié. Both are linked, in many cases, through simple
referential meaning. In the case of naturally marked signifiants this leads to a
linguistic environment where new items can be introduced through a process
called Markedness Agreement. In other words, if the signifiant is (linguistically)
marked on the level of form, the signifié also tends to be (cognitively) marked,
and vice versa. The ontological status of the entity referred to must therefore also
be taken into account (cf. Maes 1997 and Palmer & Woodman 1998 for a similar
problem). There are, it seems, certain designated entry points for such
innovations, and, accordingly, such actuations of linguistic change. Actuation,
i.e. real change in the linguistic system, however, only occurs when the marked
status of both signifiant and signifié are somewhat reduced to a 'normal' level,
and the innovation is actualized in different contexts, styles, and so on. Both
concepts, actuation and actualization, though different in terms and content, seem
inextricably entwined and will be regarded as inseparable in the present paper.
80 BERGS AND STEIN

These ideas will be exemplified from a detailed study of the development

of the system of English relative particles. In particular, the first developments of
the wh-series will be scrutinized from this point of view. Further reference will
be made to similar developments of do-periphrasis and genitive case markers. In
all of these instances, we find a referentially highly marked context that triggers
(in a somewhat iconic way) the use of marked syntactic structures.

1. Markedness agreement: agreement in what?

Markedness Agreement as a concept has been proposed by Andersen
(2001). He claims that "in normal agreement in case, number or gender, the rules
produce syntagms that are homogeneous in markedness and conform to the
Principle of Markedness Agreement" (26). According to this principle, drawn
from observations of ritual behavior, text structures, lexical, grammatical, and
phonological parallelisms, grounding structures of narrative discourse,
morphosyntax, morphophonology, and phonology, marked items tend to occur
in marked environments, while unmarked ones tend to occur in unmarked
contexts. For instance, nasality is commonly regarded as marked, and in
American English nasal vowels typically occur before nasal consonants, while
nonnasal vowels occur elsewhere (before nonnasal consonants and not before
consonants). This kind of allophonic variation has also been termed Markedness
Assimilation, as a hitherto unspecified (vowel) phoneme is either marked or
unmarked depending on context, i.e. it is assimilated to a given context. In
morphosyntax the matter is more difficult, naturally, as correlates of allophones
are hard to find. The expectation, however, is that in a marked environment, a
marked linguistic form can be found.
What is a marked environment, then? To illustrate this problem, a few
phenomena from English syntax will be adduced. First, consider subject-verb
inversion. This, no doubt, can be classified as a marked structure. But this
structure is not only triggered by some marked linguistic environment (such as
introductory neg-adverb structures: "Not only did I..."), but also by marked
communicative needs. With inverted structures, the speaker wants to convey
some marked, salient meaning. Take the by now classic "Into the room came
Chomsky". Here, we do not find a marked linguistic environment—it is the
speaker's point of view, the 'extra meaning element' that is marked, and which
triggers the inverted structure. Schmidt (1980) adduces similar cases from Old
English, where the 'semantic content' of the relevant passages triggers otherwise
unexpected inversions. In (1) it may be the author's wish to emphasize the
unexpectability of Wulfheard's death just at such a moment, just as Present-day
English cleft sentences are often used to express the unexpected (cf. Dorgeloh
1997), as in (2).
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 81

(1) [just when he was having such a great success] Py geare forpferde
Wulflieard "that year died Wulheard" (ASC A837, quoted from
Schmidt 1980:127)

(2) It was Clinton who addressed the topic first, not Schröder.

Another example comes from the English case marking system. The s-genitive
of 'the Queen's English' can probably be interpreted, nowadays, as more marked
than the prepositional of-genitive of 'the people's English'. Without going into
too much detail: the s-genitive is more restricted in its use and can be interpreted
as more costly in terms of processing effort in some constructions. Compare
(3.a) and (3.b).

(3) a. the man that I met yesterday, but who disappeared in the crowd's
hat
b. the hat of the man that I met yesterday, but who disappeared in
the crowd

While (3.a) leaves the addressee of the utterance (whether hearer or reader) at sea
for the whole of three clauses—Is it an s-genitive construction or a normal NP?
Which is the head? Which is the dependent?—(3.b) is quite transparent. Here the
head is clause initial and followed by the very complex (i.e. 'heavy') modifying
NP. The s-genitive, however, tends to be used with [+animate] or [+human]
possessors, so that we have a conflict of interest in the realm of syntax and
semantics. There are more complex structures and complex environmental
conditions that not only involve such features as animacy, but also [±proper
noun], [±clause], etc. that come into play. It can be argued that we are
experiencing a shift in the ranking of these constraints at the moment. While
[+human] and [+animate], both simple semantic features, have been prominent
triggers of the s-genitive so far, grammatical weight seems to be gaining in
importance, that is, if the resulting sentence is too complex to process, [+human]
is overruled by [+clause], and the of-genitive is used instead (note similar
tendencies in Heavy Argument Shift; cf. Wasow 1997). In other words, before
we postulate a driving force of (referential or semantic) markedness in language
change we need to keep in mind that in many cases a multitude of factors have to
be considered, and that syntax often overrides semantics. With this caveat in
mind we will provide an analysis of the influence of marked referents, i.e. of the
ontological status of the signifié, in the development of English relative particles.
82 BERGS AND STEIN

2. The problem
The Present-day English system of standard relative markers (that, which,
who, whose, whom, 0) has its origin in Middle English. The Old English
relativizers pe and se were abandoned, and pat, the nominative-accusative
singular neuter form of se, was generalized for all genders and cases and so took
over as the sole relative marker in the middle of the thirteenth century (Fischer
1992:296). Which, who, whose, and whom were in principle also available in the
linguistic system from the beginning of the Middle English period onwards, but
each of these relative particles has a very interesting history of its own. Apart
from their differences in frequency, each of these items developed at different
times and also developed in relation to the complex system of possible
antecedents that we know today. Who, for instance, took the longest to be fully
introduced into the linguistic system—quite unexpectedly, considering Keenan &
Comrie's Accessibility Hierarchy (1977:66), according to which subject
relativization in general is realized earlier than the relativization of direct objects,
indirect objects, etc. However, this implicational hierarchy cannot be utilized to
its full extent as the new wh-series in Middle English does not represent a
primary, but only a secondary relativization strategy. Nevertheless, it can be
argued that even unexpected secondary strategies that run counter to the
Accessibility Hierarchy serve to delimit the explanatory power of Keenan &
Comrie's theory (96).
The present paper will investigate the origin and early history of the
individual relative particles in detail, with special focus on the status of
markedness that can be ascribed to them (Romaine 1982). It will be argued that
the wh-series originated in the need to represent a clearly marked referential
object (antecedent) in a somewhat iconic way, through the mechanism referred to
above as Markedness Agreement (Andersen 2001).
This development is interesting for a theory of actuation and actualization
insofar as it shows that new forms can originate (be actuated) in (syntactically,
cognitively, socially or otherwise) salient, marked contexts, and that new forms
can be gradually actualized in the linguistic system as their markedness declines.

3. A short history of English relative particles

"[I]n the thirteenth century that stood practically alone as a relativiser"
(Fischer 1992:296; for a discussion of the origin of that, see Traugott 1972:153).
The Old English relativization strategies se (the inflected demonstrative), pe (an
uninfected particle), or a combination of both were no longer popular, and the
wh-series had not yet been fully introduced. That was subject to an "extensive
generalization" (Traugott 1972:152) and was used uniformly for all cases,
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 83

genders, and types of antecedents; cf. (4), where that is used for the [+sing]
[+fem] [+nom] antecedent þi sistor.

(4) [...] pi sistor, pat bispekez þi deth, to quelle pe heo hath ipouƷt
"your sister, who is planning your death, to kill you she has thought
of' (Saint Kenhelm, ca. 1300)

The wh-series is an exaptation of some Old English interrogative pronouns and

indefinites, hwœt, hwylc, hwa, etc. (Traugott 1972:153). Which was probably the
first of these to be fully introduced (first OED citation 1175), followed by whom
(1175) and whose (Montgomery 1989:115). Who is the one that took the longest
(1426, cf. Rydén 1983:126) and "one of the cruxes in the history of English
syntax is the long delay in the emergence of anaphoric relative who" (ibid.). It
occurred first in epistolary use, viz. in fixed letter-closing formulas, often in
wishes for God's blessings or absolution; cf. (5), (6).

(5) by the grace of God, how haue you in his blyssid kepyng (Paston
1452)
(6) blissid be Jhesu, who preserue yow bothe body and sowle (Cely
1481/82)

Although which can still be found in such formulas at the beginning of the
fourteenth century (7), it "steeply decreases towards the end of the century"
(Rydén 1983:128).

(7) by þe grace of God, which euere haue yow in his kepyng (William
Paston I, 1430)

In her extensive study of relative markers in Middle Scots, Romaine

(1982) pointed out that the emergence of the wh-series as a whole was strongly
influenced by stylistic factors. Wh-relative pronouns are first found in formal
styles and contexts. This is due to two distinct factors: (a) the imitation of Latin
patterns {qui, quod), which were regarded as markers of elaborate style and
speech, and (b) the high saliency of these new forms. Wh-forms were mainly
used in the most salient positions, where they could achieve the greatest effect. In
the following, the notion of saliency will be further scrutinized.
It would appear that there are two basic approaches to explicating our
notion of 'saliency'. The first of these might be called 'culture-based' or static, the
second, 'event-based' or dynamic. To start with the latter, there are degrees of
newness, unexpectedness, or surprise value relative to pre-existing contents of
84 BERGS AND STEIN

our cognition. The more accruals to a cognitive state are expected, the less
marked they are in respect to content. The more candidates for integration into
cognition (typically, 'events') are unexpected or run counter to the addressee's
presuppositions, the harder they are to integrate in cognition, and the costlier they
are in processing time, the more marked they are.
The first-mentioned, static notion is discussed in some detail by Lyons
(1977:570-635). The most unmarked context for Lyons is the ego, hic et nunc
situation. It is ego, the first-person narrator that seems most natural and
unmarked—often characterized by a lack of morphological marking on first-
person verbs, the hic, the 'here'-grounding of discourse, and the nunc, the
present moment—again often characterized by verb forms that are
morphologically unmarked in the simple present tense. Furthermore, several
semantic-pragmatic and cultural factors come into play, the marked features "two
or more participants", "volitional", "agent high in potency", or "affirmative" as
distinct from the unmarked alternatives "one participant", "nonvolitional", "agent
low in potency", and "negative" (Hopper & Thompson 1980). Semantic features
such as "proper vs. common noun", "definite vs. indefinite", "concrete vs.
abstract" also have to be considered (see Andersen 2001:24-33 for extensive
listings of unmarked vs. marked features). The markedness values reflect
cognitive complexity and are often represented in linguistic forms. The more ego,
hic et nunc a situation is, the less marked are the linguistic forms used in it.
However, any real-life speech situation is marked in one way or another as any
real-life speech situation involves the saliency of one or more of its participants.
And in any case, a most unmarked situation would have to be defined as such by
all its participants.
It might, for instance, be important for the speaker to use the expressive
function of language in order to "make the utterance appropriate to his attitude
towards, or his emotional involvement in, what he is talking about" (Lyons
1977:583). It can be argued that this is a function that leads to change from
above. In such cases the speakers want to convey two things: first, they want to
express a particularly affective stance towards the situation or the object referred
to, and, secondly, they want to differentiate themselves from those speakers that
are not expected to follow this trend. Here language mirrors the real-life situation
in a (somewhat abstract) iconic fashion. The object referred to is, at least in the
speaker's mind, marked, and has a special ontological status, and thus also
receives a marked linguistic form (cf. Maes 1997; Palmer & Woodman 1998).
As for the emergence of the wh-series in English, this is also confirmed by the
fact that these pronouns first occur in a marked style, such as artistic prose and
deliberately stylized texts. They were meant to be noticed. According to
Andersen, innovative forms tend to co-occur first with universally marked
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 85

features, such as [+subordinate], [+pronoun], [+writing], [+formal]. When we

take a closer look we find that who originated as a relativizer of divine
antecedents: whatever the precise semantic features of deities (e.g., + or -
animate, + or - definite, + or - abstract; this point may need to be clarified by
historical theologians), it seems intuitively clear that God per se has a marked
status (see below).
After this first stage, we find some fossilization of who-constructions.
They are no longer used as genuine, iconic representations of highly marked
referents (such as God), but rather occur in fixed expressions. These still have a
marked status, but nevertheless belong to a different class. There was not a wide
choice of letter-closing formulas and blessings at that time (Davis 1965, Schafer
1995), so that who, after its first occurrence in accordance with Markedness
Agreement, became more and more naturalized in this context. At a third stage, a
generalization of possible antecedents can be observed. Not only are God, saints,
and religious antecedents triggers of anaphoric relative who, but also good
friends, noblemen, the King, and other worthies. It is obvious that this
generalization does not eliminate the marked status of the construction
completely, but it reduces the amount of markedness by some degree (note the
gradual, rather than discrete nature of markedness). Still, socially marked
referents deserve linguistically marked symbols, but the degree of markedness
that triggers the form is remarkably lower. Gradually its marked status is reduced
even further, until the only constraint that is left is today's [+animate] (and even
this constraint seems currently to be subject to erosion). We can thus posit the
following stages in Table 1 below.

A The most marked contexts trigger the God, saints, and other religious antecedents
most marked item trigger who
 Fossilization of stage A Generalization of "God, who keep you in hys
blessing"
 Loss of saliency; expansion of the Noblemen, good friends, and worthies
range of possible antecedents; trigger who
reduced markedness
D Further relaxation of constraints; [+human] triggers who
constraints become more grammatical
and abstract; grammaticalization and
further reduction of markedness
Table 1.
86 BERGS AND STEIN

As with reanalysis, the stages of this development are quite difficult to

document in detail. Certain features of these stages, however, are more readily
accessible than others. Stage A is probably the easiest one to document. In a
survey of the Paston letters (Davis 1971) we found 124 tokens of who, whose,
and whom. The first occurrence of who dates from 1426 in a formulaic
expression like (5) or (6). The first occurrence of whose was in 1425 in a
nonformulaic context. Whom occurred first in 1430 as a prepositional
complement in a nonformulaic passage, (8). Its first occurrence without a
preposition was in 1462, in a formulaic context, (9).

(8) ... all yowr seid lettres to deluyere to my clerk, to wham I prey yow
to gyve feith ... (William Paston I, 1430)
(9) ... hit is so pat Ser John Falstof, wham God assoyle, wip opur, was
sum tryme by Ser Herry Inglose enfeffed of trust...

Of the 31 tokens of who in the letters, only four are in nonformulaic

expressions. We find two headless relatives (1472, 1478) and the first 'normal'
occurrence as late as 1481 with a socially prominent antecedent (Mastyre Baley);
cf. (10).

(10) and also Mastyre Baley, who I wende woold not haue balkyd this
pore loggeyng to Norwyche wardys (Edmond Paston II, 1481)

The fourth occurrence is in a nonformulaic expression, where the

antecedent is the King of Scotland.
The 37 tokens of whose, however, are of rather mixed nature. Twenty-two
are in nonformulaic contexts, 15 in formulaic expressions involving God, a
decedent, etc. The first occurrence is in a nonformulaic context and the
antecedent is not even of high social status (see above). Thus we may conclude
that whose already had undergone the changes postulated above and was, at that
time, at a stage of purely syntactic conditioning. Note, by the way, that even in
Present-day English whose can refer both to animate and inanimate antecedents,
a fact that underlines its special status within the system.
Whom may be regarded as in between these two poles. The 56 occurrences
are divided almost evenly between 27 formulaic contexts and 29 nonformulaic
ones. However, in thirteen of the 29 cases the relative pronoun refers to entities
of high social status. Thus we can conclude that whom is still at a stage where it
is used in formulaic expressions and with reference to entities worthy of high
regard. However, slowly but surely other uses are creeping in. It should also be
mentioned that in this period, who and whom only refer to animate antecedents,
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 87

and that which and that can both still be used in the same functions. Even in
1469 John II refers to the Archbishop by whych, Margaret Paston in 1482 refers
to the priests who are to take care of her funeral arrangements both by that and
which.
Stage  (fossilization) can be seen in example (11). Here, the relative
clause is separated by the VP from its antecedent {my granddame), a
construction that—apart from its doubtful grammatical status—lets the relative
clause appear to be an afterthought added mainly for the sake of decorum. This
move reflects the fossilized status of these formulae, that is, they are no longer
meant literally.

(11) my grandam is dyssessyd, whom God assoyle (Edmond Paston II,

1479)

Compare also (12) and (13). Here we can see that in this phrasal
construction the overt morphology does not match with the underlying cases: in
both examples it is the subject that is relativized and thus should receive the
relativizer in its nominative form, i.e. who. Instead we find oblique whom(e).
This also shows (or at least hints at) the formulaic and fossilized status of the
expression.

(12) by pe grace of God, whom haue yow in hys kepyng (William Paston
III, 1478)
(13) by godes grace, whome have you in his kepyng (Stonor, 1479)

Stages  and D are very difficult to document. In our case, the collection
seems too limited to provide a full analysis. It is commonly assumed that it was
only in the sixteenth and seventeenth centuries that the wh-series fully developed
its present-day constraints (cf. Grijzenhout 1992). Statistical analyses of fifteenth
and sixteenth-century usage would have to show to what extent which can no
longer be used with animate antecedents, and in what way who and, to a certain
extent, whom lose their marked status.
The same development can also be observed in the collection of Cely
letters, though here the latitude of variation seems to be much smaller. In
principle we find all relative pronouns with animate antecedents, i.e. Present-day
English constraints have not set in yet, but the distribution and relative
chronology of first occurrences seem more or less the same as in the Paston
corpus.
The year 1478 has the first occurrence of who in a formulaic context with
the divine antecedent and of whom with a human antecedent; to whom occurs in
88 BERGS AND STEIN

1479. Whose is used more widely and first occurs in 1479 with an inanimate
antecedent (cf. above). Of the twenty-two occurrences of who (in its several
spelling variants), all appear in formulaic use with divine antecedents. Some of
these occurrences, however, strongly suggest the first characteristics of
fossilization and loss of saliency:

(13) wyth the grace [of] God hytt shall not be long erst, who hath yow 
hys kepyng. (William Cely, 1479)

Here the head of the relative clause is separated from the antecedent by the
whole VP. This construction is not unknown in Middle English and also appears
elsewhere, but nevertheless, as it is very rare in comparison to the standard
construction, it seems to have the character of afterthought or casualness
mentioned above. Compare also (14).

(14) and that God knows, how perserue (Harold Stawntoyn, 1480)

Despite the obvious spelling problem, the case seems clear: half of the
formula has simply been omitted in this short business note—a sign of its lack of
semantic and pragmatic strength. Of the twelve occurrences of whose, one has an
inanimate antecedent, seven appear with human antecedents, four with divine
antecedents (all of which, however, are governed by prepositions). This suggests
that whose is already on the way to its modern function, and that it has
completely lost its [+divine] constraint, if it ever had one. Whom, again, is
difficult to interpret. Of the seven tokens, six have human antecedents, one has a
divine antecedent. At first sight, this suggests loss of the deity constraint. Of the
six human antecedents, however, four may have a feature [+respect], so that this
seems to be the stage of expansion, where the deity constraint is weakened (stage
C). This seems particularly clear in (15).

(15) that knowith the blessid Trynyte, whom I beseche to preserve you
into good helthe. amen. (Richard Ryisse, 1479)

The holy Trinity seems to be an unusual formula in this context. Maybe it

was used to regain some of the saliency via expressiveness (the letter, no. 66,
though short, seems very emotional anyway and documents the author's survival
of the plague, a fact that naturally calls for thankfulness). But compare also (16).

(16) and ther whe tarryd tyll the Kyngys dowter whos kyrstynd, hos name
ys Bregyt (Richard Cely II, 1480)
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 89

Here, whose («hos»), has a high-status antecedent, the King's daughter,

and seems to reflect the development through the relaxation of constraints listed
under point (C) in Table 1.
This four-stage progression can be observed in a number of developments
in the history of English, including at least do-periphrasis, genitive case marking,
inversion, glottalization, and infinitive marking.

4. Periphrastic do
Our next example comes from the development of do-periphrasis. Stein
(1990) has established a framework that enabled us to look at the origin and
development of do-periphrasis from a semantic-pragmatic point of view. It is
claimed that two different factors contributed to the development of periphrastic
do. First we find remarkableness, and secondly there is a contrastive use that is
based on remarkableness, but constitutes a much narrower restriction. What
actually, then, makes something remarkable? ffiibler (1998:133) says quite
simply, "Assuming that the speaker gets emotionally involved entails the
assumption that the propositional content of the statement, or some parts of it,
shows characteristics that in some way or another motivate his/her involvement
....[T]he contents must show qualities that can be classified as remarkable." In
other words, it is again the ontological status of the signifié or context that is
marked or unmarked and thus triggers the respective linguistic form. It is
obviously not the linguistic environment that first triggers new forms. However,
Hiibler is also at pains to stress that his notion of remarkableness is not the same
as Stein's. Whereas Stein assumes contrastivity to be the source of
remarkableness (that χ happened is remarkable only because y could also have
happened, but did not), for Hiibler it "is merely a condition for attributing to the
periphrastic form the function of expressing the speaker's involvement" (ibid.).
However, it seems clear that the latter approach does nothing but state that
something is remarkable because the speaker thinks it is remarkable, and that
both approaches fall short when it comes to the question of the initial motivation.
Only when we take outside evidence as support can we explain why something
should be remarkable at all. Remarkableness is in the final analysis nothing but
the quality of a fact, its ontology, which lets it appear as marked. This can
include contrast (something that is unexpected naturally must be marked), but
also marked signifiés. This refers to situations which Stein (1990:64) called the
authority-type of periphrasis; see (18) and (19).

(17) our saviour Christ therefore did promise (quoted from Stein
1990:64)
90 BERGS AND STEIN

(18) As the scripture sayeth, that "God through faithe dothe puryfye &
make cleane all hartes" (quoted from Hübler 1998:134)

Here again, just as with the relative pronouns discussed above, the referent
of God, holy scripture, or general association with the deity triggers the use of
marked forms. There seems no need to take the issue up again. Hübler, then,
gives an account of a phenomenon that comes very close to Markedness
Agreement, but without mentioning markedness. He cites one of the Cely letters
as an example where neither remarkableness nor contrastivity seems to be
present:

( 19) Fyddyrmore, plese hytt yowre mastyrschyppys to vnderstond that I

doo send yow at thys passage be Wylliam Smyth, packer of wullys, a
letter whereyn ys enclossyd ij letters of payment, one of John
Flewelen ... and anoder of Roger Bowser (quoted from Hübler
1998:135).

Nevertheless, if we look at the broader context of the utterance, we note

that this letter was written by a businessman for business purposes. And in this
context, according to Hübler, a functional category 'remarkableness' might well
be applied, even though at first sight this is neither a case of remarkableness nor
contrastivity. From our point of view, however, the functional category
'remarkableness' seems the same as a marked environment. Even today business
letters can hardly be called an unmarked text type. Instead they brim with highly
marked constructions, vocabulary, and style (a high frequency of the passive, for
instance). Thus, we can say that again a marked context, or marked signifié
triggered the occurrence of marked forms.
What this second case shows is that markedness and Markedness
Agreement can often be determined only when a very broad context and the
social situation are taken into account. Also, the speaker or writer, and his or her
audience should be considered. Especially in the case of remarkableness some
introspection seems to be called for in order to determine what might have led to
the use of the marked form. After all, all the speakers had other expressions at
their disposal, which they deliberately decided not to use. And it is from this
perspective that markedness and the motivation(s) for the use of marked forms
should be examined.

Addendum: Heavenly language.

It has been stated above that in many cases reference to God, saints, and
other religious entities plays a certain role in choosing the right linguistic form,
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 91

probably because of their ontologically marked status, and that this might be the
source, or rather a potential entry point, for innovations, and thus also for the
actuation of linguistic change. In this brief addendum we would like to adduce
further evidence for this claim from two Southeast Asian languages, Thai and
Burmese. These languages have very complex classifier systems. On one hand,
classifiers must be present to individuate the noun (as these denote substances,
not entities, as in many European languages) before it can be enumerated, on the
other hand, classifiers may be used as modifiers in the noun phrase, together
with determiners and adjectives. In these cases, they are used to materialize the
noun to make it modifiable. Shape, for instance, is one salient characteristic that
may be expressed within the classifier system, as with salient one-dimensional
objects in Thai, e.g., sen "long, flexible", phôm saam sen "three hairs" (Foley
1997: 237). But apart from this, we also find distinctions drawn in status and
worthiness (see Table 2, adopted from Foley 1997:237; the same phenomena can
be found in Shona, cf. Palmer & Woodman 1998). Both languages reserve
special quantifiers for religious objects (note inter alia some interesting iconic
principles in the Thai ρhrá?oŋ-?oq system in Table 2). The relationship between
religious value and linguistic form, however, is far from being universal (many
other languages show no such pattern at all). Nevertheless, there can be a
connection between a social structure and a linguistic system. Burmese and Thai
are spoken in societies that are highly stratified, much like those of medieval
Europe (cf., for instance, the gradual world picture, developed by Aquinas, in
which everything is ordered in relation to God). Much work in this area needs to
be done, however, before further conclusions can be drawn.

Sacred
phrá?oq for the Buddha, deities, and royalty
?οη for the Buddha, deities, royalty, and monks (weaker alternate)
rûup for priests, monks, and idols
thân for persons of high social rank (teachers, ministers, and
lesser nobility...)
naay for men of some social standing
ηηαη for women of some social standing
khon for ordinary persons
ton for beings of supernatural faculties (with sinister implications)
chîak for tame elephants
tua for any kind of animal or bird
Profane
Table 2. Thai categories of respect.
92 BERGS AND STEIN

hsu for Buddhas, relics, and idols

pǎ for deities, saints, monks, and royalty
û for people of high status, e.g., teachers and scholars
yau? for ordinary people
uŋ for animals, ghosts, corpses, depraved people, and children
Table 3. Burmese categories of respect

One last remark seems in order: Even in Present-day English marked

forms are commonly used in connection with God, e.g., the capital initials in
God and Lord. In some cases we even find capitalization in referring pronouns:
"I am the One Who is" (Ego sum qui sum). The latter usage, however, seems to
depend on the speaker's or writer's attitude toward the matter. Perhaps the
motivation is respect for other people's faith, rather than respect inherent in the
object referred to, as other religions and their symbols are also capitalized
(Buddha, Jewish, Islam, Allah, etc.).
We can conclude that religious entities, or religions themselves, in many
cases represent marked contexts that require marked signifiants, via Markedness
Agreement. The ontological, i.e. extralinguistic status of a given signifié, then,
plays a significant role in the actuation and actualization of linguistic change.

REFERENCES
Andersen, Henning. 1989. "Markedness theory—the first 150 years".
Markedness in Synchrony and Diachrony ed. by Olga Miseška Tomic,
11-46. Berlin: Mouton de Gruyter.
Andersen, Henning. 2001. "Markedness and the theory of linguistic change".
This volume, 21-57.
Battistella, Edwin. 1996. The Logic of Markedness. Oxford & New York:
Oxford University Press.
Davis, Norman (ed.). 1971. Ρ aston Letters and Papers of the fifteenth century.
Oxford: Clarendon Press.
Davis, Norman. 1965. "The littera troili and English letters". RES NS 16,
63.233-244.
Dorgeloh, Heidrun. 1997. Inversion in Modern English. Amsterdam &
Philadelphia: John Benjamins.
Fischer, Olga. 1992. "Syntax". Cambridge History of the English Language,
vol. II ed. by N. Blake, 207-398. Cambridge: Cambridge University Press.
Foley, William. 1997. Anthropological Linguistics. An Introduction. Oxford:
Blackwell.
Grijzenhout, Janet. 1992. "The change of relative that to who and which in late
seventeenth-century comedies". NOWELE 20.33-52.
Hanham, Alison (ed.). 1975. The Cely Letters 1472-1488. (EETS No. 273.)
Oxford: Oxford University Press.
 MARKEDNESS IN ACTUATION AND ACTUALIZATION 93

Hopper, Paul J. & Sandra A. Thompson. 1980. "Transitivity in grammar and

discourse". Language 56.251-299.
Hübler, Axel. 1998. The Expressivity of Grammar. Grammatical devices
expressing emotion across time. Berlin: Mouton de Gruyter.
Keenan, Edward & Bernard Comrie. 1977. "Noun phrase accessibility and
universal grammar". Linguistic Inquiry 8.63-99.
Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press.
Maes, Alfons. 1997. "Referent ontology and centering in discourse". Journal of
Semantics 14.207-235.
Montgomery; Michael. 1989. "The standardization of English relative clauses".
Standardizing English: Essays in the history of language change ed. by
Joseph Trahern, 113-138. Knoxville: University of Tennessee Press.
Palmer, Gary & Claudia Woodman. 1998. Ontological Classifiers as Heuristic
Systems, as Seen in Shona Class 3 Nouns. (LAUD Paper No. 427.) Essen:
Linguistic Agency.
Romaine, Suzanne. 1982. Sociohistorical Linguistics. Cambridge: Cambridge
University Press.
Rydén, Mats. 1983. "The emergence of who as relativiser". Studia Linguistica
37:2.126-134.
Schafer, Ursula. 1995. "The Late Middle English Paston Letters. A grammatical
case in point for reconsidering philological methodologies". Proceedings of
the Conference of the German Association of University Teachers of
English, vol. XVII, 313-323. (Anglistentag 1995, Greifswald.) Tübingen:
Max Niemeyer.
Schmidt, Deborah Ann. 1980. "A History of Inversion in English". Unpublish­
ed Ph.D. Dissertation. The Ohio State University.
Stein, Dieter. 1990. The Semantics of Syntactic Change. Aspects of the evolution
of do in English. Berlin: Mouton de Gruyter.
Timberlake, Alan. 1977. "Reanalysis and actualization in syntactic change".
Mechanisms of syntactic change ed. by Charles Li, 141-180. Austin:
University of Texas Press.
Traugott, Elizabeth Closs. 1972. A History of English syntax: A
transformational approach to the history of English sentence structure. New
York: Holt, Rinehard, Winston.
Wasow, Thomas. 1997. "Remarks on grammatical weight". Language Variation
and Change 9:1.81-106.
 ON THE ACTUALIZATION OF THE PASSIVE-TO-ERGATIVE
SHIFT IN PRE-ISLAMIC INDIA

Vit Bubenik
Memorial University of Newfoundland, St. John's, Canada

0. Introduction
This paper explores the role of 'vertical contact' (diglossia or polyglossia)
in the actualization of the passive-to-ergative shift in Pre-Islamic India. It is
based on my analysis of a corpus of fourteen literary (poetic) texts in late Prākrit
(called Apabhramsa) written in North West India between the seventh and
eleventh centuries A.D. (Bubenik 1998). The perusal of current writings on
reanalysis and actualization reveals that the role of 'vertical contact' has largely
been ignored by theoreticians of morphosyntactic change.

1. Functionalism and markedness

Lightfoot (1979, Chapter 2) summarizes the 'cataclysmic' view of
reanalysis whereby smaller changes accumulate in a language until they effect a
major change. Earlier, Timberlake argued against this approach by pointing out
that 'surface' changes lack motivation until reanalysis takes place (only after
reanalysis are the norms accommodated to the new structure of the grammar).
According to Timberlake (1977:168), reanalysis is made possible "by the
potentially ambiguous character of surface output", and actualization occurs as
the "consequence" of reanalysis. The latter takes place through "the elimination
of rules or subrules in the norm that are evaluated as unmotivated with respect to
the productive systemic principle established by the reanalysis".
Harris & Campbell's approach (1995, Chapter 4) differs from
Timberlake's in one important respect. Unlike Timberlake, who claims that
reanalysis always involves ambiguity, Harris & Campbell maintain that this is
not always the case. Instead, they argue that reanalysis always involves the
speakers' recognition of multiple analyses; without excluding 'true' (intentional?)
ambiguity, they maintain that in most instances two analyses simply coexist. At
variance with most recent parametric treatments (van Kemenade & Vincent
1997), which describe reanalysis in terms of a construction A that is reanalysed
96 VIT BUBENIK

as  (a new parameter is set up), they point out that only some reanalyses replace
the old analyses, but in others two (or more) analyses continue to exist side by
side. In time, very often, these two syntactic doublets may gradually become
independent of each other. According to Harris & Campbell (1995:116) two
scenarios may obtain: none of the variants is marked, and they are treated as
more or less equal alternatives; or, one variant is marked, "considered odd by
speakers". More specifically, during an extension of reanalysis, it is often the
case that one morphosyntactic variant commands more attention, is felt as
something different, marked. Gradually the innovative form or construction
becomes more familiar while its older counterpart comes to be felt as an archaic,
i.e. highly marked form or construction, which is eventually lost. The path of
spread through the community follows an S-curve, with the change in
markedness occurring at its mid-point.
The stance on markedness of Harris & Campbell is remarkably close to
that of Functional Grammar, as proposed by Dik (1989:41). According to Dik
the markedness value of a linguistic item is not fixed and immutable, but rather it
may vary (i) with the environment in which it is used and (ii) with the frequency
with which it occurs; put succinctly, (i) what is marked in one environment, may
be unmarked in another; and (ii) when frequent use is made of marked forms,
they gradually lose their markednes. Dik describes this loss of markedness and
labels it 'markedness shift'. It is to be understood as a historical process whereby
an originally marked item loses its marked character and ultimately makes room
for a new marked form. Two principles are claimed to underlie this process: (i)
the need for especially expressive items to achieve special effects in
communication, and (ii) the tendency to overexploit such items and thus subject
them to a process of 'inflation'. The schema in Figure 1 illustrates this process of
markedness shift:

Figure 1. Markedness shift (after Dik 1989:42).

At stage 1 there is an opposition between an unmarked form 1 and a

marked form E2. At stage 2 the marked form E2 has become unmarked, and has
ousted E1 in usage. At stage 3 a newly created marked form E 3 has been
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 97

introduced, so that the original opposition is restored. In 1989 Dik argued that
ergative systems can arise through markedness shift operating on the
active-passive opposition of a nominative language (242-246). I tested his
proposal on the data summarized in Figure 2.
A number of intermediate forms are found between Old Indo-Aryan krta
"made" and its ultimate reflex in Hindi kiyā (< kiau < kida < kita < krita < krta).1
Tena in stages 1 and 2 is the instrumental form of sa "he". In stage 3 (Hindi) the
ergative case is expressed analytically by cliticizing the ergative postposition -ne
to the oblique form of the pronoun yah "he, she; that". The source of the ergative
postposition -ne is not clear (the Old Indo-Aryan instrumental suffix -ena ends
up as the suffix -e/-ĩ in some New Indo-Aryan languages). At least one
intermediate form is found between Middle Indo-Aryan tena and New Indo-
Aryan us=ne; in early New Indo-Aryan texts tin (< tena) is used in honorific
contexts and the oblique form tihi elsewhere.

Figure 2. Markedness shift in the history of Indo-Aryan (after Bubenik 1998:134).

Old Indo-Aryan could form passives from transitive verbs; the unmarked
expression for past events was the aorist akārsat or the reduplicative perfect
cakāra "he did" (with different aspectual values progressively blurred during the
Old Indo-Aryan period); the marked expression was the nonfinite passive
construction tena krtam "[it is] done by him", based on the past participle in -ta.
At a certain point in time, a markedness shift occurred in the passive construction
of Old Indo-Aryan. The originally marked passive became less and less marked
and eventually ended up as the unmarked construction, which pushed the active
construction out of use altogether (the perfect form was discontinued by the end
of the Old Indo-Aryan period, and the sigmatic aorist akāsi survived until the
Middle Indo-Aryan period in Pāli and Ardha-Māgadhī). In later Middle Indo-
Aryan, although there was no active construction for the expression of past
completed events, the construction with the ία-form nevertheless followed the

1
In accordance with the author's practice, syllabic r is notated as r.
98 VIT BUBENIK

rules applying to the nonfinite passives of Old Indo-Aryan: if there is agreement

in the verb form, it is triggered by the Goal Subject (= patient), the agentive
phrase need not be present, and in general all rules involving subjects will
operate on the patient rather than on the agentive phrase. On the other hand, the
main reason for speaking of a 'passive construction'—viz. the existence of an
opposition with a corresponding active one—disappeared since by late Middle
Indo-Aryan times the construction tena kiyau (< kida < krta) "he did" was fully
incorporated into the inflectional paradigm of active tenses (karai "s/he does",
karisai "s/he will do"). Furthermore, the demarked passive transitive
construction is in conflict with the normal markedness relations, and we may
expect strong pressure to reinterpret the agent of the demarked passive
construction as subject and the patient as object.
A number of objections may be raised against the scenario in Figure 2. The
three stages that span more than two thousand years are obviously very 'discrete'
and present essentially a too schematic scenario of typological change. Stage 2 is
based on the unwarranted assumption that Middle Indo-Aryan passed through a
stage being an ergative language without a passive construction. My corpus
demonstrates that the ordinary "be"-passive and the inherited finite passive were
available throughout the Middle Indo-Aryan period, and that the new analytic
passive construction with the auxiliary jânâ "go" appeared at its end. Thus we
have to add at least one more stage between 2 and 3 to obtain a realistic picture of
typological changes during the Middle Indo-Aryan period; see Figure 3.

Obsolete Unmarked Marked

Early MIA akāsi tena kida (Pass ~ Erg) tena kida āsi
("be"-Pass)
Late MIA tena kiyau (Erg) tena kiyau gayau
("go"-Pass)
NIA us=ne kiyâ (Erg) us=se kiyā gayà
("go"-Pass)
Figure 3. Coexistence of passive and ergative construction during
the Middle Indo-Aryan period.

Strictly speaking, the full-fledged passive construction was established as late as

the New Indo-Aryan period (in the seventeenth century), when new
postpositions marking the agentive phrase in the passive construction {-se or
-ke dvārâ) developed, distinct from those used in the ergative construction;
contrast Hindi us=ne kiyā "he did [it]" with us-se kiyâ gayā "[it] was done by
him".
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 99

Another untenable assumption of the functional scenario, which relies

exclusively on markedness relations, is the claim that during the formative stage
2 the marked form E2—after losing its markedness—drove the unmarked form
E1 out of business (Dik 1989:42). Furthermore, Dik did not specify its status in
terms of markedness; given its 'oddity' we should entertain the idea that this
form was actually (doubly) marked. Even more seriously, it will be shown that
in diglossic and polyglossic societies it is very difficult (or practically
impossible) to drive an archaic or obsolete form 'out of business'. Here the
functional scenario is in need of subtler reasoning involving the generalization of
an innovation across the social and stylistic categories of a language.
In spite of these shortcomings, the functional theory of markedness makes
a useful typological prediction about the rise and disappearance of ergative
systems in a language or language family. Their development is explained in
terms of well-established principles such as subject and object assignment,
markedness relations and case marking; and the theory tries to account in a
natural way for the phenomena that seem to relate the ergative construction to the
passive construction.
Nevertheless, to view morphosyntactic change as arising in order to fill a
gap in the grammar is hardly satisfactory. Actually, one could argue that there
was no hole to fill in the grammar of Middle Indo-Aryan dialects (contrast
Figures 2 and 3). In Section 2 it will be shown that the real momentum in the
development of the ergative construction came from the ambiguity of the
ancestral construction with the past participle, which could be interpreted in two
ways; and in Section 3 I will argue that the functional approach to the passive-to-
ergative shift has to accord significantly larger space to the triggering phenomena
of reanalysis inherent in the factors of 'horizontal' and 'vertical' contact.
While 'horizontal' contact (e.g., in a Sprachbund) has for the most part
been acknowledged in historical linguists' models of change, the 'vertical'
contact—i.e. the stylistic, bidialectal, and diglossic factors that expose the Low
variant to influence from the High variant of the same language and vice
versa—has been far less prominent in theoretical writings. Harris & Campbell
(1995:86) are unsympathetic to admitting diglossic accounts of the history of
English (advocated by Lightfoot 1991) into their version of reanalysis cum
actualization, and claim that the term 'diglossia' was intended to distinguish "the
relatively rare examples" of distinct language varieties serving distinct social
functions. However, linguists investigating speech communities that end up at a
certain period of their history with more or less stable diglossia—Hellenic,
Arabic, or Indic—cannot avoid the issue of 'vertical' conditioning of reanalysis.
The best known case is that of Byzantine Greece, which may be described
as 'triglossia' or the coexistence of "three distinct sets of registers of literary
100 VIT BUBENIK

language" (Browning 1989:51). The educated Byzantines could resort to the

High variant of classicizing Attic Greek on special occasions and to the
intermediate variant of 'literary koine' in technical and official writings; or they
could make various concessions to what they really spoke, as they did in works
of popular edification. During the eleventh-twelfth centuries the Low variant
acquired its own identity as 'popular koine'. To exemplify some of the options
faced by the Byzantines on the level of morphology and syntax one can mention
the augmentation of prefixed verbs in the imperfect and the aorist. The High
variant places the augment after the prefix, e.g., kat-e-dékheto "he was
accepting", but the other variants could place it before the prefix, e-kata-dékheto]
and there are also doubly augmented forms, e-kat-e-dékheto, documenting the
uncertainty in the minds of diglossic (polyglossic) speakers of Byzantine Greek.
(Standard Modern Greek uses the augment only when it is accented, with some
residual problems arising in pairs such as katd-laba "I understood" vs. kat-é-
laba "I seized"). Numerous examples of syntactic doublets and triplets are found
in analytic formations which started replacing earlier synthetic aspectual forms
during the early Byzantine period (fifth-seventh centuries). For instance, the
High variant would use the Attic reduplicating perfect, as in gégrapha "I have
written", while the lower variants could avail themselves of eimi gégraphös
(copula plus the active perfect participle) or ékhö gramménon ("have" plus the
mediopassive perfect participle) or ékhö grápsas ("have" plus the active aorist
participle). The last two analytic formations have survived to the present time, the
former in a specialized, resultadve meaning with certain verbs, while in the latter
the 'illogical' active aorist participle has been replaced by the infinitive, yielding
ékhö grápsei.
In Arabic-speaking countries one can distinguish several 'registers' or
'stylistic levels' between the two extremes of 'standard' or 'Neo-Classical' Arabic
and the 'plain, colloquial' (or 'vernacular of the illiterate'). The middle level,
called āmmiyyatu 'l-muOaqqafin "common language of the educated", emerged
in Egypt under the impact of European civilization. Currently, this variant is the
informal language used by educated Arabs; one of its most important
'koineizing' properties is its mutual comprehensibility to speakers of different
Arab nationalities (cf. Versteegh 1984).

2. Reconstructing sociolinguistic variation in Pre-Islamic India

Most theory-driven approaches to the study of diachronic syntax will have
difficulty with the inherently 'untidy' state of affairs in the literary corpus that has
come down to us from the four Pre-Islamic centuries in India (seventh-eleventh
centuries). It is not only the existence of numerous morphosyntactic doublets in
Apabhramsa texts, but also the latters' coexistence with literary texts harking
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 101

back to previous linguistic stages of Prākrits and Sanskrit. Any investigation of

the far-reaching morphosyntactic changes during the late Middle Indo-Aryan
period has to take into account the vertical influence of the prestigious Classical
Sanskrit on the Prākrits, the intermediate variant, and Apabhramsa (the Low
variety). On the other hand, one also observes increasing influence from
Apabhramsa, the Low variant, on the Prākrits and Sanskrit. The study of the
ordinary genetic transmission in the Middle Indo-Aryan dialectal continuum
('evolutive' grammatical change) cannot be divorced from its accompanying
socio-cultural and stylistic factors. (For an example balancing these two
dimensions, see Sch0sler 2001:181) It is not easy to visualize all these processes
simultaneously. Figure 4 is offered as a crude synoptic diagram putting some of
these matters together (see also Bubenik 1998:223). The ordinary genetic
transmission (→) represents only fundamental morphosyntactic changes in
nominal and verbal systems. This genetic transmission can be thought of as
consisting of several 'teleological' drifts: the loss of old finite tenses and the
erosion of the declensional system between the early and middle periods, and the
passive-to-ergative shift and the rise of the postpositional case at the end of late
Middle Indo-Aryan. The ascending arrow ( ↑ ) represents the elevation to literary
status of a former spoken variety; the broken ascending arrow represents the
learned transmission of Classical Sanskrit (comparable to that of Ancient Greek
and Hebrew in medieval Europe); the two vertical arrows ( ↑ ↓ ) stand for the
bidirectional influence from the High on the Low variety (sanskritization) and
vice versa (prakritization); and finally, the vertical arrow ( ↑ ) indicates the use of
regionalisms (i.e. the influence from a spoken variety on the written language).
All the phenomena represented by nonhorizontal arrows are essentially vertical
contact phenomena resulting from the diglossic (polyglossic) character of
medieval Indic society; they will play a crucial role in our explications of the
actualization patterns of the above drifts. Figure 4 also captures the fact that
diglossic relationships during the Middle Indo-Aryan period changed from
earlier Sanskrit-Prakrit diglossia to later Sanskrit-Prākrit-Apabhramsa
'triglossia'. (The linear conception of triglossia may be in need of further
elaboration; instead of viewing Prâkrit(s) as the intermediate variety, a strictly
binary conception of diglossia would entail viewing this situation as 'double'
diglossia—Sanskrit-Apabhramsa and Prākrit-Apabhramsa.)
The upper time limit of the eleventh century for our investigation has been
set arbitrarily. Post-Islamic India (after the twelfth century) emerges with a new
High variety rivaling Sanskrit, viz. Persian (Dari), the lingua franca of Eastern
Islam. Its pervasive influence on New Indo-Aryan languages has so far been
studied only lexically. To the received 'classic' diglossic relationship, islamized
strata of the Indian population added 'superposed' bilingualism in Persian, to
102 VIT BUBENIK

follow Fasold's refinement of Ferguson's theory of diglossia (1984:54). The

current linguistic situation in India has been described as 'horizontal' diglossia in
Hindi and Urdu by Kelkar (1968).

Verb: loss of old finite tenses → passive-to-ergative shift

Noun: erosion of the old declensional system → postpositional case
Figure 4. Sociolinguistic aspects of the history of Indo-Aryan languages.

Linguistic properties of the High variant were codified by Pānini

(sixth-fifth century B.c.) in his famous grammar of (late Vedic) Sanskrit (called
Astâdhyàyï). Relying on a longstanding grammatical tradition and his native
intuitions Pānini described the variety spoken by the upper echelon of society
(the brahmans) in the northwestern corner of India. We know that there were
other contemporary regional and social varieties of the same ancestral Indo-
Aryan language spoken in other parts of India. Suffice it to mention Pānini's
illustrious contemporaries (?), Buddha (*486) and Mahâvīra (*468), born in
eastern ksatriya clans outside the brahmanical region of Àryâvartta, who spoke
the eastern varieties of Indo-Aryan. Buddhism and Jainism showed distinct
preference for the eastern regional dialects ancestral to literary dialects such as
Buddhist Pāli and Jainist Ardha-Māgadhi. For a number of religious, cultural,
and social reasons, Paninian Sanskrit became the unified stable norm for most of
the writing activity in the High variant during the whole Middle Indo-Aryan
period. Close scrutiny of the Medieval Sanskrit corpus reveals the existence of a
'classicizing' or 'purist' effort to emulate an ever distant Pāninian model, on one
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 103

hand, and various concessions made in the direction of the native code of its
authors, on the other. The 'quality' of nonnative Sanskrit in brahmanical works
varies considerably, and Dandin's celebrated Dasakumāracarita 'The Tale of the
Ten Princes' (seventh century) is probably the only medieval work that uses the
synthetic perfect as the inferential mode in accordance with Panini's description
of it. At the end of this continuum one could place the so called Buddhist
'hybrid' Sanskrit. It was based on a regional Middle Indic variety which was
extensively sanskritized (Edgerton 1953). Among its salient features are genitive
forms (of u- and /-stems) such as bhiks-usya "of the monk" vs. Paninian bhiks-
oh, modeled on the regional form bhikkh-ussa (similarly, mun-isya "of the sage"
vs. Pāninian mun-eh, modeled on mun-issa). In morphosyntax one finds the
ambiguous clitic form -se "to him/her", which is never used in Pāninian
Sanskrit (dehi-se "give him/her" would be realized as dehi asmai "give him" or
dehi asyai "give her").
Of immediate significance for our further inquiry into the passive-to-
ergative shift is the emergence of the analytic perfect construction in Classical
Sanskrit prose (also in Pāli) of the type kr-ta-vān asmi or aham (lit.: "having-
made-one.MASC I-am") "I have done" (and its feminine counterpart kr-ta-vat-ï
asmi or aham). Its full incorporation into late Sanskrit conjugation can be
understood as one of the solutions to the ambiguity of the construction with the
past participle accepted by the High variant. We saw above that the ambiguity of
Early Middle Indo-Aryan tena kida "[it is] done by him" ~ "he did [it]" was
alleviated by the appearance of the "go"-passive, which allowed for an
unambiguous active interpretation of the former passive construction. The High
variant did it differently, recycling the past passive participle as a past active
participle by means of the possessive suffix -vān (in Old Indo-Aryan this suffix
could be attached to adjectives and nouns, e.g., bhág-a "good fortune" → bhága-
vān "possessing a happy lot, fortunate"). In syntactic terms, the solution of the
High variant was in keeping with the nominative-accusative typology of Old
Indo-Aryan, whereas the solution offered by the Low varietie(s) resulted in the
ergative-absolutive typology of late Middle Indo-Aryan. One of the major
contributing factors in the latter process was the elimination of the nominative vs.
accusative contrast in Apabhramsa (nar-ah vs. nar-am > nar-u "man"). The
resulting absolutive form allowed for the active (= ergative) interpretation of the
former passive construction when its agent is specified; see (1).

(1) a. naru mānyau

man.ABS kill.pp
"The man was killed" (passive interpretation).
104 VIT BUBENIK

b. maĩ nar mānyau

I.INSTR man.ABS kill.pp
"I killed the man" (active interpretation).
(2) naram māritavān asmi
man.ACC kill.pp.poss I-am
"I killed the man".

The active counterpart to (l.b) in the High variant is given in (2). Nevertheless,
depending on the context, even the Apabhramsa construction in (l.b) could be
interpreted passively (cf. Section 3.2), and it was only during the New Indo-
Aryan period, after the crystallization of the "go"-passive and the establishment
of two different postpositions, as argued above, that this ambiguity was fully
sorted out; contrast Apabhramsa maĩ kiyau "[It was] done by me" ~ "I did [it]"
with Hindi maï=ne kiyā "I did [it]" vs. maĩ-se kiyā gayā "[It was] done by me".

3. Multiple analyses in the actualization of the passive-to-ergative shift

This section will explore the issue of the possible passive or ergative
interpretation of the same syntactic structure containing the past participle in
different semantic, pragmatic and discourse environments. In some instances the
two analyses continued side by side while in others the ergative reanalysis
replaced the old passive analysis. The following environment factors will be
considered: (i) agents differing in animacy, (ii) information structure, (iii)
discourse type, and (iv) predicates differing in illocutionary force.

3.1 The status of an argument low in animacy

Where there is a single instrumental-marked inanimate noun in the
construction with the past participle we are often dealing with a real instrument
rather than an agent.

(3) cauhĩ cayâri turāgama ghâiya (Rittha 3.7.4)

four.INSTR four horses wound.pp
"The four horses were wounded with four [arrows]".

Here the passive interpretation is more likely than the ergative intepretation "The
four arrows wounded the four horses". The unlikelihood of an argument low in
animacy in the function of ergative subject is also observable in causative
constructions such as (4).
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 105

(4) munivara-vayanem diyavaru vāsiu (Pc 28.10.1)

muni-best-word.INSTR brahman-best live.CAUS.pp
"The brahman was converted by the word of the sage" (rather
than "The word of the sage caused the brahman to live").

3.2 An agent in the marked postverbal position

These instances are to be interpreted passively because of the well-known
tendency to place the agent of passives in the marked sentence-final position. In
functional sentence perspective, the passivization can be understood as one of the
means of assigning the function of focus to the agent (of course there are other
means that do not involve permutation of word order, such as sentence accent
and the particles used in New Indo-Aryan languages). Out of the six possible
permutations of the 'unmarked' sequence Ag(ent) Go(al) PP—assuming that
Apabhramsa tended towards the SOV type of most New Indo-Aryan
languages—two allow for a passive interpretation, namely Go PP Ag and PP Go
Ag. Generalizing from limited statistics (for the declarative mode) based on the
three cantos 81-83 of Puspadanta's Harivamsapurdna (tenth century A.D.), it
appears that Go PP Ag and PP Go Ag are the least common options; see Table 1
(based on Bubenik 1998:128, 148).

Word order Reading Number of tokens

Ag Go PP 57
Ag PP Go 40
Go Ag PP 26
Go PP Ag passive 9
PP Ag Go 10
PP Go Ag passive 8
Table 1. Word order in sentences with past participle in Harivamśapurāna.

An example of PP Go Ag is provided in (5). It goes without saying that (5) can

also be interpreted actively when focusing on the agent in an appropriate context,
i.e. "[It was] Krsna [who] protected the gokula".

(5) parirakkhiu gokula kanhë (Rittha 5.8)

protect gokula Krsna
"The gokula was protected by Krsna".

3.3 Dialogue
In the first and second persons we can a priori maintain that the
spontaneous use of language in dialogue does not favor the marked passive
106 VIT BUBENIK

interpretation of the source passive construction. Consider the (late) Old Indo-
Aryan passive construction in (6.a) and its intermediate Middle Indo-Aryan
descendants (6.b and 6.c) and eventual ergative reflex in contemporary Hindi
(6.d).

(6) a. tatah Ketumatyā aham grham riitah (ΟΙΑ)

then Ketumatl.INSTR ISG.NOM house.ACC take.pp.M
"Then I was taken home by Ketumatī".
b. tā keumaië haü gharaho riïya (Kc 6.12.1)
then K.INSTR 1SCABS house.GEN/DAT take.pp
 *tâ keumaië majjhu gharaho nïya
then K.INSTR ISG.GEN/DAT house.GEN/DAT take.pp
d. phir ketumati-ne mujhe ghar-me liyd (Hindi)
then K=ERG 1SG.DAT/ACC house=LOC take.pp
"Then Ketumatl brought me home".

Notice the atypical morphology of the 1SG object in (6.b): haü "I" can be
categorized as the nominative case (from the point of view of Sanskrit, the High
variant) or as absolutive (from the point of view of Apabhramsa, whose system
of personal pronouns possessed an oblique form maĩ Acc/Instr). Clearly, during
this 'proto-ergative' stage the interpretation of (6.b) could swing either way
(towards the marked passive or unmarked active interpretation) depending on the
context. Postulating the existence of a later intermediate construction in (6.c) with
the ISG object in the GEN/DAT, we reach the state of affairs when the unmarked,
active interpretation becomes the only one possible. The final 'improvement' on
the previous state of affairs regards the analytic case marking in contemporary
Hindi.
Returning to the ambiguous stage (6.b), two more points are in order.
Speakers of early Apabhramsa could avail themselves of the marked synthetic
passive form in the first and second persons; for the above reasons, these
instances are sporadic in our Middle Indo-Aryan documents (e.g., munijjami "I
am recognized", thuvvahi "you are praised", suvvahi "you are heard"). This
would make the active interpretation of (6.b) more likely. On the other hand,
diglossic speakers (of Sanskrit and Prākrits) always had recourse to the source
construction in the High variant in (6.a), and we may assume that they would
favor the marked passive interpretation of (6.b).
Instances of genuine indeterminacy between the ergative and the passive
interpretation obtain in contexts where the speaker may be presenting a given
content from the point of view of his listener(s) or from his own viewpoint; see
(7).
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 107

(7) aharn-iha, bhadda! tumhehim nitthāriu bhav'uyahihi

1SG=PTCL fellows ye.INSTR save.pp being-water.LOC
devi niya do vi hatthau (Sc 755.1-5)
give.GER own two PTCL hands
"Dear fellows, you saved me from the ocean [of samsara] by
giving [me] [your] two hands" (presented from the point of view
of the addressee)
"Dear fellows, I was saved by you ...." (presented from the point of
view of the speaker).

In dialogue it was possible to avoid the ergative construction by compounding as

shown in (8.a). In this case the pronominal speaker-agent does not have to be
realized by the oblique form mal (ACC/INSTR); the compound of the type krta-
kàrya (lit.: "the having done-deed one") agrees with the subject in the
nominative. In the context of dialogue in Apabhramsa this construction is a nice
example of the influence from Sanskrit, the High variant; see (8.b). An ordinary
ergative counterpart to (8.a) is given in (8.c).

(8) a. haũ kaya-kukajja (Hv 81.16.5)

1SG.NOM done-bad-deed
"I have done the bad deed" or rather "I am the one who has done a
bad deed".
b. aham krta-kāryah (Sanskrit)
ISG.NOM done-object (= gerundive "to-be-done")
"I have accomplished my object" (= "I am satisfied").
 mal kukajja kaya
1SG.OBL bad-deed do.pp
"I have done the bad deed".

In functional terms (8.a) assigns the marked value of focus to the speaker-agent;
this reading is absent from (8.c) under normal sentential-stress conditions. In
short, (8.a) has to be evaluated as a convincing example of the influence from the
High variant; more specifically, we are dealing with an imitation of the Sanskrit
descriptive compound (karmadhâraya) krta-kukârya, prakritized kaya-kukajja.

3.4 Verbs of speaking

Verbs of speaking fall into two groups, those which pattern only ergatively
{kahiu "told", pucchiu "asked", both as in Hindi) and those which pattern both
ergatively and nominatively (yolliu "said"—in Hindi bolna patterns nomina-
tively, pajãpiu "said", pabhaniu "addressed", vuttu "said"). The addressee may
108 VIT BUBENIK

be in the absolutive or the 'oblique' case. The verbs which have the addressee in
the absolutive are vuttu, pucchiu, and pabhaniu; those which have the addressee
in the oblique case are volliu, kahiu, and pajâpiu. The verbs that pattern both
ergatively and nominatively have the addressee in the absolutive case only when
the speaker is realized by the instrumental; if there is no addressee they may
resort to the absolutive marking for the speaker. Verbs which can have the
addressee in the genitive/dative are free to realize the speaker by either the
instrumental or the absolutive. Table 2 shows the statistical distribution of these
syntactic constructions in the artistic idiolect of Svayambhüdeva (eighth century):

Speaker "speak" + PP Addressee Gloss No. tokens

INSTR volliu GEN/DAT "said" 2
ABS volliu GEN/DAT "spoke" 5
INSTR volijjai (finite passive) 4
INSTR kahiu GEN/DAT "told" 1
INSTR kahijjai (finite passive) GEN/DAT "said" 2
INSTR pajâpiu 3
ABS pajâpiu GEN/DAT "said" 10
INSTR vuttu ABS "said" 10
ABS vuttu 1
INSTR vuccai (finite passive) ABS 14
ABS vuccai (finite passive) 1
INSTR pucchiu ABS "asked" 7
INSTR pabhaniu ABS "addressed" 7
ABS pabhaniu "said" 2
Table 2. Verbs of speaking in Svayambhūdeva (Pc 1-38, Rittha 1-7).

The competition between the nominative and the ergative patterning of volliu (5 :
2), pajâpiu (10 : 3), and vuttu (1 : 10) could be taken as indicative of the ongoing
shift towards ergativity in Western India. It would appear that in the unmarked
context of the third person the shift lagged behind that in the marked context of
the first and second persons; here we may surmise that the innovation was
motivated communicatively in that it ran its complete course in the most salient
environment. Further complication is introduced by the fact that certain verbs of
speaking appear also in the finite passive form (limited to 3SG): vuccai (< ΟΙΑ
ucyate), vollijjai (innovative in Middle Indo-Aryan), kahijjai (< ΟΙΑ kathyate) "it
is said". Here I suspect that the passive morphology is due to influence from the
Sanskrit passive construction tena ucyate "[it] is said by him". To judge by the
limited statistics in Table 2, the sanskritizing construction with vuccai (14
tokens) was actually more common than its ergative counterpart with the past
participle (10 tokens). In both types—the ergative and the sanskritizing
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 109

passive—there appeared isolated instances of the agent phrase expressed by the

absolutive case (padihāru vuttu "the doorkeeper said" and var vuccai "the hero
says"). Without further statistical information we can perhaps regard them as
symptomatic of the code mixing that is characteristic of diglossic situations.
More specifically, the diglossic speaker could refer to past events with a
nominative and a finite form, i.e. sa avoca "he said" (as in early Middle Indo-
Aryan) instead of tena vuttu (lit.: "by him said"). In the finite passive vuccai one
cannot exclude contamination with the active form vacai (so_ vacai x tena vuccai
= so vuccai). But these matters should also be studied with due attention to
transitivity; one observes that in Hindi, bol-nâ (< Middle Indo-Aryan voll-)
"speak, talk" patterns ergatively when used as a transitive verb (us=ne jhūth
bold "he spoke falsehood" = "he lied"), but nominatively elsewhere (vah
mujh=se hindi me bold "he spoke to me in Hindi").
In dialogue, a marked context in relation to the unmarked context of third-
person narrative, we do not find instances of this markedness 'disagreement';
here the absolutive forms (haü "I", tuhu "you") are found only in the unmarked
present tense; and conversely the marked oblique forms (maĩ "by me", pal "by
you") are found in the marked past tense. See the pertinent Apabhramsa
examples in Table 3.

lSG agent 2SG agent

Present tense
haü pal pucchaũ "I ask you" tuhü maĩ pucchahi "you ask me"
I you. OBL ask you I.OBL ask
UU UU
Past tense
tuhü maĩ pucchiyau "I asked you" haü paï pucchiyau "you asked me"
you I.OBL ask.PP I you.OBL ask.PP
M M M M
Table 3. Markedness relations in Apabhramsa.

We can thus conclude that the passive-to-ergative shift was actualized

earlier and more widely in the context of the discourse-participant persons than
in the third person. This point applies a fortiori to causative structures, where in
the first and second persons the spontaneous use of language does not favor the
passive interpretation:
110 VIT BUBENIK

(9) pādiu haü calanehĩ deva (Pc 30.11.7)

fall.CAUS.pp 1SCABS feet.LOC lord
"Lord you made me fall at [your] feet" (literally, "I was caused to
fall...")

3.5 Interrogative mood

In addition to dialogue, the active interpretation of the past-participial
construction is more likely in the interrogative mood (in any person). For
instance, the passage in (10) is to be interpreted "Have you seen [my] beloved,
while passing in front [of you]?" rather than "Has [my] beloved been seen by
you ...".

(10) ditthi pia paĩ sãmuha janti (Vikr 4.45)

see.pp.FEM beloved.FEM you.OBL in-front passing

The same is true—a fortiori—of causative constructions, as in (11), where the

reading "Or did someone show me an illusion?" is more likely than "Or was I
caused to see an illusion by someone ?"

(11) indiyâlu kĩ vā kina-νί darisiu (Sc 604.3)

illusion Q or someone.INSTR see.CAUS.pp
"Or did someone show me an illusion?"

3.6 Causative constructions

Given the increased valency of the causative past participle, the active
interpretation is preferable even when factors such as sentence-final agent would
favor a passive interpretation with a noncausative past participle; cf. (12).

( 12) ghosana puri devāviya kâsë (Rittha 5.11.1 )

proclamation city give.CAUS.CAUS.pp Kamsa
"Kamsa let proclaim [in] the city"

Here, in spite of the causer being in postverbal clause-final position, the active
interpretation is preferable to the passive one (lit.: "it was caused to be
proclaimed... by X").
Generalizing from the examples presented in Sections 3.1-3.6, we can
conclude that in (late) Middle Indo-Aryan the passive-to-ergative shift was
actualized earlier and more widely in contexts with agents high in animacy and
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 111

with agents in sentence-initial position (i.e. in unmarked environments). The

development also proceeded with differences for discourse participants (first and
second vs. third person) and for illocutionary force (interrogative vs. declarative
mood). In these two instances the passive-to-ergative shift spread from marked
to unmarked contexts; here the innovation appears to be communicatively
motivated in that it occurred first in the most salient and monitored environments.
Direct speech and questions structured ergatively undoubtedly offer the speaker
an expressive and communicative advantage compared to the passive structure;
the spread of this innovation to the unmarked environment of narration was
presumably delayed and limited. When evaluating the passive-to-ergative shift in
causative predicates, one has to recall the pervasive influence of the High variant
in most literary works. Classical Sanskrit favored passive causatives (statistics
are available in Bubenik 1998:168-178), but many of them are interpretable
actively in accordance with the four previous principles.
On the whole, the changes described above are in keeping with Andersen's
Principle of Markedness Agreement (2001:28), which predicts innovations will
spread earliest in environments with equivalent markedness value and
subsequently spread to the complementary environments with opposite
markedness value, as in Table 4.

More widely, earlier More limited, later

with agents high in animacy with agents low in animacy
in agent-initial sentences in agent-final sentences
with discourse participants in the third person
in interrogative mode in declarative mode
in causative predicates in noncausative predicates
Table 4. The linguistic contexts in the actualization of the passive-to-ergative shift.

4. Noniconic relationships between morphology and semantics

Before addressing the issue of the noniconic relationships between
morphology and semantics ('markedness disagreement') of nonfinite forms
(gerundive and gerund), the following comments on the morphology of Middle
Indo-Aryan finite verb forms will be helpful. Unlike Old Indo-Aryan with its
active vs. middle voice dichotomy (the passive was limited to the third person of
the present and the aorist), Middle Indo-Aryan conjugates its present passive
actively (kar-ijj-ai "it is done" vs. kar-ai "he does") and its past active passively
(tena kiyau "he did [it] < ΟΙΑ tena krtam "[it is] done by him"). The
unambiguous passive interpretation of the finite passive kar-ijj-ai—in spite of its
active suffix—is warranted by the suffix -ijj. Compared to the corresponding
112 VIT BUBENIK

Old Indo-Aryan ancestral form kri-ya-te we are here dealing with a reduction of
markedness (or demarking) in that kri-ya-te was doubly marked by ablaut and
the suffix. (In early Middle Indo-Aryan there was also an archaism from Old
Indo-Aryan whereby—to indicate the higher degree of personal involvement
—one might conjugate the active verb with the earlier mediopassive forms, e.g.,
lahate "he takes for himself.) Similarly, the ambiguous construction tena kida
(passive ~ ergative) of early Middle Indo-Aryan was disambiguated in late
Middle Indo-Aryan by the rise of the "go"-passive; hence (in Apabhramsa) tena
kiyau "he did [it] vs. tena kiyau gayau "[it] was done by him" (cf. Figure 3).
In the passive-to-ergative shift operating on the past participle it is no
surprise that even the gerundive, the modal counterpart of the past participle, was
ultimately affected as in (13). The passive interpretation "His wife Padmâvatī
will be carried away by a vicious elephant" would be consistent with a diglossic

(13) pomāvai taho bhāmini gaena nevevī

Padmāvatī his wife elephant.INSTR carry.GERVE
dutthë harivi tena
vicious.INSTR take.GER he.INSTR

or polyglossic speaker approximating the Sanskrit gerundival phrase gajena

nayitavya "is to be carried by an elephant". A monoglossic speaker of
Apabhramsa whose grammar had already reanalysed the passive gerundive as
the active future tense would be more likely to use (13) to describe the event
"The vicious elephant will carry away his wife Padmāvatī".
With the gradual demise of the locative absolute construction during the
Middle Indo-Aryan period the passive interpretation of the gerund became
readily available. In these instances the gerund is usually embedded in the
passive matrix clause to guarantee the coreferentiality of the agent of the matrix
and that of the gerundial clause; cf. (14).

(14) mantrinā punar aham ahūya abhyadāyisi

minister.iNSTR again I call.GER tell.MED/PASS.AOR.lsG
"Having been called again by the minister I was told" (Dasa
125.5-6)

The unmarked active interpretation of the gerund āhūya, i.e. "having called", is
unlikely because of the marked finite medio-passive aorist abhyadāyisi in the
matrix clause. If the gerund were collocated with abhihito'ham (tell.PP I), it could
be interpreted actively ("The minister having called [me] again told me"), if
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 113

uttered by a diglossic speaker whose grammar already possessed a fully

developed ergative construction.

5. Syntactic ambiguity in late Middle Indo-Aryan absolute constructions

The merger of the instrumental and locative during the late Middle Indo-
Aryan period rendered the locative absolute construction of earlier Prākrits and
Sanskrit dysfunctional. More specifically, the phonological contrast between
instrumental and locative singular was lost as a result of the raising of mid
vowels in α-stems; i- and u-stems (and feminine α-stems) could distinguish the
instrumental from the locative in the singular, but not in the plural. Consequently,
the earlier absolute construction involving the passive participle with the Goal
Subject in the locative and the agentive phrase in the instrumental stopped being
viable. To use a fictive example, the Old Indo-Aryan passive construction in
(15.a) is to be translated "when the son was (had been) killed by the man". Its
(sanskritized) Apabhramsa equivalent narena mārie sue (man.INSTR
kill.pp.iNSTR/LOC son.INSTR/LOC) in (15.b) can be interpreted actively as "when the
man (had) killed the son"; however, if the agent is marked with the (non-
Sanskrit) genuine Apabhramsa suffix of the instrumental/locative, the sentence in
(15.c) becomes ambiguous (given the relatively free word order in Apabhramsa):

(15) a. narena mārite sute (Sanskrit)

man. INSTR kill.pp. LOC son. LOC
"when the son was killed by the man"
b. narena marie sue (sanskritized Apabhramsa)
man. INSTR kill.pp. INSTR/LOC son. INSTR/LOC
"when the man (had) killed the son"
 nare marie sue (Apabhramsa)
man. INSTR/LOC kill.pp. INSTR/LOC son. INSTR/LOC
"when the man (had) killed the son" or "when the son (had)
killed the man"

Needless to say, (15.c) would be disambiguated in context. Nevertheless, in

functional terms the loss of the morphological contrast between instrumental and
locative rendered the Sanskrit locative-absolute construction dysfunctional. With
the progressive decay of morphological case towards a single oblique form, a
compromise solution in late Middle Indo-Aryan was to adopt another
unambiguous case for the absolute construction, and the genitive case could fulfil
this role. (The genitive absolute was also available in Sanskrit in a specialized
meaning of 'disrespect'). Thus if the Goal Subject is expressed by the genitive,
as in (16), the construction is no longer ambiguous.
114 VIT BUBENIK

(16) nare mâriaho suaho (Apabhramsa)

man.INSTR/LOC kill.PP.GEN SOn.GEN
"when the man (had) killed the son"

However, recourse to ambiguity would not suffice to explain the fact that our
diglossic writers had apparently no problem using the Old Indo-Aryan locative
construction throughout the late Middle Indo-Aryan period, nor actually even
increasing its domain. The highly marked absolute construction involving the
passive imperfective participle is one of the most salient Sanskritisms in
Apabhramsa texts such as Svayambhüdeva's Paumacariu (seventh century).
The morphology of the latter form, -ijj-anta, is peculiar in combining the
inherited passive marker -ya- > -ja- and the suffix of the active participle -anta,
which replaced the Sanskrit mediopassive suffix -mana: Sanskrit dï-ya-mâna >
Apabhramsa di-jj-anta "being given". All the examples I have come across in
Paumacariu are of the following type:

(17) giyahim gijjantaehim (Pc 21.14.7)

SOng.INSTR/LOC.PL sing.PASS.PART.INSTR.PL
"(while the sprinkles of santal-scented water were being given
and) the songs were being sung"

The highly marked extralinguistic context for this highly marked grammatical
construction (involving the imperfective passive participle) is the mahotsava-
śobhā "a splendid great festival" of the royal wedding ceremony of Rām and
Sītā arranged by the king Dasaratha.
The late sanskritizing poet Haribhadra (twelfth century) produced even
'triglossic' constructions involving the Sanskrit mediopassive suffix -mana
attached to the Apabhramsa passive stem in -ijj in combination with the Prakrit
locative form of the head noun dān-ami (with the pronominal suffix-ãmi) instead
of the nominal locative suffix dān-e; cf. (18). The environment of this

( 18) dijjamāna dānammi (Sc 468.3)

give.PASS.MED-PASS.PTCPL gift.LOC
"while the gifts were being given"

passage is equally highly marked. In the preceding verse (467) we learn that the
queen Sahadevī has given birth to the future cakravartin, the supreme ruler of
the three worlds. Our verse runs as follows (in my literal translation):
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 115

While the bards were reciting,

the singers were singing,
the mangalas were being made by those paying homage,
the trumpets were blown,
the king bestowed the name of Sanatkumara on his son,
the vessel of supreme bliss for humanity. (Sc 468)

Here the poet describes an event where a lot of Sanskrit would be heard; the
mangalas are conventionalized blessings, uttered in Sanskrit, of the type jaya
jaya "victory". The statement that "they were being made" could refer to their
modification by well wishers (e.g. jayo'stu "may [the lord] be victorious").
Another highly marked extralinguistic context for the absolute construction
kijjantihĩ mangalihĩ "while the mangalas were being made" is found in verses
741-744 containing a description of an audience at the cakravartin Sanatkumāra,
attended by the young gods themselves disguised as brahmans; cf. also the
comments by Bergs & Stein (2001:90-92) on the relationship between religious
value and linguistic form in Burmese and Thai.

6. Conclusions
We saw above that markedness relations as articulated by Andersen (2001)
were significantly involved in the actualization of the passive-to-ergative shift
observable in the late Middle Indo-Aryan corpus. In Section 1 I pleaded for the
overall functionalist approach to this process pinpointing, however, weaknesses
inherent in functionalist typological scenarios, such as Dik's (1989) diachronic
markedness shift (surveyed in Figures 1-3). Timberlake's ideas (1977) about the
role of ambiguity in reanalysis, and the role of speaker's recognition of multiple
analyses offered by Harris & Campbell (1995) were of fundamental importance
in my further work. Smith (1995, 2001) has argued for functional explanation of
reanalysis rooted in sentence-processing strategies, such as parsability and
recoverability, in terms of implicational hierarchies. Harris (MS) has added related
concepts of inferability and predictability to the list. Exploiting these notions, one
could maintain that the overuse of the passive construction in Sanskrit
contributed to its reanalysis as an ergative construction in late Prākrits.
In Section 2 I argued that social, cultural, stylistic, and communicative
factors resulting from the vertical contact among the three variants of the same
Indo-Aryan language were among the triggering determinants of the passive-to-
ergative shift. These factors, external to the teleological language drift, could be
reconstructed on the basis of a large literary corpus in Sanskrit, Prākrits and
Apabhramsa (sketched in Figure 4). It became clear that multiple analyses can
often be linked with different sociolinguistic and stylistic registers.
116 VIT BUBENIK

In Section 3 we saw that the passive-to-ergative shift proceeded with

differences conditioned by the animacy or focality of the agent. Further
differences were introduced by factors of discourse type and illocutionary force.
It appeared that the progress of the passive-to-ergative shift in these instances
(summarized in Table 4) was in keeping with the Principle of Markedness
Agreement proposed by Andersen (2001). In addition to linguistic factors
affecting a differential progess of this shift we noticed a number of vertical
factors stemming from the influence of the High variant. In dialogue the
ambiguous construction with past participle (in 4.b) could be interpreted either as
a marked passive or an unmarked active construction, depending on sentence
processing strategies. Of equal importance, however, diglossic speakers in
Sanskrit and Apabhramsa always had recourse to the source construction in the
High variant (in 4.a), and they might favor the marked passive interpretation.
Another phenomenon of this vertical influence was seen in the activating strategy
developed in Classical Sanskrit whereby the ambiguous past participle could be
recycled as the past active participle in (2). Yet another aspect of the same
vertical influence was seen in the preference given to compounding over the
diathetically ambiguous construction with the past participle in (6). Its source
was the descriptive compound used lavishly in Sanskrit. Another type of
influence emanating from the High variant was seen in the appearance of the
finite passive morphology of the type tena vuccai "[it] is said by him" in contexts
where the ergative construction tena vuttu "he said [it]" would normally be used.
In the constructions involving the modal nonfinite forms (gerundives) the
passive interpretation would be more likely in the case of diglossic speakers
drawing on the source construction in the High variety (in (12)). A monoglossic
speaker of Apabhramsa in whose grammar the passive gerundive had already
been reanalysed as the active future tense would favor the active interpretation.
Another corollary of this process was seen in the perseverance of the locative
absolute construction in Apabhramsa in spite of its potential ambiguity (in (14)).
It remains a big question what was the role played by individual speakers
of Apabhramsa in sorting out all these multifarious stylistic and sociolinguistic
patterns endemic to diglossic and polyglossic societies in the cumulative fashion
consistent (or inconsistent) with the overall teleological drift(s). How can change
be both teleological and hierarchical? In Section 4 we saw that language-internal
drift and influence from vertical and horizontal contact can run in opposite
directions: the general drift towards ergativity was counteracted by repeated
pushes towards nominativity stemming from the High variant. (Similar
observations regarding the opposite direction (or partly opposite paths) of
evolutive changes and contact changes have been made by other scholars, e.g.,
Romaine (1984) in her study of the adoption of wh-relativization, which was
 PASSIVE-TO-ERGATIVE SHIFT IN INDIA 117

first established in the most complex styles.) Other examples of sanskritization

counteracting the typological drift were seen in the use of absolute constructions
in (15) and (16), which combine Sanskrit syntax with Apabhramsa and Prākrit
morphology. Here one can surmise that an impulse for their use in poets' minds
came from their use in descriptions of highly marked situations such as a
wedding ceremony, the birth of a son, or an audience with the monarch. Those
are precisely occasions when a lot of Sanskrit would be used.
Taken at face value, our Apabhramsa data attest to Pre-Islamic India's
move from style-shifting to diglossia (Sanskrit-Prakrit) to triglossia
(Sanskrit-Prākrit-Apabhramsa). Let it be mentioned that in addition to the
triglossic situation during the period of the seventh to eleventh centuries one
should also address an issue of the diglossic relationship between codified and
colloquial forms of the above three varieties as a result of the activity of Ancient
and Medieval grammarians: Pānini (fifth century B.C.), Vararuci (before 600
A.D.), and Hemacandra (eleventh-twelfth centuries A.D.). A daunting task in this
respect is a measured assessment of the degree of sanskritization of our primary
medieval documents. Certain of them were exposed to sanskritization by their
authors, but many Sanskrit features were introduced by subsequent copyists and
redactors in the course of the textual tradition. Needless to say, to separate these
out requires painstaking philological work. The prestige attached to the High
variant in diglossic and polyglossic societies can hardly be exaggerated.
To sum up, when writing the linguistic history of medieval speech
communities it is impossible to ignore strong sociolinguistic indexes or to deal
with them only by making reference to more central linguistic categories. This
neglected aspect of the actualization problem acquires paramount importance in
the diachronic study of diglossic and polyglossic societies, which so far has not
been undertaken on a large scale. The sociolinguistic history of the Middle Indo-
Aryan period remains to be written.

PRIMARY LITERATURE
Daśa Dandin's Daśakumāracarita (Pācinian Sanskrit, seventh century)
Hv Puspadanta's Harivamsapuruna (Apabhramsa, tenth century)
Kc Kanakāmara's Karakandacariu (Apabhramsa, tenth-eleventh century)
Pc Svayambhūdeva's Paumacariu (Apabhramsa, eighth century)
Rittha Svayambhūdeva's Ritthanemicariu (Apabhramsa, eighth century)
Sc Haribhadra's Sanatkumāracarita (Apabhramsa, twelfth century)
Vikr Kàlidâsa's Vikramorvasiya (Sanskrit with Apabhramsa passages, fifth
century)
118 VIT BUBENIK

REFERENCES
Andersen, Henning. 1989. "Understanding linguistic innovations". Language
Change: Contributions to the Study of its Causes ed. by Leiv Egil Breivik
& Ernst Håkon Jahr, 5-27. Berlin & New York: Mouton de Gruyter.
Andersen, Henning. 2001. "Markedness and the theory of linguistic change".
This volume, 21-57.
Bergs, Alexander T. & Dieter Stein. 2001. "The role of markedness in the
actuation and actualization of linguistic change". This volume, 79-93.
Browning, Robert. 1989. History, Language and Literacy in the Byzantine
World. Northampton: Variorum Reprints.
Bubenik, Vit. 1998. A Historical Syntax of Late Middle Indo-Aryan
(Apabhramsa). Amsterdam & Philadelphia: John Benjamins.
Dik, Simon C. 1989. The Theory of Functional Grammar. Dordrecht: Foris.
Edgerton, Franklin. 1953. Buddhist Hybrid Sanskrit Grammar and Dictionary.
New Haven: Yale University Press.
Fasold, Ralph. 1984. The Sociolinguistics of Society. Oxford: Basil Blackwell.
Harris, Alice  & Lyle Campbell. 1995. Historical Syntax in Cross-Linguistic
Perspective. Cambridge: Cambridge University Press.
Harris, Alice C. MS. Issues in Diachronic Syntax. (Unpublished paper dated
1997).
Kemenade, Ans van & Nigel Vincent (eds.). 1997. Parameters of Morpho-
syntactic Change. Cambridge: Cambridge University Press.
Kelkar, Ashok R. 1968. Studies in Hindi-Urdu, vol. I. Introduction and Word
Phonology. Poona: Deccan College.
Lightfoot, David W. 1979. Principles of Diachronic Change. Cambridge:
Cambridge University Press.
Lightfoot, David W. 1991. How to Set Parameters: Arguments from Language
Change. Cambridge, Mass.: The MIT Press.
Romaine, Suzanne. 1984. "Towards a typology of relative clause formation
strategies in Germanic". Historical Syntax ed. by Jacek Fisiak, 437-470.
(Trends in Linguistics, 23.) Berlin: Mouton de Gruyter.
Sch0sler, Lene. 2001. "From Latin to Modern French: actualization and
markedness". This volume, 169-185.
Smith, John C. 1995. "Perceptual factors and the disappearance of agreement".
Linguistic Theory and the Romance Languages ed. by John C. Smith &
Martin Maiden, 161-180. Amsterdam & Philadelphia: John Benjamins.
Smith, John C. 2001. "Markedness, functionality, and perseveration in the
actualization of a morphosyntactic change". This volume, 203-223.
Timberlake, Alan. 1977. "Reanalysis and actualization in syntactic change".
Mechanisms of Syntactic Change ed. Charles N. Li, 141-180. Austin &
London: University of Texas Press.
Versteegh, Kees. 1984. Pidginization and Creolization: The Case of Arabic.
Amsterdam & Philadelphia: John Benjamins.
 THE USE OF ADDRESS PRONOUNS
IN SHAKESPEARE'S PLAYS AND SONNETS

ULRICH BUSSE
University of Osnabrück, Germany

0. The development of address pronouns in Early Modern English

It has often been noted that in the Early Modern English period a number
of important linguistic changes have been underway. Although linguistic
change is a continual process, for reasons of accountability boundaries between
periods have been established. Conventionally, the Early Modern English
period is dated from 1500 to 1700. Thus, Shakespeare's writing career
(1589-1613) falls roughly in the middle of this period. Shakespeare as an
author, and a dramatist in particular, makes use of the variability of language,
as exemplified in such alternatives as: thou/you, you/ye, -th/-s, will/shall,
relative pronouns which/the which, use or non-use of do, etc. In Shakespeare's
time all of these were viable alternatives, however, they could be used quite
purposefully as means of social, stylistic, or rhetorical variation. Stein (1974:7)
in his analysis of inflections in the Shakespeare Corpus draws attention to the
special nature of the corpus. In his opinion the Shakespeare Corpus poses two
major problems, first by being a literary corpus and, secondly, by the
uncertainties of its textual transmission, both of which problems give rise to a
dialectic of grammatical description and literary interpretation.
Despite the varied history of transmission for individual plays, much of
the variation mentioned above is textually stable and cannot simply be
attributed to the hands of different scribes or compositors. For this reason, it
can be accounted for in the framework of sociohistorical linguistics. Romaine
(1982), Milroy (1992), and others have provided the theoretical basis for
introducing into the domain of historical linguistics the modern techniques
employed in sociolinguistics to explain language variation and change.
Employed for the variation of Early Modern English pronoun usage, this does
not imply that modern parameters of sociolinguistics such as socioeconomic
factors (e.g., Labov 1994) can immediately be transferred to Early Modern
England with its rank system representing an intermediate stage between
120 ULRICH BUSSE

medieval society with its three estates and the modern upwardly mobile class
society.
The replacement of thou by you in the course of some 500 years is an
interesting case in point. The S-curve model of linguistic change (see
Aitcheson 1991:83 ff., Labov 1994:65 ff., Ogura & Wang 1996) accounts for
the frequencies of incoming and recessive variants during language change.
The replacement of thou by you starts very slowly in the thirteenth century,
reaches its peak in the sixteenth and seventeenth centuries, and then slowly
recedes from the eighteenth century on, except in special genres and registers.
At the beginning of the Early Modern English period, in 1500, there is an
overlap in function between second person singular and plural forms, but also
between individual Τ forms {thou, thee, thy, thine) and Y forms (ye, you, your,
yours). At the end of the period, about 1700, the only forms left in Standard
English are you, your, and yours. Görlach (1999:l0f.) gives a concise summary
and a succinct explanation of the various processes:
The loss of thou/thee and the rise of ye/you which left ModE with the single form
you to express case and number is partly a syntactic phenomenon, but mainly a
matter of pragmatics. ... While the motivation of the change [to you] was mainly
social, the choice of thou involved, in the decisive period between 1550 and 1620,
various stylistic aspects, all of which survived only in peripheral form after 1620.

In Early Modern English these morphosyntactic variants are conditioned

by various intralinguistic and extralinguistic factors. Thus, sociolinguistically
you and thou can, in a loose sense of the term, be seen as being in
complementary distribution, whereas the choice between thy and thine is often
explained in terms of phonological constraints (see Schendl 1997). So are
usually the variants ye and you, with ye showing a somewhat defective
distribution in Shakespeare's time (see Busse 1998). Therefore, Lutz
(1998:201) rather concludes that "the selection of the Middle English plural
object form you as the only remaining form of direct address in Standard
English was due to the interplay of several external and internal factors."

1. On the use of address pronouns in the Shakespeare Corpus

If two linguistic forms in a specific language co-occur at a given time, it
is from the point of view of economy in language very likely that they are not
semantically identical, at least in their connotations. However, Rissanen
(1986:98) is right when he demands that "the change of a variant in identical
environments should not change the referential or descriptive meaning of the
expression." We may then assume that the variants are neither in free variation
nor in complementary distribution, but that there is a certain overlap in
 ADDRESS PRONOUNS IN SHAKESPEARE 121

function between them, so that they should be viewed "as if on a sliding-scale"

(Wales 1983:116).
It is a truth acknowledged in many studies that the choice of the
pronouns of address is not only a grammatical problem, but that it constitutes a
meaningful choice in terms of sociolinguistics. Grannis (1990:109) points out
that "the use of thou rather than you, or the other way around, obviously has an
important communicative function, although it is not always clear what this
function is." For this reason he suggests the approach of microsociolinguistics,
which investigates language "at the individual and small-group level" (ibid.).
While on one hand it is correct to advise critics that we must not assume
Shakespeare's usage of the pronouns of address is an exact mirror image of that
of the society around him, it is on the other hand indispensable to have a
knowledge of the communicative value of the "language-coins", i.e. the
pronominal system and its social grammar in the historical context at large. But
at this point historically interested sociolinguists are in the dilemma of being
"at the mercy of their corpus", which consists only of written material, so we
must be careful not to draw circular conclusions. Thus, before we implement
sociolinguistic or even sociological methods, as Brown & Gilman (1960) do, to
determine the pragmatic force of the pronouns of address, we should try to
examine whether other constraints than these alone could possibly determine
or at least influence their selection.

1.1 The problem of standard use and the concept of markedness

Scheler (1982:40) points out that a count of all the singular and plural
forms of address in Shakespeare results in a ratio of 2 : 5 in favor of you. The
original nominative form ye has mostly been replaced by the former oblique
form you, which, however, predominates in the Authorized Version of the
Bible (1611), generally acknowledged to show very conservative usage, but
has apparently become archaic elsewhere.
Quirk (1974:50) quite rightly points out that it is an oversimplification to
say that by 1600 "you was polite, formal usage, but thou was familiar or
insulting". Instead he suggests treating them in terms of markedness. "You is
usually the stylistically unmarked form: it is not so much 'polite' as 'not
impolite'; it is not so much 'formal' as 'not informal'." Whereas Ilson (1971:65
f.) is of the opinion that the theory of markedness is of only limited value for
the study of Elizabethan drama, and Shakespeare in particular, Mulholland
(1967:36) also takes up the term 'markedness'. On a contextual or attitudinal
basis the marked term in this dyad is taken as an affective index. Thus, in order
to meet the requirements of linguistic decorum (in Shakespeare's plays) in a
given situation depending mostly on the intersocial relationship of the
122 ULRICH BUSSE

characters, one form of address almost triggers or elicits its expected response.
In this sense of a mutual expectancy the unmarked form can be fully predicted.
Mulholland (ibid.) draws our attention to the fact that "with regard to the
expected/affective forms, the majority forms must be established before the
force of the 'marked' term can be recognised and used for character study".
With reference to Mcintosh's (1963) paper on As You Like It, where he takes
thou as the unmarked form and you as the marked, Mulholland shows that such
grammatical clues to character may be quite misleading if the system works
the other way around. For this reason she is right in demanding "that the norms
must be established before the problem can be adequately dealt with" (41). As
already pointed out earlier, these norms of the microcosm of Shakespearean
plays need not necessarily be a true counter-image of English usage around
1600.
By investigating depositions from the ecclesiastical court of Durham
from the 1560s, Hope (1994) challenges the argument that drama mirrors
"real" usage of the pronouns of address. His work on the court records
indicates that there thou is clearly the marked form, and that as early as 1560.
But by drawing on Barber (1981) and his own work on Marlowe he doubts
whether thou can be regarded as the marked form in drama: "so as late as the
1590s in drama, it is possible to find writers who do not seem to have 'marked'
one of the forms over the other" (148). If this observation is correct, then it is
not only a matter of chronology but it would have the further
semantic-pragmatic implications "that Shakespeare's dramatic usage, if it
bears any relation to 'real' Early Modern [English] usage at all, preserves
modes of usage which have long disappeared from everyday speech" (ibid.).
As regards the evaluative norms within Shakespeare's plays, Mulholland,
who based her findings on an analysis of Much Ado About Nothing and King
Lear, and Barber (1981), working on Richard HI, come to different results in
terms of the usualness of you as the "generally accepted form of the pronoun in
use in the upper classes, except from father to daughter and, possibly, from
women to their female servants" (Mulholland 1967:42). While Barber
(1981:287) concedes that also for Richard III you can be regarded as the
normal, stylistically unmarked form among the upper classes, the statistical
material brings some counterevidence. For the exchanges between characters
of noble rank in Richard III the examples of "Thou in fact outnumber examples
of singular You in the ratio of about 54 to 46" (286). He counts 568 examples
of thou and 491 instances of singular you.
In this respect it seems useful to distinguish between descriptive
(statistical) norms—the majority form—and evaluative (correctness or
appropriateness) norms. It is conceivable that the thou vs. you distinction
 ADDRESS PRONOUNS IN SHAKESPEARE 123

might be exploited for a number of purposes, and the data would reflect more
than one dimension of variation. Thou, in particular, might signal archaizing
usage in (some) historical plays. To have this value, the thou form would not
have to be used consistently as long as it occurred in pragmatically prominent
contexts.
Adding to this, such context-independent counts of pronoun usage in
individual plays do not take us very far. If, for example, we have a play with a
very high proportion of formal scenes and another with a much higher rate of
quasi-intimate scenes, it is obvious that the language structures will also be
different.

2. Corpus study
2.1 Shakespeare's plays
On the basis of Spevack (1968-1980) the total figures for the two
pronouns, their oblique forms and compounds for the 38 plays in the
Shakespeare Corpus are as follows: there are 13,186 Τ forms (1.578%) and
22,400 Y forms (2.681%). The ratio of Τ forms to Y forms is 0.59. In their
relative frequencies the two pronouns and their variants differ by 1.103% in
favor of you.
My data on the distribution of address pronouns according to the
parameters of date of composition and genre suggest the following: in terms of
address pronouns the Shakespeare Corpus can be divided in two parts, viz. an
early part leading up to 1600, or more precisely 1596/97, and a later part after
1600. This is to say that from The First Part of King Henry the Fourth on, the
Τ forms are outnumbered by Y forms. Brainerd (1979:7) has proven that for
the Τ forms the two factors of genre and date taken together provide a
statistically highly significant result of Ρ < 0.005.
Admittedly, these results are impaired, although in a consistent way, by
the fact that the Y forms have not been separated into those functioning as
second person singular and those acting as second person plural, because such
a division was impossible to make on the basis of a concordance or an
electronic text. On the basis of a control corpus it can positively be assumed
that about 20% of the Y forms are plural. This leaves about 18,000 Y singular
forms and hence reduces their relative frequency to ca. 2% (see Busse MS).
The differences between Tragedy, Roman Plays, and Romance are not
very significant. Surprising, and quite contrary to what could be expected
according to the body of critical literature presented in Section 1.1, is the
extremely low frequency of Τ forms in the Comedies, which exhibit the lowest
rates of all genres; see Table 1 below.
124 ULRICH BUSSE

This result was confirmed by an additional test carried out on the basis of
Mitchell's (1971) British Drama Corpus spanning the two-hundred-year period
from 1580 to 1780. Throughout this period, and despite considerable
vacillation among individual authors, there is a constant "surplus" of Y forms
in Comedy in comparison with Tragedy; see Figure 1.

Rel. freq. Τ Rrel. freq. Y

Histories 1.965% 2.030%
Tragedies 1.772% 2.551%
Comedies 1.446% 3.046%
Table 1. Pronoun distribution according to genre.

Put on a cline, the data from Mitchell, which are based on 57,580
occurrences of second person pronouns from sixty-two plays by twenty-nine
dramatists, corroborate my findings for the Shakespeare Corpus: the genre of
Comedy provides a higher degree of Y forms than Tragedy in each of the
subperiods.
A comparison of drama with other nonliterary text types based on
Taavitsainen (1997) has revealed that in Fiction and Autobiography the use of
thou lingers on through all phases of Early Modern English, while in History,
Biography, Diary, and Travelogue it is a marginal form from the very
beginning of the period; see Table 2 (from Taavitsainen 1997:239).

Figure 1. The percentage of Τ and Y forms from 1580 to 1780 according to genre.
 ADDRESS PRONOUNS IN SHAKESPEARE 125

Type of text EModE 1 EmodE 2 EModE 3 Total

1500-1570 1570-1640 1640-1710
Fiction 20 29 53 102
Autobiography 9 4 10 23
History 0 0 2 2
Biography 0 1 0 1
Diary 0 0 3 3
Travelogue 0 0 0 0
Table 2. Absolute numbers of second person singular pronouns in the Helsinki Corpus.

The figures in this table suggest that Fiction and—at some

distance—Autobiography are the only genres to make use of thou throughout
all phases of Early Modern English. Interestingly, for Fiction the numbers rise
with time, and for Autobiography there is a marked drop in Early Modern
English 2 followed again by a rise. The 'unusual' behavior of Fiction in Early
Modern English 3 can be explained by the fact that all the examples here stem
from Aphra Behn's Penny Merriments (1685) "a parody with exaggerated
emotionality and comical imitation of religious language" (241).

2.1.1. The distribution of address pronouns in verse and prose. The fact that
blank verse in Elizabethan drama constitutes a linguistic and literary norm has
often been stated by critics. For this reason the poetic device of blank verse is,
among other functions, a stylistic means for social characterization. In this
tradition, persons of rank speak in blank verse in normal speech situations. If
verse thus constitutes a norm, a switch to prose indicates a deviation from it.
Persons speaking in prose, whether constantly, or only temporarily, are either
not capable, because of their inferior social status, or not willing to comply
with this norm.
General reference works on Shakespeare and case studies such as
Tschopp (1956:23-24) state that lower class characters, in particular peasants,
artisans, soldiers, etc. only speak in prose, which in turn can be interpreted as
an indicator of social class. The servants of noblemen, on the other hand,
usually speak in verse, e.g., Oswald in King Lear. The porter in Macbeth is an
exception serving to indicate drunkenness. Children and fools also normally
speak in prose. Noblemen temporarily turn to prose under exceptional
circumstances such as situations of extreme stress or strain, e.g., the mad King
Lear on the Heath, Ophelia while mentally deranged, or the somnambulant
Lady Macbeth. Furthermore, prose can also be used in asides, where the
speeches are given off the record and are intended not to be overheard by those
 126 ULRICH BUSSE

talked about, e.g., the dialogue between Kent and Gloucester in the opening
scene of King Lear, where the bastardy of Edmund is revealed.
On the other hand, rhymed passages with couplets, often at the end of a
scene, also mark a deviation from the norm. In some of Shakespeare's early
plays whole scenes are written in verse; for instance, the exchange between
Romeo and Juliet (1, 5, 93-106) has the form of a sonnet.
The alternation of verse and prose could then, in the broadest meaning of
the term, be interpreted as a discourse or textual marker. A switch between
these two dramatic media can convey a change of mood or attitude, topic of
discourse, social setting, etc. as outlined above.
With due attention to the fact that there are different layers of prose style,
and that prose fulfils different dramatic functions in Comedy and Tragedy, and,
in addition, that Shakespeare went through an artistic development throughout
his writing career, the following working hypothesis could be formulated:
despite these differences it can be generalized that prose is an indicator of
social inferiority. On the basis of this, it can further be assumed that this social
division of the two dramatic media has a bearing on the forms of address that
are being exchanged. Thus, if one of the two address pronouns is marked or, in
other words socially restricted, a difference in frequency in either verse or
prose should result.
However, the expectation that the Τ forms would predominate in the
Comedies with their typical lower-class personnel and a higher proportion of
prose passages has been refuted, for the Comedies show the highest incidence
of Y forms (see Table 1).
In the following, the plays will be categorized into the genres of History
Plays, Tragedies, Roman and Greek Plays, Comedies and Late or Problem
Plays, and the Τ forms and Y forms will be accounted for according to their
occurrence in a verse context or in a prose context.
For the corpus as a whole, the ratio between words occurring in verse and
prose contexts is: 76.73% in verse and 23.27% in prose. However, in
comparison to this, the overall proportions of verse and prose vary quite
considerably from genre to genre. While the Histories, Tragedies, and Roman
and Greek Plays show a fairly similar distribution of verse and prose, ranging
between 82 and 84% of vcrsc and 16 to 18% of prose, the Comedies, not
surprisingly, differ sharply from these as they exhibit the highest ratio of prose
(43%) in the corpus. The Late or Problem Plays occupy an intermediate
position.
As regards the address pronouns, Table 3 shows that 10,473 Τ forms
(79.44%) occur in a verse context, and only 2,711 (20.56%) in a prose context.
In comparison to these numbers, the Y forms feature more prominently in
 ADDRESS PRONOUNS IN SHAKESPEARE 127

prose contexts, viz. 15,731 Y forms (70.23%) in verse as opposed to 6,669

(29.77%) in prose.

Τ Verse % Τ Prose % Y Verse % Y Prose %

Hist. 3,469 82.41 692 17.59 4,106 82.20 1,114 17.80
Trag 1,625 83.39 344 16.61 2,227 76.32 784 23.68
Rom. 1,991 86.99 316 13.01 2,809 82.09 683 17.91
Com. 1,534 61.03 910 38.97 2,777 54.97 2,686 45.03
Late 1,854 80.36 449 19.64 3,812 74.25 1,402 25.75
Total 10,473 2,711 15,731 6,669
Table 3. Absolute and relative frequencies of address pronouns in verse and prose by genre.

In Section 1.1 the advantages and disadvantages of treating the two

pronouns in terms of markedness were critically appraised. Recently Andersen
(2001) has pointed out that the concept of markedness "has developed into a
cluster of (dis)similar concepts, adapted for application in a variety of
approaches to language, analytic and descriptive, empirical and deductive,
functional and formal". In his opinion, markedness relations obtain in
diachronic linguistics in every variety of change "from its inception to its
completion, both in the relations among variants and in the relations that define
the plethora of categories that typically condition the gradual process by which
newer forms replace older correspondents" (ibid.). In an earlier paper (1990)
on the notion and direction of drift in the sense developed by Sapir, Hjelmslev,
and Coseriu, Andersen has shown that "certain elements of the theory of drift
make it understandable why markedness relations in language would structure
the actu[aliz]ation of linguistic change" (13). As examples he lists a number of
correlations that hold between morphological categories such as present tense
vs. preterite, grounding distinctions such as main clauses vs. subordinate
clauses, genre categories, as for example prose vs. poetry, spoken vs. written
media, casual vs. formal style, etc., in which the first category in each dyad is
described as being "more compatible with innovation" than the latter (see
Table 4). He then goes on to say that these observations can be rephrased in
terms of markedness. His reasons to assume a correlation between innovation
and unmarked contexts are the following: "this means that the correlation we
wish to understand is between the markedness value of the two variants [...]
and the equivalent markedness values of a variety of contexts in which they are
distributed" (18). In a second step he tries to justify the correlation of
unmarked with unmarked and marked with marked categories by the finding
that "it seems to be the case that the opposite terms of any feature or variable
which is not being exploited for communicative purposes, will be distributed in
128 ULRICH BUSSE

the most orderly fashion possible, which is, in such a way as to maximize
homogeneous syntagmatic combinations."
When this model is applied to the problem of the distribution of the two
address pronouns in the media of verse and prose, the higher incidence of the
Y forms in prose can then be accounted for. So, if the yous in Shakespearean
drama occur with a higher frequency in prose than in verse this could be an
indicator of their being the stylistically unmarked form correlating with prose
as the unmarked genre (in contrast to poetry) and for this reason being "more
compatible with innovation".
Having found so far that for the use of address pronouns in Shakespeare's
plays the factors date of composition and genre—or rather their
correlation—and their distribution in verse or prose contexts are of importance,
it is tempting to test whether all, or at least some of these factors could
possibly be attributed to the use of pronouns in his nondramatic works. For this
reason, Shakespeare's usage in the Sonnets and his other poems shall be
examined next.

2.2 Address pronouns in Elizabethan poetry

In the previous section the predominance of the second-person plural
pronoun you in the prose parts of Shakespeare's plays was interpreted with
recourse to the theory of drift and the concept of markedness propounded by
Andersen (1990). Battistella's (1996) monograph sums up the numerous
definitions of the concept since the 1930s. He traces "how both structuralist
and generative theories have expanded markedness as a way of characterizing
linguistic constructs and as part of a theory of language" (4). However,
Battistella also points out that the word markedness has lost much of its
terminological status because the notion of "markedness has been applied to a
wide variety of data [...]. The earliest applications of markedness analyzed
phonological correlations [...] and grammatical opposites [...]. Others have
extended markedness to connect grammar and discourse categories as well"
(16). As an example of the incorporation of discourse categories, Givón
(1990:945-986) can be mentioned. In his opinion three major criteria can be
used to distinguish the marked from the unmarked category in a binary
contrast:

(a) Structural complexity'. The marked structure tends to be more complex—or

larger—than the corresponding unmarked one.
(b) Frequency distribution: The marked category (figure) tends to be less frequent,
thus cognitively more salient, than the corresponding unmarked one (ground).
(c) Cognitive complexity'. The marked category tends to be cognitively more
complex—in terms of attention, mental effort or processing time—than the
unmarked one. (Givón 1990:947)
 ADDRESS PRONOUNS IN SHAKESPEARE 129

Table 4, cited from Andersen (1990), links different morphosyntactic,

pragmatic, and stylistic categories with their respective openness to linguistic
innovation. The table sums up observed chronological differences in a
language-particular (Polish) long-term change of the auxiliary verb "to be",
first investigated by Teodozja Rittel, which may have begun around 1300
"from sentence clitics, regularly placed in clause-second position, to verbal
desinences marking person and number" (3):

More compatible Less compatible

with innovation (unmarked) with innovation (marked)
Morphological categories present tense preterite
indicative mood conditional
singular number plural
plural number dual
third person other persons
first person second person
Grounding distinctions main clauses subordinate clauses
asyndetic clauses syndetic clauses
initial lexical noun initial pronoun
Genre categories prose poetry
expository prose artistic prose
secular content religious content
Media spoken written
Styles casual formal

Table 4. Linguistic categories and their openness towards innovation.

From this table he concludes that each category forms a binary contrast or
opposition in markedness, "with the unmarked term of each pair in the left-
hand column and the marked term to the right. The table, in short, attests to a
strong correlation, in this development, between the markedness of different
conceptual, grammatical, and textual contexts and their compatibility with
innovation" (12). What is interesting, though, is the fact that any indications of
sociolinguistic variation are absent in the table. This is due to the fact that they
were absent in the data on which Andersen reports, as writing, for most of the
600-year-long progression of this change was a privilege of the Polish elite,
which was probably too small to be sociolinguistically differentiated. In this
particular case "one might guess that in a society where there is no particular
use for sociolinguistic indexes, variation rules simply make reference to more
central linguistic categories" (18).
In the light of the findings of Section 1.1 it seems nonetheless
worthwhile to put the theory and its predictive force to the test in the case of
130 ULRICH BUSSE

the highly sociolinguistically determined variation of pronominal address by

investigating pronominal use in Elizabethan poetry to substantiate the findings
on pronoun distribution in verse and prose contexts in Shakespeare's plays.
According to the interconnection of unmarked linguistic categories and their
openness to innovation it should be the case that in Shakespeare's Sonnets,
with poetry being a marked category, the pronoun thou as the marked one in
the dyad should predominate. For this reason a short glance will be taken at
Shakespeare's Sonnets, other Elizabethan sonneteers, and poems other than
sonnets as a control corpus, to see if Andersen's model holds true for the
distribution of thou and you in Elizabethan poetry. However, it has to be
acknowledged that blank verse in drama and verse in poetry belong to different
literary genres with different exigencies.

2.2.1 Quantitative evidence from Shakespeare's nondramatic writings. In the

Shakespeare Corpus the figures for the Τ forms and Y forms in the 38 plays
and the poems, considered in the aggregate, are as in Table 5.

Τ forms % Y forms % Τ: Υ
The 38 plays 13,186 37.05 22,400 62.95 0.589
Poems 1,094 74.12 382 25.88 2.864
Table 5. The number of Τ forms and Y forms in the Shakespeare Corpus.

The difference between the plays and the nondramatic works in their use of the
two forms is highly significant, as the plays exhibit a T : Y ratio of 0.589 as
opposed to 2.864 for the nondramatic works. This extreme difference in
distribution tells us that the two pronouns have entirely different status in
drama and in poetry. The figures for each of the works or cycles, in the case of
the Sonnets, is presented in Table 6.
Table 6 clearly reveals that except for A Lover's Complaint all the poems
or poem cycles definitely favor Τ forms, however, with different degrees of
salience, ranging from 95% in The Rape of Lucrece to only 68% in the
Sonnets. As the Sonnets constitute the largest group they will be investigated
more closely. For a quantitative analysis of the Sonnets the studies of Berry
(1958), Jones (1981), and GUrr (1982) provide further data on the number and
distribution of the pronouns.

2.2.2 Qualitative analysis of pronoun distribution in Shakespeare's Sonnets.

A bird's-eye view of the sum of address pronouns, as shown in Table 7, clearly
reveals that thou poems are much more numerous than you poems. Within the
 ADDRESS PRONOUNS IN SHAKESPEARE 131

Date Τ forms % Y forms %

Venus and Adonis 1592/93 173 85.64 29 14.36
The Rape of Lucrece 1593/94 210 95.02 11 4.98
The Passionate Pilgrim 1599 22 91.66 2 8.33
The Phoenix and Turtle  1601 6 100 — 0
The Sonnets 1593-1609 676 67.87 320 33.03
A Lover's Complaint 1602/08 7 25.93 20 74.07
Total 1,094 74.12 382 25.88

Table 6. The number of Τ forms and Y forms in Shakespeare's nondramatic works.

sequence of the Sonnets two different parts need to be recognized: "in the
sonnets to the Fair Youth [1-126] the dominant pronoun is thou, occurring in
69 of the 126 sonnets (54.75%), but you also appears in 34 sonnets (27%)"
(Jones 1981:80). On the other hand, in the sonnets addressed to the Dark
Mistress, or to the conventionally-termed Dark Lady, which form a less
coherent group, you is never used to address the Dark Lady. In the whole
sequence there are 86 thou poems as opposed to only 34 you poems. The thou
poems also clearly outnumber those in which the Young Man or the Dark Lady
is addressed indirectly by using a third-person pronoun.
This use of you and thou has been at the center of a number of articles,
but as far as I can see, these have mostly focussed on the seeming
(ir)regularities of pronoun switching, rather than on the broader genre-typical
considerations or aspects of linguistic change which will serve as point of
departure for my investigation.
Some critics have indeed been puzzled by the frequent use of thou in the
Sonnets', Gurr (1982), for instance, considers the shifts in the 126 sonnets to
the Young Man a "remarkable display of inconsistency" (12) and wonders why
in the first twelve sonnets the Young Man is constantly addressed with thou. In
the notes to his article he mentions that "the anomalous use of the pronouns"
was first noted by the German scholar Karl Goedeke in 1877, and that from
then on much inconclusive speculation about the reasons for the pronoun
switching has followed. For instance, Archer (1936:544) has tried to account
for these changes in terms of rhyme and euphony. While Finkenstaedt
(1958:456) also considers these factors important, Jones (1981:80) and Gurr
(1982:12-13) regard them as supplementary factors at best. Berry (1958) asks
the question,
What is the difference in poetic result between a 'thou' sonnet and a 'you' sonnet?
[...] We can, in other words, expect some significance in the fact that one sonnet
may be built around an 'I-thou' relationship, while another sonnet centres around
an 'I-you' relationship" (138, 140)
132 ULRICH BUSSE

He states that in his Sonnets, Shakespeare uses thou for more distanced rela­
tionships between the lyrical I and the addressee, whereas you is used for

thou you he/she Zero Mixed

1-4 5
6-12 13
14 15-17
18 19
20 21
22 23
24
25
26-32 33
34-51 52-55 56
57-59
60-62 63 64-66
67-68
69-70 71-72
73-74 75-76
77-79 80-81
82 83-86
87-93 94
95-97 98
99 100
101
102-104 105 105
106
107-110 111-115 116 116
117-118 119 119
120 121 121
122 123-124 123-124
125-126
127
128 129
130
131-136 137
138
139-143 144-145 146
147 148
149-152 153-154

Table 7. Pronoun distribution in Shakespeare's Sonnets, by number.

closer personal relationships, from which he draws the somewhat surprising

conclusion that "'You' is, then, more intimate, 'thou' more formal—the
opposite of what might be expected according to some imaginary Elizabethan
Fowler of Correct Usage" (143). Finkenstaedt criticises this viewpoint,
 ADDRESS PRONOUNS IN SHAKESPEARE 133

remarking that "it is dangerous to make statements about the emotional value
of you and thou in Shakespeare's Sonnets without reference to Elizabethan
prose usage" (456). In reply to this, Berry (1959) defends his results by
refuting the thesis that the pronoun usage of other contemporary writers or
different genres could falsify his thesis on Shakespeare's pronoun usage:
It is precisely my belief that poets do or, at least, can use a language in a way other
than it is used by writers or speakers of prose [...] personal pronouns—it follows
that these are, or can be, apprehended and used in a poetic way. [...] 'each existence
has its own idiom' (196).

Finkenstaedt (1963:166-167, footnote) epitomizes their controversy on

pronoun use in poetry by dismissing Berry's claim of poets' individual
language use as a 'Humpty-Dumpty-Theory' of literature.
To my mind, these contradictory statements could possibly be reconciled by
assigning the genres of poetry and prose different degrees of openness towards
innovations. By adopting this stance it would be perfectly normal to

have almost simultaneously two different 'Elizabethan Fowlers of Correct

Usage', one for poetry, and another for prose style. This is of course not to say
that the emotional content of the pronouns varies from genre to genre. It seems
indeed out of the way to assume that you functions as a marker of nearness in
one genre and as a marker of distance in another.
The latest and also the most comprehensive work on Shakespeare's
Sonnets by Vendler (1997) does not really touch upon the role of the address
pronouns from a structural point of view, but provides detailed and insightful
interpretations for each of the 154 sonnets.
Finkenstaedt (1963) is right when he argues that relegating the problem
of pronoun usage to the individual level is not the point, but by comparing
Shakespeare with contemporary sonneteers one can arrive at less speculative
and emotional postulates on pronoun use in poetry than, for example, Berry
did.

2.2.3 Other Elizabethan sonneteers. As the investigation of pronoun usage in

Shakespeare's Sonnets only makes sense when it is compared to that of his
contemporaries, the previous findings will be put into the larger perspective of
other Elizabethan sonneteers to answer Finkenstaedt's question,
Wie weit setzt sich im Englischen die alte Konvention der Sg.-Anrede im Gedicht
fort, in welchem Umfang erscheint das you der taglichen Rede, und wie weit
entspricht die Pronomenverwendung der Dichtung jener der Umgangssprache?
(1963:166).
134 ULRICH BUSSE

In so doing, some similarities but also differences between Shakespeare's

pronoun usage and that of his contemporaries become apparent; see Table 8.

thou you she Zero Mixed

Anon. Zepheria 33 — 1 3 3
Barnes Parthenophil 36 1 33 17 17
Constable Diana 30 5 18 11 11
Daniel Delia 15 — 31 5 4
Drayton Idea 19 — 16 14 2
Fletcher Licia 6 9 17 5 16
Griffin Fidessa 3 — 33 24 2
Linche Diella 11 2 15 10 1
Lodge Phillis 3 — 16 13 6
Percy Coelia 4 — 8 2 6
Sidney Astrophel 12 3 39 45 9
Smith hloris 5 — 22 10 10
Spencer Amoretti — 16 50 13 9
Tofte Laura 35 7 43 27 8
Watson Tears Offancy 3 — 22 27 1
Shakespeare Fair Youth 69 34 6 15 1
Shakespeare Dark Mistress 17 — 5 5 —
Total: 1089 tokens 296 78 371 238 106
Percentages 27.18 7.16 34.08 21.85 9.73

Table 8. Pronoun distribution in Elizabethan sonnet sequences. (Jones 1981:83;

the last two lines are my addition—UB).

An interesting divergence between Shakespeare and his contemporaries

is the fact that he is much more careful to avoid pronominal inconsistency in
one poem.
If the figures for the other Elizabethan sonneteers are compared to those
of Shakespeare, another point of departure is that Shakespeare's
contemporaries prefer the third-person pronoun (360 tokens) as the usual form
of indirect address of the mistress rather than direct address by means of
second person pronouns. "This indirect approach has the general effect of
turning the Mistress into an object of contemplation, analysis and commentary,
rather than her appearing to be a mute partner in a dramatic dialogue" (Jones
1981:74).
In addition to this, the two different modes of direct dramatic address
(you, thou) and indirect reference {he, she) are often functionalized within a
single sonnet for textlinguistic purposes. Jones has found that, for instance,
Fletcher and Spenser often build sonnets around the following contrast: in the
first twelve lines, which are devoted to narrative description, they use third
 ADDRESS PRONOUNS IN SHAKESPEARE 135

person pronouns, but then they switch to direct address of the Mistress in the
final couplets (76).
Stein (1985) has found similar evidence for the discourse marker functions
of the third-person endings -s and -th in texts from the turn of the sixteenth
century. With recourse to Halliday, he attributes to them a textual function in
the global organization of discourse "in the sense of differentiating between the
two different media of narrating and reporting" (284). For the diachronic
perspective he also applies the concept of markedness, concluding that
one would have to posit s as the marked form for the earliest texts studied and th
for the later texts studied. The fact that, in the printed text, s eventually follows the
colloquial spoken language is reflected in the markedness formulation by a
markedness reversal: in the later texts—Deloney and Nashe—it is th which carries
the extratextual information besides designating person, tense and mood (294).

The ratio of 296 thous : 78 yous clearly shows that in Elizabethan sonnets thou
is the majority form in those cases where the Mistress is addressed directly.
The fact that you is absent in quite a number of sonnets cannot be an instance
of chance. Jones attributes the numerical dominance of thou in Elizabethan
sonnets to a number of factors. If the intersocial and textual functions of the
pronouns at the end of the sixteenth century are grouped under keywords, the
binary categories shown in Table 9 may be established.

THOU YOU
• in elevated utterances as the pronoun of • neutral, prosaic and unemotive form
poetic convention
• as an archaic and poetic mode • as socially polite pronoun signalling
greater politeness or deference
• private pronoun: prayer, intimate and • pronoun of public address
familial relationships
• old-fashioned in nonpoetic usage • normal social intercourse
• becoming restricted to ritualistic and • conversational and prosaic pronoun
poetic contexts (invocation) signalling a less elevated poetic manner
• in emotive utterances • neutral, unemotive form
• conventional literary pronoun • realistic (more usual) pronoun
• (Sonnets): individual addressed as a • (Sonnets): addressed to some patron or
fictive convenience, often in the context mistress or affecting to be addressed to
of make-believe emotions some flesh-and-blood woman

Table 9. Functions of thou and you in Early Modern English literature

(based on Jones 1981:76-78).

Thus, thou "is the 'high' term of invocation, and it is also the emotive and
familiar term. With thou the poet can shift from adulation to familiarity to
insult with one pronoun, depending on the emotional context established"
136 ULRICH BUSSE

(Jones 1981:77). This double function of thou has also been recognized by
Finkenstaedt (1963:169), who establishes a thou1 as an impersonal
conventional pronoun in poetry and an intimate thou2. Yet, on the other hand,
matters are complicated, because you can also be used in a double function as
the neutral default and as the socially polite pronoun.
In his appraisal of the communicative functions of the two pronouns,
Jones offers the following convincing conclusions why you is generally
avoided as pronoun of address in Elizabethan sonnets other than Shakespeare's.
In his opinion, you is "the neutral, prosaic and unemotive form. [...] This
emotional and social neutrality appears to be the major reason why most
sonneteers, writing consciously in an elevated manner and ostensibly not for
the public ear, avoid you, the prosaic term reminiscent of normal social
intercourse" (77; my emphasis).

2.2.4 Elizabethan poems other than sonnets. For poetry other than sonnets
I can only offer random evidence from secondary sources, especially
Finkenstaedt (1963) and Jones (1981). Patchy as the data may be, a tentative
conclusion may, nonetheless, be drawn. In my opinion, future analyses of
Elizabethan poetry should take into account sociopragmatic genre constraints
such as those described by Biber & Finegan (1992), who in their diachronic
study of text types compare written 'literate' genres such as essays, fiction, and
personal letters to speech-based 'oral' genres, e.g., dialogue from plays and
from fiction, because the different degrees of privacy and/or orality and the
literate or formal, more "written" character of a text could, among other
factors, also trigger the selection of address pronouns.
For instance, Jones (1981:78) states that Drayton, who avoids you in his
sonnets, uses it in his odes and elegies, and Daniel does so too in many of his
poetic epistles. Finkenstaedt (1963:167-169), in his discussion of Harington's
epigrams, finds the following numerical distribution of address pronouns: 257
of the epigrams contain 32 real plural addresses, including those to the
reader(s) where it is not clear whether a single reader or the reading public in
general is meant, 108 contain only thou, 73 only you, and 44 show a change.
For Donne (169-172) he concludes that his pronoun use is highly
conventionalized, and that metre, in contrast to rhyme, is not an important
factor. A case in point for the dichotomy of private (more oral) vs. public
(more written) seems to be the following:
In alien Widmungsgedichten steht you; in den Letters to Several Personages steht
you bei allen mit vollem Namen genannten Personen. Bei den nur mit den
Anfangsbuchstaben des Namens adressierten poetischen Episteln steht fast nur thou
(170).
 ADDRESS PRONOUNS IN SHAKESPEARE 137

Thus, the avoidance of you in the sonnet sequences (other than Shakespeare's)
could be attributed to their more private nature as opposed to more overtly
public poems such as odes, epistles, and epigrams, which quite to the contrary
feature you more strongly.
Finkenstaedt is right in demanding that any study trying to scrutinize
pronominal usage in poetry, and pronoun switches in particular, should seek to
elucidate whether the addressee remains the same, and/or the matters of
content (literary conventions) demand a shift, or formal criteria such as rhyme,
metre, and euphony necessitate a change. In the more recently developed
framework of pragmaphilology this implies that all "the contextual aspects of
historical texts, including addressers and addressees, their social and personal
relationship, the physical and social setting of text production and text
reception, and the goal(s) of the text" (Jacobs & Jucker 1995:11) have to be
taken into account.

3. Summary and conclusion

From the data on Shakespeare's plays and his Sonnets and those of his
contemporaries the following conclusions on the correlation between pronoun
use in drama and poetry and the markedness of address pronouns may be
drawn.
While Andersen (2001) and others admit that the notion of markedness
has never been clearly defined and has accordingly developed in a number of
different directions, quantitative evidence alone has proved that it is difficult to
conceive of the stylistically marked category as the less frequent one (see
Section 2.2). You was found to be the majority form in drama, in particular in
Comedy, as opposed to poetry. Owing to this, frequency and stylistic value or
appropriateness are relative criteria depending on the regularities and
necessities of different text types. In both cases thou is the stylistically marked
pronoun in the dyad. As a result of this, it would be useful to distinguish
between descriptive (statistical) norms, in terms of frequency, and evaluative
norms in terms of correctness or appropriateness.
In the explanation of language change, markedness not only plays an
important role in the progression of change, but also on the synchronic level.
In the case of Early Modern English address pronouns, sociolinguistic and
textlinguistic factors do have to be taken into account to accommodate
pronoun shifts in a society keenly aware of social indexes.
However, as it was not the prime objective of this paper to unravel the
textlinguistic intricacies involved in the triggering of pronoun switches in
individual sonnets, the above commentaries and explanations may suffice, but
the evidence presented in this article and the arguments of the previous studies
138 ULRICH BUSSE

seem to validate the initial hypothesis on the direction of drift in that it could
be proved that despite vacillation between individual authors, the stylistically
marked pronoun of the dyad (thou) predominates in poetry as the form of
direct address. As a genre, poetry with its preference for more traditional
address forms seems to be more constrained by literary conventions than, for
instance, the different text types of prose writing, or nonliterary colloquial
usage. Within poetry a cline from more overtly public, colloquial "written
orality"—preferring you—to more private, artistic, conventionalized and
formal "truly written" texts—preferring thou—could be shown. This cline
would account for the appropriateness of pronouns in terms of genre
conventions, distance between author and addressee, etc. (see Section 2.2.1.5).
For this reason, pronoun use in Elizabethan poetry corroborates the
dichotomies put forward by Andersen (see Table 4), because three of the
variables, namely 'genre categories', 'media', and 'styles', provide supporting
evidence. On the other hand, it seems unfounded to maintain (as Berry did)
that in terms of distance the two pronouns function antagonistically in poetry
and in prose writing or colloquial language.
Despite the differences that exist between blank verse in drama and verse
in poetry, primarily exemplified by sonnets, their similar preference for the
pronoun thou "in the higher poetic style" (Abbott 1870:154) has been
confirmed.
The investigation has shown that the concept of markedness can be
applied to explain linguistic change, and especially its embedding, as a
diachronic process, but in addition to this, by making a synchronic cut around
1600, the concept can also be brought to fruition in constructing a typology of
texts. That is to say by making a link between a statistically more or less
probable form (thou) and its stylistic value as the marked term in the dyad the
following text typology in terms of thou-fulness can be arrived at:

• In authentic texts (depositions) thou is the numerically less frequent and

stylistically marked form as early as the 1560s (see Hope 1994). Hope's
findings are supported by those of Taavitsainen (1997), as she has found
that, excepting Fiction and to a lesser extent also Autobiography, thou only
plays a marginal role in other nonfictional text types from 1500 on.
• In Shakespearean drama, the History Plays and the Tragedies are the most
thou-ful plays; in the Comedies you is more frequent. This correlation
holds for the entire period (see Mitchell 1971).
• As concerns their distribution in the media of verse and prose (in drama),
thou predominates in verse and you in prose.
 ADDRESS PRONOUNS IN SHAKESPEARE 139

• This distribution has been confirmed for poetry as an elevated register,

which in comparison to drama shows the highest incidence of Τ forms.
• More overtly public poems prefer you, more private ones, thou.1

REFERENCES
Abbott, Edwin A. 1870. A Shakespearian Grammar. London: Macmillan
(Reprinted 1972, New York: Haskell.)
Aitcheson, Jean. 1991. Language Change: Progress or Decay? Second
edition. Cambridge: Cambridge University Press.
Andersen, Henning. 1989. "Markedness theory—the first 150 years".
Markedness in Synchrony and Diachrony ed. by Olga Mišeska Tomic,
11-46. (Trends in Linguistics. Studies and Monographs, 39.) Berlin & New
York: Mouton de Gruyter.
Andersen, Henning. 1990. "The structure of drift". Historical Linguistics 1987.
Papers from the 8th International Conference on Historical Linguistics (8.
ICHL, Lille, 31 August-4 September 1987) ed. by Henning Andersen &
Konrad Koerner, 1-20. (Current Issues in Linguistic Theory, 66.)
Amsterdam & Philadelphia: John Benjamins.
Andersen, Henning. 2001. "Markedness and the theory of linguistic change".
This volume, 19-55.
Archer, C. 1936. "'Thou' and 'you' in the sonnets". Times Literary Supplement
27.06.1936,544.
Barber, Charles. 1981. "'Thou' and 'you' in Shakespeare's Richard III". Leeds
Studies in English, New Series, 12.273-289. Reprinted in A Reader in the
Language of Shakespearean Drama ed. by Vivian Salmon & Edwina
Burness, 163-179. Amsterdam & Philadelphia: John Benjamins, 1987.
Barber, Charles. 1997. Early Modern English. Second edition. Edinburgh:
Edinburgh University Press.
Battistella, Edwin L. 1996. The Logic of Markedness. New York & Oxford:
Oxford University Press.
Berry, Francis. 1958. "'Thou' and 'you' in Shakespeare's sonnets". Essays in
Criticism 8.138-146.
Berry, Francis. 1959. "Pronouns in poetry". Essays in Criticsm 9.196-197.
Biber, Douglas & Edward Finegan. 1992. "The linguistic evolution of five
written and speech-based English genres from the 17th to the 20th
centuries". History of Englishes. New Methods and Interpretations in
Historical Linguistics ed. by Matti Rissanen et al., 688-704. Berlin & New
York: Mouton de Gruyter.
Brainerd, Barron. 1979. "Pronouns and genre in Shakespeare's drama".
Computers and the Humanities 13.3-16.
Brown, Roger & Albert Gilman. 1960. "The pronouns of power and
solidarity". Style in Language ed. by Thomas Sebeok, 253-276.
Cambridge, Mass.: The MIT Press. Reprinted in Communication in Face to
1
Iwould like to thank Henning Andersen for his comments on an earlier version of this paper,
from which the present version has greatly profited.
140 ULRICH BUSSE

Face Interaction ed. by J. Laver & S. Hutcheson, 103-127. Harmonds-

worth: Penguin.
Busse, Ulrich. 1998. "Stand, sir, and throw us that you have about ye. Zur
Grammatik und Pragmatik des Anredepronomens 'ye' in Shakespeare's
Dramen". Betrachtungen zum Wort: Lexik im Spannungsfeld von Syntax,
Semantik und Pragmatik ed. by Eberhard Klein & Stefan J. Schierholz,
85-115. Tübingen: Stauffenburg.
Busse, Ulrich. MS. "The function of linguistic variation in the Shakespeare
Corpus. A study of the morpho-syntactic variability of the address
pronouns and their socio-historical and pragmatic implications".
Evans, G. Blakemore with the assistance of J. J. M. Tobin (eds.). 1997. The
Riverside Shakespeare. Second edition. Boston: Houghton Mifflin.
Finkenstaedt, Thomas. 1958. "Pronouns in poetry". Essays in Criticism
8.456-457.
Finkenstaedt, Thomas. 1963. You und Thou. Studien zur Anrede im Englischen
(Mit einem Exkurs iiber die Anrede im Deutschen). Berlin: de Gruyter.
Franz, Wilhelm. 1939. Die Sprache Shakespeares in Vers und Prosa unter
Berücksichtigung des Amerikanischen entwicklungsgeschichtlich
dargestellt. Shake speare-Grammatik. Fourth edition, reprinted 1986.
Tübingen: Max Niemeyer.
Givdn, Talmy. 1990. Syntax: A functional-typological Introduction, vol. 2.
Amsterdam & Philadelphia: John Benjamins.
Görlach, Manfred. 1999. Aspects of the History of English. Heidelberg: Carl
Winter.
Grannis, Oliver. 1990. "The social relevance of grammatical choice in
Shakespeare". Deutsche Shakespeare-Gesellschaft West, Jahrbuch
1990.105-118.
Gurr, Andrew. 1982. "You and thou in Shakespeare's sonnets". Essays in
Criticism 32.9-25.
Gvozdanovic, Jadranka. 1989. "Defining markedness". Markedness in
Synchrony and Diachrony ed. by Olga Mišeska Tomic, 47-66. (Trends in
Linguistics. Studies and Monographs, 39.) Berlin & New York: Mouton de
Gruyter.
Hope, Jonathan. 1994. "The use of thou and you in Early Modern spoken
English: Evidence from depositions in the Durham Ecclesiastical Court
records". Studies in Early Modern English ed. by Dieter Kastovsky,
141-151. (Topics in English Linguistics, 13.) Berlin & New York: Mouton
de Gruyter.
Howe, Stephen. 1996. The Personal Pronouns in the Germanic Languages: A
Study of Personal Pronoun Morphology and Change in the Germanic
Languages from the first Records to the present Day. (Studia Linguistica
Germanica, 43.) Berlin: de Gruyter.
Ilson, Robert F. 1971. "Forms of address in Shakespeare, with special
reference to the use of 'thou' and 'you'". Unpublished Ph.D. Thesis,
University of London.
 ADDRESS PRONOUNS IN SHAKESPEARE 141

Jacobs, Andreas & Andreas H. Jucker. 1995. 'The historical perspective in

pragmatics". Historical Pragmatics. Pragmatic Developments in the
History of English ed. by Andreas H. Jucker, 1-33. (Pragmatics and
Beyond, New Series, 35.) Amsterdam & Philadelphia: John Benjamins.
Jones, G. P. 1981. "You, thou, he or she? The master-mistress in Shakespearian
and Elizabethan sonnet sequences". Cahiers Elisabétains 19.73-84.
Labov, William. 1994. Principles of Linguistic Change, vol. 1. Internal
Factors. (Language in Society, 20.) Cambridge, Mass.: Basil Blackwell.
Lutz, Angelika. 1998. "The Interplay of external and internal factors in
morphological restructuring: The case of you". Advances in English
Historical Linguistics (1996) ed. by Jacek Fisiak & Marcin Krygier,
189-210. (Trends in Linguistics. Studies and Monographs, 112.) Berlin &
New York: Mouton de Gruyter.
Mcintosh, Angus. 1963. "As You Like It: A grammatical clue to character".
Review of English Literature 4.68-81.
Milroy, James. 1992. Linguistic Variation and Change: On the historical
Sociolinguistics of English. (Language in Society, 19.) Oxford: Basil
Blackwell.
Mitchell, Eleanor R. 1971. "Pronouns of address in English, 1580-1780: A
study of form changes as reflected in British drama". Unpublished Ph.D.
Dissertation, Texas Agricultural and Mining University [University
Microfilms].
Mulholland, Joan. 1967. "'Thou' and 'you' in Shakespeare: A study in the
second person pronoun". English Studies 48.34-43. (Reprinted in A Reader
in the Language of Shakespearean Drama ed. by Vivian Salmon & Edwina
Burness, 153-161. Amsterdam & Philadelphia: John Benjamins, 1987.)
Ogura, Mieko & William S.-Y. Wang. 1996. "Snowball effect in lexical
diffusion: The development of -s in the third person singular present
indicative in English". English Historical Linguistics 1994. Papers from
the 8th International Conference on English Historical Linguistics. (8th
ICEHL, Edinburgh, 19-23 September 1994) ed. by Derrek Britton,
119-141. (Current Issues in Linguistic Theory, 135.) Amsterdam &
Philadelphia: John Benjamins.
Quirk, Randolph. 1974. The Linguist and the English Language. London:
Edward Arnold. (Reprinted in A Reader in the Language of Shakespearean
Drama ed. by Vivian Salmon & Edwina Burness, 3-21. Amsterdam &
Philadelphia: John Benjamins, 1987.)
Rissanen, Matti. 1986. "Variation and the study of English historical syntax".
Diversity and Diachrony ed. by David Sankoff, 97-109. (Current Issues in
Linguistic Theory, 53.) Amsterdam & Philadelphia: John Benjamins.
Romaine, Suzanne. 1982. Socio-historical Linguistics: Its Status and
Methodology. Cambridge: Cambridge University Press.
Schabert, Ina. 1992. Shakespeare-Handbuch: Die Zeit, der Mensch, das Werk,
die Nachwelt. Third edition. Stuttgart: Alfred Kröner.
Scheler, Manfred. 1982. Shakespeares Englisch. Berlin: Erich Schmidt.
142 ULRICH BUSSE

Schendl, Herbert. 1997. "Morphological variation and change in Early Modem

English: my/mine, thy/thine". Language History and linguistic Modelling:
A Festschrift for Jacek Fisiak on his 60th Birthday, vol. 1. Language
History ed. by Raymond Hickey & Stanisław Puppel, 171-191. (Trends in
Linguistics. Studies and Monographs, 101.) Berlin: Mouton de Gruyter.
Spevack, Marvin. 1968-1980. A complete and systematic Concordance to the
Works of Shakespeare. 8 vols. Hildesheim: Georg Olms.
Stein, Dieter. 1974. Grammatik und Variation von Flexionsformen in der
Sprache des Shakespeare Corpus. München: Tuduv.
Stein, Dieter. 1985. "Discourse markers in Early Modern English". Papers
from the 4th International Conference on English Historical Linguistics,
Amsterdam, 10-13 April, 1985 ed. by Roger Eaton, Olga Fischer, Willem
Koopman & Frederike van der Leek, 283-302. (Current Issues in
Linguistic Theory, 41.) Amsterdam & Philadelphia: John Benjamins.
Taavitsainen, Irma. 1997. "Genre conventions: personal affect in fiction and
non-fiction in Early Modern English". English in Transition: Corpus-based
Studies in Linguistic Variation and Genre Styles ed. by Matti Rissanen,
Merja Kytö & Kirsi Heikkonen, 185-266. (Topics in English Linguistics,
23.) Berlin & New York: Mouton de Gruyter.
Tomic, Olga Mišeska (ed.). 1989. Markedness in Synchrony and Diachrony.
(Trends in Linguistics. Studies and Monographs 39.) Berlin & New York:
Mouton de Gruyter.
Tschopp, Elisabeth. 1956. Zur Verteilung von Vers und Prosa in Shakespeares
Dramen. (Schweizer Anglistische Arbeiten, 41.) Bern: Francke.
Vendler, Helen. 1997. The Art of Shakespeare's Sonnets. Cambridge, Mass.:
The Belknap Press of Harvard University Press.
Wales, Kathleen M. 1983. "Thou and you in Early Modern English. Brown and
Gilman re-appraised". Studia Linguistica 37.107-125.
Wales, Katie. 1996. Personal Pronouns in Present-Day English. Cambridge:
Cambridge University Press.
 ACTUALIZATION PATTERNS IN GRAMMATICALIZATION:
FROM CLAUSE TO LOCATIVE MORPHOLOGY
IN NORTHERN IROQUOIAN

Marianne Mithun
University of California, Santa Barbara

0. Introduction
In their 1968 article on language change, Weinreich, Labov, and Herzog
drew attention to what they termed the embedding problem: the identification of
factors that could account for the gradual spread of changes through a language
and across communities. Since that time a number of works have traced the
progress of particular changes over grammars and populations. Among these is a
seminal study by Timberlake (1977) detailing the step-by-step spread of the
genitive case through Finnish participial clauses, and of the accusative case
through Russian negative clauses. Another study by Andersen (1987)
documents the gradual development of auxiliaries into person and number
markers on Polish verbs. Both authors point to the orderly actualization of the
changes through contexts that can be described in terms of specific semantic and
grammatical features, noting a progression from unmarked to increasingly
marked contexts. In their large-scale study of syntactic change, Harris and
Campbell propose that grammatical changes progress systematically to ever
wider contexts definable in terms of natural classes (1995:101). A number of
studies, however, have shown that change can proceed one lexical item at a time.
Wang (1966, 1977) and Labov (1981) document the word-by-word spread of
certain sound shifts. Ard (1975), Warner (1982), and Disterheft (1984) describe
the step-by-step spread of complement constructions into clauses dominated by
different matrix verbs. Warner (1983, 1990), Lightfoot (1979, chapter 2; 1991,
chapter 6; 1999:180-185), Plank (1984), Hopper and Traugott (1993:45-48),
and others discuss the development of certain English verbs into modal
auxiliaries, lexical item by lexical item. Lichtenberk (1991) traces the evolution of
certain Tobaba'ita verbs into prepositions and then conjunctions and
complementizers, demonstrating that the items involved show different stages of
 144 MARIANNE MITHUN

development. These works and others raise questions concerning the kinds of
generalizations that can be made about actualization patterns. Here it will be
shown that lexical factors need not necessarily constitute evidence against
systematicity. They can in fact contribute to the motivation of grammatical
change in principled ways.
A kind of change that offers opportunities for observing actualization
patterns is grammaticalization, by which lexical items evolve into grammatical
markers. Grammaticalization is typically gradual and leaves evidence within the
language of earlier diachronic stages. The actualization patterns to be examined
here involve the grammaticalization of a new locative category. The modern
Northern Iroquoian languages of northeastern North America contain paradigms
of locative suffixes that are pervasive in speech. Examples of locative
constructions can be seen in the sentences in (1) from Mohawk, spoken in
Quebec, Ontario, and New York State.1

(1) Mohawk locatives; Sha'tekenhatie' Phillips, Konwatiense Jacobs,

Warisose Kaierithon, speakers.

a. Kèn:'en kaniatar-ákta entewaia'tdta'.

here lake-near we will body-put it in
"We'll bury it here near the lake"

1
Examples are presented in the practical orthography adopted in all six Mohawk communities.
The practical orthography is essentially phonemic. Symbols t, and k represent plain stops
(voiced before voiced segments); ts is an alveolar affricate in Kahnawake, Kanehsatake, and
Wahta, which corresponds to an alveopalatal affricate spelled tsi in Ahkwesahsne,
Thaientaneken, and Ohsweken; s is a voiceless spirant; n, r, w, i are resonants, with i
representing the glide [y] before vowels; h is always pronounced as a distinct segment and the
apostrophe ' represents glottal stop. The vowel symbols i, e, a, o have IPA values. Digraphs
en and on represent nasalized vowels: en is a nasalized low, central vowel [^] and on is a
nasalized high, back vowel [u]. The colon : represents vowel length, the acute accent ' marks
stress with high or rising tone, and the grave accent', stress with falling tone.
In the glosses, the following abbreviations of grammatical terms are used: AGT
(grammatical agent), AND (andative), AUG (augmentative), CISLOC (cislocative), COINCID
(coincident), CONTR (contrastive), DIM (diminutive), DISTR (distributive), DUPL (duplicative),
EPENTH (epenthetic vowel), EXCL (exclusive), FACT (factual), IMPF (imperfective aspect), INDEF
(indefinite gender), INCH (inchoative), INSTR (instrumental), MASC (masculine), NEUT (neuter),
OPT (optative), PAT (grammatical patient), PF (perfective aspect), PL (plural number), PROGR
(progressive), recipr (reciprocal), REPET (repetitive), SG (singular number), STAT (stative), SUFF
(suffix), TRANSLOC (translocative).
I am grateful to the Mohawk speakers from the communities of Kahnawake, Kanehsatake,
Ahkwesahsne, Thaiendaneken, Wahta, and Ohsweken, who have generously shared their
expertise. I especially appreciate the many insightful comments provided by Kanerahtenhawi
Nicholas and Skawennati Montour of Kanehsatake, Kaia'titahkhe Jacobs of Kahnawake, and
Rokwaho Dan Thompson of Ahkwesahsne.
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 145

b. Karhá:-kon iahóhtka'we'
woods-in he would leave him there
"He would leave him in the woods."
c. Kaheht-à:ke ionkwatehiá:ron.
field-at we grew up
"We grew up in the country."

The other Northern Iroquoian languages show similar patterns.

The locative constructions have evolved from clauses, but the evolution is
not complete. Varying stages of development can be seen with different lexical
items and in different lexical contexts, allowing us to infer pathways along which
the innovation has been spreading through the grammar.

1. Stimulus to reanalysis: structural ambiguity

Three lexical categories can be identified in Iroquoian languages on the
basis of morphological structure: verbs, nouns, and particles.
Verbs consist minimally of a pronominal prefix, verb stem, and aspect
suffix; cf. (2).

(2) Morphological verbs; Sha'tekenhatie' Marion Phillips, Lazarus Jacob,

speakers.
a. k-atkétskwa-s . ronwd-rio
l.SG.AGT-get.up-IMPF INDEF.PL.AGT/MASC.SG.PAT-kill.STAT
"I get up" "they have killed him"
b. wak-iita'-s
l.SG.PAT-sleep-iMPF
"I sleep"

Basic nouns consist of a prefix, noun stem, and suffix. The prefix indicates
the gender of the referent or its possessor, and the suffix simply identifies the
word as a noun; cf. (3).

(3) Morphological nouns.

a. ka-nákt-a'
NEUT.I-bed-NOUN.SUFF
"bed"
146 MARIANNE MITHUN

b. o-tshá:t-a! . rao-nákt-a'
NEUT.II-Cloud-NOUN.SUFF MASC.POSSESSOR-bed-NOUN.SUFF
"cloud" "his bed"

Particles are by definition morphologically unanalysable; cf. (4)

(4) Morphological particles.

a. ó:nen . wish
"now" "five"
b. tsi
"how, where, that, so"

The pronominal prefixes on verbs specify the core arguments of the clause.
They show grammatical case distinctions originally based on semantic factors.
Participants who actively and voluntarily instigate actions such as getting up or
jumping, or states such as residing somewhere, are represented by grammatical
agent prefixes (Paradigm I). Those who are not in control but are significantly
affected by situations such as sleeping, falling, or being ill are represented by
grammatical patient prefixes (Paradigm II). A third paradigm of pronominal
prefixes represents agent/patient combinations in transitive verbs such as "kill"
or "touch". Verbs denoting inherent states such as "be big" or "be good" appear
with Paradigm I prefixes, while those denoting resultant states like "be wet" or
"have eaten" appear with Paradigm II prefixes; cf. (5).

(5) Pronominal prefixes with stative verbs.

a. ka-kowá:n-en . ró:-k-on
NEUT.I-be.big-STAT MASC.II-eat-STAT
"it is big" "he has eaten"
b. io-ná:naw-en
NEUT.II-be.wet-STAT
"it is wet"

Despite the semantic basis of the pronominal categories, the choice of prefix
paradigm is categorical and lexicalized with each verb stem. Speakers cannot
switch from agent forms (Paradigm I) to patient forms (Paradigm II), or vice-
versa, for semantic effect. With some verb stems, semantic change over time has
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 147

obscured the original basis for a particular prefix choice, but the semantic
foundations underlying the system as a whole are easy to discern.
The prefixes on nouns show some formal resemblance to those on verbs,
but the resemblance is not exact, as can be seen by comparing the verbal and
nominal neuter singular prefixes in Table 1; for the symbols, see footnote 1.

Verbs Nouns
Paradigm I ka- ka- before consonant and i (a + i → en)
w- ø before a, e, en
/- ø before o, on
Paradigm II io- o- before consonant and i, a, o, on (V —» 0)
iaw- aw- before vowels e, en
Table 1. Verbal and nominal prefixes: neuter singulars.

The verbal and nominal prefixes also differ in function. Both indicate
gender, but only the pronominal prefixes on verbs distinguish grammatical
relations. The gender prefixes on nouns are simply lexicalized as part of the noun
word and are invariant across syntactic contexts. In (6), for example, the prefix
o- (NEUT.II.) on ohkwá.ri "bear" remains unchanged though the noun functions
syntactically as a grammatical agent in (6.a) and as a grammatical patient in (6.b).

(6) Invariant gender prefixes on nouns; Warisose Kaierithon, speaker.

a. Wa'-t-há:-t-a'-ne' kí:ken o-hkwá:ri.
FACT-DUPL-MASC.I-Stand-INCH-PF this NEUT.II-bear
"The bear stood up."
b. S-a-konwa-ia't-isdk-h-a' ne o-hkwá:ri.
REPET-FACT-3PL/3.NEUT-body-search-AND-PF the NEUT.II- bear
"They went back to look for the (dead) bear."

In addition to the pronominal prefix and aspect suffix, verbs may include a
variety of other affixes, as well as an incorporated noun stem, like -ia't- "body"
in (6.b) above or no'ts- "tooth" in (7) below.

(7) Noun incorporation; Konwatiense Jacobs, Skawennati Montour,

speakers.

Wa-honwa-no'ts-ot-à:ko-'.
FACT-MASC.PL/MASC.SG-tOOth-Stand-REVERSIVE-PF
"They pulled his teeth out."
148 MARIANNE MITHUN

The incorporated noun does not function as a syntactic argument: it simply

qualifies the predication. As can be seen in (7), it is not represented within the
pronominal prefix complex that specifies the core arguments. Incorporated nouns
also do not affect the choice of pronominal paradigm. The verb -ri "be ripe,
cooked", for example, appears with Paradigm II prefixes whether or not it
contains an incorporated noun, and whether the noun itself would appear with a
Paradigm I or Paradigm II prefix; cf. (8).

(8) Pronoun paradigm selection: governed by the verb stem.

a. ió:-ri
NEUT.PAT(ii)-be.ripe.STAT
"it is ripe, cooked"
b. io-hi-á-ri
NEUT.PAT(ii)-fruit-EPENTH-be.ripe.STAT
"the fruit is ripe" ká-hi "fruit"

The syntactic functions of the three word classes are for the most part as
would be expected. Particles serve as demonstratives, adverbs, numbers, con­
junctions, discourse markers, etc. Morphological nouns serve as arguments of
clauses. Morphological verbs serve as predicates. But verbs can do more. Since
they contain pronominal reference to their core arguments, they can constitute
fully grammatical, independent clauses in themselves, as in examples (2), (5),
(7), and (8) above. They can also function syntactically as nominals, providing
descriptive labels for arguments without overt nominalizing morphology.

(9) Mohawk verbal nominals.

a. ka-wistó-ht-ha!
NEUT.I-be.COld-CAUSATIVE-IMPF
"it chills" = "refrigerator"
b. w-aten-nhoh-a-niiont-âhkkw-cï
NEUT.I-MTODLE-door-EPENTH-hang-INSTR-IMPF
"it door-hangs with it" - "hinge"
 io-hsk-óhar-e'
NEUT.II-bridge-hang-STAT
"it is bridge-suspended" = "suspension bridge"
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 149

Many morphological verbs, like those in (9), have become conventionalized as

lexical nominals. Speakers recognize them first as names of entities. Other verbs
are used sometimes as predicates, sometimes as nominals, like kakowá:nen "it is
big/the big one" in (5). Still others are used rarely if ever as nominals, like
katkétskwas "I get up" in (2).
Among the verbs are some that describe spatial location, as in (10).

(10) Spatial verbs; Skawennati Montour, Awenhráthen Deer, speakers.

a. te-hi-até-kh-en
DUPL-MASC.DU.i-RECiPR-be.adjacent-STAT
"they two are adjacent to each other" = "they are neighbors"
b. Iah ki í:non tha'-te-iakw-ate-nonhs-áter-e'.
n o t jUSt far CONTR-DUPL-l.EXCL.PL.I-RECIPR-hoUSe-be.apart-STAT

"Our houses were not far apart."

Like other verbs, spatial verbs may become lexicalized as nominals, as in (11).

(11) Lexicalized spatial verb.

te-ha-honht-a-né:k-en
DUPL-MASC.I-ear-EPENTH-be.side.by.side-STAT
"his ears are side by side" = "cottontail rabbit"

Some spatial verbs incorporate a noun indicating the object located, like -nonhs-
"house" in (10.b) and ahonht- "ear" in (11) above.
Others incorporate a noun that provides a point of reference. Some of these
have evolved away from full status as verb roots toward locative suffixes. In
natural speech, they are seldom used as the main predicates of sentences. If the
point of a sentence is to indicate location, a verb of position serves as the main
predicate. The answer to "Where is my shirt?" is the sentence in (12).

(12) Predication of location.

Ka-ronto isher-á:-kon io-hrénit-on.
NEUT.I-closet-EPENTH-in NEUT.II-hang-STAT
"It's hanging in the closet."

Given the translations of the sentences in which they occur, it might at first
appear that the original locative verbs have evolved into the functional equivalent
150 MARIANNE MITHUN

of locative adpositions "in the middle of', "beyond", "beside", "under", "in", and
"at" (as in Tobaba'ita) and ultimately into locative case markers.

(13) Locative case? Sha'tekenhatie' Phillips, speaker.

Ka-ná:t-a-kon kwi' ni-'teron:-tahkwe'.
NEUT-tOWn-EPENTH-in right MASC.DU.AGT-reside-IMPF.PAST
"They used to live in the village, you know."

The function of the locative markers is subtly different, however. In Iroquoian

languages, the only arguments marked for grammatical role are the pronominal
prefixes within verbs. Independent nominals within the sentence, whether they
represent grammatical agents, patients, or obliques, carry no formal marker of
their grammatical relation. The locative constructions are no exception. The
locative verbs have evolved into locative nominalizers: they create nominals that
are labels for places. Because of their meanings, such nominals function most
often to locate the events or states predicated by the clause, but the locative
markers are not relational. The sentence in (14) was the answer to a question
"Do you know the Kanehsatake reserve?" The word kaná:takon does not mean
"in the village", but rather "the village place".

(14) Locative nominalizer; Sha'tekenhatie' Phillips, speaker.

Né: ki'k ní:' ka-ná:t-a-kon ni-t-ienté:r-i.
it.is just the-myself NEUT.I-town-EPENTH-place.in PART-1.SG.AGT-
know-STAT
"I myself just know the village."

Formal traces of the verbal origins of the locative constructions are still
easy to see. They show internal structures reminiscent of those of incorporating
verbs. As seen earlier, morphological verbs often serve as syntactic nominals
without overt nominalizing morphology. If deverbal nominals are incorporated
into other verbs, however, they must be formally nominalized. The terms
karón:t "bureau, closet" and kà:sere "car" were both originally coined from
verbs. When incorporated, they always carry a nominalizing suffix, -tsher- for
"bureau" and -ht for "car"; cf. (15)-(16).
Locative constructions generally show the same restriction as incorporating
verbs: associated nominals must be formal nouns, either noun roots or
nominalized stems; cf. (17).
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 151

(15) Overt nominalization under noun incorporation: "bureau".

a. ka-ron:t-o
NEUT.I-tree-stand
"bureau"
b. wa'-ke-ronto-'tsher-a-hni:non
FACT-1.SG.AGT-bureaU-NOMINALIZER-EPENTH-buy.PF
"I bought a bureau."

(16) Overt nominalization under incorporation: "car".

a. kà-:sere
NEUT.I-drag
"car"
b. a-ho-'seré-ht-a-ni-'
OPT-3.MASC/3.MASC-vehicle-NOMINALIZER-EPENTH-lend-PF
"He would lend him the car."

(17) Locatives with nominalizers -tsher- and -ht-.

a. karonto'tsherákta ka'serehtákta
"near the dresser" "next to the car"
b. karonto'tsheró:kon ka'serehto:kon
"under the dresser" "under the car"
 karonto'tsherá:kon ka' seréhtakon
"in the bureau/closet" "in the car"
d. karonto'tsherà:ke ka'serehtà:ke
"on the dresser" "on the car"

Verbs with incorporated nouns also show special phonology. If a verb root
begins in a consonant, and the preceding incorporated noun stem ends in a
consonant, then an epenthetic -a- is inserted between them, as can be seen in
(15.b) and (16.b) above. The same epenthetic stem joiner -a- appears in locative
constructions. In (17.c) it joins ronto'tsher- to -kon, and sereht- to -kon.
The stative aspect suffixes that appear on verbs show a variety of forms,
among them -en-, -on, -e', -i, and zero. The locatives show similar endings:
-(i)hen "between", -ti "beyond", akta "near, beside",  "underneath", -kon
"inside", and -a'ke "in, at, on". The locatives appear with other affixes specific to
verbs as well, such as the distributive suffix -hson. In (18) this distributive can
be seen with a prototypical verb, and in (19) with locatives.
152 MARIANNE MITHUN

(18) Distributive -hson on a verb.

Wa-honwati-hseré-hson.
FACT-3.PL/3.PL-chase-DiSTR.PF
"They chased them around, all over the place."

(19) Distributive -hson on locatives; K. Phillips, K. Jacobs, speakers.

a. Ahsat-a-kón-hson te-ho-tstikaw-en-hátie'-s.
(NEUT.I)-shadow-EPENTH-place.in-DiSTR DUPL-MASC.PAT-travel-
STAT-PROGR-DISTR
"He travels along in the shadows."

b. O-wis-ríké-hson n-ia'-e-tákh-e'
NEUT.II-ice-place-DISTR PART-TRANSLOC-INDEF.AGT-run-IMPF- DISTR
"She ran, slipping and sliding across the ice."

The reanalysis of locative verbs as nominals has included a shift in

referentiality. In verbs serving as predicates, only the pronominal prefixes are
referential: ka-kowa:nen "it is big". In these locative constructions, as in other
nominals, the full word is referential: kanâ:takon "the village place". The
difference is mirrored in formal shifts in the shape and selection of the prefixes.
As seen earlier in Table 1, many of the prefixes on nouns differ slightly in form
from their counterparts on verbs. A number of the verbal prefixes, for example,
contain initial glides that are absent from the nominal prefixes. Before α-stems,
the NEUT.I verbal prefix is w-, but the noun prefix is zero. Though the locative
constructions originated as verbs, they now appear with the nominal prefixes.

(20) Neuter verb and noun prefix.

Verb Noun
a. w-akera'-s b. aten'èn:r-a'
NEUT.I-Stink-IMPF (NEUT.l)-fence-NOUN.SUFF
"it stinks" "fence"
 aten'enhr-ákta
(NEUT.I)-fence-near
"near the fence"

Prefixes for other genders and persons show the same distribution: those on
locative constructions match those used on nouns.
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 15 3

(21) Masculine plural rati- on verb; Seth Newhouse in Hewitt (1899: 256.3).
Né kerí liken rati-náker-e'
it.is here it.is MASC.PL.AGT-reside-STAT
né tsi rati-nonhs-ó:t-on
the where MASC.PL.PAT-house-stand-DISTR.STAT
"This is where they lived, the place where their houses were

(22) Masculine plural raoti- on nominals; Konwatsi'tsaienni Phillips,

speaker.
a. raoti-nákt-a'
MASC.PL-bed-NOUN.SUFF
"their bed"
b. Raoti-nónhs-kon ni-on-sa-ha-tákhe'.
MASC.PL-h0USe-in PART-FACT-REPET-MASCAGT-run-PF
"He went running back into their house."

The locative constructions also show a shift in the government of prefix

paradigm. In verbs, the choice of pronominal prefix paradigm is governed by the
verb root, whether or not there is an incorporated noun. In locative constructions,
the choice is now generally governed by the noun, as can be seen by comparing
the rows and columns in Table 2.
As derivational nominalizers the locatives function to create lexical items.
Speakers do not generally assemble locative nominals anew as they speak, but
rather retrieve them ready-made from their lexicons. As lexical items, they may
be coined for special purposes or develop specialized meanings.

(23) Specialized meanings.

a. o-honts-ó:kon ohóntsa' "earth"
NEUT.II-earth-place.inside
"in the earth" = "cellar"
b. aten'en:r-a-kon ateríen:ra' "fence"
(NEUT.I)-fence-EPENTH-place.inside
"inside the fence" = "yard"
 o-hson'kar-à:ke ohsòn:kare' "board"
NEUT.II-board-place
"on the boards" = "floor"
154 MARIANNE MITHUN

d. o-thorè:=ke' iothoire' "it is cold"

NEUT.II-be.cold-place
"where it is cold" = "north"

Especially common among lexicalized locative constructions are placenames.

(24) Highly lexicalized nominals: placenames.

a. S-ka-hnéht-a-ti (source of the name of Schenectady)
REPET-NEUT.I-pine-EPENTH-place.beyond
"on the other side of the pines" = "Albany, New York"
b. Ka-hnaw-à:ke
NEUT.I-rapids-place
"at the rapids" = "Kahnawake, Quebec" (Mohawk community)

Like other derivational morphology, the locative suffixes are not syntactically
obligatory. Though they appear in large numbers of placenames, many other
placenames do not contain them, such as Tekahson'kahró:rens "Hogansburg"
("they split planks", site of a sawmill), Kanón:no "New York City" and
Ken'taróhkwi "Kingston".
Since the locative constructions are derived nominals, they can undergo
further derivational processes applicable only to nominals. The Northern
Iroquoian languages contain sets of enclitics that are added only to lexicalized
nominals, whatever their internal morphological structure. One is the
augmentative =kowa, visible with a basic noun in (25.a) and with locative
nominals in (25.b) and (25.c).

(25) Augmentative clitic on nominals.

a. ka-honwei-a'=kó:wa
NEUT.I-boat-NOUN.SUFF=AUG
"ship"
b. Ka-rh-a-kon=hkó:wa rón:n-e'-s
NEUT.I-trCC-EPENTH-placc.in=AUG MASC.PL-go-DISTR.IMPF
"They were walking around in the great forest."
 Ka-hent-a!ke-hkó:wa ni-t-konwati-ia't-énha.
NEUT.I-field-place=AUG PART-CISLOC-3.PL/3.PL-body-carry.PF
"They got them from the big field."
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 155

Locative construction Noun

kanatákta "near the town" kand:ta' "town"
karahkwákta "near the moon" karáhkwa' "moon"
kanaktdkta "beside the bed" kandkta' "bed"
ahsirákta "next to the blanket" áhsire' "blanket"
ahskwen'nákta "next to the porch" áhskwèn:na' "porch"
ahsonhtákta "near the wall" ahsónhta' "wall"
onontákta "near the hill" onón:ta' "hill"
o'nerohkwdkta "next to the box" o'neróhkwa' "box"
okwirákta "near the tree" ó:kwire' "tree"
kanaktó:kon "under the bed" kandkta' "bed"
kentskaró:kon "under the rug" kéntskare' "rug"
kanekotó:kon "under the stairs" kanekó:ta' "stairs"
ahsiró:kon "under the blanket" dhsire' "blanket"
ahskwen'nó:kon "under the porch" ahskwèmna' "porch"
ahskó:kon "under the bridge" áhskwa' "bridge"
otsható:kon "in the clouds" otshá:ta' "cloud"
okwiró:kon "under the tree" ó:kwire' "tree"
ohnekó:kon "under water" ohné:ka' "liquid"
kaná:takon "in town" kaná:ta' "town"
kahiónhakon "in the creek" kahiónha' "creek"
kanáktakon "in bed" kandkta' "bed"
ahsónhtakon "inside the wall" ahsónhta' "wall"
a'thé:rakon "inside the basket" à:there' "basket"
a'd:rakon "in the net" à:'are' "net"
ohròmwakon "in the ditch" ohròn:wa' "ditch"
o'neróhkwakon "in the box" o'neróhkwa' "box"
oién:takon "in the wood" ó:iente' "wood"
kahehtà:ke "in the garden, field" kahéhta' "field"
kaniataràike "on the lake" kaniá:tara "lake"
karonhià:ke "in the sky" karòn:ia' "sky"
ahsonhtà:ke "on/in the wall" ahsónhta' "wall"
a'therà:ke "on the basket" à: there' "basket"
ahsirà:ke "on the blanket" dhsire' "blanket"
onontà:ke "on the hill" onón:ta' "hill"
onawa'tstà:ke "in the mud" onawà:sta' "mud"
ohahà:ke "on the road" oháha' "road"
Table 2. Pronominal paradigm selection governed by noun.

Another nominal clitic is the residential =hronon', which derives names for
inhabitants of places. Because of its meaning, it appears particularly often with
locative nominals.
156 MARIANNE MITHUN

(26) Locative with residential clitic.

ie-ronhi-dke-hró':norí
INDEF.AGT-sky-place=RESIDENTIAL
"resident of the sky place" = "angel"

A typical component of grammaticalization processes is an extension of

meaning, often to more general or abstract uses. The Northern Iroquoian locative
constructions exhibit the common extension from the domain of space to that of
time.

(27) Extension to time.

a. ent-ákta b. ako-hserà:ke
day-beside winter-NOMINALIZER-in
"Saturday" "wintertime"

The languages thus show the development of a new grammatical category.

Certain locative verb constructions containing incorporated nouns have been
reanalysed as locative nominals, and as a result the original verb roots in those
constructions have been reanalysed as locative nominalizers.

2. Actualization
The locative markers appear to constitute a paradigm. Yet individual
locative morphemes show different degrees of grammaticalization. Some of the
locatives appear to be closer to their verbal origins than others. The locative -ti

(28) Locative without associated noun; Warisose Kaierithon, speaker.

Ni-t-io'karà:-'on iako-ió't-e'
PART-CISLOC-NEUT.II-be.night-INCH-STAT INDEF.PAT-WOrk-IMPF
so is it night she works
s-ká:-ti ne ion-hsit't-à:ke
REPET-NEUT.I-on.other.side-STAT the NDEF-foot-place
it is on one side the where her foot is
one foot
te-ion-tatkarén:r-on o-wirà:-a.
DUPL-INDEF.AGT-rOCk-STAT NEUT.I-child-NOUN.SUFF-DIM
she is rocking it baby
"All night she worked rocking the cradle."
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 157

"beyond, on the other side" still occurs on its own as a predicate, without an
associated noun. The word ská:ti in (28) has the form of a regular verb without
incorporation.
Both -ti "beyond" and -(i)hen "in the middle of' still occur with affixes
otherwise found only on verbs. The locative construction formed with —ti in the
first line of (29) contains a partitive prefix and the verbal pronominal prefix w-.

(29) Locative -ti with verbal prefixes; Konwatsi'tsaienni Phillips, speaker.

ísi' nónhskwati she's
isi' na'-w-ahskw-ati she's
far PART-NEUT.I-bridge-EPENTH-place.beyond customarily
way on the other side of the bridge
niio.re' niekonnéhtha' ne tionnhónhskwaron
so.it.is.far there.they.wandered the they.have.jowels
"The cows would pasture way on the other side of the bridge."

The -ti construction in (30) contains a repetitive prefix, which also occurs
otherwise only in verbs.

(30) Locative -ti with repetitive; Seth Newhouse, speaker (Hewitt

1899:270.1).
S-ka-nhóh-a-ti
REPET-NEUT.I-door-EPENTH-place.beyond
place on the other side of the door
i-on-sa-ki-at-aweia't-e'
TRANSLOC-FACT-REPET-MASC.DU.AGT-MTODLE-enter-PF
there they two entered again
"They two went back into the other room."

The -(i)hen locative construction in (31) also contains prefixes that are part of
the regular verbal morphology, the coincident and the duplicative.

(31) Locative -then with verbal prefixes; Niioronhia'a Mae Montour.

sha' -te-ka-nekot-í:hen
COINCID-DUPL-NEUT.I-ladder-place.in.the.middle.of
place between the two equal halves of the ladder
158 MARIANNE MITHUN

sh-á-h-e- ....
COINCID-FACT-MASC.I-gO-PF
as he went
"When he was halfway up the ladder

The locatives -akta "beside" and -kon "in" no longer occur as predicates in
Mohawk, but cognates -akwt and -kew have been recorded functioning as
predicates in Tuscarora, the Northern Iroquoian language most distantly related
to Mohawk. The first, recorded in this use by Rudes, was translated "be beside"
or "be near".

(32) Tuscarora -akwt "be beside" (Rudes 1999: 42).

na'-á'.-k-t-akwt
PART-OPT- 1 -DU-be.beside.PF
"that the two of us be near one another"

The second, recorded a century ago by J.N.B.Hewitt, was translated "lie

inside".

(33) Tuscarora -kәw "lie inside" (Hewitt in Rudes & Crouse 1987:79).
we-hra-kәw-hә-h
TRANSLOC-MASC.AGT-lie.within-DISTR-IMPF
"he is lying there inside"

On the basis of the substantial textual material recorded by Hewitt at the end of
the nineteenth century, we can see that even then the use of these morphemes as
predicating verbs was rare.
Furthest advanced along the path from verb to nominalizer is the general
locative -a'ke. In many contexts -a'ke "place of' behaves like the other locatives.
It still shows traces of a verbal origin, requiring overt nominalizers on associated
deverbal noun stems. But it also shows evidence of an evolution toward status as
a simple noun ending: it never functions as a main predicate, always occurs with
the prefixes appropriate for nouns, never governs the choice of prefix paradigm,
and appears with nominal enclitics. Certain -a'ke constructions show an even
further evolution.
While many noun stems in Iroquoian languages appear both in independent
nouns and incorporated in verbs, some appear in only one context
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 159

or the other. The Mohawk roots akehr- and -ks- both mean "dish"; the first
occurs in nouns but not verbs, while the second occurs in verbs but not nouns.

(34) Noun roots akehr- and -ks- "dish".

a. akè:r-a
(NEUT.I)-dish-NOUN.SUFF
"dish(es)" (noun)
b. a-hshe-ks-a-hér-hahs-e'
OPT-2.SG/3.PL-dish-EPENTH-Set-BENEFACTIVE-PF
"you should set down a dish for them" - "... serve them" (verb)

The distribution of the two stems in locative constructions is interesting. Most of

the locative morphemes occur with the incorporable root -ks-, like verbs. The
locative -a'ke, however, occurs with the root akehr-, which usually forms the
basis of nouns.

(35) Locative constructions.

Grammatical Ungrammatical
kaksákta "beside the dish" (*akehrâkta)
kaksóikon "under the dish" (*akehró:kori)
kdksakon "in the dish" (*akè:rakon)
akehrà:ke "on the dish" (*kaksà:ke)

The general locative -a'ke appears to be simply an ending added to nouns. (In
stressed syllables, a coda glottal stop stimulates creaky voice over the preceding
vowel, lowers the tone, then disappears: *akehra'ke —» akehrà:ke.) In fact some
forms show that the general locative has been reanalysed from a single
morpheme -a'ke to a sequence of noun suffix -a' plus locative enclitic =ke. In
earlier formations, the shape of the general locative is -a'ke no matter what the
shape of the noun suffix in the corresponding independent noun, as in (36). In
more recent formations, the noun suffix remains, and only =ke is attached; cf.
(37).

(36) Original locative -a'ke regardless of the shape of the noun suffix.
kén-tsi-on ken-tsi-à:ke
NEUT.I-fish-NOUN.suFF NEUT.I-fish-place
"fish" "on the fish"
160 MARIANNE MITHUN

(37) More recent formations with =ke following noun suffix.

a. ka-ríht-on' ka-riht-ôn:=ke
NEUT.I-police-NOUN.SUFF NEUT.I-police=place
"policeman" "police station"
b. Oston Oston-hronön:-ke
Boston Boston=RESIDENTiAL=place
"Boston" (loan) "Bostonians' place" = "U.S."

As noted above, the locatives, like verbs, generally occur with formal noun
stems, either noun roots or noun stems containing overt nominalizers.

(38) Locatives with nominalizers.

a. o-nonhs-a-tokenht-i'-'tsher-ákta
NEUT.II-house-EPENTH-be.holy-STAT-NOMINALIZER-place.beside
"next to the church"
b. o-nonhs-a-tokenht-i'-'tsher-ó:kon
NEUT.I-house-EPENTH-be.holy-STAT-NOMiNALiZER-place.under
"under the church"
 o-nonhs-a-tokenht-i'-'tsher-á'.-kon
NEUT.I-hoUSe-EPENTH-be.holy-STAT-NOMINALIZER-EPENTH-place.in
"inside the church"
d. ka-hwehn-o-'tsher-ákta
NEUT.I-island-be.in.water.STAT-NOMINALIZER-beside
"near the island"

But the nominalizer is not always present before the general locative enclitic.

(39) General locative =ke without nominalizers.

a. o-nonhs-a-tokenht-i-ke
NEUT.II-house-EPENTH-be.holy-STAT=place
"at the church"
b. ka-wehn-o:-ke
NEUT.I-island-be.in.water.STAT=place
"on the island"
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 161

c. sh-io-rh-ón'=ke
COINCID-NEUT.II-dawn-STAT=place
"time when it has dawned" = "this morning"

The locative appears to have simply been attached as an ending to these

lexicalized nominals, even though they are morphological verbs.
The term for "Tuesday" is interesting in that it shows multiple layers of
derivation, with an inner occurrence of the earlier locative -a'ke and an outer
occurrence of the later locative =ke.

(40) Layers of locative derivation.

rati-ronhi-a,ke-hronön:-L·
MASC.PL.I-sky-place=RESIDENTIAL=place
"time of the residents of the heavens" = "day of angels" = "Tuesday"

The form =ke now alternates with an allomorph -hue under phonological
conditioning: =ke occurs after glottal stop, and -hne occurs everywhere else.
The diachronic origin of =hne can no longer be discerned, and it shows no traces
of an earlier verbal origin. It may be added to any lexicalized nominal whatever
its internal morphological structure, and it never requires a nominalizer.

(41) Locative -hue on deverbal nouns without nominalizer.

a. ate-khw-a-hrá-hne
(NEUT.I)-MIDDDLE-food-EPENTH-set=place
"on the table"
b. ka-nonhs-és-hne
NEUT.I-house-be.long.STAT=place
"at the longhouse"
 an-itskw-a-hrá=hne
(NEUT.I)-MIDDLE-buttocks-EPENTH-set=place
"on the chair"
d. akai-ón-hne
(NEUT.I)-be.old-STAT=place
"secondhand store"
162 MARIANNE MITHUN

e. onhwentsi-a-kai-ön:-ne
(NEUT.II)-land-EPENTH-be.old-STAT=place
"Europe"
f. Ken-tsi-a!-kowá-hne
NEUT.I-fish-NOUN.SUFF=AUG=place
"place of the big fish" = "Fort Covington, New York"
g. onkwe - honwè - :ne
(NEUT.I)-person=real=place
"Indian land, reserve"
h. ronwa-ia't-a-nentakt-on=ne
3.PL/MASC.SG-body-EPENTH-Stick-STAT=place
"time of their having nailed him to the cross" = "Friday"

The nominal clitic -hne also never governs the choice of prefix paradigm. It is
added even to nominals with no prefix at all, and no other noun morphology.

(42) Locative -hne on other nominals.

a. Warí-hne . kweskwés=hne
Mary=place pig=place
"at Mary's" "pigpen"
b. Sosé=hne d. kitkít-hne
Joseph=place chicken=place
"Joseph's time" = "Wednesday" "chicken coop"

The locatives thus exhibit varying degrees of evolution from verb to

nominalizing enclitic. In many instances -(i)hen "middle" and -ti "place beyond"
show more verbal features than -akta "place beside", - "place under", -kon
"inside", and -a'ke "place"; and =ke/hne "place" show few verbal features at all.
But the actualization pattern is not just a straight path. Individual locative
morphemes show different patterns of behavior in different lexical items.
As we havc seen, it is now usually the noun that governs the choice of
prefix paradigm in locative constructions. But in some constructions the choice is
still governed by the locative morpheme. The nouns onón:ta' "hill" and
ohson:kara' "board" appear with the Paradigm II prefix o- on their own, but
locative constructions containing -ti "beyond" and -ihen "middle" show the
Paradigm I prefix ka-.
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 163

(43) Government of prefix by locative; Kanerahtenhawi Nicholas, speaker.

a. η-ia' -ka-non:t-a-ti
PART-TRANSLOC-NEUTJ-hill-EPENTH-place.beyond
"the other side of the hill"
b. sha'-te-ka-hson'kar-í:hen
COINCIDENT-DUPL-NEUT.I-board-middle
"the middle of the board"

The prefix-locative combinations seen here and earlier are regularly recurring,
idiomatic constructions: the partitive plus locative -ti, the repetitive plus locative
-ti, and the coincident plus duplicative plus locative -ihen. All systematically
appear with Paradigm I prefixes. Lexicalization is a significant factor in both the
frame of these structures and in the inventory of locative constructions in the
language. Not all combinations of locative markers and nominals exist. To say
"in the middle of the rock", for example, a periphrastic construction is used.

(44) Periphrastic construction; Kanerahtenhawi Nicholas, speaker.

ahsén:nen ne o-nén:i-a'
middle the NEUT.II-rock-NOUN.SUFF
"in the middle of the rock"

Similar idiosyncrasies appear with -akta "beside". In combination with the

noun onón:ta' "hill", it yields kanontákta "beside the hill" with prefix choice
governed by the locative. With the noun onén.ia' "rock", the usual form is
onenidkta "beside the rock", but kanenidkta is heard as well. As often happens
when doublets are in competition, differences in patterns of usage can sometimes
be discerned. With the noun o'neróhkwa' "box", the usual form is
ka'nerohkwdkta "beside the box", but o'nerohkwákta is also heard. Speaker
Kanerahtenhawi Nicholas observes that the first is used in contexts of
immediacy and the second in contexts of greater remoteness. If I am carrying
something heavy, she would use the command in (45) to ask me to set it down
right next to the box immediately in front of her.

(45) Immediacy; Kanerahtenhawi Nicholas, speaker.

Kèn'.'en ka-'nerohkw-dkta í-ts-ien.
here NEUT.I-box-beside PROTHETIC-2.SG.AGT-set
"Set it down right here next to the box."
164 MARIANNE MITHUN

If on the other hand she found something last week, and I asked her where she
had found it, she might respond as in (46).

(46) Greater remove; Kanerahtenhawi Nicholas, speaker.

O-'nerohkw-ákta wake-tshénri-on.
NEUT.II-box-beside 1 .SG.PAT-find-STAT
"I found it next to the box."

The locatives -kon and -a'ke show the expected government of the prefix choice
by this noun: o'neróhkwakon "in the box", o'nerohkwà:ke "on the box".
But -kon "in" shows variation of its own. The term for the muddy bottom
of a river is onón:wa'. To refer to the area in the muddy bottom, both
kanón:wakon and onón:wakon are used, with different prefix preferences for
different speakers. Another noun, o'nónhkwa' "bottom, seat of pants", always
appears with the Paradigm II prefix o- on its own, but the term for "area in the
bottom", as in the bottom of a barrel, is always ka'nónhkwakon.
Even -a'ke shows variation. The noun otstèn.ra "rock" appears with the
Paradigm II prefix o- on its own, but with the Paradigm I prefix when associated
with the general locative: kentstenhrà:ke "on the rock". The model for this prefix
choice appears to be another morphological verb commonly used to refer to rock:
tkentstèn:rote' "there it rock stands".

(47) otstèn.ra "rock"

ken-tstenhr-à:ke t-ken-tstènr-ot-e'
NEUT.I-rock-place CISLOC-NEUT.I-rock-stand-STAT
"(place) on the rock" "standing rock" = "rock"

Because the locative nominalizers create lexical items, derived forms may
remain in the language after the bases on which they were formed go out of use.
The term onontoharà:ke "on top of the hill" occurs frequently, but the base,
onontóhare', does not occur.
Individual lexical items also show varying degrees of phonological
crosion, a typical concomitant of grammaticalization. The term for "in the house",
a very frequent word, shows erosion not found in other words with parallel
morphological structure. Its expected form would be kanónhsakon, but the
unstressed epenthetic vowel has been lost, reducing it to kanónhskon. Erosion is
also common in placenames, highly lexicalized constructions whose internal
morphological structure can fade quickly with use. The name for "Montreal" is
Tiohtià:ke, with an ending typical of nominalized locatives, but the base of the
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 165

form is no longer entirely clear. Parts are suggestive, and speakers have various
ideas about possible earlier forms, but it is clear that some of its substance has
been lost.

3. Motivating f orces
The verb-to-locative shift shows different degrees of development with
different locative markers. It has apparently progressed the furthest with =ke/hne
"place", slightly less far with kon "inside", - "underneath", and -akta "near,
beside", and still less with (i)hen "middle", -ti "beyond", and several others. We
can now ask whether the pattern of spread reflects any general principles.
In a number of ways the shift seems to have proceeded from the unmarked
to the marked, in accord with hypotheses proposed by Timberlake (1977) and
Andersen (1987, 2001:30-37). The most grammaticalized marker, =ke/=hne, is
the most general in meaning: "place". It is by far the most frequent locative in
natural speech, perhaps more frequent than all other locatives combined. It is also
the only locative to show significant allomorphy, a feature cited by Greenberg
(1966) as characteristic of unmarked elements.
Another feature associated with markedness by Givón (1990:945-966) is
degree of cognitive complexity. This feature too is pertinent to the change at
hand. In a survey of locative constructions in twenty-six languages, Svorou
(1994, 1999) found that "asymmetry in the degree of grammaticalization parallels
the cognitive asymmetry observed with respect to the frontal axis, vertical axis,
and in/on terms" (1999 handout). She cites work by Clark (1973) and Miller and
Johnson-Laird (1976) demonstrating that vertical axis terms have greater
cognitive salience, and thus greater conceptual simplicity, than frontal axis terms.
She notes that the degree of cognitive salience or conceptual simplicity matches
the order of acquisition of spatial terms by children in a number of languages,
resulting in the order:

on > in > under > beside > back > front

The match between this pattern and the progression of grammaticalization of the
Northern Iroquoian locative markers is remarkable.
In fact the motivation behind both the shift and its spread can be
understood in terms of general cognitive and communicative factors. We know
that grammaticalization typically begins with items that are relatively general in
meaning. (Semantically general markers are by nature typically simplex
cognitively.) The generality of meaning engenders high frequency of use, since
such markers are applicable to large numbers of contexts. (Cognitive simplicity
and especially frequency of use naturally lead to early acquisition by children.)
166 MARIANNE MITHUN

Human beings tend to routinize recurring operations, whether it be riding a

bicycle or expressing ideas. The automation of frequently occurring expressions
results in a loss of awareness of their internal structure. A verb like karhd:kon "it
is in/among the trees", used repeatedly to identify a kind of place, would come to
represent not a complex description but a single concept, the label for a forest.
Such constructions appeared in the same syntactic contexts as morphological
nouns, and began with a prefix identical to those found on nouns: ka-. The
semantic, syntactic, and morphological ambiguity set the scene for a
reinterpretation of the original clausal description ("it is among the trees") as a
nominal ("place among the trees" > "forest"). Morphemes recurring frequently at
the ends of these nominals, like kon here, were reanalysed as nominalizers. The
shift in analysis was accompanied by a shift in referentiality. In locative verbs,
the only referential element was the pronominal prefix ("it is among the trees"),
but after reanalysis, the entire word was used referentially ("forest"). New
locative constructions began to appear with prefix shapes appropriate for nouns,
even when these differed in shape from the prefixes used on verbs, and the
prefixes on existing forms began to be remodeled to conform to their new status
as nominals. As parallels with noun incorporation faded, new forms were
derived by adding locative suffixes to existing independent nomináis, without
formally nominalizing them first or altering their prefixes.
An important aspect of the evolution is the fact that it shows different
degrees of development with different markers and in different lexical items.
Actualization that proceeds morpheme by morpheme and word by word would
appear to confound attempts at description in terms of general features. We know
that idioms often preserve relicts of earlier forms and constructions, because they
are lexicalized: speakers retrieve them as units from memory rather than creating
them anew each time they speak. Words can constitute particularly tightly
lexicalized units. In the Northern Iroquoian languages, the lexicalization of
locative constructions has preserved earlier grammatical patterns, even after new,
productive patterns have evolved for deriving additional forms.
But lexicalization can also facilitate change. The integrity of lexicalized
locative constructions first set the stage for reanalysis by reducing the salience of
their internal structure and components. The most strongly lexicalized
expressions were those used the most often. These forms served as models
according to which speakers derived new vocabulary. The process of
grammaticalization was thus accelerated for those locative morphemes used the
most often, such as the general =ke/-hne "place", and within the areas of the
lexicon most often enriched by derivation, such as placenames.
 FROM CLAUSE TO LOCATIVE MORPHOLOGY IN IROQUOIAN 167

Nichols and Timberlake describe grammar as a network with areas of

varying degrees of rigidity.
The relatively fixed areas of a grammatical network are then exemplars with
characteristic properties; exemplars are pieces, perhaps minimal units, of text.
Because a grammar based on exemplars necessarily underspecifies usage, new tokens
of text may arise by extrapolating from given exemplars. These new tokens of usage
can be conventionalized as exemplars (grammaticalized, if you will), setting the
stage for more innovations. (1991:130)

Lexicalization can be a powerful force in fixing areas of such a network, creating

conventionalized exemplars, pieces of text which serve as models for further
formations. Seen in this light, the morpheme-by-morpheme and word-by-word
actualization of the grammaticalization of the locative construction turns out to be
principled after all.

REFERENCES
Andersen, Henning. 1987. "From auxiliary to desinence". Historical
Development of Auxiliaries ed. by Martin Harris & Paolo Ramat, 21-52.
Berlin: Mouton de Gruyter.
Andersen, Henning. 2001. "Markedness and the theory of linguistic change".
This volume, 21-57.
Ard, William Josh. 1975. "Raisings and word order in diachronic syntax".
Unpublished Ph.D. dissertation, University of California, Los Angeles.
Bavin, Edith. 1990. "Locative terms and Warlpiri acquistion". Journal of Child
Language 17.43-66.
Bruyas, Jacques. 1862. Radical Words of the Mohawk Language. New York:
Cramoisy Press.
Clark, Herbert. 1973. "Space, time, semantics, and the child". Cognitive
Development and the Acquisition of Language ed. by T. E. Moore, 28-63.
New York: Academic Press.
Disterheft, Dorothy. 1984. "Irish complementation: a case study in two types of
syntactic change". Historical Syntax ed. by Jacek Fisiak, 89-106. Berlin:
Mouton de Gruyter.
Dromi, E. 1979. "More on the acquisition of locative prepositions: an analysis of
Hebrew data". Journal of Child Language 6.547-562.
Givón, Talmy. 1990. Syntax: a Functional-Typological Introduction, II.
Amsterdam & Philadelphia: John Benjamins.
Greenberg, Joseph H. 1966. Universals of Language. (Janua Linguarum, Series
Minor, 59.) The Hague: Mouton.
Harris, Alice & Lyle Campbell. 1995. Historical Syntax in Cross-Linguistic
Perspective. Cambridge: Cambridge University Press.
Hewitt, Jonathan Napoleon Brinton. 1899-1900. "Mohawk cosmology". US.
Bureau of Ethnology Report 21.255-339.
Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization.
Cambridge: Cambridge University Press.
168 MARIANNE MITHUN

Johnston, Judith R. 1984. "Acquisition of locative meanings: 'behind' and 'in

front of". Journal of Child Language 11.407-422.
Lichtenberk, Frantisek. 1991. "On the gradualness of grammaticalization".
Approaches to grammaticalization ed. by Elizabeth Closs Traugott & Bernd
Heine, vol. 1, 37-80. Amsterdam & Philadelphia: John Benjamins.
Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge
University Press.
Lightfoot, David. 1991. How to Set Parameters: Arguments from Language
Change. Cambridge, Mass.: The MIT Press.
Lightfoot, David. 1999. The Development of Language: Acquisition, Change,
and Evolution. Malden, Mass.: Blackwell Publishers.
Miller, George A. & P. N. Johnson-Laird. 1976. Language and Perception.
Cambridge, Mass.: Harvard University Press.
Nichols, Johanna & Alan Timberlake. 1991. "Grammaticalization as
retextualization". Approaches to Grammaticalization ed. by Elizabeth Closs
Traugott & Bernd Heine, vol. 1, 129-146. Amsterdam & Philadelphia: John
Benjamins.
Plank, Frans. 1984. "The modals story retold". Studies in Language 8.305-364.
Rudes, Blair A. 1999. Tuscarora-English, English-Tuscarora Dictionary.
Toronto: University of Toronto Press.
Rudes, Blair A. & Dorothy Crouse. 1987. The Tuscarora Legacy of J. N. B.
Hewitt. (Canadian Ethnology Service Paper No. 108.) Ottawa: Canadian
Museum of Civilization.
Svorou, Soteria. 1994. The Grammar of Space. Amsterdam & Philadelphia:
John Benjamins.
Svorou, Soteria. MS. "Semantic constraints in the grammaticalization of locative
constructions". Paper presented at the conference "New Reflections in
Grammaticalization", Potsdam, Germany, 1999.
Timberlake, Alan 1977. "Reanalysis and actualization in syntactic change".
Mechanisms of Syntactic Change ed. by Charles N. Li, 141-180. Austin:
University of Texas Press.
Warner, Anthony. 1982. Complementation in Middle English and the
Methodology of Historical Syntax. University Park, Pa.: Penn State
University Press.
Warner, Anthony. 1983. Review article on Lightfoot (1979). Journal of
Linguistics 19.187-209.
Warner, Anthony. 1990. "Reworking the history of the English auxiliaries".
Papers from the 5th International Conference on English Historical
Linguistics ed. by Sylvia Adamson, Vivien Law, Nigel Vincent & Susan
Wright, 537-558. Amsterdam & Philadelphia: John Benjamins.
Weinreich, Uriel, William Labov & Marvin I. Herzog. 1968. "Empirical
Foundations for a Theory of Language Change". Directions for Historical
Linguistics ed. by Winfred P. Lehmann & Yakov Malkiel, 95-188. Austin:
University of Texas Press.
 FROM LATIN TO MODERN FRENCH:
ACTUALIZATION AND MARKEDNESS

Lene Schøsler
KøbenhavnsUniversitet

0. Introduction
In this paper I will present and discuss changes in French from the areas of
morphosyntax and syntax. I will attempt to assign to these changes a place
within the framework of the theory of markedness, labelled the "formal"
approach, proposed by Andersen (1990, 2001). Andersen (2001) claims that
"markedness relations can be observed in every variety of linguistic change,
from its inception to its completion, both in the relations among variants and in
the relations that define the plethora of categories that typically condition the
gradual process by which newer forms replace older correspondents". Do the
changes I will discuss confirm or disconfirm Andersen's theory of actualization?
Or is the alternative offered by Timberlake (1977, 1999), which Andersen
labeled the "substantive approach", to be preferred? In the following I will
mainly focus on the question of how changes occurred, that is, I will focus on
the process of actualization, usually without discussing why the changes
occurred, i.e. without discussing the reanalysis process.

1. Morphosyntax
In this Section I present the main changes in the nominal declension
system from Latin to Modern French. I especially look at the question of
whether the results of the changes fit into Andersen's theory of markedness.

1.1 The Declension system from Latin to Modern French

During the evolution of the noun case system from Latin to the modern
Romance languages the many declension types broke down, and the six cases
had either disappeared already in the oldest sources (Italian, Spanish,
Portuguese) or were reduced to a two case system in medieval times, followed
by the reduction to a system without case (French, both langue d'oïl and langue
d'oc, and Rhaeto-Romance); only one language, Rumanian, still has a
170 LENE SCH0SLER

functioning case system (with genitive-dative forms vs. nominative-accusative

forms). See Tables la and lb for a presentation of the reduction of the nominal
declension system in Northern France. Without going into details concerning the
possible factors leading to the loss of the case systems, I will follow the
reduction process in French, focusing on the actualization, on the basis of a
corpus of 5 million words.1
The evolution of case in Old French will be described from the following
points of view: word class (Section 1.2) and dialect and chronology (Section
1.3). Finally, I consider the evolution in terms of markedness (Section 1.4).

1.2 Word class

With respect to word class, loss of case starts earlier in proper nouns than
in common nouns, earlier in feminine nouns than in masculine nouns, earlier in
adjectives than in nouns, earlier in nouns than in articles, earlier in nominal
classes and articles than in pronouns (personal and some relative pronouns still
have case declension in Modern French), earlier in nouns referring to
nonhumans than in nouns referring to humans (see Reenen & Sch0sler 1997).
The relations between the singular and the plural are unclear for several
reasons that I will not discuss here (e.g., conflicting case and number markers;
for more details, see Sch0sler 1984). What is clear, however, is that the singular
conserves case more frequently than the plural. Thus, case distinction in feminine
nouns is only found in the singular. Moreover, stem allomorphy distinguishing
case occurs in the singular only (more details below). With

Masculine, Second Declension Singular Plural

Case system 1: Latin
Nominative mūrus mûrī
Accusative ūr mūrōs
Genitive mūri mūrdrum
Dative mūrö mūris
Ablative mūrö mūris
Case system 2: Old French
subject case murs mur
oblique case mur murs
Case system 3: Modern French
unique form mur murs

Table la. Reduction of the nominal declension system in Northern France, masculine.

For a detailed discussion of the changes and of the possible causes leading to these changes,
see Schøsler (1984), Reenen and Schøsler (1986, 1988, 1997).
 FROM LATIN TO MODERN FRENCH 171

Feminine, First declension Singular Plural

Case system 1 : Latin
Nominative rosa rosae
Accusative rosam rosas
Genitive rosae rosârum
Dative rosae rosis
Ablative rosā rosīs
Case system 2: Old French
subject case rose roses
oblique case rose roses
Case system 3: Modern French
unique form rose roses

Table lb. Reduction of the nominal declension system in Northern France, feminine.

respect to declension, the most conservative classes are the semantic subclass of
nouns referring to humans and the grammatical class of articles. This hierarchy
of noun classes shows up more or less clearly from late Latin and in all Romance
languages. Thus, stem allomorphy is found, as a kind of redundant case
marking, only in nouns denoting humans, of the Latin third declension, as in
nominative imperator > OFr. emperere, accusative imp eratorem > empereor,
whereas late Latin already had abolished this type of case marking in nouns
denoting nonhumans, as in mansio > *mansionis > maisons, accusative
mansionem > maison.2
It is probably the same tendency that shows up in the use of prepositions
with human objects in several Romance languages as opposed to nonhuman
objects; see the examples in (1), most of them quoted from Meyer-Lübke III.
Clearly enough, there is, in the Romance languages, a special tendency towards
overt marking of nouns denoting humans, both in the function of subject and in
the function of object, whereas nouns denoting nonhumans are less clearly
marked.

(1) a. Sp. no he visto a mi hermano; no lejío he visto "I have not seen
my brother; I have not seen him".
b. Sp. quiero a Maria; la quiero "I love Maria; I love her".
 Sp. fueron a buscar a un médico "they fetched a doctor"; cf.
Meyer-Lübke III:372.
d. Rum. iubesc pe frate "I love my brother"; Meyer-Lübke III:374.
e. It., Sicilian dial. l'aviti visu a me/rati? "have you seen my
brother"; cf. Meyer-Lübke III:373.

2
These forms are discussed in detail in Reenen and Sch0sler (1988), in which also a general
discussion of the formation and evolution of these forms is found.
172 LENE SCH0SLER

f. Rhaeto-Rom., Vallader dial. Nus vezzain ad Annina "we see

Annina"; cf. Liver 1991:86.

Thus, we find in Old French a clear hierarchy with nouns denoting

humans being conservative with respect to declension, and nouns denoting
nonhumans, innovative. Proper nouns, however, are an exception: they lost their
case marking early, in spite of the fact that the proper nouns we are discussing
here refer to humans and thus might be expected to be conservative in their case
behavior.

1.3 Dialects and chronology

In French dialects, the loss of case starts in the west and spreads to the east
and the north at a slow rate from around 1000 to 1400. Dialect and chronological
differences are more important than linguistic hierarchy, which I find hard to
describe in terms of markedness, as proposed by Andersen. There is extensive

Figure 1. The progressive loss of declension reflected in thirteenth-century French dialects.

 FROM LATIN TO MODERN FRENCH 173

loss of declension in the very first Western Romance texts (ca. 1050), but
declension is still strong in the northern and eastern parts of Northern France
generally in the thirteenth century. The declensional system is totally abandoned
by the end of the fourteenth century. This dialectal and chronological movement
is well documented and illustrated in the maps of the two Atlases published by
Dees et al. (1980, 1987); see, e.g., Dees et al. 1980, map 206, Figure 1 above.
There is a high score of absence of declension marking ('case errors') in the
western dialects, whereas the eastern and northeastern dialects conserve the
declension marking. The map shows the proportion of omitted flectional -s in the
nominative singular of masculine nouns, without distinction of semantic
subtypes.
In this connection it is legitimate to ask questions such as whether the
changes occurring are evolutive changes (i.e. language-internal changes) or
contact changes (i.e. influenced from outside)? Are evolutive changes more
important than contact changes, or is it the other way round? A question of a
different, but very difficult kind is: which were the centers of prestige in
Northern France that possibly influenced the rate of the changes? I will not go
into details on the dialectal and chronological distributions and their causes here,
but refer to the presentation and discussion in Sch0sler (1984).

1.4 Markedness
What I will focus on now is whether loss of case occurs according to an
expected hierarchy or not. I reproduce in Table 2 the hierarchy from Andersen
(2001:31).

Unmarked Marked
(a) proper common
(b) human nonhuman
(c) animate inanimate
(d) concrete abstract
(e) singular plural
(f) definite indefinite
Table 2.

Now, let us try to establish a hierarchy of morphological and semantic features

of the Old French pronominal and nominal system. I will base my evaluation of
markedness on the morphological nominal system and on the function of the
forms. The case form that spreads in Old French is the unmarked oblique form.
It is unmarked, first since it is multifunctional as opposed to the marked
nominative form, which has only subject-related functions, and secondly because
174 LENE SCH0SLER

there is often syncretism between the two case forms, for instance always in the
feminine plural, but also in other cases. And syncretisms have the oblique form.
Thus, the oblique form, being unmarked, is expected to spread in unmarked
contexts. This does not seem to be fully in accordance with the hierarchy
established, even if we accept that in French the hierarchy is reversed compared
to the one quoted in Table 2. In fact, Andersen suggests that hierarchies may be
language specific, and we can reasonably claim that in French the categories
proper, human, etc. are marked, and the other series unmarked. But even then
proper nouns remain a problem, because proper nouns lose case first instead of
last. If we base the hierarchy of Old French on the rate of case loss, we find the
following hierarchy:

Case loss earlier Case loss later

(a) human proper nouns common nouns
(b) feminines masculines
(c) adjectives substantives
(d) nouns determiners
(e) nouns and determiners pronouns
(0 nonhuman nouns human nouns
(s) plurals singulars
Table 3.

This hierarchy is less transparent than the one in Table 2. If we assume the
oblique form was the unmarked form, and if we assume this form spread first to
unmarked contexts, then the unmarked contexts seem to be rather heterogeneous,
the main problem being the position of human proper nouns. Human proper
nouns and a few human nouns seem to possess similar semantic features, so one
would like them to form one group. Additionally, they constitute the only
instances in which the nominative form was preserved rather than the oblique
form. Nevertheless, they behave very differently with respect to declension, and
are consequently placed in different colums in Table 3. Andersen has pointed out
to me that the oblique form might not be a simple unmarked form. It might be the
case that the plural oblique form in -s is so strongly associated with plural
number, that the singular -s is marked and the plural -s unmarked, and that the
non-s-forms have the polar values. This interpretation would not preclude the
exceptional lexicalization of individual human proper nouns in -s. If we do
consider proper nouns a special case and break down the hieararchy as proposed
by Andersen, the hiearchy appears indeed more homogeneous.
Linguists usually do not examine the distribution of language changes
outside the areas they consider directly relevant to the phenomenon under
analysis. I have carried out a modest check of an Old French prose text from the
 FROM LATIN TO MODERN FRENCH 175

end of the thirteenth century, Le Roman de Tristan en Prose. In this text I have
tested the evolution of case declension in a context which is not directly relevant
to the change examined. I think that if Andersen's actualization theory is correct
in claiming that language evolution occurs according to a (language specific)
hierarchy relevant for all changes, then the evolution of case might be different at
different text levels, for instance in main clauses as opposed to subordinate ones,
and in direct discourse as opposed to epic, or narrative, passages, even if I
cannot see any syntactic or pragmatic reasons for this difference. However, I
have problems with the attribution of markedness to genres and registers in older
languages. One could imagine that direct discourse is unmarked and closer to
ordinary speech. On the other hand, one could imagine, to the contrary, that
direct discourse is marked since written text in medieval times is essentially an
artificial language, which causes spoken language inside a written text to be
doubly marked. Can we directly project our feelings of what is marked and what
is unmarked onto older languages? However, whether we conceive of direct
discourse as marked or unmarked there should always be a difference
concerning the use of case, as long as we accept the hierarchical theory of
actualization of Andersen. I have analysed 1171 noun phrases in three sections
of Le Roman de Tristan en Prose, of which 689 occur in narration, and 482 in
direct discourse. What I find is presented in Table 4a and Table 4b.

Direct discourse Narration

case preserved loss of case case preserved loss of case
401 = 83.2 % 81 = 16.8 % 615 = 89.3 % 74 = 10.7 %

Table 4a. Variation in nominal declension in two different registers

in the Roman de Tristan en prose.

Main clauses Subordinate clauses

case preserved loss of case case preserved loss of case
298 = 80.5 % 72 = 19.5 % 227 = 87.0 % 34 = 13.0 %

Table 4b. Variation in nominal declension in the Roman de Tristan en prose

according to clause type.

It is quite clear that loss of case occurs more frequently in direct discourse than
in narration, and that this result is significant (X2 = 9.08, ρ < 0.01). So the
unmarked case form progresses first in what is intuitively felt as the register
closest to ordinary speech, i.e. direct discourse.
When I examine loss of case in main clauses and subordinate clauses in a
section of the same text, I find a total of 631 noun phrases, and again the
 176 LENE SCH0SLER

difference is striking: loss of case occurs more frequently in main clauses than in
subordinate clauses, and that result again is significant (X2 = 4.53, ρ < .05). This
confirms the spreading of the unmarked case form in the unmarked clause type.
Let us conclude this section. Inside the nominal declensional system I find
the spread of the unmarked, oblique form, following a clear hierarchy from
-human, -definite to +human, +definite contexts, with the outstanding exception
of human proper nouns, which semantically belong to the latter category, but
formally belong to the former. If we consider the syntactic and pragmatic
conditions of actualization, the unmarked oblique form clearly progresses earlier
in what can be considered unmarked contexts, direct discourse (vs. narration)
and main clauses (vs. subordinate clauses).

2. Syntax
In this section I will study the process and possible interaction of the
following changes in Old and Middle French: 1. word order, 2. loss of nominal
declension, 3. loss of verbal inflection, 4. loss of null subjects (Section 2.1). A
second problem to be studied concerns tense and aspect (Section 2.2).

2.1. Four interacting changes

During the Old and Middle French period four possibly related changes
took place. In the following sections I will briefly recount the facts (Section
2.1.0), consider the possible interrelations of the changes (Section 2.1.1), and
finally evaluate the actualization changes in terms of markedness (Section 2.1.2).

2.1.0 Four Changes. The four possibly related changes are, from (a) to (d):
(a) The shift of word order type, from Late Latin OV 3 to (Old French)
VO, turns Old French into a V2-language. Classical French (seventeenth
century) adopts a fixed SVO order in declaratives. It has been argued (see Vance
1997, with references) that late twentieth-century Modern French is no longer a
V2 language.
(b) The nominal two-case declension was lost in the period from ca. 1000
to 1400 depending on the dialect, as described in the previous section. I have
argued (Sch0sler 1984) that the most important sentence function of the
declension system was to mark the distinction between the grammatical subject
(and elements related to the subject) on one hand and the different grammatical
oblique functions on the other hand. This distinction is marked by means of
fixed word order at least from the period of Classical French.

3
Here I refer to the traditional view, rightly contested, however, by Pinkster (1991).
 FROM LATIN TO MODERN FRENCH 177

(c) Early Old French verbal paradigms distinguish all six persons with
specific inflections in the present tense indicative; late Old French has at least
three distinct persons (2SG, 1PL, and 2PL); Modern Standard French has only two
distinct persons (1PL and 2PL). In nonstandard French there is only a single
distinct form left (see the tables in Vance 1997:215).
(d) Latin is a language with no obligatory grammatical subject; in other
words, it permits Pro-drop. In many Modern Romance languages, e.g., Spanish,
Portuguese, and Italian, the nonexpression of the subject pronoun is the norm;
others, e.g., French, Franco-Provençal, Rhaeto-Romance, and some Occitan and
Northern Italian dialects, do not permit Pro-drop. However, this second group of
Romance languages all had null subjects in their medieval stages.
Let us consider different propositions concerning causality. Numerous
diachronic studies have proposed different causal relations among the four
changes mentioned.
(a)-(b) Word order and nominal declension.
It has been proposed that the fixation of word order, which took over the
functions of the nominal declension system, in fact caused the loss of this
system, as the latter became superfluous. However, this very popular causal link
is in conflict with the facts, one of the evident problems being that the fixation of
word order first appeared after the loss of declension (for details, see e.g.
Sch0sler 1984).
(c)-(d) Obligatory subjects and verbal inflection.
It has been argued that rich verbal inflection is a condition for the existence
of null subjects (for a detailed discussion, see Vance 1997). However, it is
difficult to determine exactly what constitutes 'rich inflection': does Old French
have rich inflection? Chinese has no verbal inflection of persons, yet it has null
subjects. German certainly has a rich verbal inflection, but it does not permit Pro-
drop. All statistics of null subjects in Old and Middle French (see e.g. Vance
1997:322, 350) show that null subjects are more frequent in Middle French, i.e.
at a period of reduced verbal inflection compared to Old French, so the proposed
causal link between 'rich inflection' and null subjects does not fit the facts.
Moreover, there is negative statistical evidence concerning the correlation of null
subjects and inflection of the persons that have continuously overt flexion, i.e.
1PL and 2PL (Sch0sler 1991, Vance 1997, chapter 6). Interestingly enough, null
subjects vary according to the type of sentence and the grammatical person (see
Vance 1997, chapter 6), e.g., the first and second persons are reluctant to have a
null subject, whereas the third, especially the impersonal third person, and all the
persons of the plural favor null subjects Vance 1997:304-305). These statistics
show no clear correlation between the two phenomena of null subjects and
inflection.
178 LENE SCH0SLER

2.1.1 Word order modifications. Let us try to describe in detail some of the
modifications of word order that occurred in Medieval French.
In Early Old French, the anteposition of any nonsubject and nonclitic
complement in declarative main clauses provoked the postposition or
suppression of the grammatical subject. This is a clear case of V2. These rules,
however, never applied in subordinate clauses. Neither postposition nor
suppression of the grammatical subject was general in Old French subordinate
clauses. It is thus possible to imagine that V2 never spread from main clauses to
subordinate clauses. Consequently, it is possible to describe main clauses as
innovative in comparison to subordinate clauses with respect to the word order
change from (Latin) XOV to (Old French) V2.
Now let us consider overt subjects. Recall that Latin had null subjects, and
Modern French has obligatory subjects. Old French shows a special, well-
known, intermediate pattern, as mentioned above: it permits Pro-drop much more
frequently in main clauses than in subordinate clauses (Vance 1997, Chapter 6).
Surprisingly enough, the number of overt subjects in Middle French subordinate
clauses diminishes, but it is still more frequent than overt subjects in main
clauses. How can we account for these differences in distribution? The logic of
the argument suggests that we consider subordinate clauses innovative in
comparison to main clauses with respect to the generalization of subject
pronouns. Moreover, it has been proposed (Vance 1997:182) that inversions in
Old and Middle French subordinate clauses (which are not due to the V2 rules,
as mentioned above) follow the same rules for stylistic inversion as Modern
French main clauses. If this is correct, we are forced to accept conflicting
tendencies with respect to innovation in the two types of clauses, and this is not
very satisfactory.

2.1.2 Markedness. Let us present the nature of the four changes in terms of
markedness.
1. V2 word order first appears in main clauses, i.e. in the unmarked type,
and never spreads to subordinate clauses. Accordingly, inversion of the subject
follows the V2 pattern only in main clauses. In subordinate clauses we find a
different pattern. Generativists implicitly classify the subordinate pattern as
innovative (Vance 1997). In terms of markedness, though, it is more satisfactory
to consider the subordinate pattern conservative and to consider its word order a
simple continuation of the Late Latin word order. Against this claim stands the
genuinely innovative presence of subjects spreading from subordinate clauses to
main clauses. In Modern French, V2 is now disappearing. Can this be accounted
for as another spread of specific subordinate sentence features to main clauses?
Or must we consider that in French, word order innovations only affect main
 FROM LATIN TO MODERN FRENCH 179

clauses? The high frequency of subjects in subordinate clauses stands against the
latter claim.
2. The loss of the nominal declension system has been described in terms
of markedness in the preceding section. It does not seem to have any influence
on word order changes.
3. The loss of the verbal inflection in all persons but 1PL and 2PL has been
considered the direct cause of the obligatory status of overt subjects. However, it
appears clearly from the facts mentioned above that the two changes, loss of
inflection and loss of Pro-drop, are not linked. The loss of inflection in Old and
Middle French concerns lSG, 2SG, 3SG, and 3PL of the present indicative (and
subjunctive), the imperfect indicative, and the conditional. In no tenses at all does
it abolish the inflection of 1PL and 2PL, except in nonstandard Modern French.
Put differently: the loss of personal inflection occurs earlier in the third person
than in the interlocutor persons, earlier in the singular than in the plural. In other
words, in unmarked contexts before marked contexts. The loss is earlier in the
most unmarked of all tenses: the present tense, and earlier in the indicative than
in the subjunctive. So this evolution is well described in terms of markedness of
grammatical processes.4
4. The loss of Pro-drop starts in subordinate clauses. The grammatical
persons first losing Pro-drop are first and second, referentials, the last (in
Classical French) are third nonreferential and a few special cases of 1PL and 2PL
with definite reference, mentioned by Maupas (seventeenth century, according to
Vance 1997:323). After the seventeenth century, the use of subject pronouns is
grammaticalized. This progression of change does not easily fit into a hierarchy
of markedness.

2.1.3 Conclusion. This short study of (possible) interrelations between the four
changes show that some changes are easily described in terms of markedness,
but that innovation does not seem to spread uniformly according to a markedness
hierarchy. It is, however, possible to claim that at least some changes mainly
affect the unmarked contexts (word order changes in main clauses); still, the loss
of Pro-drop is a problem for a description in terms of markedness.

4
I must admit, however, that I find it difficult to describe changes of complex oppositions,
such as the distinction of the six grammatical persons, in terms of markedness. Binary
oppositions, such as the Old French case opposition and the distinction between main clauses
and subordinate clauses, are much more easily described in terms of markedness. I have
followed Benveniste (1966) in my categorization of grammatical persons.
180 LENE SCH0SLER

2.2 Tense and Aspect

Old French epic literature is characterized by a special—and very
chaotic—distribution of verbal tenses, typical of epics in older language stages.
If we try to put some order in the chaos, we find the following patterns: the main
events (foreground) are presented in passé simple, historical present or, less
frequently, in imparfait; secondary events (background) and states are presented
in passé simple, or, less frequently, in imparfait. Passé simple is the unmarked
tense in the sense that it can replace all other forms. The few cases of passé
composé are best interpreted as variants of the historical present, not as forms of
the past (see Table 5a).

foreground: passé simple, pres. hist., passé composé

—————————————————————————————►
background: passé simple, imparfait

Table 5a: Old French: tense and aspect

Modern French has a different distribution of verbal tenses. Passé simple

is marked as a highly literary form that highlights foregrounded events, whereas
imparfait is the form of secondary events (backgrounded) or states. On the other
hand imparfait is also the form of dramatic, foregrounded events; this is the
variant called 'imparfait pittoresque'. In terms of markedness, Modern French
imparfait forms can replace the other tenses, so it is clearly the unmarked form.
Passé composé has undergone a change from being a form of the present to
being a genuine form of the past tense, replacing the passé simple in speech and
often in literature (see Table 5b).

foreground: passé composé, passé simple, pres. hist.,'imparfait pittoresque'

background: imparfait

Table 5b: Modern French: tense and aspect

How did these changes take place? We consider the different text types as
reflexes of different stages in the language: at one extrcmc we have direct
discourse of prose texts, which is considered to be close to natural speech—I
found this hypothesis confirmed above, see Section 1.4; at the opposite extreme
we have narration in poetic texts, which is considered far from natural speech.
We will then expect direct discourse in prose texts to be closer to the tense and
aspect system described in Table 5b, and epic passages in poetic texts to mirror
the system of Table 5a. I have examined the tense and aspect system in a text
 FROM LATIN TO MODERN FRENCH 181

from the thirteenth century, Aucassin et Nicolete (Sch0sler 1973), which is a

special case since it combines passages of prose and poetry and passages of
direct discourse and narration. In this text I do find the expected distributions
according to register: poetry passages conform to the system of Table 5a, prose
passages conform to the system of Table 5b. In the poetry passages narration
conforms more to the old system, direct discourse less so. In the prose passages
narration conforms less to system 5b, direct discourse conforms more to system
5a. Studies of other texts show the same pattern. The facts then confirm that the
evolution takes place as proposed here, from structure 5a to structure 5b
according to a hierarchy in which innovation spreads from the informal register
(direct discourse, prose) to the formal register (narration, poetry), i.e. from
unmarked to marked contexts, since poetry and narration can reasonably be
considered marked with relation to prose and direct discourse (see Table 6). Put
differently, the four registers examined here in one and the same text recapitulate
the diachronic evolution, which progresses according to the markedness
hierarchy described by Andersen.

Unmarked Marked
(a) Old French passé simple imparfait, présent historique,
passé composé
(b) Modem French imparfait passé composé, présent
historique, passé simple
(c) Genre prose poetry
(d) Register direct discourse narration

Table 6. Tense and aspect over periods, genres, and registers.

I conclude that the changes of tense and aspect proceeded according to the
hierarchy proposed by Andersen. The verbal forms are labeled marked or
unmarked according to my evaluation of whether the forms have an exclusive,
more limited, i.e. marked function as opposed to an unmarked function.
Consequently the terms are defined according to my evaluation of their
distribution. This distribution has been additionally checked against manuscript
variations (Sch0sler 1994). The evaluation of the markedness of genres and
registers is harder: can we apply our intuition of what is marked or unmarked to
older stages of the language? Here I accept the hierarchy described by Andersen.
It seems to fit nicely with the facts, but as mentioned in Section 1.4, the
markedness hierarchy of genres and registers is not uncontroversial.
I have not investigated here tense and aspect in relation to sentence type. It
is known, however (see Sch0sler 1973, 1994), that the constraints on tense and
aspect in subordinate clauses of Modern French were not yet present in Old
182 LENE SCH0SLER

French. The grammaticalization of the imparfait form as the form of background

information and, accordingly, as the general form of subordinate clauses, spreads
from the Middle French period. I would find it interesting to discuss the possible
interaction or clash of hierarchically spreading changes and grammaticalization
processes, but such a discussion would exceed the limits of this paper.

3. Conclusion
I have presented a series of three major changes from the areas of
morphosyntax and syntax. Some of the changes do in fact proceed according to
Andersen's theory of actualization (see Section 2.2). Some only partly confirm
the theory (see Sections 1 and 2.1). Some of the cases, however, do not easily fit
into a description based on markedness (see Sections 1.3 and 2.1.2, subsection
4). On the other hand, the spreading of changes according to register differences
such as poetry, direct discourse and narration nicely confirm Andersen's
actualization theory. Thus my investigations have been enriched by the inclusion
of the markedness discussions, and I have been able to evaluate the changes
from a more general and, by implication, more interesting point of view than
would have been feasible without the actualization discussion started by
Andersen and Timberlake. However, during the analysis of the three changes, I
have found the following main problems for the actualization theory proposed by
Andersen.
1. As far as the declension system is concerned, a serious problem for a
general, language specific, hierarchy of markedness values concerns the category
of human common nouns and human proper nouns, which follow opposite
tendencies in the evolution of case. I am as yet unable to propose any intelligent
explanation of this fact.
2. As far as the type of sentence is concerned, the status of the subordinate
clause, as innovative or conservative, is unclear. I have found that subordinate
clauses are conservative as far as the preservation of the two case system
(Section 1.4) and the nonacceptance of V2-word order (Section 2.1) are
concerned. On the other hand, subordinate clauses are innovative as far as the
loss of Pro-drop is concerned, and possibly also in the case of so-called stylistic
inversion (Section 2.1.1).
3. Moreover, I find it difficult to handle grammaticalization phenomena like
the grammaticalization of subject pronouns and specific aspect rules in
subordinate clauses in terms of markedness: how does grammaticalization
interact with the hierarchically dependent changes?
4. Additionally, how do we integrate dialect (presupposing contact)
influence in a theory of actualization (see Section 1.3)? In a community with
dialects of clearly different levels of prestige it is possible to attribute different
 FROM LATIN TO MODERN FRENCH 183

values to different dialectal features. In a community like Old French, where no

dialect had particular prestige (pace the traditional philologists!), it is difficult to
handle dialectal influences in terms of markedness. Each speech community has
speakers of more and less prestige, but we are ignorant as to which were the
centers of prestige in Old French dialects, and, more generally, in the Middle
Ages. As I see it, dialectal influence and markedness belong to two quite
different domains of linguistics, that of linguistic relations (dialect) and the
metadomain of linguistic description. How do we relate them? And will our lack
of knowledge ever make it possible for us to relate them?
Still, I am convinced that in many cases of linguistic change, a better
understanding of the process is gained by describing the spreading of the
changes, i.e. the actualization, in terms of markedness.

SORCES
The electronic corpus of 5 million words established at the Free University,
Amsterdam. For details, see Dees et al. 1980, 1987.
Aucassin et Nicolete chantefable du III siècle, ed Mario Roques. (2e éd.)
Paris: E. Champion, 1929.
Le roman de Tristan en prose, éd. Joël Blanchard & Michel Quéreuil. Paris: H.
Champion, 1976.

REFERENCES
Andersen, Henning. 1990. "The structure of drift". Historical Linguistics 1987.
Papers from the 8th International Conference on Historical Linguistics ed.
by Henning Andersen & Konrad Koerner, 1-20. Amsterdam &
Philadelphia: John Benjamins.
Andersen, Henning. 2001. "Markedness and the theory of linguistic change".
This volume, 21-57.
Bauer, Brigitte. 1987. "L'évolution des structures morphologiques et syntaxiques
du latin au français". Travaux de Linguistique 14/15.95-107.
Benveniste, Emile. 1966. "La nature des pronoms". Problèmes de linguistique
générale, vol. I, 251-257. Paris: Gallimard.
Charpin, François, 1991. "Ordre des mots et identification de l'objet". Stemma
1.25-34.
Dees, Anthonij et al. 1980. Atlas des formes et des constructions des chartes
françaises du 13e siècle. (Beihefte zur Zeitschrift für romanische Philologie,
178.) Tübingen: Max Niemeyer.
Dees, Anthonij et al. 1987. Atlas des formes linguistiques des textes littéraires
de l'ancien français. (Beihefte zur Zeitschrift für romanische Philologie,
212.) Tübingen: Max Niemeyer.
Goyens, Michèle. 1994. Emergence et évolution du syntagme nominal en
français, Berne: Peter Lang.
Liver, Ricarda. 1991. Manuel pratique de Romanche. Sursilvan - Vallader.
(Romanica Rætica, 4.) Cuira: Edizun Lia Rumantscha.
184 LENE SCH0SLER

Lodge, R. Anthony. 1997. Histoire dun dialecte devenu langue. Paris: Fayard.
Original edition: French, from Dialect to Standard. London & New York:
Routledge, 1993.
Marchello-Nizia, Christiane. 1995. L'évolution du français. Ordre des mots,
démonstratifs, accent tonique. Paris: Armand Colin.
Meyer-Lübke, Wilhelm. [1899] 1972. Grammatik der Romanischen Sprachen,
III. Romanische Syntax. Hildesheim: Georg Olms.
Pinkster, Harm. 1991. "Evidence for SVO in Latin?". Latin and the Romance
Languages in the Early Middle Ages ed. by Roger Wright, 69-82. London &
New York: Routledge. (Paperback reprint, University Park: Penn State
University Press, 1996).
Reenen, Pieter Th. van & Lene Sch0sler. 1986. "Le système des cas et sa
disparition en ancien français". Actes du XVIIe Congrès de Linguistique et
Philologie Romanes, 79-114. Marseille: Université de Provence.
Reenen, Pieter Th. van & Lene Sch0sler. 1988. "Formation and evolution of the
feminine and masculine nominative singular nouns in Old French li
maison(s) and li charbons". Historical dialectology: Regional and social ed.
by Jacek Fisiak, 505-545. Berlin & New York: Mouton de Gruyter,
Reenen, Pieter Th. van & Lene Sch0sler. 1997. "La déclinaison en ancien et en
moyen français, deux tendences contraires". Le moyen frayais, Philologie et
linguistique. Approches du texte et du discours (Nancy, septembre 1994),
595-612. Paris: Didier.
Sch0sler, Lene. 1973. Les temps du passé dans Aucassin et Nicolette. L'emploi
du passé simple, du passé composé, de l'imparfait et du présent "historique"
de Vindicatif. Odense: Odense University Press.
Sch0sler, Lene. 1984. La déclinaison bicasuelle de l'Ancien Français, son rôle
dans la syntaxe de la phrase, les causes de sa disparition. (Etudes Romanes
de l'Université d'Odense, 19.) Odense: Odense University Press.
Sch0sler, Lene, 1985. "L'emploi des temps du passé en ancien français. Etude
sur quelques textes manuscrits". Razo 5.107-119.
Sch0sler, Lene. 1986. "L'emploi des temps du passé en ancien français. Etude
sur les variantes manuscrites du Charroi de Nîmes". Actes du IXe Congrès
des Romanistes Scandinaves ed. by Elina Suomela-Harma & Olli
Valikangas, 341-352. (Mémoires de la Société Néophilologique de Helsinki,
44.)
Sch0sler, Lene. 1991. "Les causes externes et internes des changements morpho-
syntaxiques". Acta Linguistica Hafniensia 23.83-112.
Sch0sler, Lene. 1994. "Did Aktionsart ever 'compensate' verbal aspect in Old
and Middle French?" Tense, Aspect and Action. Empirical and Theoretical
Contributions to Language Typology ed. by Carl Bache et al., 165-184.
Berlin & New York: Mouton de Gruyter
Sch0sler, Lene. 1995. "New methods in textual criticism: The case of the
Charroi de Nîmes". Medieval dialectology ed. by Jacek Fisiak, 225-276.
Berlin & New York: Mouton de Gruyter.
Timberlake, Alan. 1977. "Reanalysis and actualization in syntactic change".
Mechanisms of syntactic change ed. by Charles Li, 141-180. Austin, Tex.:
 FROM LATIN TO MODERN FRENCH 185

University of Texas Press.

Timberlake, Alan. 1999. "Change as kairos and khronos". Paper presented in the
Workshop on Patterns of Actualization in Linguistic Change at the
Fourteenth International Conference on Historical Linguistics, Vancouver,
August 1999.
Vance, Barbara S. 1997. Syntactic Change in Medieval French. Verb-Second
and Null Subjects. Dordrecht: Kluwer.
 MARKEDNESS, CAUSATION, AND LINGUISTIC CHANGE:
A SEMIOTIC PERSPECTIVE

MICHAEL SHAPIRO
Brown University

1. Instead of prolegomena: a philosopher's-eye view of language

Philosophers have not been much interested in the material aspect of
language, apart from the fact that there must be one. In fact, one could assert
that philosophers have not begun to approach the real phenomenon of
language.1 But some of what linguists say may sound naive to a philosopher.
Perhaps a bridging of the gap between the two disciplines can be essayed here
by way of approaching the several points I would like to make about
markedness, causation, and linguistic change.
Philosophers have been exercised about the ontological status of
meaning. When a linguist (of the semiotic-structuralist persuasion) makes
signans and signatum (the material and the intelligible part, respectively, in
Jakobson's gloss) the two parts of a sign, this seems to be treating them as
being ontologically on a par. But to a philosopher, the signatum is, though
equally real, of a very different ontological type. The signans would remain the
same material thing even if it had no significance (although then it would
probably not have been produced). But the signatum is only a potentiality
consisting in the interpretability of the signans. That is why a philosopher
would want to say that the signans is the sign—i.e. what is interpret­
able—though it is a sign (and hence a signans) only because it is interpretable
(nonarbitrarily). The signatum, then, is (using Peirce's terminology) the
immediate object of the sign or signans, answering to the immediate
interpretant.
l
This essay's specific formulation of the issues (like Shapiro 1991) owes much to my
correspondence with T. L. Short (beginning in 1981) and to a careful study of his writings on
Peirce, especially the semeiotic (see References). I am also grateful to Henning Andersen for
having invited me to participate as a discussant in his Workshop on Actualization Patterns in
Linguistic Change and to contribute to this volume, as well as for his encouragement during
the several stages leading up to its publication. I am, of course, solely responsible for all and
any shortcomings.
188 MICHAEL SHAPIRO

That is why, rightly or wrongly, linguists' talk of 'content system' and

'expression system' (à la Andersen 1984) makes a philosopher of language
uneasy. While it is true that the material aspect of language tends to develop
into a diagrammatization of its meaning, yet the meaning is not there (as either
a Platonic object or as thoughts or concepts in the minds of language users)
independently of the 'expression system' or system of signantia. There is no
meaning or content except in the interpretability of the material signantia.
There is no discursive thought, no concepts except in our learned capacity to
use the material signantia—whether in production for others to interpret or in
interpreting the linguistic acts of others, or in that internalized speaking to
oneself that we call 'thinking'. That, at least, is the doctrine of such diverse
philosophers as Plato, Wilfrid Sellars, and Peirce. So there is not a content
system independent of the expression system, on this philosophy of language:
there is only a shared set of rules for (1) forming expressions and (2)
interpreting them. And when expressions are interpreted they are not replaced
(e.g., in the mind of the interpreter) by content. For there is no 'content' that
can stand by itself in that way. Instead, on Peirce's view, decoding is
translation, i.e. from one material signans to another (whether in the same or a
different expression system) or from material signans to actions, feelings, or
habits that are not themselves signs in the same sense. Now these actions,
feelings, and habits cannot themselves be what is meant by 'content system'.
The content system must be, instead, either the rules of interpretation by which
the translation is made or the immediate objects represented by the original
signs and grasped or apprehended in this interpretation or translation of these
signs (into new signs or into such ultimate interpretants as habits, actions, or
feelings).
Philosophers of language do not ordinarily mention phonemes or even
morphemes. They tend to begin with complete words, sentences, arguments
(Peirce's semes, phemes, and delomes). Aristotle says in his De Interpretatione
(16al): "First we must define the terms 'noun' and 'verb', then the terms
'denial' and 'affirmation', then 'proposition' and 'sentence'. (In the next work,
the Prior Analytics, he goes on to treat arguments.) Continuing (16a20): "By a
noun I mean a sound significant by convention, which has no reference to time,
and of which no part is significant apart from the rest". And (16b6): "A verb is
that which, in addition to its proper meaning, carries with it a notion of time.
No part of it has any independent meaning, and it is a sign of something said of
something else". Finally (16b26): "A sentence is a significant portion of
speech, some parts of which have an independent meaning ...." Now Jakobson
and Halle (1971) define morphemes "as the ultimate constituents endowed
with proper meaning" and as "the smallest semantic vehicles". But a
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 189

morpheme might be only part of a word, since, e.g. in English, un- and -ly as
prefix and suffix, respectively, contribute the same thing to the meaning of the
various words to which they are affixed, though they never mean anything if
left to stand alone. Philosophers of language follow Aristotle in making the
word their basic or simplest unit of analysis, even though it is not clear that
Aristotle's definitions of nouns and verbs really distinguishes them from
morphemes that are not words. What does Aristotle mean by "no part is
significant apart from the rest?" For un- has a significance, albeit no replica
signifies anything by itself (except in metalinguistic utterances like "All the
uns are the same"). If we return to what Aristotle says about meaning, we get
no help (16a3): "Spoken words are the symbols of mental experience and
written words are the symbols of spoken words. Just as all men have not the
same writing, so all men have not the same speech sounds, but the mental
experiences, which these directly symbolize, are the same for all, as also are
the things of which our experiences are the images". Why does ungainly
symbolize experience and not -ly? Of course, there is no ly-ness found except
as part of ungainliness, loveliness, etc., but neither is any ungainliness found
except as part of ungainly boys, ungainly horses, etc.
The problem of how to draw the line between word meaning and
morpheme meaning (in the case of those morphemes that are less than whole
words) involves the more general problem of determining what meaning is. In
the last passage quoted from Aristotle he appears to have no notion that
language might be a determinant of experience. That is mostly true, being only
slightly mitigated by what he says about concept formation (e.g., at Posterior
Analytics II, 19). But putting that problem aside, he clearly locates meaning 'in
the head', or, more accurately, in individuals' experiences of the world. Why
not in the world directly? Or why not in neither, but in some realm of abstract
entities—if not Plato's timeless forms, then in Poppers "World Three" of
cultural artifacts? In any case, philosophers since Aristotle have been much
exercised about the ontology of meanings. And one may wonder whether, on
that point, even despite their failure to distinguish morpheme meaning from
word meaning, philosophers might not well be in advance of linguists.
For example, when linguists conventionally speak of language as a
"bridge between meaning and sound" and identify meaning with the
"nonlinguistic real or imagined world, the things we talk about", this is still
cruder than Aristotle. Perhaps for much of linguistics it does not matter that we
have no clear idea of what meaning is: as long as we can express the meanings
of sentences, words, and morphemes in other words—as long as such
translations are available—we can express points about which things mean
what, which mean the same, which are different ways of meaning the same
 190 MICHAEL SHAPIRO

thing, what meaning a morpheme contributes to the meaning of a word, etc.

We can make such points without ever saying what meaning is or whether
there are any such things as meanings at all.2 But a general theory of language
would seem to require a clear conception of meaning—or whatever it is that
language, to be more than mere sound, is supposed to 'express'.
Linguists like Hjelmslev (following Saussure) define language as
mediating "between two realms of substance—the things about which we
speak and the physical tokens" and then proceed to distinguish expression and
content. So content would appear to be the things we talk about, or what would
otherwise be called 'meaning'. Now we might speak of Napoleon, of trees, of
this piece of paper, so content would seem to include physical things and
persons. For this reason a philosopher may be confused when a linguist says
that language must comprise "two distinct sign systems, a system of content
signs and a system of diacritic signs". Since content is not 'in' language at all,
except for that part of it that makes up the sounds speakers make in speaking,
by 'content sign' a linguist cannot mean content. Diacritic signs, then, are other
aspects of language. But is there a real difference? The smallest units of the
content system are morphemes, and the diacritic signs are what Jakobson calls
'distinctive features'. But morphemes are made up (of phonemes that are in turn
made up) of distinctive features. Why call these two distinct sign systems,
rather than signs (morphemes) and their material parts (distinctive features)?
The signantia of the diacritic signs are the familiar Jakobsonian
distinctive feature terms, and these signs all have the same signatum, viz.
otherness. But to a philosopher this is tantamount to saying that their sole
function is to be different from one another. And is that, in itself, a semiotic
function? To be such as to be recognized as "this is that and not the other" is
not the same as signifying otherness. A complementary problem affects the
content system, since content signs are said to have, for the most part, no
signantia of their own apart from those of the diacritic system. Which is to say

2
To say as Peirce sometimes said that meaning is translation is to say that there is no such
thing as meaning. As Wilfrid Sellars puts it: to say that German rot means red is simply to say
that from a certain point of view rot and 'red' are to be classed together. To say they have the
same meaning does not mean that there is some third thing (their meaning) that they have: it
means only that they belong to the same class of words. But what defines these classes is the
function or use of their members in the representative languages of those members. Hence
Wittgenstein's "meaning is use". But all of this is too nominalistic unless we understand, with
Peirce, that use or function presupposes the reality of laws, rules, habits. Use and function also
implicate the future. Hence it is possible to identify meaning also with either rules or future
effects. As Peirce says, the reality of a rule is not exhausted by the present: its reality consists
in its influence on actual events, including future events. The meaning of a legisign has to be
explicated in terms of general rules or habits, but the meanings of its replicas in terms of future
effects, actual or potential.
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 191

that the expression system constitutes the sign vehicles of the content system:
the content signs are syntagms of diacritic signs (phonemes). And diacritic
signs themselves are not really signs at all: they are only the vehicles of
morphemes and, hence, they reduce to the distinctive features that constitute
them. Instead of two systems of signs we have a distinction between a system
of material constituents of signs and the signs formed out of these materials.
The attempt to make these into two separate systems of signs leads linguists to
invent signs without significance (diacritic signs) and signs without identity
(content signs).
The reason linguists divide language into these "two basic subsystems" is
that they are two distinct systems, albeit not two systems of signs. As is shown
clearly in the case of nonsense words, a native speaker recognizes phonemes of
his own language even when they signify nothing, whereas of a language he
does not know, he cannot distinguish phonemes from inarticulate noises or, at
best, he cannot identify different occurrences of the same phonemes. So there
is a system at that level and then another governing possible combinations and
the interpretation of those phonemic syntagms that constitute morphemes,
words, etc. Both systems together constitute the semiotic phenomenon of
language, but there is no advantage (much less necessity) to viewing each
separately as a system of signs.
Now suppose we maintain that morphemes are signs distinct from the
distinctive features that constitute them (which sounds like a flat
contradiction). Then how are morphemes to be identified? If one says that
content signs form oppositions, which make up the system of meanings in
language, strictly on the basis of their signata, this cannot be to say that we
distinguish one morpheme or word from another because of what they
severally mean. We distinguish them from one another by the distinctive
features in which they differ. Whether they mean the same or different things
depends on the rules of their interpretation (i.e. on the habits of interpretation
common to the individuals of the community that speaks that language: the
interpretants such habits determine may be emotional, energetic, or logical). In
light of this, a philosopher would have some difficulty understanding a
linguist's assimilation of encoding and decoding to inference, specifically by
taking the content (not content sign) as a premiss and the 'message', i.e. the
expression (i.e. linguistic sign), as the conclusion. This is at best an analogy,
since genuine inference is from a set of sentences or thoughts to another
sentence or thought. But even granted this analog, there is a problem: in what
shape is the content possessed (as 'premiss') before it is encoded? Is it a piece
of the world that is encoded? A piece of experience? Probably neither, since,
prelinguistically, neither answers to verbalization. But the world or experience
192 MICHAEL SHAPIRO

as thought of is already relative to language. 'Encoding' is really speaking what

is on one's mind. But the thought that is thus expressed is already in words
albeit unspoken words. Peirce like Plato said we think in words. But how can
our initial verbalization, whether overt or in thought only, be an inference from
anything? Now Peirce did speak of our first judgments (perceptual
judgments)—first in the sense of being inferred from preceding judgments—as
being formed in a process that in some respects is like inference in that it is a
limiting case of inference, but the type of inference he meant is abductive not
deductive. The perceptual judgment or first verbalization is elicited by
sensations, i.e. physiological stimuli, or by resultant sensory images, but it does
not encode these: it is abductive precisely because it posits its own object, the
supposed cause of the stimuli. This is crucial. As a first verbalization it does
not put into words something not already in words. That cannot be done.
Instead, first words introduce, create, posit their own objects. These may in
some manner correspond to or fit antecedent realities, but as posited they are
what answers to words. Hence, instead of deductive encoding we have
abductive positing. And that, one could say, puts meaning where it belongs,
viz. in language or as existing only as relative to language, and not in
antecedent or prelinguistic reality or experience. Meaning, then, is inseparable
from the interpretability of words. It involves potential interpretation.
Reference is determined by the causal factors eliciting thought or utterances,
and truth is a fit of meaning to reference, where 'fit' is to be explained in terms
of the telos of speech.
Decoding, on the other hand, is interpretation of utterances by thought,
which is not replacing expression by nonlinguistic content but is a translation
of words into other words or, ultimately, it is a change in the interpreter's
('decoder's') emotions, actions, or dispositions to act. This is not to deny that
the hearer must form a certain hypothesis about what the speaker meant (a
genuinely abductive aspect of 'decoding'), nor the converse possibility that the
speaker did not express himself fully or accurately.
This analysis would lead to the following postulates. There is no meaning
apart from habits of interpretation. There are no linguistic signs that do not
have a meaning. The supposed analogy between the expression system and the
content system is really an analogy between the material aspect of language
and its meaning, i.e. how its uses are to be interpreted. One could speak of the
immediate object instead of the interpretant; but the dynamic object is the
referent.
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 193

2. Nominalism and realism in linguistics

Philosophers have always thought of nominalism as a doctrine, not as a
practice. They may therefore be excused for having trouble seeing the relation
of nominalistic linguistics to the doctrine of nominalism, which is that the
former is a way of doing linguistics to which doctrinal nominalists could not
object, but that would seem deficient to those who are doctrinal realists. For if
there are no classes in reality, but they exist in name only, as doctrinal
nominalists claim, then any way of dividing up phenomena, including
linguistic phenomena, is as good—or at least as true—as any other. And by
'nominalistic linguistics' I mean the practice of imposing an arbitrary taxonomy
on linguistic phenomena.
This use of terms and concepts from the history of philosophy to make
headway in linguistic theorizing may be interesting but also possibly
confusing, the latter for the following reason. The linguistic phenomena
classified might include linguistic universals (the Peircean 'types') as well as
linguistic individual events (the Peircean 'tokens'). And one who is familiar
with the nominalist/realist distinction as a matter of doctrine only might
naturally suppose that by 'nominalist linguist' is meant one who denies the
reality of linguistic universals. That, of course, would be an application of the
nominalist doctrine to linguistic phenomena; but that, one can see now, is
distinct from nominalist linguistics as a practice or method. Nominalism as a
practice would not necessarily deny that universals are real; rather, it consists
in deciding their classification arbitrarily—both their classification into
subtypes, if they are segregated from individuals, and whether to so segregate
them. Even their classification as real or unreal would be quite arbitrary.
The Chomskyan search for deep structure and generative principles looks
relatively realist from a doctrinal point of view.3 For whether or not surface
phenomena are conceptualized in terms of types as well as tokens, the deep
structure and principles look like universals, and especially so the way
Chomsky and his followers speak of them. Chomsky and his school are
nominalist linguists, not realist linguists, because their taxonomy of surface
phenomena—the phenomena they wish to explain as following from deeper
principles—is arbitrary. (It would follow that the hypothetical structure must
be arbitrary too, for it is justified only by its capacity to explain those
phenomena.)

3
In using the label 'Chomskyan' I intend to let it refer not only to Chomsky himself (see now
Chomsky 2000 for the latest tergiversations) but to all the latter-day offshoots of
transformational-generative grammar as well—even those like Natural Phonology and
Morphology or Optimality Theory (see Kager 1999 for a handy summary) that claim to be
founded on principles that diverge from Chomskyan linguistics.
 194 MICHAEL SHAPIRO

'Realism', of course, is used to designate the opposite of phenomenalism

as well as the opposite of nominalism. With respect to doctrine exclusively, not
method, Jakobson and his structuralist continuators (like Andersen and me)
look like phenomenalists in contrast to Chomsky and his followers, since the
former seem much more concerned with the description of what is here being
called surface phenomena, whereas the latter plunge quickly to the (putative)
underlying realities that explain them. One could say that Chomsky is in error
for proceeding too quickly: after all, how can he abduce explanatory realities
when he is wrong about the explanandum? But this is not so simple an issue as
that. For if the classification of phenomena is to be real, not nominal, then it is
often impossible to know what that classification is until the underlying
realities have been identified. As an example from a domain other than
language, consider whether it was possible to know that rusting, fire, and
metabolism should be classed together as members of the same natural kind
before they were all explained as different forms of oxidation. The circle here
is like the hermeneutic circle: the explanans and the explanandum are found
together, not first one and then the other.
But there is another way of looking at this which can be identified,
mutatis mutandis, with that of semiotic neostructuralism in linguistics. 4
Realism in contradistinction to nominalism (doctrinally) is connected with
teleology—or so, at least, Peirce appears to have thought. A natural class is one
the members of which exist because each satisfies the same idea. That idea has
a certain potency, and hence the class exists independently of anyone's having
named it. This idea is consistent with the argument of the preceding paragraph
according to which some natural classes may be those classes entailed by a true
explanatory theory. But it is not limited to cases where the explanatory
structures lie beneath the surface phenomena. Suppose language qua
phenomenon has a history, and suppose that history can be understood by
postulating goals not involving any underlying mechanisms. For example,
linguistic change might be seen as tending toward a more adequate
diagrammatization. Then we have a teleological basis for identifying natural
linguistic classes, namely those that we have to attend to in order to understand
language as diagrammatization. (This too involves a hermeneutic circle:
neither the right description of the process nor the goal that explains it can be
discovered without also discovering the other.)
If the preceding is a roughly correct account of the linguistic practice of
semiotic neostructuralism, then it would seem that one who espouses the latter
4
By 'semiotic neostructuralism' as applied to the study of language I mean the doctrine and
method that emanate from an amalgamation of Jakobsonian linguistics with Peircean
semiotics.
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 195

is in method, if not in doctrine, a realist as opposed to a nominalist, but a

phenomenalist as opposed to a realist, and a teleologist. 5 One may doubt
whether a semiotic neostructuralist is a phenomenalist in doctrine. For such a
linguist does not deny, in fact, he presupposes that there are realities beyond or
beneath language but for which his teleological account of linguistic change
would make no sense. That is, there must be flesh-and-blood bodies that speak
and listen, and it is their desires and needs that explain why ever more
adequate diagrammatization is an inevitable if unintended goal. If the research
program subtended by semiotic neostructuralism can be made to work, then it
will indeed conflict with Chomskyan linguistics—and prove superior to it.
Here is why.
Chomsky has a rather mechanistic view of language, for all that he
understands that the freedom to compose sentences that are original,
unpredictable, and yet intelligible is different from the unoriginal, predictable
products of strictly mechanical action. His view is mechanistic nonetheless
because he simply posits underlying structures by which sentences are to be
generated. Possibly in a wider perspective, Chomsky is no more reductively
mechanistic than a semiotic neostructuralist, in a wider perspective, is a
phenomenalist. For he no doubt admits (or would admit) that the linguistic
universals in our brains are not just there, period, but evolved, with the brain's
evolution, as chance variants that were 'selected' by the principle of
reproductive success. Similarly, the intentions or needs or felt urgencies to
speak or to achieve certain outcomes might explain—but only in a context
wider than Chomskyan linguistics—why language's generative mechanisms
are used in this way rather than in that. But if we focus simply on the linguist's
study, as diversely conceived by Chomsky and the semiotic neostructuralist,
then there is this difference: for the one, the teleology of language is excluded
from linguistic explanation, while for the other it is the very stuff of
explanation. For the one, linguistic phenomena conform to a describable
structure of highly abstract laws, while for the other linguistic phenomena
exhibit an intelligible if less abstract, more complicated structure. For the one,
the system is a given, and any changes in it are accidental, while for the other
development is essential to language—development is more the reality than is
any one system of rules—and that development is also intelligible and not
merely given.

5
T. L. Short (p.c.) points out that there is a methodological use of 'phenomenalist'; for
instance, classical thermodynamics is often called 'phenomenal thermodynamics', not because
its proponents are phenomenalists in philosophical doctrine but because it formulates the laws
of thermodynamics without reference to the atomic theory of matter, which, with Boltzmann et
al., was found to explain and quantify those laws.
 196 MICHAEL SHAPIRO

That is the conflict. The reason the semiotic neostructuralist approach is,
if it is successful, superior is that it can be used to explain the very evolution of
the brain-mechanism or linguistic capacities and universals that Chomsky can
at best describe. That is, given creatures somewhat sociable, exchanging signs
as their way of life, then the survival value of their communicating more
elaborate and precise diagrams would explain the retention of those fortuitous
variations, say, in brain structure that promote exactly such powers of
expressible diagrammatization. That is, the principle of this evolution will be
itself linguistic, and continuous with the principles of postbiotic, strictly
linguistic evolution. The thought here is not unlike that which refuses to
postulate linguistic intentions separate from the capacity to exercise those
intentions. Just as there could be no desire to speak without an ability to speak,
so also there could be no evolution of linguistic capacities—even, or
especially, at the physiological level—except among those who, already
speaking to one another, will more likely survive as a species if they speak
more effectively. Thus, instead of a neurophysiological explanation of
language, we have a linguistic explanation of the higher cortex (and probably
not just the speech centers either, since so many of our capacities for sensation
and action would be bootless without our capacities for speech).

3. Semiosis and linguistic change: efficient and final causation

Peirce's distinction between legisigns and replicas can be used to good
account in lifting some of the confusion that surrounds linguistic change,
which is the end-directed evolution of a system of legisigns.6 Replication is the
end-directed use of already developed legisigns. In this process the legisigns
(or rules of replica formation) do not function as efficient causes precisely:
indeed, it is doubtful whether a rule or general type could ever be an efficient
cause. But neither are they teloses (Gk. telē) of replication. The purpose of
replication is communication (conveying information, issuing commands,
expressing emotions, etc.). Thus, legisigns are not replicated simply for the
sake of being replicated. They could be efficient causes of acts already
explained by final causes—except for one thing. They could be efficient causes
because final causes require the cooperation of efficient causes. Suppose I want
Jones to close the door. I look around for means to do so. One means is
replicating the English sentence, "Jones, close the door!" If that were the only
means then, given my purpose, one can suppose that the availability of that
legisign causes me (like a mechanical push) to replicate it. (But this is
6
Conceptual change is the end-directed evolution of the rules of interpretation of symbols,
sometimes with concomitant changes in the symbols themselves. Conceptual change then
determines linguistic change, but in general this is not necessary to linguistic change.
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 197

wrong—why in a moment.) But the availability of alternative legisigns (e.g.,

"For God's sake, Jones, close the door!" or "Jones, dear fellow, I feel a draft.")
means I must choose, and so those legisigns are not efficient causes. Legisigns
cannot be efficient causes at all. In the first place, the efficient causes that must
cooperate are those motor reflexes and the like that make my tongue wag, my
mouth open and close, or my hand type these words. Secondly, legisigns are
general types and hence can never be efficient causes. The upshot of this is that
legisigns both exist for a purpose (they have evolved to make communication
possible or to facilitate communication that was already possible) and are used
when we act for the purpose of communicating. Thus, already existing
legisigns are subsidiary final causes: we make such-and-so sounds or marks in
order to replicate certain legisigns, and we replicate those legisigns in order to
communicate something.7 There is, therefore, an important difference between
(1) legisigns developing and (2) legisigns being used.
Talk about final causation is often accompanied by contrasting references
to efficient causation. An efficient cause is a particular event or condition that
compels its effect. The effect follows the cause in accordance with a general
law (a law of efficient causation). A final cause is not a particular event or
condition and does not compel its effect. Suppose a man is seen bounding
down a steep incline. Why? Possibly because the man was pushed. That would
be an efficient cause. But perhaps the man acted in order to catch a goat. 'To
catch a goat' is the final cause; it is not a particular event and did not compel
the behavior.
Final causation is consistent with efficient causation, indeed requires it.
Men cannot bound 'goatwards' if their muscles do not relax and contract,
compelling movement of limbs. Presumably, then, the two types of cause
explain different phenomena—or complementary aspects of the same
phenomenon.
To explain something by a final cause is teleological explanation.
Teleology is the doctrine that teleological explanations are sometimes
legitimate, that some phenomena can only be explained teleologically, and that
7
Notice that when we say things just for the sake of saying them, then legisigns may be truly
final causes. But we need to distinguish three cases. The availability of certain meanings (=
rules of interpretation of symbols) might intrigue me: so I want simply to express those ideas.
Or it might be the legisigns themselves that intrigue me: poets (like the Russian futurist
Mayakovsky) and composers (like Mozart) are said to have been fond of repeating certain
(nonsense or foreign-language) words simply for the sake of their sound rather than their sense.
Or it might be the truth we wish to state for its own sake, and in that case the final cause is the
agreement of certain legisigns with an independent reality. In any case, replication of legisigns
can be an end in itself, and in that case the legisigns are essential to one's ultimate purpose in
speaking. That is to say, we would have a different purpose or none at all if we did not have
those legisigns.
 198 MICHAEL SHAPIRO

final causes are real. Teleological explanation was introduced deliberately by

the Greek philosophers, primarily Plato and Aristotle, in explicit contrast to
already well-established conceptions of causation—those that Aristotle
identified as 'efficient' and that we can identify as 'mechanistic'. And already
with Plato, it was recognized that this new form of explanation would be
rejected by those who think (a) that everything can be explained by causes that
compel or (b) that nothing that does not compel its effect could explain it.
In particular, what teleology was invented to explain is the existence of
order—in human affairs, in individual actions, in plant and animal life, in the
cosmos—wherever that order is inexplicable mechanistically. The point of
teleology is to explain the emergence of order out of chaos. By contrast, the
mechanistic world view of modern science admits none but efficient causes.
However, not all forms of explanation in modern science conform to the
mechanistic idea, even in its broadest and most up-to-date sense, but do
approximate to the Aristotelian idea of explanation by final causes.
Teleological theories are thus the best, or only, explanations of certain
important classes of phenomena. Hence, we have good reason to suppose that
final causes are real.
If this sounds too apodeictic for some readers' taste, it is probably due to
the fact that teleology is badly understood.8 An aid in dispelling some of the
mist surrounding teleology is Peirce's idea of certain processes as 'finious', a
neologism he coined for fear that "teleological is too strong a word to apply to
them" (7.471).9 These are nonmechanistic processes that "act in one
determinate direction and tend asymptotically toward bringing about an
ultimate state of things" (ibid.). The importance of nonteleological finious
processes is that they explain how teleological phenomena are possible. One
might say that they remove the mystery from teleology. Operating with the
notion of finiousness imposes an obligation on the analyst: a hierarchical
ordering of nonmechanistic explanations, some of which are merely finious,
and some of which are teleological.
If one is to arrive at such an ordering following Peirce's conception, then
it will be necessary to take into account his definition of final causation:

8
Perhaps especially by linguists—like Lass (1997) and Laboν (1994); see Short 1999 for a
demolition of the former's antiteleological stance. As for the latter, his "Plan of the Work as a
Whole", set out on the book's very first page, already betrays a fundamental misunderstanding
of causation: it presents the organization of a projected three volumes into (respectively)
"Internal factors", "Social factors", and "Cognitive factors"—as if these 'factors' were
categorically distinct from each other (they are, of course, all 'internal').
9
Citations in this form (volume and paragraph separated by a dot) are to Peirce's Collected
Papers.
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 199

... we must understand by final causation that mode of bringing facts about
according to which a general description of result is made to come about, quite
irrespective of any compulsion for it to come about in this or that particular way;
although the means may be adapted to the end. The general result may be brought
about at one time in one way, and at another time in another way. Final causation
does not determine in what particular way it is to be brought about, but only that
the result shall have a certain general character. (1.211; cf. 1.204)

Any finious process is the result of fortuitous variation plus a principle of

selection. These processes are everywhere observable in populations of
individuals, whether molecules or living things. Other processes, equally
finious, might be found within the actions of a single individual (not
necessarily human).10 It is the nature of finious processes that their particular
outcomes cannot be predicted; all that we can predict is their general tendency.

4. Markedness in a theory of change

Peirce understood a final cause as being a possibility—sometimes he said
"idea", but that is not to be understood in a subjective sense as existing in some
person's thought—that has a tendency to become actual, one way or another:
"... every general idea has more or less power of working itself out into fact;
some more so, some less so" (2.149).
It is in this sense that markedness must be viewed as a final cause in
linguistic change.11 When the question of causation is posed in terms of
efficient and final causes—and ideological processes distinguished from
finious—then the claim that, rather than markedness principles, it is
"perceptual factors and processing strategies [that] may influence the
development of linguistic structures" (Smith 2001:207) will be seen for what it
is, a category mistake.
This mistake results from the apriorism that underlies how contemporary
linguists commonly understand markedness (e.g., in Optimality Theory, but
not only). On this view (partly reflected in Smith's contribution), markedness is
simultaneously conflated with and pitted against notions like "sentence
processing" or "perceptual strategies" as if markedness were an efficient cause,
i.e. categorically of a piece with the latter. Lending support to skepticism
regarding the relevance of markedness (and emanating directly from what I

10
With respect to the deliberate conduct of human beings, the principle of selection is a type of
outcome they have in mind, and which they consciously apply in choosing among the
alternatives available to them. In other words, what we have in this case is purposefulness.
Since an analysis of purpose would take us even farther afield, I refer the reader to the
admirably clear exposé in Short 1999.
11
In the event I understand Andersen's conception of markedness (2001) to be compatible with
this view. For a discussion of final and efficient causes in linguistic change that takes part-
whole relations into account, see Shapiro 1991:16ff.
200 MICHAEL SHAPIRO

would now call the Apriorism Fallacy) is the perceived difficulty of assigning
universal or immutable markedness values, even though markedness is
invariably context-sensitive and dependent on the existence of choice between
variants.
The question Why? as applied to linguistic change does not have a
homogeneous answer. The problem of assigning markedness values is not
solely the burden of linguists: it falls on language users as well. Linguistic data
always contain the germ of ambiguity, of differing interpretations, and it is
only by trial and error that the finious process of reaching a definitive
markedness assignment proceeds. This process is necessarily always historical
and not given apriori because at any given time linguistic habits, like all other
habits, have a structure, and this structure is always in statu nascendi. But the
important thing is that an assignment will be reached.
Language users do not need to wait for linguists to decide what is marked
and what unmarked in order to be influenced by markedness considerations in
making innovations and (tacitly) agreeing that some innovations qualify for the
(social) status of full-fledged changes: they do it willy-nilly because they are
impelled to by the power of the idea. Or as Peirce put it: "... it is the idea that
will create its defenders and render them powerful" (1.217).

REFERENCES
Andersen, Henning. 1984. "Language structure and semiotic processes".
Arbejdspapirer, udsendt af Institut for Lingvistik, K0benhavns
Universitet 3.33-54.
Chomsky, Noam. 2000. New Horizons in the Study of Language and Mind.
Cambridge: Cambridge University Press.
Jakobson, Roman & Morris Halle. 1971. Fundamentals of Language. Second
edition. The Hague: Mouton.
Kager, René. 1999. Optimality Theory. Cambridge: Cambridge University
Press.
Labov, William. 1994. Principles of Linguistic Change. Volume 1: Internal
Factors. Oxford: Blackwell.
Lass, Roger. 1997. Historical Linguistics and Language Change. Cambridge:
Cambridge University Press.
Peirce, Charles Sanders. 1965-1966. Collected Papers, Volumes 1-8 ed. by C.
Hartshorne et al. 2nd printing. Cambridge, Mass.: Harvard University
Press.
Shapiro, Michael. 1991. The Sense of Change: Language as History.
Bloomington: Indiana University Press.
Short, T. L. 1981a. "Peirce's concept of final causation". Transactions of the
Charles S. Peirce Society 17.369-382.
Short, T. L. 1981b. "Semeiosis and intentionality". Transactions of the Charles
5. Peirce Society 17.197-223.
 LINGUISTIC CHANGE: A SEMIOTIC PERSPECTIVE 201

Short, T. L. 1981b. "Semeiosis and intentionality". Transactions of the Charles

S. Peirce Society 17.197-223.
Short, T. L. 1982. "Life among the legisigns". Transactions of the Charles S.
Peirce Society 18.285-310.
Short, T. L. 1983. "Teleology in Nature". American Philosophical Quarterly
20.311-320.
Short, T. L. 1989, "Why we prefer Peirce to Saussure". Semiotics 1988 ed. T.
Prewitt et al., 124-130. Lanham, Md.: University Press of America.
Short, T. L. 1996. "Interpreting Peirce's interpretant: A response to Lalor,
Liszka, and Meyers". Transactions of the Charles S. Peirce Society
32.488-541.
Short, T. L. 1998a. "Jakobson's problematic appropriation of Peirce". The
Peirce Seminar Papers ed. Michael Shapiro, 3.89-123. New York &
Oxford: Berghahn Books.
Short, T. L. 1998b. "What's the use?" Semiotica 122.1-68.
Short, T. L. 1999. "Teleology and linguistic change". The Peirce Seminar
Papers ed. Michael Shapiro, 4.111-158. New York & Oxford: Berghahn
Books.
Short, T. L. MS. "Peirce's theory of signs". To appear in The Cambridge
Companion to Peirce ed. C. Misak. Cambridge: Cambridge University
Press.
MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION IN THE
ACTUALIZATION OF A MORPHOSYNTACTIC CHANGE

John Charles Smith

University of Oxford

0. Introduction
Both markedness and functionality have been canvassed and questioned as
significant influences on morphosyntactic change.1 In this paper, I survey some
data concerning the differential disappearance of agreement between a past
participle and a direct object in the Romance compound past tenses formed with
the auxiliary "have" and attempt to assess the influence of each of these factors
on the progression of the change.

1. Markedness and morphosyntactic change

Andersen (2001:30-37) claims that markedness relations can be observed
in every variety of linguistic change, from its inception to its completion, both in
the relations among variants and in the relations that define the categories that
typically condition the gradual process by which newer forms replace older
correspondents. Specifically, morphosyntactic change is usually alleged to take
place in unmarked environments before marked ones. For instance, changes
which affect main clauses before subordinate clauses, such as the change of
word order from SOV to SVO in many Germanic languages, have long been
seen as examples of the unmarked context being in the vanguard of change (see
Hock 1991:332, and, for some more recent discussion, Matsuda 1998; Bybee et
al. (1994:296) further note that "new grammaticalizations of tense and aspect
tend to arise in main, asserted clauses"). Some insight into this phenomenon may
1
An earlier version of some of the discussion of markedness in this paper appeared as Smith
(1999). I am grateful to Henning Andersen for his invitation to take part in the Workshop on
"Patterns of Actualization in Language Change", held in Vancouver in August 1999 as part of
the Fourteenth International Conference on Historical Linguistics, which gave me the
opportunity to amplify and refine my original ideas; to Martin Maiden and Suzanne Romaine
for their valuable comments on a subsequent presentation of some aspects of this analysis to
the Oxford Romance Linguistics Seminar; and to Nigel Vincent for helpful discussion and
much good advice. Any failings are mine alone.
204 JOHN CHARLES SMITH

be provided by Lightfoot's concept of 'Degree-0 Learnability', whereby "primary

linguistic data are restricted to data occurring in an unembedded binding
Domain" (Lightfoot 1991:38)—in other words, a child acquiring a language has
access only to the binding Domain of main clauses.
However, the notion that the unmarked environments are the initial locus
of actualization was given particular prominence by Timberlake (1977), in his
discussion of two diachronic processes. The first is the disappearance of a
subject-to-object raising rule in Finnish complex sentences involving a participial
complement clause, which leads to the underlying subject of the participial clause
always appearing in the genitive as opposed to any of the cases normally
associated with objects. The second is the replacement of the Russian adverbal
object genitive found with certain verbs by the accusative. The former change
(156) is "actualized earlier for constituents which are relatively more subjectlike"
(pronouns before nouns; agentive or animate nouns before non-agentive or
inanimate nouns). The latter change (160) takes place according to a "hierarchy
of individuation, the extent to which an object is considered as an individual"
(proper nouns before common nouns; animate nouns before inanimate nouns;
concrete nouns before abstract nouns; singular before plural). Timberlake
(168-169) concluded that the changes are "accomplished earlier in contexts that
are unmarked for the innovation and later in contexts that are marked for the
innovation". However, "the concept of markedness, or naturalness, must be
understood with reference to the particular change involved".
More recently, Andersen (1990) has provided further evidence for the
view that morphosyntactic change is actualized first in unmarked contexts, in his
examination of a development in Polish whereby forms of the auxiliary verb
"be" become verbal endings. Once again, this development takes place earlier or
more readily in contexts which can broadly be defined as 'unmarked'—not only
morphological contexts (present tense before preterite, indicative mood before
conditional mood, singular number before plural number, plural number before
dual number, third person before other persons, first person before second
person), and grounding contexts (main clauses before subordinate clauses,
asyndetic clauses before syndetic clauses), but also genres (prose before poetry,
expository prose before artistic prose, secular content before religious content),
media (spoken language before written language), and styles (casual before
formal) (see Andersen 1990:10).
It is not the aim of this paper to provide a lengthy catalogue of
morphosyntactic changes which take place earlier in an unmarked environment;
but, as a final example of such a change, we may note that analogical feminine
forms of originally epicene adjectives in French (such as grand "big") appear to
be generalized first of all in postnominal position (the unmarked position for an
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 205

adjective in this language) and only later in prenominal position (see Pope 1934:
§780).
The prevailing view, then, is that language change is sensitive to
markedness in the way described above. However, this hypothesis has not
always been regarded as unproblematic. Ferguson (1996a:243, 1996b: 175), for
instance, discussing changes in the system of personal verb inflection in
Germanic, notes that in Old English the three persons of the plural fall together,
whilst in Old Swedish, the three persons of the singular fall together. He
comments:
It is a sober reminder of the inadequacy of current notions of markedness or
naturalness that of the two languages, beginning from roughly the same structure
and both 'simplifying', one collapsed the three persons of the plural and the other
the three persons of the singular.

And Hendriks (2000:165) observes that the merger of the attributive and
predicative forms of Old Japanese occurs in dependent clauses before main
clauses; in this analysis, "Japanese is providing us with a counter-example to the
frequently made comment that subordinate clauses preserve older forms and lag
behind matrix clauses in terms of syntactic change".
In addition, there is an abiding problem with any analysis based on
markedness: the definition of the concept is not unproblematic, and in many
cases (although not all) the choice of one or other member of an opposition as
the marked term can appear arbitrary. If we are to make use of the notion, we
should certainly try to define it in an independently motivated way; Timberlake's
suggestion that "the concept of markedness, or naturalness, must be understood
with reference to the particular change involved" (1977:169), although echoed by
subsequent work in natural morphology (compare the notion of 'system-
dependent naturalness' developed by Wurzel (1984) and Dressier (1985)), can
all too easily lead to circularity. Qualitative unmarkedness is defined by a number
of well-known criteria, summarized by Battistella (1990:26) as "optimality,
breadth of distribution, syncretization, indeterminateness, simplicity, and
prototypicality". Optimality refers to the fact that "When certain segments or
certain feature values imply others in language after language, those values are
taken to be unmarked" (26). As far as distribution is concerned, "Unmarked
terms are distinguished from their marked counterparts by having a greater
freedom of occurrence and a greater ability to combine with other linguistic
elements" (26)—the characteristic referred to by Croft (1990:77) as 'versatility'.
The unmarked term is also the one that occurs in positions of absolute
neutralization. Syncretization means that "Unmarked categories tend to be more
differentiated than marked ones" (27). By the criterion of simplicity "unmarked
206 JOHN CHARLES SMITH

elements are less elaborate in form than their [marked] counterparts", and by that
of prototypicality, they are "experientially more basic" (27). Similarly, Croft
(1990:72-84) defines qualitative markedness on the basis of 'structural' criteria
(Battistella's 'simplicity') and 'behavioral' criteria, the latter subdivided into
'inflectional' (Battistella's 'syncretization'), 'distributional' (Battistella's 'breadth
of distribution') and 'cross-linguistic' (Battistella's 'optimality'). In addition,
higher frequency is generally assumed to be a quantitative indicator of
unmarkedness (see especially the discussion in Greenberg 1966:64). Bybee
(1985:117-118) further suggests that items which occur more frequently in texts
or discourse have greater 'lexical strength'—that is, they are more firmly
entrenched in the mental representation of the lexicon.
If all the criteria agree, there will be no problem; if they conflict, it will
sometimes be difficult to tell which member of an opposition is marked.

2. Functionality and morphosyntactic change

Functional accounts of language change take as their basic premiss that the
need to preserve information is an influence on how language develops. This
notion is particularly associated with phonology. It has been invoked to explain,
inter alia, chain shifts and the symmetry of vowel systems (Martinet 1955), the
avoidance of 'homophonic clash' (Gilliéron & Roques 1910; Gilliéron 1918;
and, more recently, Samuels 1987), and a variety of grammatical phenomena
(see, for instance, Givón 1979). There is generally less disagreement about the
meaning of functionality than about the meaning of markedness. Functionalists
see language as a system of communication specifically adapted to that purpose;
changes in the system will of necessity be sensitive to that purpose. This is not to
say that the definition of 'functionalism' has been uncontroversial; see the
discussion in Croft (1995), Newmeyer (1998), and, in a specifically diachronic
context, Croft (2000:87-144).
For formalists, the rejection of functionalism has often been an article of
faith (but see the essay in reconciliation undertaken by Newmeyer 1998).
However, many nonformalists, such as Labov (1994:547-599) and Lass
(1997:352-365), are also critical of functional explanations. Labov examines the
role of functional considerations in phonological variation. After discussing
work by Poplack (1980; 1981) on the deletion of final [-s] in Puerto Rican
Spanish and by Guy (in unpublished work) and Scherre & Naro (1991) on
comparable phenomena in Brazilian Portuguese, he concludes (568):
Given phonological and morphological variation, the functional hypothesis predicts
a tendency for speakers to choose one variant or the other in a fashion that will
preserve information. Most of the results cited show the opposite: in the stream of
speech, one variant or the other is chosen without regard to the maximization of
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 207

information. On the contrary, the major effects that determine such choices are
mechanical: phonetic conditioning and simple repetition of the preceding structure.

It is important to realize that Labov's strictures essentially concern the

(non)effect of morphological considerations on stable phonological variation. He
observes that, in diachrony, other changes may emerge which appear to
compensate for the loss of the phonological distinction and allow meaning to be
maintained. For instance, when a sound change leads to the loss of a flexion
indicating person or number, these categories may come to be encoded
syntactically, by the use of a pronoun or of a determiner or quantifier which can
vary for number. However, the transition from one stage to the next is a complex
process which involves 'probability matching'.
Lass (1997:352-365) takes a rather more abstract stance, arguing that
functionalism has no predictive power, and is therefore essentially an "a
posteriori fudge" (358). Amongst other things, he points out that some
'homophonic clashes' go through, whilst others, which are equally plausible, fail
to occur, and that it appears impossible to establish a principled etiology of such
cases.
It is beyond the scope of this paper to present a detailed survey of every
functionalist argument that has been advanced in order to account for linguistic
change. I shall be concerned in what follows with a special case of
functionalism, the claim that perceptual factors and processing strategies may
influence the development of linguistic structures. This idea has found favor with
many scholars, including Vincent (1976), in his discussion of the evolution of
Latin word order, and Fodor (1981), in her more general examination of the
relationship between overt surface elements ('fillers') and phonologically null
elements which form part of the same chain ('gaps').

3. Markedness, functionality, and object-participle agreement

3.1 Agreement hierarchies
In Late Latin, the construction HABERE ("to have") + past participle is
reanalysed as a present perfect tense/aspect form (for some discussion of this
process, see Vincent 1982). A consequent actualization is the disappearance of
participial agreement. As pointed out in Smith (1989), the rule required to
account for this agreement after the reanalysis of HABERE + past participle has
taken place is highly opaque and very much a 'black sheep' amongst other
agreement rules in Romance, and these circumstances will favor its elimination.
However, the disappearance of agreement is differential. In Smith (1991; 1993b;
1995b), I noted that the data concerning object-participle agreement in Romance
enable us to establish a number of implicational hierarchies (in the sense of
208 JOHN CHARLES SMITH

Greenberg 1963) of the form: "if, in a given language or dialect, the past
participle agrees with a direct object of type X, then it will also agree with a
direct object of type Y". These hierarchies are presented below (the notation X >
Y is to be interpreted to mean that agreement with X implies agreement with Y,
but not necessarily vice versa).

1. Position of Direct Object:

Following > Preceding
2. Identity of Preceding Direct Object:
Topics
Interrogatives > Relatives > Clitic Pronouns
Exclamatives
3. Person of Clitic Pronoun:
First Person
Second Person > Third Person Nonreflexive
Third Person Reflexive
4. Number and Gender of Third-Person Nonreflexive Clitic Pronoun:
Masculine Plural > Feminine Plural > Feminine Singular

The resulting patterns of agreement are illustrated in (1)-(7) below from Catalan,
a Romance language which has dialects exemplifying all the above hierarchies
(see Smith 1995a). Examples exhibiting agreement between past participle and
direct object are in bold type. Parentheses enclose the number and gender
specification of controllers that do not vary for gender, but have inherent gender.
Square brackets enclose the number and gender specification of controllers that
neither vary for gender nor have inherent gender, but in the given example are
deemed to have the gender indicated.
These synchronic patterns are matched by diachronic data concerning the
differential disappearance of the agreement and may be considered a synchronic
reflection of this diachronic process (see Smith 1995b).

(1) General Agreement

He vistes les pellicules, "I-have seen-F.PL the films (F.PL)."
Quines pellicules he vistes? "What films (F.PL) have-I seen-F.PL?"
Les pellicules que he vistes "The films (F.PL) which I-have seen-F.PL"
Us he vistes. "YOU-[F.PL] I-have seen-F.PL."
Els he vists. "Them-M.PL I-have seen-M.PL."
Les he vistes. "Them-F.PL I-have seen-F.PL."
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 209

(2) Agreement with preceding direct objects

He vist les pellicules. "I-have seen-M.SG the films (F.PL)."
Quines pellicules he vistes? "What films (F.PL) have-I seen-F.PL?"
Les pellicules que he vistes "The films (F.PL) which I-have seen-F.PL"
Us he vistes. "YOU-[F.PL] I-have seen-F.PL."
Els he vists. "Them-M.PL I-have seen-M.PL."
Les he vistes. "Them-F.PL I-have seen-F.PL."

(3) Agreement with relative and clitic-pronoun direct objects

He vist les pellicules. "I-have seen-M.SG the films (F.PL)."
Quines pellicules he vist? "What films (F.PL) have-I seen-M.SG?"
Les pellicules que he vistes "The films (F.PL) which I-have seen-F.PL"
Us he vistes. "YOU-[F.PL] I-have seen-F.PL."
Els he vists. "Them-M.PL I-have seen-M.PL."
Les he vistes. "Them-F.PL I-have seen-F.PL."

(4) Agreement with clitic-pronoun direct objects

He vist les pellicules. "I-have seen-M.SG the films (F.PL)."
Quines pellicules he vist? "What films (F.PL) have-I seen-M.SG.?"
Les pellicules que he vist "The films (F.PL) which I-have seen-M.SG"
Us he vistes. "YOU-[F.PL] I-have seen-F.PL."
Els he vists. "Them-M.PL I-have seen-M.PL."
Les he vistes. "Them-F.PL I-have seen-F.PL."

(5) Agreement with third-person clitic-pronoun direct objects

He vist les pellicules. "I-have seen-M.SG the films (F.PL)."
Quines pellicules he vist? "What films (F.PL) have-I seen-M.SG?"
Les pellicules que he vist "The films (F.PL) which I-have seen-M.SG"
Us he vist. "YOU-[F.PL] I-have seen-M.SG.
Els he vists. "Them-M.PL I-have seen-M.PL."
Les he vistes. "Them-F.PL I-have seen-F.PL."

(6) Agreement with third-person feminine clitic-pronoun direct objects

He vist les pellicules. "I-have seen-M.SG the films (F.PL)."
Quines pellicules he vist? "What films (F.PL) have-I seen-M.SG?"
Les pellicules que he vist "The films (F.PL) which I-have seen.-M.SG"
Us he vist. "YOU-[F.PL] I-have seen-M.SG."
Els he vist. "Them-M.PL I-have seen-M.SG."
Les he vistes "Them-F.PL I-have seen-F.PL."
210 JOHN CHARLES SMITH

(7) General non-agreement

He vist les pellicules. "I-have seen-M.SG the films (F.PL)."
Quines pellicules he vist? "What films.F.PL have-I seen-M.SG?"
Les pellicules que he vist "The films.F.PL which I-have seen-M.SG"
Us he vist. "YOU-[F.PL] I-have seen-M.SG."
Els he vist "Them-M.SG I-have seen-M.SG."
Les he vist "Them-F.PL I-have seen-M.SG."

3.2 Markedness and the morphosyntactic hierarchies

Can these data be accounted for in terms of markedness, as defined in
Section 1? There is little problem with hierarchy 1 (the règle de position);
preceding direct objects are clearly more marked than following direct objects in
a language with canonical VO order. (In Smith (1999:205-206), I discuss
Bréal's claim that this pattern of agreement in French reveals morphosyntactic
change taking place first of all in an unmarked context2—although he does not
use the term markedness?) Hierarchy 4 also accords with markedness principles
as far as gender is concerned, since agreement is more frequent in the (marked)
feminine, but, when we turn to number, we find that it conflicts with them, in as
much as agreement is less likely in the (marked) plural.
Some other aspects of the hierarchies seem even more problematic for a
theory of change based on markedness.
With reference to hierarchy 2, it is not obvious that topics, interrogadves,
and exclamatives are less marked than relatives. Here, the criteria to some extent
conflict. Relatives are more frequent than the other types of preceding direct
object; however, unlike topic, interrogative, and exclamative structures, relative
clauses are almost by definition subordinate clauses, and, according to this
criterion, marked environments (compare the discussion of word-order change in
Section 1).
Nor is it clear, in respect of hierarchy 3, that reflexive third-person clitic-
pronouns are less marked than their nonreflexive counterparts. According to
2
Mais comme il a fallu du temps pour opérer ce changement, comme les anciens tours sont
longs à se perdre, et comme la moindre dérogation au train ordinaire leur est un prétexte pour
se maintenir, le changement en question ne s'est imposé qu'avec la construction la plus
fréquente, celle que nous sommes habitués à considérer comme la construction normale.
Partout ailleurs, la langue se montre fidèle à l'ancienne grammaire." (Bréal 1897:224-25)
3
The term was coined, according to Andersen (1989:21) in 1930, as recorded in the
correspondence between Jakobson and Trubetzkoy. See also Greenberg (1966:62, note 3), who
traces the first published use of the term in phonology to a 1931 paper by Trubetzkoy and its
first use in relation to grammatical categories to a 1932 paper by Jakobson, also noting that
"earlier adumbrations of these ideas in reference to inflectional categories are to be found in
certain Russian grammarians".
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 211

qualitative criteria, the reflexive forms are more marked: they exhibit gender and
number syncretism; they are defective, in that they lack a subject form; and they
are less 'versatile', in that their reference is highly restricted (they must be
coreferential with the subject of the verb). However, reflexive forms are more
frequent than nonreflexive forms4 and might therefore be regarded as
quantitatively unmarked.
More serious, in respect of the same hierarchy, is the fact that third-person
clitic pronouns are arguably less marked than their first and second-person
equivalents. Greenberg (1966:84-85) examines data from a number of languages
which "lead one to posit, tentatively at least, a hierarchy in which the third person
[is] the least marked, and the second person the most marked, with the first
person intermediate". The behavior of the Romance pronouns supports this
view—the first and second-person forms exhibit syncretism of masculine and
feminine gender and of accusative and dative case, whilst the third-person forms
show greater 'versatility' (for instance, they may be anaphoric or deictic). Yet it
is precisely with the third-person forms that agreement is most resilient. An
alternative, discourse-based, view might be put forward, in which discourse
participants are less marked than nonparticipants, and in which the speaker, as
the necessary participant in every utterance, is less marked than the hearer,
yielding a hierarchy 'First Person > Second Person > Third Person'. Such a
view underlies the work of Bühler (1934:79-148) and Benveniste (1956), and is
explicitly articulated in the animacy/agency hierarchy of Silverstein (1976) (see
also Dixon 1994:84-90). But, even if this second analysis can be maintained, it
will be at best a Pyrrhic explanation: we shall have salvaged a markedness
account of the Romance phenomena, whilst demonstrating the Protean nature of
markedness.
Finally, hierarchy 2 provides a further problem for a universal definition of
markedness. Timberlake (1977:156) argues that pronouns are less marked than
nouns with respect to the changes in Finnish participial complement clauses. Yet
agreement in Romance between a participle and a direct object which is a noun
disappears earlier than agreement between a participle and a pronoun direct
object. The 'Pronoun > Noun' markedness hierarchy is therefore either invalid or
not universal.
4
See the frequency counts, based on 500,000-word corpora, of Juilland & Chang-Rodriguez
(1964) for Spanish (8038 se, as opposed to 6555 non-reflexive third-person conjunctive
pronouns: ratio 1.23:1); Juilland, Brodin & Davidovitch (1970) for French (4637 se, as against
4279 non-reflexive third-person conjunctive pronouns: ratio 1.08:1); and Juilland & Traversa
(1973) for Italian (5618 si, as opposed to 4333 non-reflexive third-person conjunctive
pronouns: ratio 1.30:1).
212 JOHN CHARLES SMITH

A possible solution to this last point can be found in recent work by

Aissen (1999), who recasts Silverstein's animacy/agency hierarchy in terms of
Optimality Theory (itself a theory of markedness—see Kager 1999:3).
Specifically, she claims that apparent markedness reversal is to be accounted for
in terms of formal harmonic alignment of interrelating scales, so that what is
unmarked in an unmarked context is marked in a marked context. (There are
similarities here with the theory of strong and weak variants propounded by
Orešnik (1999), in a theory which relates markedness to environment, such that
complex constructions ('strong variants') spread from complex environments
and simple constructions ('weak variants') spread from simple environments,
although the diachronic element is absent from Aissen's discussion, and the
conceptual framework is somewhat different.) Harmonic alignment of the
grammatical function scale 'Subject > Object' with the 'Pronoun > Noun'
markedness hierarchy, which is part of the Silverstein scale, will result in
pronoun subjects being less marked than noun subjects, but noun objects being
less marked than pronoun objects. This provides a relatively integrated account
of Timberlake's Finnish data and the Romance participial agreement hierarchy
2—the change described by Timberlake takes place first of all with pronouns
because it involves subjects; object-participle agreement in Romance disappears
first of all with nouns because it involves objects. Note that this account is
independently motivated and considerably less ad hoc than Timberlake's
statement that "the concept of markedness, or naturalness, must be understood
with reference to the particular change involved" (Timberlake 1977:168-169).
So far, so good. However, Aissen's approach cannot in itself salvage a
markedness account of the Romance object-participle agreement data, as
harmonic alignment of the Silverstein person scale 'Local (i.e., 1 or 2) > 3' with
the grammatical function scale 'Subject > Object' will still predict that third-
person objects are less marked than first or second-person objects and leave us
with the problem of accounting for hierarchy 3. It must also be pointed out that
harmonic alignment is not a panacea. There seems to be no justification at all, for
instance, for aligning the number markedness scale 'Singular > Plural' with the
grammatical function scale 'Subject > Object'. Even if singular subjects are less
marked than plural subjects, as may be plausible, this is simply a reflection of the
fact that singular nouns are less marked than plural nouns, regardless of their
grammatical function. It is certainly not the case that markedness reversal holds
in this instance, and that plural objects are less marked than singular objects. It is
not independently clear when harmonic alignment applies and when it does not.
Harmonic alignment must therefore be stipulated, and we are forced back to the
imprecise notion that the definition of markedness is in some sense specific to
the particular construction involved.
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 213

3.3 Functionality and the morphosyntactic hierarchies

In earlier work, I suggested that each of the hierarchies in Section 3.1. had
a functional explanation, rooted in sentence-processing strategies, which can be
summarized as follows (see Smith 1996; 1997:1100-1101).
Hierarchy 1. A direct object which precedes the verb is not in canonical or
unmarked position, so that a sentence containing such an object will be more
difficult to parse than one in which the direct object follows the verb. Agreement
of the participle may therefore facilitate processing by serving to 'flag' both the
immediately following empty object position and the number and gender of the
item elsewhere in the sentence to which it should be linked.
Hierarchy 2. In the case of Topics, Interrogatives, and Exclamatives, all the
information required in order to determine the referent of the direct object is
present in the same sentence as the participle. Similar arguments could be
advanced in the case of relatives—they require an antecedent, which is normally
found in the same matrix sentence. However, here the risk of ambiguity is
greater, as there may be more than one plausible antecedent. Clitic pronouns, on
the other hand, are much more difficult to process—they may be deictic, and so
require pragmatic resolution, and can present problems even when endophoric
(that is, anaphoric or cataphoric). Moreover, in the Romance languages, when a
clitic ends in a vowel, this is often elided before the initial vowel of the auxiliary
"have", giving rise to identical surface forms for clitics with referents of different
numbers and genders. Here, too, then, the likelihood of participial agreement can
be correlated with the difficulty of processing the sentence.
Hierarchy 2. First and second-person clitics are unambiguously deictic, the
first-person form always denoting the speaker and the second-person form the
addressee. It is clear that in these cases the referent can almost always be
recovered pragmatically, and the functionality of participial agreement is
therefore minimal. Similarly, the referent of a reflexive pronoun is by definition
identical with the subject of the verb and is therefore automatically recoverable
from the context. It is with nonreflexive third-person clitics that agreement will
be most functional, and it is therefore not surprising that we find it maintained
longest in these contexts.
Hierarchy 3. In varieties of Romance in which the likelihood of agreement
with a third-person clitic-pronoun direct object is dependent on the number and
gender of this item (some varieties of Catalan, Sardinian, and Rhaeto-Romance),
the gender of the singular pronoun preceding the auxiliary "have" can be
retrieved only through the inflection of the participle, whilst the gender in the
plural is evident from the surface form of the pronoun itself. In these
circumstances, agreement with the object pronoun is functional in the singular,
but redundant in the plural; it is therefore not surprising that it should be more
214 JOHN CHARLES SMITH

prevalent in the former case. It is less easy to account for the apparently
differential disappearance of agreement from the plural; but it may be that
feminine plural agreement is more resilient under the influence of agreement with
the feminine singular.
Finally, we may note that, in the languages from which object-participle
agreement has disappeared completely (Spanish, Portuguese, Rumanian), the
third-person accusative clitic pronouns have distinct forms for each number and
gender when they occur as the direct object of a compound past tense. In these
circumstances, object-participle agreement is redundant, as it conveys no
information which cannot be obtained from other items in the sentence.
All in all, then, it seems that the existence of the hierarchies in Section 3.1
is consistent with a functional account of the actualization, based on perceptual
strategies and sentence-processing. It is important to stress that the functional
motivation for the hierarchies is diachronic, not synchronic. Romance verbs in a
simple tense do not exhibit object agreement; and yet sentences containing such
items rarely present insurmountable parsing problems, regardless of the position
and identity of the object. Synchronically, therefore, the functionality of
participial agreement, where it occurs, is highly marginal. But, of course, my
claim is not that agreement with a preceding direct object was introduced in order
to facilitate parsing (apart from anything else, such a claim is inconsistent with
the data); rather that the marginal functionality of such agreement has nonetheless
been a factor in its differential disappearance—that is, given a tendency for this
type of agreement to disappear, agreement will be lost first in contexts where it
has less functional value. In other words, functionality is here acting as a brake
on actualization. In Smith (1995:169), I summed up the position as follows:

[T]he principle of 'recoverability' does not constrain the SYNCHRONIC phenomenon

of agreement; it constrains the DIACHRONIC process of the disappearance of (certain
types of) agreement. The differential agreement patterns observed are synchronic
reflexes of this constraint on a diachronic process.

3.4 Beyond the hierarchies, 1 : Perseveration

There is nonetheless one intriguing datum concerning object-participle
agreement in Romance which poses a problem for an analysis based entirely on
perceptual strategies. In Gascon, the pattern of agreement is entirely in keeping
with the hierarchies set out in Section 3.1. Agreement almost never takes place
with a following direct object, except in the conservative Lavedan dialect of
Arrens; see Bouzet (1963:43), Darrigrand (1974:200, 209, 214), Rohlfs
(1977:223), and Map 2499 of the Atlas linguistique de la Gascogne (Séguy &
Ravier 1973; henceforth ALG), which shows this type of agreement as normal
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 215

(four examples out of six) at only one point,5 and sporadic (one example out of
six) at four others.6 On the other hand, agreement almost always takes place
with a preceding clitic-pronoun direct object—at only one ALG locality7 does
agreement fail to occur with such an item. Agreement is variable when the direct
object is a relative. However, different types of relative behave in different ways
with respect to agreement. Relatives whose antecedent is a phrase with a noun as
its head are much less likely to trigger agreement than relatives whose antecedent
is a pronominal element which may itself vary for both number and gender. This
claim is made by many secondary sources, including Darrigrand (1974:
200-201) for Gascon in general, and Bouzet (1963:42-43) and Hourcade
(1986:173-74) for the Béarnais dialect. Confirmation is provided by Map 2498
of the ALG. Of the three sentences (8)-(10), quoted here, as in the text of the
ALG, in French:

(8) la chanson que j'ai chantée "the song that I have sung"
(9) c'est celle que je vous ai dite "it's the one I said"
(lit. "it's the one that I you have said")
(10) la lettre que j'ai lue "the letter that I have read"

it is in the responses to (9), where the antecedent itself (French celle; Gascon la)
is an item which may vary for both number and gender, that agreement is most
widespread, being found at 28 points.8 Agreement in (8) and (10), in which the
antecedent is a phrase with a noun as its head, is much rarer—for (8) it is found
at eight points9 and for (10) at only six points. 10 This pattern of agreement is in
5
6950 (Arrens, Hautes-Pyrénées).
6
681N (Vielle-Saint-Girons, Landes); 691NE (Sainte-Suzanne, Pyrénées-Atlantiques); 696
(Gerde, Hautes-Pyrénées); 696S (Lesponne, Hautes-Pyrénées).
7
699SE (Casau, Val d'Aran).
8
549N Saint-Yzans (Gironde); 643NE Grézillac (Gironde); 650E Castelnau-de-Médoc
(Gironde); 665S Villeneuve (Landes); 668SE Montaut-les-Créneaux (Gers); 672 Parentis-en-
Born (Landes); 672NO Biscarrosse (Landes); 676NO Aire-sur-1'Ador (Landes); 6760 Geaune
(Landes); 679 Lombez (Gers); 681 Soustons (Landes); 682 Tartas (Landes); 682N Ygos
(Landes); 688N Esclassan (Gers); 689N Galan (Hautes-Pyrénées); 691NE Sainte-Suzanne
(Pyrénées-Atlantiques); 692S Bedous (Pyrénées-Atlantiques); 693 Aas (Pyrénées-Atlantiques);
693NE Ferrires (Hautes-Pyrénées); 694 Nay (Pyrénées-Atlantiques); 694E Pontacq (Pyrénées-
Atlantiques); 695 Cauterets (Pyrénées-Atlantiques); 6950 Arrens (Hautes-Pyrénées); 696E Uzer
(Hautes-Pyrénées); 699NE Melles (Haute-Garonne); 760 Léguevin (Haute-Garonne); 762NE
Sainte-Foy-d'Aigrefeuille (Haute-Garonne); 781NO Boussan (Haute-Garonne).
9
549 Cissac (Gironde); 658 Lectoure (Gers); 669 Gimont (Gers); 692S Bedous (Pyrénées-
Atlantiques); 693NE Ferrières (Hautes-Pyrénées); 6950 Arrens (Hautes-Pyrénées); 696S
Lesponne (Hautes-Pyrénées); 762NE Sainte-Foy-d'Aigrefeuille (Haute-Garonne).
10
693NE Ferrières (Hautes-Pyrénées); 693NO Bielle (Pyrénées-Atlantiques); 694E Pontacq
(Pyrénées-Atlantiques); 6950 Arrens (Hautes-Pyrénées); 696 Gerde (Hautes-Pyrenées); 762NE
Sainte-Foy-d'Aigrefeuille (Haute-Garonne).
216 JOHN CHARLES SMITH

keeping with the notion that variation (and hence change) may be conditioned by
perseveration, or what Labov (1994:568) terms "simple repetition of the
preceding structure".

3.5 Beyond the hierarchies, 2: Genre, medium, and style

Last, we should discuss to what extent the distribution of object-participle
agreement conforms to the genre, medium, and style hierarchies adumbrated by
Andersen (1990:10). In other words, is there a correlation between the
markedness of the nongrammatical context and the likelihood that agreement will
take place?
This may well be the case for genre in at least some Romance languages.
Discussing the decline of object-participle agreement in Old Spanish,
Macpherson (1967:253-254) notes that agreement disappears from prose texts
earlier than it does from verse; and in his survey of the phenomenon in Old
Italian, Lucchesi (1963:255) observes that agreement is more likely to be
maintained in Dante's Divina Commedia than in the same author's prose works.
As far as Old French is concerned, Busse (1882:20, 25, 33) finds that agreement
is less frequent in popular poetry than in learned verse during the twelfth
century; but Wehlitz (1887) maintains that no such distinction exists in the
thirteenth.
As for medium, there is evidence that, in French, agreement is lost earlier
in speech and retained longer in writing. French authorities as diverse as Foulet
(1968:105) and Tesnière (1965:581) have claimed that the règle de position is
essentially bookish, whilst for Price (1971:233), it is "an artificial rule that is
widely ignored in speech and—even if only by inadvertence—occasionally in the
written language as well". Of course, this is to some extent a reflection of the
inaudibility of agreement with many past participles, as opposed to its systematic
visibility in writing (see the discussion in Smith (1993a)). Nonetheless,
agreement is audible with a number of past participles; and the impression that
we are dealing with a genuine difference between the spoken and the written
language is reinforced by the fact that the discrepancy appears to be more evident
when the direct object is higher on hierarchy 2; Dauzat (1926:108), Muller
(1964:44), Bonnaud (1984:26), and Hagège (1987:39), for instance, all
specifically claim that agreement with a preceding relative (as opposed to
agreement with a preceding clitic pronoun) is less likely in spoken French. On
the other hand, Alcover (1908:126) claims that written Catalan exhibits less
object-participle agreement than spoken Catalan. Alcover, who has a
nationalistic axe to grind, attributes the apparent markedness reversal in this case
to Castilian influence, a proposal which is open to question (see Smith 1995a).
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 217

However, the evidence regarding loss of agreement in casual, as opposed

to formal, style is not so easy to interpret. As early as 1576, the French
grammarian, Antoine Cauchie, defined nonagreement as "populaire" and
agreement as "plus érudit" (see Rickard 1968:279). But Cohen (1963:224ff)
claims, albeit anecdotally, that participial agreement in modern French tends to
thrive amongst the working class and be slightly less common in the speech of
intellectuals; and, according to the survey of Italian undertaken by Hall
(1958:97), teachers are less likely to make the agreement with a preceding
relative direct object, whilst white-collar workers are more likely to do so.
(However, the reactions of other professions are inconclusive, and Hall's sample
of forty-six informants is perhaps too small to enable any worthwhile
conclusions concerning social distribution to be made.) Brinker (1984:243-247),
on the other hand, in a more recent and thorough survey of Italian, finds that, on
the whole, speakers with a higher level of education are more likely to exhibit
participial agreement. It is clear that more information is needed, and that
extensive sociolinguistic surveys of this phenomenon need to be undertaken
before any valid generalizations can be made.
Finally, we may note that Nicoli (1983:371-372), in an aside in his
grammar of the Milanese dialect, claims that agreement with a following direct
object in standard Italian is more likely to occur in the trapassato remoto (or
'past anterior', which, with the auxiliary in the passato remoto, or simple past, is
both morphologically and stylistically marked) than in other compound past
tenses.

4. Conclusion
Evidence from agreement between past participle and direct object in the
Romance compound past tenses formed with "have" points to the following
conclusions:
This particular instance of actualization appears to be sensitive to
functionality, rather than markedness, in respect of morphosyntactic
environments. (When there is no conflict between a functional account and one
based on markedness, then the two may be regarded as complementary—we
may accept both factors as contributing, at least potentially, to the pattern of
change, and are not obliged to choose between them. But in this case the two
hypotheses cannot account equally well for all the data.)
Although the functional hypothesis seems to hold most of the time, there is
at least one morphosyntactic environment—relative clauses in Gascon—where a
counterfunctional pattern of agreement appears to predominate. The data here
lend some support to the notion that perseveration may also be an influence on
the actualization of morphosyntactic change.
218 JOHN CHARLES SMITH

As regards genre, medium, and style, the actualization does seem to be

sensitive to markedness in a preponderance of cases, with the change tending to
take place first of all in the unmarked context.
These conclusions raise some further questions.
Markedness, functionality, and perseveration all appear to be potential
factors in the actualization of morphosyntactic change. What determines the role
of each in a given instance? Without further motivation, the claim that only some
morphosyntactic change has a functional explanation is open to the same
objections that have been levelled against the 'markedness hypothesis' in this
paper—that it is essentially ad hoc. If a given type of explanation is valid only
some of the time, can the notion 'some of the time' be clarified in a principled
way? Or are we bound to accept the view espoused by Lass (1997:357), who
sees functional explanations as "epiphenomena of particular analyses,
unspecifiable in advance"?
What, if any, is the relationship between these factors? For instance, one
way to 'square the circle' and reconcile accounts based on functionality and
markedness would be to suggest that recoverability is unmarked and
nonrecoverability marked. Although perhaps intuitively plausible, such a
proposal seems ad hoc and not independently motivated. Might there be any
justification for it?
These issues must be the object of further research.

REFERENCES
Aissen, Judith. 1999. "Markedness and subject choice in Optimality Theory".
Natural Language and Linguistic Theory 17.673-711.
Alcover, A. M. 1908. "Concordansa del participi ab el terme d'acció". Primer
congrès internacional de la llengua catalana, Barcelona, octubre de 1906,
124-128. [Barcelona]: Horta.
Andersen, Henning. 1989. "Markedness theory—the first 150 years".
Markedness in Synchrony and Diachrony ed. by Olga Mišeska Tomic,
11-46. Berlin & New York: Mouton de Gruyter.
Andersen, Henning. 1990. "The structure of drift". Historical Linguistics 1987:
papers from the 8th International Conference on Historical Linguistics
(Lille, 31 August-4 September 1987) ed. by Henning Andersen & Konrad
Koerner, 1-20. (Current Issues in Linguistic Theory, 66.) Amsterdam &
Philadelphia: John Benjamins
Andersen, Henning. 2001. "Markedness and the theory of linguistic change".
This volume, 21-57.
Battistella, Edwin L. 1990. Markedness: The Evaluative Superstructure of
Language. Albany, N.Y.: State University of New York Press.
Benveniste, Emile. 1956. "La nature des pronoms". For Roman Jakobson:
Essays on the occasion of his sixtieth birthday ed. by Morris Halle, H. G.
Lunt, H. McLean, H. &  H. van Schooneveld, 34-37. The Hague:
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 219

Mouton. Reprinted in Emile Benveniste, Problèmes de linguistique générale,

251-257. Paris: Gallimard, 1966.
Bonnaud, Pierre. 1984. Mémento abrégé des trois types grammaticaux
fondamentaux de l'auvergnat. (Annales du CRDP de Clermont-
Ferrand—Documents régionaux.) Clermont-Ferrand: CNDP-CRDP.
Bouzet, J. 1963. Syntaxe béarnaise et gasconne. Pau: Marrimpouey jeune.
Bréal, Michel. 1897. Essai de sémantique: science des significations. Paris:
Hachette.
Brinker, Jacques H. 1984. Problemi dell'accordo del participio passato
nell'italiano moderno. Groningen: Regenboog.
Bühler, Karl. 1934. Sprachtheorie: die Darstellungsfunktion der Sprache. Jena:
Fischer.
Busse, Johannes. 1882. Die Congruenz des Participa Praeteriti in activer
Verbalconstruction im Altfranzösischen bis zum Anfang des XIII. Jahr-
hunderts. (Inaugural-Dissertation, Georg-Augusts-Universität, Göttingen.)
Göttingen: Kaestner.
Bybee, Joan L. 1985. Morphology: a study of the relation between meaning and
form. (Typological Studies in Language, 9.) Amsterdam & Philadelphia:
John Benjamins.
Bybee, Joan, Revere Perkins & William Pagliuca. 1994. The Evolution of
Grammar: tense, aspect and modality in the languages of the world.
Chicago: University of Chicago Press
Cohen, Marcel. 1963. Nouveaux regards sur la langue française. Paris: Editions
sociales.
Croft, William. 1990. Typology and Universals. Cambridge: Cambridge
University Press.
Croft, William. 1995. "Autonomy and functionalist linguistics". Language
71.490-532.
Croft, William. 2000. Explaining Language Change: an evolutionary approach.
London: Longman.
Darrigrand, Robert. 1974. Initiation au gascon. Deuxième édition revue et
corrigée. Orthez: Per Noste.
Dauzat, Albert. 1926. La langue française: sa vie, son évolution. Paris: Stock.
Dixon, R. M. W. 1994. Ergativity. (Cambridge Studies in Linguistics, 69.)
Cambridge: Cambridge University Press.
Dressier, Wolfgang U. 1985. Morphonology: the dynamics of derivation. Ann
Arbor: Karoma.
Ferguson, Charles A. 1996. 'Variation and drift: loss of agreement in
Germanic". Sociolinguistic Perspectives: papers on language and society
1959-1994 by Charles A. Ferguson ed. by Thorn Huebner, 241-260. New
York & Oxford: Oxford University Press. Reprinted in Towards a Social
Science of Language: papers in honor of William Labov ed. by Gregory R.
Guy, Crawford Feagin, Deborah Schiffrin & John Baugh, 173-197.
(Current Issues in Linguistic Theory, 164.) Amsterdam & Philadelphia: John
Benjamins, 1996.
220 JOHN CHARLES SMITH

Fodor, Janet Dean. 1981. "Does performance shape competence?" Philosophical

Transactions of the Royal Society of London, series B, volume 295 (no.
1077), 285-295.
Foulet, Lucien. 1968. Petite Syntaxe de l'ancien français. Paris: Champion.
Gilliéron, Jules. 1918. Généalogie des mots qui désignent l'abeille. Paris:
Champion.
Gilliéron, Jules & Mario Roques. 1910. "Études de géographie linguistique XII.
Mots en collision. A: Le coq et le chat". Revue de philologie française
24.278-288.
Givón, Talmy. 1979. On Understanding Grammar. New York: Academic
Press.
Greenberg, Joseph H. 1963. "Some universals of grammar with particular
reference to the order of meaningful elements". Universals of Language
(Report of a conference held at Dobbs Ferry, New York, April 13-15, 1961)
ed. by Joseph H. Greenberg, 58-90. Cambridge, Mass.: M.I.T. Press.
Greenberg, Joseph H. 1966. "Language universals". Current Trends in
Linguistics: Vol. III, Theoretical Foundations ed. by Thomas A. Sebeok,
61-112. The Hague & Paris: Mouton.
Hagège, Claude. 1987. Le français et les siècles. Paris: Odile Jacob.
Hall, Robert A. Jr. 1958. "Statistica sintattica: l'accordo del participio passato
coniugato con avere". Lingua nostra 19.95-100.
Harris, Alice C. & Lyle Campbell. 1995. Historical Syntax in Cross-Linguistic
Perspective. Cambridge: Cambridge University Press.
Hendriks, Peter. 2000. "Kakari particles and the merger of the predicative and
attributive forms in Old Japanese". Historical Linguistics 1995. Vol. 1:
General Issues and Non-Germanic Languages. Selected Papers from the
Twelfth International Conference on Historical Linguistics (Manchester,
August 1995) ed. by John Charles Smith & Delia Bentley, 155-167.
(Current Issues in Linguistic Theory, 161.) Amsterdam & Philadelphia: John
Benjamins.
Hock, Hans Henrich. 1991. Principles of Historical Linguistics. Second edition.
Berlin: Mouton de Gruyter.
Hourcade, A. 1986. Grammaire béarnaise. Pau: Los Caminaires.
Juilland, Alphonse & E. Chang-Rodriguez. 1964. Frequency Dictionary of
Spanish Words. The Hague: Mouton.
Juilland, Alphonse, Dorothy Brodin & Catherine Davidovitch. 1970. Frequency
Dictionary of French Words. The Hague & Paris: Mouton.
Juilland, Alphonse & Vincenzo Traversa. 1973. Frequency Dictionary of Italian
Words. The Hague & Paris: Mouton.
Kager, René. 1999. Optimality Theory. Cambridge: Cambridge University
Press.
Labov, William. 1994. Principles of Linguistic Change. Volume 1: Internal
Factors. Oxford: Blackwell.
Lass, Roger. 1997. Historical Linguistics and Language Change. Cambridge:
Cambridge University Press.
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 221

Lightfoot, David. 1991. How to Set Parameters: arguments from language

change. Cambridge, Mass.: The M.I.T. Press.
Lucchesi, Valerio. 1963. "L'accordo fra participio passato e oggetto nei tempi
perifrastici retti da 'avere' nel volgare antico (secc. XIII-XIV)". Atti e
memorie dell'Accademia toscana di scienze e lettere 27.191-278.
Macpherson, Ian R. 1967. "Past participle agreement in Old Spanish: transitive
verbs". Bulletin of Hispanic Studies 44.241-254.
Martinet, André. 1955. Économie des changements phonétiques. Bern: Francke
Matsuda, Kenjirô. 1998. "On the conservatism of embedded clauses". Historical
Linguistics 1997: selected papers from the 13 th International Conference on
Historical Linguistics (Düsseldorf 10-17 August 1997) ed. by Monika S.
Schmid, Jennifer R. Austin & Dieter Stein, 255-267. (Current Issues in
Linguistic Theory, 164.) Amsterdam & Philadelphia: John Benjamins.
Muller, Charles. 1964. "Les Français et leurs participes". Le Français dans le
monde [avril-mai 1964], 43-44.
Newmeyer, Frederick J. 1998. Language Form and Language Function.
Cambridge, Mass.: The M.I.T. Press.
Nicoli, Franco. 1983. Grammatica milanese. Busto Arsizio: Bramante.
Orešnik, Janez. 1999. Krepke in sibke dvojnice ν skladnji. Ljubljana: Slovenska
Akademia Znanosti in Umetnosti.
Pope, Μ. Κ. 1934. From Latin to Modern French with Especial Consideration
of Anglo-Norman: phonology and morphology. Manchester: Manchester
University Press.
Poplack, Shana. 1980. The notion of the plural in Puerto Rican Spanish:
competing constraints on /s/ deletion". Locating Language in Time and
Space ed. by William Labov, 55-68. New York: Academic Press.
Poplack, Shana. 1981. "Mortal phonemes as plural morphemes". Variation
Omnibus ed. by David Sankoff & Henrietta J. Cedergren, 59-72. (Current
Inquiry into Language, Linguistics, and Human Communication, 40.)
Carbondale & Edmonton: Linguistic Research Inc..
Price, Glanville. 1971. The French Language: present and past. London:
Edward Arnold.
Rohlfs, Gerhard. 1977. Le Gascon: études de philologie pyrénéenne. Troisième
édition. Tübingen: Niemeyer and Pau: Marrimpouey jeune.
Rickard, Peter 1968. La langue française au seizième siècle: étude suivie de
textes. Cambridge: Cambridge University Press.
Samuels, M. L. 1987. "The status of the functional approach". Explanation and
Linguistic Change ed. by Willem Koopman, Frederike van der Leek, Olga
Fischer & Roger Eaton, 239-250. Amsterdam: John Benjamins.
Scherre, Maria Marta Pereira & Anthony J. Naro. 1991. "Marking in discourse:
'birds of a feather"'. Language Variation and Change 3.23-32.
Séguy, J. & X. Ravier. 1973. Atlas linguistique de la Gascogne. Vol. VI. Paris:
C.N.R.S.
Silverstein, Michael. 1976. "Hierarchy of features and ergativity", Grammatical
Categories in Australian Languages ed. by R. M. W. Dixon, 112-171.
Canberra: Australian Institute of Aboriginal Studies.
222 JOHN CHARLES SMITH

Smith, John Charles. 1989. "Actualization reanalyzed: evidence from the

Romance compound past tenses". Synchronic and Diachronic Approaches
to Linguistic Variation and Change (Georgetown University Round Table
on Languages and Linguistics, 1988) ed. by Thomas J. Walsh, 310-325.
Washington, D.C.: Georgetown University Press.
Smith, John Charles. 1991. "Problemi dell'accordo del participio passato
coll'oggetto diretto nei tempi composti coniugati con avere in italiano, con
speciale riferimento ai dialetti". Tra Rinascimento e strutture attuali: saggi di
linguistica italiana (Atti del Primo Convegno delia Società Internazionale di
Linguistica e Filología Italiana, Siena, 28-31 marzo 1989, volume 1) ed. by
Luciano Giannelli, Nicoletta Maraschio, Teresa Poggi Salani & Massimo
Vedovelli, 365-371. Torino: Rosenberg & Sellier.
Smith, John Charles. 1993a. "The agreement of the past participle conjugated
with avoir and a preceding direct object: a brief history of prescriptive
attitudes". Authority and the French Language: papers from a conference at
the University of Bristol ed. by Rodney Sampson, 87-125. Münster: Nodus.
Smith, John Charles. 1993b. "La desaparición de la concordancia entre participio
de pasado y objeto directo en castellano y catalán: aspectos geográficos e
históricos". Actas del Primer Congreso Anglo-Hispano: tomo I, Lingüística
ed. by Ralph Penny, 275-285. Madrid: Castalia.
Smith, John Charles. 1995a. "Agreement between past participle and direct
object in Catalan: the hypothesis of Castilian influence revisited". Linguistic
Change under Contact Conditions ed. by Jacek Fisiak, 271-289. (Trends in
Linguistics—Studies and Monographs, 81.) Berlin & New York: Mouton de
Gruyter.
Smith, John Charles. 1995b. "Perceptual factors and the disappearance of
agreement between past participle and direct object in Romance". Linguistic
Theory and the Romance Languages ed. by John Charles Smith & Martin
Maiden, 161-180. (Current Issues in Linguistic Theory, 122.) Amsterdam &
Philadelphia: John Benjamins.
Smith, John Charles. 1996. "Surfonctionnalité et hyperanalyse: l'accord du
participe passé dans les langues romanes à la lumière de deux théories
récentes". Faits de langues 8.113-120.
Smith, John Charles. 1997. "Types and tokens in language change: some
evidence from Romance". Language History and Linguistic Modelling: a
Festschrift for Jacek Fisiak on his 60th birthday ed. by Raymond Hickey &
Stanisław Puppel, 1099-1111. (Trends in Linguistics—Studies and
Monographs, 101.) Berlin & New York: Mouton de Gruyter.
Smith, John Charles. 1999. "Markedness and morphosyntactic change revisited:
the case of Romance past participle agreement". The Emergence of the
Modern Language Sciences: studies on the transition from historical-
comparative to structural linguistics in honour of E. F. Konrad Koerner ed.
by Sheila Embleton, John E. Joseph & Hans-Josef Niederehe, 499-511.
Amsterdam & Philadelphia: John Benjamins.
Tesnière, Lucien. 1965. Eléments de syntaxe structurale. Deuxième édition.
Paris: Klincksieck.
 MARKEDNESS, FUNCTIONALITY, AND PERSEVERATION 223

Timberlake, Alan. 1977. "Reanalysis and actualization in syntactic change".

Mechanisms of Syntactic Change ed. by Charles N. Li, 141-177. Austin &
London: University of Texas Press.
Vincent, Nigel. 1976. "Perceptual factors and word order change in Latin".
Romance Syntax: synchronic and diachronic perspectives ed. by Martin
Harris, 54-68. Salford: University of Salford.
Vincent, Nigel. 1982. "The development of the auxiliaries HABERE and ESSE in
Romance". Studies in the Romance Verb: essays offered to Joe Cremona on
the occasion of his 60th birthday ed. by Nigel Vincent & Martin Harris,
71-96. London & Canberra: Croom Helm.
Wehlitz, Hermann. 1887. Die Congruenz des Participa Praeteriti in activer
Verbalconstruction im Französischen (vom Anfang des XIII. bis zum Ende
des XV. Jahrhunderts). (Inaugural-Dissertation, Universitat Greifswald.)
Greifswald: Abel.
Wheeler, Max W. 1988. "Catalan". The Romance Languages ed. by Martin
Harris & Nigel Vincent, 170-208. London: Routledge.
Wurzel, Wolfgang U. 1984. Flexionsmorphologie und Natürlichkeit. Berlin:
Akademie- Verlag.
 ACTUALIZATION
AND THE (UNI)DIRECTIONALITY OF CHANGE

HENNING ANDERSEN
University of California, Los Angeles

0. Introduction
One of the central notions in a speaker-oriented theory of linguistic change
is that of 'actualization', the observable manifestation of grammar innovations in
speech. Given the focus on observable change in this collection of papers, it
seems right to include here a few lines to characterize the notion of
'actualization', define its place in such a theory of change, and compare it with
similar notions in other theories of change with which it contrasts. This is the
topic of the first part of this paper.
Once it is understood that all change progresses through synchronic
variation, the central issue for a speaker-oriented theory of change is to explain
how synchronic variation turns into change that has a determinate direction. In
the second part of this chapter I show how the theory of Markedness that is
discussed in several of the preceding papers explains why the actualization of all
linguistic change—historical changes, typological drifts, and universal tendencies
alike—typically proceeds in a determinate direction.

1. Actualization: observable innovations in usage

The term actualization has been current in historical linguistics for a couple
of decades—principally, surely, thanks to the memorable title of Timberlake's
1977 paper "Reanalysis and actualization in syntactic change". Speaking in
general terms, the notion 'actualization' implies the view that every innovation in
usage actualizes—that is, makes manifest in observable reality—some
corresponding innovated element of grammar. More narrowly, 'actualization' is
mostly associated with reanalysis. This is the way Timberlake applied the notion
in his title as well as in the exposition of his paper, and this is the way it is
mostly understood by those historical linguists who recognize reanalysis as a
significant source of linguistic change, and who share the understanding that
226 HENNING ANDERSEN

such structural innovations in grammar precede and explain the innovations in

usage by which they are actualized.
It is notable that the notion of actualization—which is the only observable
part of linguistic change—has played no role in most theorizing about change.
Compare, for instance, the theories of Halle (1962), Weinreich, Labov & Herzog
(1968), Lightfoot (1991, 1999), and Harris & Campbell (1995), which are
outlined briefly in Section 1.1.

1.1 The role of actualization in theories of change

In Halle's theory (1962), the rule additions of mature speakers are naturally
directly manifested (one could say 'actualized') in usage. But when subsequently
learners produce a reanalysis of the grammar, this change in grammatical
structure entails no observable change in usage, that is, it has no actualization.
Weinreich, Labov & Herzog (1968) subscribed to a similar hypothesis,
viz. that innovations first arise in usage and are then 'embedded' in grammar.
Accordingly they identified what they called a "problem of embedding", but no
'problem of actualization'.
Lightfoot (1999) believes that reanalysis occurs as a result of geographical
or social dialect mixture (158). His theory does not explain how dialect
differences can arise in the first place. But he supposes that if dialect contact
occurs, it is likely to produce conflicting cues for the setting of parameters (166)
and thus to prompt learners to "grow" competing grammars. In such instances of
"internalized diglossia" (92), as time goes by, one or the other of the competing
structures carries the day statistically in the speech community—and a change
has occurred. In Lightfoot's theory, then, as in Halle's and Weinreich's, major
grammar change is preceded by and prompted by changes in usage. Lightfoot
explicitly hypothesizes that actual observable change is "piecemeal, gradual, and
chaotic" (105)—though it still somehow statistically describes an S-curve (102)
as one of two competing grammars spreads through a population of speakers,
eventually eclipsing the other. The reference to the S-curve is a notable
concession to empiricism in this work, but the hypothesis that actual change is
chaotic is in direct conflict with the kind of observations of change to which this
volume is dedicated: change in usage progresses not in a chaotic, but in an
orderly, grammatically conditioned step-by-step fashion.
In an only slightly more realistic vein, Harris and Campbell (1995) sketch
a theory of change which they illustrate with a great variety of changes, but
which explicitly recognizes "only three mechanisms of syntactic change",
reanalysis, extension, and borrowing (50). Fortunately, as they develop their
exposition, the authors are led to acknowledge the existence of other kinds of
 ACTUALIZATION AND UNIDIRECTIONALITY 227

change that are essential to any descriptively adequate account of language

history.
There is the phenomenon they call "exploratory expressions", by which
speakers draw on a "set of universally available syntactic constructions" to go
beyond the bounds of the codified grammatical means of their language (54).
There are also therapeutic changes, "steps taken to rid the grammar of'
dysfunctional homophony (182), which they oddly relabel "compensation"
(318-320).
They also mention 'actualization' (77-88). They first describe this in the
spirit of Timberlake (1977:141) as "the gradual mapping out of the consequences
of ... reanalysis" (80) and then as a process or "period" of adjustment attendant
on reanalysis (81). But then they apparently change their minds, finding "that
each example of change under actualization [is] itself either an extension or an
additional reanalysis" (80). And although the word actualization is used
frequently in their subsequent exposition, it varies in meaning, sometimes
subsuming "extension", sometimes apparently being interchangeable with
extension (see the tokens of actualization p. 80 with footnote 23 and pp. 81, 97,
104, 105, 114, 178, 214, and passim). In one place it is stated that the
actualization of syntactic change may include phonological change (81).
Harris and Campbell emphasize the gradualness of syntactic change
(39-40, 48-49), and some of their case histories illustrate syntactic changes that
are attested as chronological progressions. But in their failure to come to grips
with the fundamental difference between actualization and the three "mechanisms
of change" they recognize, they miss the opportunity to integrate actualization
—the observable part of language change—into their theory. Quite apart from
this, the fact that they have to supplement their three "mechanisms of change"
with several additional kinds of change that are not "mechanisms" (but may
include still other "mechanisms") is a clear indication that their theory, as it is laid
out, is neither exhaustive nor coherent. Harris and Campbell's work is justly
praised for its rich contribution to the study of syntax. But they themselves have
found the 'three-mechanism theory' wanting, and it is time to proceed towards a
more comprehensive, explanatory theory of change.

1.2 Some basic concepts for the description of change

This is not the place for a full-scale presentation of such a theory. But in
order to speak intelligibly about actualization it seems necessary to clarify the
notion relative to other aspects of change. There is probably no simpler way of
doing this than by drawing some elementary distinctions and analysing a few
commonplace types of change into their component elements.
228 HENNING ANDERSEN

Let us first of all distinguish between diachronic correspondences and

changes. A diachronic correspondence is the relation between homologous
elements (be it of grammar or of usage) belonging to two chronologically
separate synchronic states in a linguistic tradition. Most accounts of linguistic
change limit their concerns to diachronic correspondences—which are not
changes, but the results of change. Such accounts often do not pay attention to
the chronological unfolding of changes and often do not recognize speakers as
agents of change. Changes, by contrast, are the historical events in a linguistic
tradition by which practices of speaking vary over time. Changes can be
observed as they occur, though they are often not noticed as changes by the
members of a speech community. Changes are always manifested in synchronic
variation, and past changes can commonly be found to be reflected in synchronic
alternations, or attested in written records, in such a way that it is possible to
define diachronic correspondences; and these may enable the language historian
to infer what sort of changes have occurred. No account of such historical events
can attain descriptive, let alone explanatory, adequacy without reference to
speakers.
Secondly, let us recognize that 'change' (no less than 'diachronic
correspondence') is a metalinguistic notion, regardless of whether the word is
employed by a native speaker who observes that 'young people nowadays' do
not speak the way we used to, or it is used as a quasi-technical term by the
historical linguist. Furthermore, 'change' is a pretheoretical notion with no fixed
extension. This is why it is sometimes an issue whether a given historical event
sequence in a tradition of speaking is to be understood as a series of changes, as
one single change, or as part(s) of a change.1 And this is why it can be a
problem—for instance, where a language is poorly attested—whether to define
as 'a change' some minor alteration that may be part of a larger picture or,
alternatively, a larger development, extrapolated from scant evidence. On an ad-
hoc basis historical linguists sometimes get around the vagueness of the count-
noun change by distinguishing micro and macro changes, or short-term and
long-term changes, or changes and drift. I return to this issue in Section 2.1
Thirdly, let us recognize that every change in a tradition of speaking is
analysable into smaller events. In every change we examine, we can discern or
infer what may be called subchanges, sometimes understandable as a single act
of innovation, but mostly composed of a stream of numerous, practically
identical acts of innovation, made in speech acts by individual speakers and
hearers. Changes of different type consist of distinct combinations of
1
I imagine most language histories offer examples of such problems. For instance, the
Common Slavic Second and Third Velar Palatalizations have been viewed as a sequence of
changes, as a single change, and as part(s) of a drift toward tonality (palatality) harmony.
 ACTUALIZATION AND UNIDIRECTIONALITY 229

subchanges, logically consecutive, but in actuality always overlapping in time.

One can think of such combinations as change scenarios.2
Among the subchanges the most interesting ones, generally speaking, are
the initial (or primary) innovations, which define scenarios of different type (the
names of these will be capitalized in the following exposition); see Section 1.3.
But the most common component parts of change scenarios are the subchanges
adoption, reanalysis, and actualization', I discuss these in some detail in Section
1.4.
In the following I will repeatedly have occasion to refer to the synchronic
variation of newer and older elements. In speaking of the atomic events in which
speakers make meaningful choices between such elements I will refer to them as
I-variants and O-variants, with Τ for "innovated, incoming" and 'O' for "older,
outgoing".

1.3 Change scenarios

In this section I briefly describe a few change scenarios.
Coinage. Coinage begins with a pragmatically motivated innovation: One
or more individual speakers (i) draw on their metagrammatical or grammatical
competence to coin a new word, a new turn of phrase, or a new morphosyntactic
combination in order to achieve a communicative aim and (ii) actualize this
innovation in speech. The innovation may go unnoticed, or other speakers may
passively accept it or actively (iii) adopt it and (iv) actualize this adoption in their
own usage. If it is adopted and used widely enough, it may be (v) acquired by
new cohorts of speakers and (vi) actualized in their usage. Through such first-
language acquisition (subchange (ν)), the I-variant is for the first time integrated
in speakers' grammars, that is, analysed (reanalysed) as an integral element of a
given lexical field, syntactic repertoire, or morphosyntactic paradigm, as the case
may be (see further Section 1.4.1). Subsequently the I-variant may spread and be
generalized in the community through repeated cycles of subchanges (iii)-(vi).
There are several subtypes of Coinage, amply exemplified in the literature; cf.
Stern 1931:192-198, 282-330, Andersen 1974:19-20, 1980a:7-9, Harris &
Campbell 1995:54, 72-75.
Remedial change. Remedial change too originates with a pragmatically
motivated innovation. One or more individual speakers (i) draw on their
(meta)grammatical competence to circumvent an awkward or otherwise

2
Neither change scenarios nor the subchanges of which they are composed are 'mechanisms',
indeed there is nothing mechanical about language change. It is an important point about
change—perhaps too important to be relegated to a footnote—that no innovation can become
part of a tradition of speaking unless it is purposely adopted by bearers of that tradition and
purposely selected by them for use in speech.
230 HENNING ANDERSEN

dysfunctional expression and (ii) actualize this Remedial innovation. If the new
element is noticed and deemed useful by other speakers, the just mentioned
subchanges (iii)-(iv) and possibly (v)-(vi) will follow, and the I-variant may be
generalized through cycles of subchanges (iii)-(vi). The most common subtypes
of Remedial innovation are therapeutic ones and euphemism. Cf. Stern
1936:330-336, Andersen 1974:21-22, 26, 1980a: 10-11, Harris & Campbell
1995:318-320.
Borrowing. Borrowing is a type of contact change—in origin obviously
pragmatically motivated. One or more individual speakers (i) draw on another
tradition of speaking (or a tradition of writing) to satisfy a communicative need
and (ii) actualize this innovation in speech. If enough other speakers see the
utility of the new element, some or all of subchanges (iii)-(vi) will follow, and
the new element may be generalized through cycles of these subchanges. Cf.
Bloomfield 1935: 444-460, Andersen 1974:22, 1980a: 11-14, Harris &
Campbell 1995:120-150.
Extension. Extension is probably the best term for a type of change for
which no particular pragmatic motivation can be defined. One or more individual
speakers (i) more or less intentionally extend a received lexeme (metaphorically
or metonymically) to a new referent, or the use of an established construction to
a new lexical or syntactic context, or the application of a given morphophonemic
rule to a new morphosyntactic environment; (ii) the innovation is naturally
immediately actualized in usage. If other speakers accept the innovation, the I-
variant may spread across the speech community through subchanges (iii)-(vi),
and it may be generalized through cycles of these subchanges. Cf. Stern
1931:340-350, 199-236, 237-281, Andersen 1974:24-25, 1980a:16-18, Harris
& Campbell 1995:97-119 (but see Sections 1.1, 1.4.3).
Transference and interference. Transference and Interference are often not
distinguished from Borrowing. Like Borrowing they are types of contact change,
but they are not pragmatically motivated and are in essence kinds of extension.
In Transference, one or more individual speakers of a language L1 who
are familiar with another tradition of speaking, L2, (i) apply a lexical, syntactic or
morphosyntactic rule proper to L 2 (ii) in speaking L1. If other L1 speakers adopt
the L2 element, some or all of subchanges (iii)-(vi) will follow, and it may be
generalized through cycles of these subchanges. Transference is particularly
important for an understanding of 'superstratum effects'.
In Interference, an innovation arises when one or more speakers of L1 (i)
select an element of L1 (ii) in speaking L2; the further course of events is parallel
to that in Transference and leads to the generalization of the L1 element in L2.
Interference plays an important role in language contact situations, being
 ACTUALIZATION AND UNIDIRECTIONALITY 231

responsible for 'substratum effects'. Cf. Bloomfield 1934:461-475, Andersen

1974:27, 1980a:8, Kaufman & Thomason 1988.
Both Transference and Interference are relevant to the development of
linguistic alliances (Sprachbünde) in areas with extensive bilingualism
(horizontal contact) and to the mutual rapprochement of grammars employed in
diglossic traditions of speaking (vertical contact); cf. Bubenik 2001.
These are some of the most common types of change. Evidently each area
of grammar presents its own specific possibilities for innovation. But these
major types of change can be exemplified from all areas of grammar, including
phonology, which has not been illustrated here, but is subjected to a revealing
comparison with lexical change in Andersen (1974). Note that all the innovations
that make up the subchanges in these scenarios have been characterized here with
speakers as agents.

1.4 Adoption, reanalysis, and actualization

As illustrated in Section 1.3, each and every type of change can be
resolved into a sequence of subchanges, representing alternately covert grammar
innovations and their overt, observable actualizations in speech. The covert
subchanges adoption and reanalysis, and the overt subchange actualization enter
into the scenarios of all types of change and deserve separate comment.

1.4.1 Adoption and reanalysis. Adoptions are an essential part of all types of
change. They can be thought of as contact innovations, but unlike, for instance,
Borrowing, adoption involves linguistic elements that already occur in the
adopter's community usage.
The distinction between adoption and reanalysis made in the scenarios
above implies the widely accepted understanding that there is a fundamental
difference between, on one hand, primary language acquisition, in which the
basic structures of a speaker's grammar are formed—and a novel analysis
(reanalysis) of the base grammar may occur3—and, on the other hand, the
secondary process through which speakers, throughout their lives, on a day-to­
day basis adjust their usage competence by repairing any shortfalls it might have,
elaborate it to encode additional categories of descriptive, referential, pragmatic,
stylistic, or social content they encounter in the usage of the community, and
3
The historical linguist's interest in change has yielded us the term reanalysis for the cases
where a novel analysis arises, but we have no established term either for the analyses that
replicate previous ones or for 'degree-0' analysis in general. The neologism neo-analysis (or
neanalysis) would be suitable, but would it be adopted and generalized? Terminology aside, it
is worth emphasizing that for language learners forming a first grammar, there is only
neanalysis.Neanalysis may include reinterpretations and revisions, but reanalysis, as defined
here, is strictly a linguist's notion.
232 HENNING ANDERSEN

keep their usage rules up-to-date with observed changes in ambient usage and
with any changes in standards of appropriateness these may reflect. Whereas
reanalysis of the base grammar occurs in the course of a speaker's primary
grammar formation, adoption is achieved through a secondary modification of
the speaker's usage rules. We return to the distinction between base grammar and
usage rules in Section 1.4.4 below.
Adoption has been the subject of much discussion in the sociolinguistic
literature, where the focus of interest has been the diffusion of innovations
across social networks, the relative strength of the links of which such networks
consist, and the relations of prestige and solidarity these links imply—in short,
the questions of who adopts from whom and why (see, for instance, Milroy
1992, 1993). However, when adoption is compared with other kinds of
innovation, as it is here, it must be characterized primarily in functional terms:
Adoptions are the innovations through which speakers throughout their lifetime
make ever more elements of their community's tradition of speaking theirs, as
they grow in experience, so that they can communicate with others in accordance
with their practical needs and their perceived status and roles.
Adoptions, then, differ from reanalysis by being purposive innovations.4
Reanalysis, by contrast, occurs in the process of first-language acquisition.
This is a purposive activity, but reanalysis itself is neither intended nor
purposeful: since learners have no insight into the grammars of the speakers
whose usage they observe, they cannot aim to form grammars that differ from
their models'.
But a common feature of both adoption and reanalysis is that any
innovated element is ascribed value in relation to its older, established
counterpart. Both in adoption and in reanalysis (neanalysis; see Footnote 3) I-
variants and O-variants are evaluated in terms of the community's current
appropriateness norms, and as they are used, their usage manifests the speakers'
understanding of these values (cf. Andersen 1989:24-25).5

At one extreme, perhaps, adoptions may include grammar adjustments that are not intended,
and of whose actual consequences speakers may be unaware, namely, individual speakers'
instinctive, involuntary imitation of the usage of others. Thus some of the mutual adaptation
among members of a speech community may be purposive for some speakers, while for others
it cannot properly be termed purposive. In either case, though, it can be recognized as
functional in as much as it serves the individuals, and indirectly serves the group they form,
by signifying their solidarity with the group.
In speaking of norms it is useful to distinguish (i) statistical norms (where normal means
"frequent"), (ii) historical norms (where normal means "familiar to most speakers,
customary"), (ii) prescriptive norms (where normative means "correct according to an
established standard"), and (iv) appropriateness norms (where the corresponding adjective is
appropriate). (i) and (ii) are kinds of descriptive norms, (iii) and (iv), kinds of deontic norms.
 ACTUALIZATION AND UNIDIRECTIONALITY 233

1.4.2 Reanalysis and actualization. In the survey of change types in Section 1.3
it was suggested that the integration of every kind of (initial) innovation into a
tradition of speaking occurs through reanalysis (innovation (v) in each of the
change scenarios there). But in addition to the change types enumerated in
Section 1.3, there are changes whose very origin is in reanalysis, conventionally
called Evolutive changes.
The idea that grammar change can arise in speakers' analysis of surface
realizations presupposes that surface realizations can be structurally ambiguous,
so that speakers can construe them differently, assigning them different content
or different underlying relations or representations. This is not a new idea, but a
rather old insight, as Harris & Campbell point out (1995:31-32). But recently
there has been an emerging consensus that reanalysis—rather than only ensuing
from changes in usage, as in the theories of Halle (1962), Weinreich, Labov &
Herzog (1968), and Lightfoot (1991) (see Section 1.2) and in the change types
surveyed in Section 1.3—is a major, primary source of grammar innovations.
One strong argument in favor of this view is that if one assumes that reanalysis
occurs only after a number of changes in surface structure, many surface
changes will appear unmotivated. If, on the other hand, one supposes that
reanalysis precedes such surface changes, then the observable changes are
thereby explained as actualizations that bring the surface into line with an
innovated underlying structure (cf. Harris & Campbell 1995:77).
This view finds ample support in instances of idiosyncratic usage, which
are most naturally understood as direct actualizations of individual instances of
reanalysis.6
6
Reanalysis and actualization can often be identified in individual speakers' usage. For
example, one recent textbook of linguistics repeatedly speaks of the progression of linguistic
change as remorseless. Evidently the writer (or someone from whom he has adopted or
acquired this usage) has interpreted remorseless as a synonym of unremitting. As a
consequence of this reanalysis the writer's usage may in part replicate received usage
{remorseless in the sense "remorseless"), in part varies from received usage (remorseless in the
sense "unremitting").
Another example. The utterance in (1) suggests to the eavesdropping linguist that the
speaker in question has reanalysed the elative adverb extra as a comparative adverb. This
(1) "Nah, I don't wanna work any extra hard than I gotta."
reanalysis is naturally actualized in any sentence the speaker produces that contains extra +
adjective, but the covert innovation has overt consequences only in sentences in which extra +
adjective are combined with a standard of comparison.
Compare the fictional exchange in (2), in which Al's second remark does not show
whether he uses extra as an elative or a comparative adverb. Instead, his remark illustrates the
(2) Al. We have a lot of new orders, we gotta work hard today.
Bob. I always work hard.
Al. I know, Bob. But today you gotta work extra hard.
234 HENNING ANDERSEN

One can sketch a first approximation to a scenario of Evolutive change as

follows: One or more individual speakers (i) construe a certain type of surface
strings differently from previous cohorts of speakers (reanalysis) and (ii)
actualize their reanalysis in usage that varies from received usage in some
particular(s). The innovative usage may be (iii) adopted by other speakers and
(iv) actualized by them and may be generalized in the community through
repeated cycles of subchanges (i)-(ii) and (iii)—(iv).7 Note that the notion of
'learner' in this account is age-neutral (as in Andersen 1973). A reanalysis may
concern one or more of the following dimensions: (i) segmentation (including
phonological, morphosyntactic or sentence-syntactic constituency); (ii) valuation
(ascription of content and category labels, centricity and bond strength,
morphophonemic and phonological features); (iii) ranking (of relevant features);
cf. Andersen (1974, 1980a). It is not clear yet what constitutes structural
ambiguity in surface realizations; this remains a question for the future.
However, to be observationally adequate this skeleton scenario must be
modified to accommodate two kinds of facts. First, when a change of any type
occurs in a speech community, speakers who produce the innovative usage
normally continue to produce the older usage as well, typically according to
different appropriateness conditions. Secondly, the replacement of O-variants by
I-variants is negotiated through a period—years, decades, generations, or
centuries—of gradually changing synchronic variation that is typically
conditioned by social, stylistic, pragmatic, as well as grammatical categories.

1.4.3 Multiple analyses? To explain the regular coexistence, after reanalysis, of

O-variants and I-variants—not just in community usage, but in the usage of
individual speakers—some linguists suppose that learners perform multiple
analyses of the usage to which they are exposed, inferring distinct sets of rules
for the older and the newer usage; thus Harris and Campbell (1995:70-72,
81-89 and passim). This theory, however, fails to explain how and why the
members of a speech community would tacitly decide to collectively favor one of
two such analyses over the other. In particular it cannot explain the common
configuration in change events in which a new, statistically insignificant usage
(pattern) proves more viable than an established, statistically predominant one
and in time entirely replaces it.

sort of structural ambiguity—here of meaning, elative or comparative—that makes extra open

to divergent analyses, that is—speaking in diachronic terms—makes it possible for some
speakers to reanalyse extra as a comparative adverb. I have no evidence of this reanalysis from
other speakers' usage and do not know how widespread it is, or what it would take to make
other speakers adopt this conflation of extra and more.
7
For a simple example of this scenario, consider the expression I couldn't care less "I don't
care at all", reanalysed, actualized, adopted, and actualized as I could care less.
 ACTUALIZATION AND UNIDIRECTIONALITY 235

Secondly, this theory does not explain why the gradual replacement of one
usage by another would progress by grammatically defined steps. Harris and
Campbell theorize that the 'winning' analysis goes through stages of repeated
extension and reanalysis (1995:80).8 Here they are in general agreement with the
understanding of Garcia (1990, 1997). This scenario (Extension, as defined in
Section 1.3) probably explains some historical changes, but it is certainly not
appropriate in cases of Evolutive change. Applying it to such a change as the
development of animacy marking in Russian (see the "Introduction", pp. 1-2)
one would have to suppose that this grammatical category started out as a special
morphological marking for singular, direct-object, first-declension nouns
referring to adult, free, healthy, male humans. But this supposition raises difficult
questions. What would impel any speaker or group of speakers to set up special
morphological marking for such a narrowly delimited set of potential referents?
And if one takes this reanalysis to be a natural outcome of some universal
grammar-formation strategy, why would it then subsequently be degraded and
watered down by Extensions? And why would these Extensions lead precisely

Consider the reanalysis of anymore as "nowadays". This is a semantic reinterpretation that

was possible for a long time in sentences like (3.a), where anymore could either be understood
as part of the paradigm of anyone, anybody, anything, anywhere, etc. or not, that is, as
independent of the negation.
(3) a. My advice is, Buy land! They don't make that anymore.
b. We cook with gas anymore.
c. Anymore, you don't have to be doing something to be stopped by police. You just
have to be.
As it happened, (i) anymore was reanalysed, maybe first by one or more speakers in Northern
Ireland, maybe independently both there and in the U.S.A.; (ii) the result of the reanalysis was
actualized in sentences like (3.a), which are structurally ambiguous, as well as in innovative
usage such as (3.b). The occurrence of anymore in affirmative declarative sentences like (3.b)
increased the likelihood of a growing stream of (iii) reanalyses, (iv) actualized in ever
spreading corresponding usage (in Northern Ireland and in Oklahoma, Tennessee, Kentucky,
Iowa, and western states settled from these; cf. AHD, s.v.).
Note that at the time of the reanalysis, anymore occurred only nonfocalized (as in 3.a-b).
More recently (in my experience) it has come to occur also focalized (3.c). In purely
observational terms, one could say that it has been 'extended' to this new environment; this
would be in accordance with Harris and Campbell's theory. However, this overt innovation is a
logical consequence of anymore's reanalysis as "nowadays", a meaning which is equally
compatible with both environments. Hence the new occurrence of anymore in sentence-initial
position can be considered part of the actualization of the initial reanalysis; this would be in
accordance with the theory of Timberlake (1977) and Andersen (1969, 1973, 1974, 1980,
1989).
Still, one can wonder why there would be a time lag between the occurrence of anymore
"nowadays" in nonfocalized and in focalized position. To understand this one might suppose
that all speakers have a memory of experienced usage, and that speakers generally feel most
comfortable keeping their own usage within the bounds of what is usual in their experience.
Alternatively one can look at this as a change from below being actualized in unmarked
environments before marked. See Section 1.4.4.
236 HENNING ANDERSEN

toward consistent expressions for (the typologically widespread category of)

animacy rather than something else, say, categories that are already salient in the
language, such as singular count-nouns or first-declension masculine nouns?

1.4.4 Base grammar and usage rules. A reasonable alternative account is the
theory of Evolutive change advocated by Timberlake (1977), which Harris and
Campbell touch on briefly, but then glance off (1995:81).
This theory assumes, first of all, the existence of two layers of structure in
speakers' grammars, (i) a base system of content categories, syntactic relations,
and expression types that embodies universal principles of grammar and (ii) a
system of usage rules that are formulated in terms of this base grammar, but
make reference to more superficial categories, in part community-specific
(including pragmatic, stylistic, and sociolinguistic ones)—rules that enable
speakers to match the local and contemporary usage of the tradition of speaking
they consider theirs.9
Secondly, the theory assumes that in grammar formation, any base-
grammar analysis (whether it is innovative or not) entails the formation of
complementary usage rules. The usage rules enable speakers who have inferred
dissimilar base grammars to by and large match current customary community
usage, and they enable all speakers, whatever differences there may be among
their base grammars, to produce variable usage in accordance with the
community's appropriateness norms. In this way usage rules serve to ensure
relative uniformity of (variable) usage in synchrony and relative continuity of
usage across cohorts. (The reality of usage rules is demonstrated by the
phenomenon of hypercorrection; cf. Andersen 1973:781-782.)
Thirdly, the theory assumes that, in the case of Evolutive change, the
covariants that are directly generated by a given base grammar are unmarked in

The distinction between base grammar and usage rules corresponds very approximately to
Coseriu's distinction between 'functional system' and 'norms' (1962), which is adopted by
Andersen 1973 with the labels core system and adaptive rules. Timberlake (1977:142-145,
172) preferred Coseriu's terms system and norms, which were used for some time also by
myself. But it should be noted that Coseriu does not operate with 'speakers' grammars'; his
'norms' are explicitly characterized as a speech community's "historical norms", which is not
what Timberlake had in mind. Coseriu's 'norms' thus cannot to be equated with the usage rules
we may wish to posit for individual speakers' grammars. The term usage rule, which I suggest
here, is intended to be equally applicable with reference to individuals and collectives and is
intended to be neutral with respect to the issue(s) of norms; cf. footnote 5.
In ontogenetic terms, the complementary relation between base grammar and usage rules
may well be reinterpreted during the maturation of the individual learner-speaker. Also, it may
be different from one speaker to the next—perhaps as a result of differences in individual
language acquisition strategies—even though such differences may be practically
inconsequential with respect to individual performance. This is a psycholinguistic perspective
on grammar formation that may be amenable to investigation.
 ACTUALIZATION AND UNIDIRECTIONALITY 237

relation to the variants that are defined by its usage rules. In view of the
Principle of Markedness Agreement (Andersen 2001:27-37), this makes
I-variants that result from base-grammar reanalysis initially compatible with
unmarked usage categories and with environments defined by unmarked
grammatical categories. By virtue of their inclusive definition (see the theory of
Markedness outlined in Andersen 2001:44), the unmarked I-variants have
greater use potential and are able to substitute for the marked O-variants. Hence
over time—where there is no community resistance to change—I-variants will
tend to move in on the O-variants' privileges of occurrence, gradually replace
them, and become generalized.
Thus this theory of Evolutive change explains how a new construal of
observed usage—a base-grammar reanalysis—can provide motivation for an
observabe step-wise development through which—in Timberlake's words—"the
consequences of the reanalysis are mapped out" in usage. Once some speakers
form base grammars that are innovative in a certain respect (base-grammar
reanalysis), and their analysis is actualized in innovative usage, more speakers
are likely to arrive at that analysis, and the frequency of occurrence of I-variants
will gradually increase. While successive cohorts of learners perform this
innovative base-grammar analysis in increasing numbers, the I-variants gradually
spread—in the usage of successive cohorts of speakers—from unmarked to
marked categories across a set of environments defined by a hierarchy of
grammatical features, as illustrated in several of the papers in this volume. These
changes in usage result in part from the greater use potential of I-variants, which
allows each cohort of speakers to innovate the use of I-variants in lieu of O-
variants in some environments. In part the changes reflect a succession of usage-
rule reanalyses through which each cohort of speakers makes its predecessors'
actual use of I-variants the base line for its own tendency to give I-variants
greater play in yet more environments.
In brief, in Evolutive change (i) a base-grammar reanalysis is in part (ii)
actualized directly in innovative usage, in part realized through cycles of (iii)
usage-rule reanalysis and (iv) actualization which gradually alter the privileges of
occurrence of I-variants and O-variants.10 The direction and the dynamic of the

The distinction between base-grammar reanalysis (and actualization) and usage-rule

reanalyses (and actualizations) which is made explicit here is mostly implicit in Timberlake's
account (1977). Timberlake's focus on 'the grammar of the language' may create the impression
that in this theory of Evolutive change there is one reanalysis and one actualization and the
relation between them is chronological. On the other hand, Garcia's understanding of change as
a cyclical reanalysis-and-actualization process captures the step-wise alterations in usage rules,
but does not give due prominence to the relation between base grammar and usage rules. It
appears that a coherent account can be created only if the focus is kept on speakers' grammars,
and the complementary relation between base grammar and usage rules is acknowledged.
238 HENNING ANDERSEN

change are provided by the relation between base grammar and usage rules,
which determines the Markedness values of I-variants and O-variants.
Throughout the progression of the change, these Markedness values remain the
same, but the privileges of occurrence of the I-variants gradually expand while
those of the O-variants contract. (The extension of the inclusive term increases at
the expense of that of the included term.) However, at any given time during the
progression, the privileges of occurrence of the two variants overlap, and it is in
the environments of this overlap that speakers make the pragmatically,
stylistically, and/or socially meaningful choices between the two variants that
define the synchronic variation at that time.

1.5 Markedness and Markedness shift

The distinction between base grammar and usage rules, which is essential
to the account of Evolutive change presented in the preceding paragraphs, is
equally important for an understanding of the other types of change whose
scenarios were outlined in Section 1.3. Whereas the subchange adoption may
typically be effected through addition or modification of a usage rule, the
reanalysis through which any innovated element is integrated in new speakers'
grammars consists in either an innovative formulation of usage rules (usage-rule
reanalysis) or a simultaneous reshaping of some part of the base grammar and
the usage rules that are relevant to it (base-grammar and usage-rule reanalysis).
Previously linguists have mostly assumed that any innovative element is
marked in relation to the element it eventually replaces (Section 1.1). This is
surely true of some pragmatically motivated types of (initial) innovation (see
Section 1.3). In such cases a marked I-variant will naturally occur initially in
marked usage categories and in environments defined by marked grammatical
categories (in conformity with the Principle of Markedness Agreement;
Andersen 2001:28). When it is widely adopted in a speech community and gains
currency, it may, at some point, be judged to be unmarked—through a
revaluation by some speakers or a Markedness reanalysis by new cohorts of
learners, or both. The reinterpretation of the I-variant's value allows it to spread
to unmarked environments and thus to gain ascendancy in community usage and
eventually supersede the O-variant,
Such a change in the Markedness values of I-variants and O-variants is
aptly called a Markedness shift. Mostly this term is understood simply as the
change in frequency by which the I-variant crosses the 50% threshold and
becomes the more frequent variant. In the formulation of the preceding
paragraph, by contrast, the Markedness shift occurs in speakers' grammars
(revaluation) and between speakers' grammars (reanalysis), and it is this (covert)
 ACTUALIZATION AND UNIDIRECTIONALITY 239

change in value that explains the subsequent overt change in distribution and
frequency of occurrence.
As mentioned in Andersen (2001:33-37), marked innovations may be
characteristic of certain pragmatically motivated change types, among them the
contact changes termed changes from above and clarification changes (which
include some innovations that renew bleached expressions). But in speaking of
such innovations it should not be forgotten that I-variants may be valuated
differently not only when a Markedness shift occurs. During their entire
histories, I-variants and O-variants may be ascribed different value by different
social groups and/or different individual speakers. It is only when we speak
abstractly of 'the community grammar', that we can speak of a shift in the
Markedness values of a pair of I-variants and O-variants—abstractly—as a
single, simple event.
In other types of innovation, however, where the I-variants are in full
conformity with the productive rules of speakers' base grammars, they may
typically be unmarked from their very first occurrence, being directly generated
by the new underlying grammatical structure. One can imagine this to be
common in instances of Coinage—certainly in instances where there is no O-
variant to replace, as is the case with certain kinds of neologism. Perhaps also I-
variants that arise by Extension are typically unmarked. But the archetypical
examples of unmarked I-variants, undoubtedly, are those of Evolutive change.
As the papers in this volume show, a great deal can be learned about the
progression of different kinds of change just from a few well-described changes.

2. On the (uni)directionality of change

The central issue for a speaker-oriented theory of Evolutive change is to
explain how the speakers of a language—who have no idea of its past history
and no plans for its future—transform synchronic variation into change.
The theory that has been sketched in the preceding pages is intended
precisely to explain how the asymmetrical relations of synchronic variation
translate into the epiphenomenon of directional change as a tradition of speaking
is passed on in the changing population of a speech community.
It should be clear from the preceding exposition that the dynamic and the
direction of individual change events cannot be understood unless one
recognizes distinct types of change and acknowledges that each is composed of
distinct combinations of innovation types—see Sections 1.3 and 1.4.4—and each
hence involves its own kinds of motivation.
It will be useful, before concluding this presentation, to reemphasize the
point that was made in Section 1.2: the word change does not have a determinate
extension; not only can it be applied to event sequences of different degrees of
240 HENNING ANDERSEN

magnitude, but its content can be conceptualized on several distinct levels of

generality. The theory that has been sketched here is designed specifically for the
description and explanation of historical change in traditions of speaking. But
there are other levels of observation or construal of language in diachrony to
which this theory does not apply directly, but which are of considerable interest
to the historical linguist.

2.1 Levels of observation in diachrony

The individual speaker. The lowest level of observation is the chrono­
logical development of the individual speaker's grammar. Changes in a speaker's
grammar over the time of the individual's life, beyond the period of grammar
formation, is a potentially fruitful topic for psycholinguistic research (cf.
Footnote 9). The language historian rightly considers his object of investigation
not the speech or grammars of individuals, but the history of a tradition of
speaking, which implies its historical transmission from speaker to speaker
through language acquisition, the presumed locus of reanalysis. This does not
mean, however, that the historical linguist should ignore documented diachronic
differences in individual usage. On the contrary, data from individual speakers
can offer significant insight into the progression of historical change. This is so,
for example, when a writer's works reflect the different (variable) usage of their
respective dates of composition, as in the case of Shakespeare's use of thou and
you discussed in Busse (2001). It can also occur that there are usage differences
between different versions a writer's works (drafts and final copy, earlier and
later editions) that the historical linguist can recognize as reflections of change in
progress; cf. the changes in clitic placement in edited writings discussed in Rittel
(1975:70-71). But changes in the usage of an individual may reflect no more
than the individual's running adaptation to ambient community usage (cf. Section
1.4.2). They do not explain Evolutive change and are not explained by a theory
of Evolutive change.
The tradition of speaking. Historical change represents that intermediate
level of observation which is the language historian's domain, and which was the
topic of Sections 1.2-1.5. At this level, we can distinguish between changes of
different magnitude (cf. Section 1.2), measured, perhaps, not so much in time as
in terms of structural import. It seems reasonable to distinguish between
historical changes that do not entail a shift in the 'groundplan' of the language (as
Sapir called it) and changes that can be understood as typologically significant.
Probably Sapir and Coseriu were right that, in Evolutive change, just as
structural reanalysis is actualized in new usage, so typological reanalysis is
realized through structural changes, that is, a typological shift precedes and
explains the structural changes that manifest it; see Sapir (1921:170), Coseriu
 ACTUALIZATION AND UNIDIRECTIONALITY 241

(1968). It may be difficult to draw the line between these two categories of
change. In this volume, it seems, the papers by Bakken, Bergs & Stein, Busse,
and Smith fall into the former, those by Bubenik, Mithun, and Sch0sler, into the
latter category.
Change typologies. At a higher level of observation are the generalized
diachronic schemas that sum up the historical linguist's experience with language
histories, observed as well as reconstructed. In phonology, for instance, there is
the 'lenition schema' that shows the typical developments from voiceless tense
plosives through various degrees of phonetic reduction to zero and various
vowel-shift schemas. In sentence grammar, we have, for example, the
'grammaticalization schemas' that lead from elements with lexical content
(phrases or words) through stages of semantic simplification (bleaching) to
grammatical morphemes. Grammaticalizations are typically correlated with a
'scope-expansion schema' and a 'synthesis schema', which leads from free forms
(phrases or words) to affixes, and what one might call the 'Zipf schema', which
leads from -syllabic expressions through monosyllables, segments, and single
distinctive features to zero—though each of these schemas is widely instantiated
independently of grammaticalization. Numerous individual gram-maticalization
schemas have been defined since the first synthetic overview by Tauli (1958);
for a comprehensive inventory summing up recent progress, see Heine & Kuteva
MS.

2.2 Change schemas

Each of these 'change schemas', as we can call them, represents a
generalization about a class of changes. They naturally attract attention by their
generality and command interest especially, perhaps, by virtue of their apparent
unidirectionality.
It seems important to define their place in a theory of change.
First of all, it has to be understood that change schemas are not changes,
but generalizations over changes. (Actually it would be more precise to call them
diachronic-correspondence schemas; cf. Section 1.2). The relation between a
change schema and a historical change that conforms to it, then, is that of type
and token. Where a given change schema is fully instantiated in the history of a
language, the instantiation results from a series of historical change events, each
of which has comprised numerous cycles of reanalysis and actualization carried
out by the bearers of the given tradition of speaking, and each of which has
occurred independently of any previous or subsequent changes in the relevant
schema. For any element that fits the description of a given step in such a schema
may change to fit the description of the step to its right regardless of its past
history. And no such schema implies a prediction that any actual language
242 HENNING ANDERSEN

element that fits the description of a given step will change. Change schemas are,
to paraphrase Kurylowicz's well-known formulation ([1949] 1960:60), like the
eavestroughs, downspouts, and gutters that channel the water away when it
rains: they show what way the water will run, but not whether there will be any
rain. Often enough an actual historical change instantiates only part of a given
schema; in fact, every multi-step change schema (the lenition schema, for
instance, or the lexical verb > auxiliary > tense/aspect-marker schema) is
exemplified more widely by historical changes that instantiate some part of its
trajectory than by changes that instantiate the complete trajectory.
Secondly, change schemas are generalizations over Evolutive changes.
This is important to bear in mind in discussing whether an actual historical
change conforms to or deviates from a given schema. This point has been
emphasized by Tabor & Traugott (1998:236) in relation to the grammatical-
ization schema. They point out that known changes of affixes to nouns that run
counter to this schema (e.g., Eng. -ism > ism "any distinctive doctrine or
practice") are not Evolutive changes, but are tokens of other change types,
apparently mainly Coinage (Section 1.3).
Thirdly, change schemas are theory-internal generalizations that have the
status of analytical propositions. They are essentially classifications of data, they
do not make empirical claims, and hence they are not necessarily invalidated by
data they do not subsume. Thus the generalizations expressed by the lenition
schema are in no way weakened, let alone invalidated, by the existence of
fortition changes. Nor is the monophthongization schema invalidated by
diphthongization changes. Nor yet are the significant generalizations regarding
'syllable structure changes' (a technical term for changes that improve syllable
structure) weakened by the existence of other kinds of change that affect syllable
structure adversely; see Vennemann (1988:66). With regard to the
grammaticalization schemas, which have been the subject of much discussion
from this point of view, it appears that once the concept of grammaticalization is
adequately defined, none of the alleged 'counterexamples' that have been debated
in the literature are grammaticalizations (see, for instance Croft 2000, Traugott
MS). It is not certain at present whether there is any need for a
'degrammaticalization schema' representing instances of Evolutive change that
run in the opposite direction of grammaticalization changes. But if such a schema
turns out to be justified, that is, if there is indeed a class of such changes, they
will turn out to have entirely different motivation from the changes subsumed by
the grammaticalization schema.
Finally, regarding motivation: Since change schemas are generalizations
over changes, they have no linguistic motivation and are not explananda. But
each schema directs our attention to the changes it subsumes and the motivation
 ACTUALIZATION AND UNIDIRECTIONALITY 243

these presumably have in common. That makes change schemas a useful

instrument in the investigation of linguistic changes and their motivations.

2.3 Unidirectionality
The importance of distinguishing between actual changes and change
schemas should be obvious.
All actual historical changes reflect a multiplicity of conditioning
parameters—any linguistic change, in particular, is the product of multifarious
grammar-internal conditions plus the social conditions that enable speakers to
assign to I-variants and O-variants the changing social values they have during
the progression of a change. As a consequence of shifts in social valuation,
many linguistic changes do not run their full course; some barely get off the
ground before they peter out, some get under way in one style of speaking and
are 'reversed' when another style of speaking becomes preferred in the
community, and some changes lose their momentum along the way and leave
unchanged residue of one kind or another. Every experienced historical linguist
can cite examples of changes that could have been, but fell by the wayside, as
can be inferred from circumstantial evidence. In language change false starts and
arrested developments are nothing out of the ordinary.
Change schemas, by contrast, abstract from all this idiographic 'noise' and
allow the investigator to focus on the single motivating parameter each step in the
schema reflects. By their unidirectionality they point to the principles of valuation
that speakers apply in the process of grammar formation, including the
Markedness values speakers willy-nilly assign to any synchronic variants they
identify, which will turn these into I-variants and O-variants in case a change
occurs.
Let me illustrate this with a single example, the universal tendency for
words to become shorter, the Zipf schema.
There are several regular phonological change types through which
expressions become longer or shorter, in particular prothesis, epenthesis, and
paragoge and aphaeresis, syncope, and apocope (subsumed by the diphthongi-
zation and monophthongization schemas; see Andersen 1972). But quite apart
from phonological changes such as these, which may affect the length of
expressions in various ways and to various extents depending on how regularly
they are carried through, all languages seem to manifest a drive towards shorter
expressions, which is particularly notable for frequent expressions. Typical
examples are grammatical morphemes such as the clitic auxiliaries in English,
e.g., we have /wi-v/, we will /wi-1/ and the negaton in have not /hæv-nt/, must
not /mәs-nt/.
What does the Zipf schema show us about grammar formation?
244 HENNING ANDERSEN

Traditionally linguists who have considered this universal tendency have

been content to say that words just get worn from use, and frequent words are
exposed to "exceptional wear" (Anttila 1972:187). But words are not things that
can get worn from frequent handling. Word tokens, once they are uttered, are not
exposed to wear, and there is no evidence at all that the types (or mental images)
that such tokens instantiate get worn in the process of being instantiated.
One might of course just read Zipf s law, that "high frequency is the cause
of small magnitude" (1935:31), as an indication that "people shorten the
linguistic expressions that are most commonly used" (Newmeyer 1998:254). But
this is really not an explanation, but a restatement of what has been observed. It
very reasonably places responsibility for the shortening on people, but there is
no reason to suppose that speakers keep running-frequency counts that would
enable them to single out individual expressions for shortening, nor is there any
evidence that this is how shortenings come about.
Alternatively one might 'explain' the Zipf schema 'simply' by pointing out
that an efficient code should minimize the average message length and use less
bandwidth for more frequent messages (thus Hamming 1980, cited in Tabor &
Traugott 1995:267). This 'explanation' assumes speakers know something about
code efficiency, which is an interesting proposition. But it begs the question of
what form such tacit knowledge might take.
However, if it is true that all change arises out of variation, we can interpret
the changes the Zipf schema subsumes as follows: (i) language expressions
develop variation in length; (ii) of any two variants, the shorter will normally be
unmarked; hence (iii) if the synchronic relation is translated into change, the
shorter will be the I-variant and the longer, the O-variant—that is, if one of the
variants is generalized, it will be the shorter one.
Let us assume that all expression types can occur in utterance contexts that
call for (pragmatically conditioned) more or less reduced realizations. If this is
so, then the more numerous the tokens of a given expression (the greater its
relative frequency), the greater the likelihood that such a morpheme will have
reduced realizations in usage, and the more likely it is that these will be
recognized by learners and be incorporated into speakers' grammars by being
reanalysed as reduced allomorphs.
Although the reduced variants arise under specific conditions, any time
these conditions are not identified precisely (reanalysed) in language acquisition,
the longer variants will be marked in relation to the shorter variants. This is so
because a longer variant is realized only in full tokens, whereas a shorter variant
may be realized both in tokens of the reduced and in tokens of the full variant.
Thus all the segments of must not /most nat/ are realized only in ['most 'nat],
whereas those of /mәs-nt/ are realized both in ['mәsnt] and in ['mәst 'nat].
 ACTUALIZATION AND UNIDIRECTIONALITY 245

This explains the well-established fact that the basic-derived relation between
lento and allegro forms typically is reversed, so that lento forms become derived,
and the basic allegro forms in the end are generalized. Another important
corollary of this default Markedness relation between shorter and longer
expression variants is the finding that deletion rules are less widely employed in
languages (less 'natural') than addition rules (Dressier 1980, Andersen 1980b);
that is, shorter variants are preferred as basic, and longer variants, as derived
allomorphs.
In this interpretation, phonetic forms are not exposed to wear, and no
intentional shortening of expressions by speakers is alleged. Change is not
intended, but arises in the diachronic projection of the values of synchronic
variants. Does this account assume speakers know how to achieve code
efficiency? In a sense it does: The default assignment of Markedness values to
shorter (u) and longer () variants is surely based on an innate principle of
grammar formation, which is a kind of procedural knowledge. The principle
clearly favors code efficiency. The opposite default value assigment would be
dysfunctional.
Comparable insights into principles of grammar formation can be achieved
by exploiting other change schemas in a focused examination of the changes they
subsume in relation to the kinds of synchronic variation these changes project.

3. Conclusion
The theory of change that has been sketched in the preceding pages has
been developed in the spirit of Edward Sapir's often cited conception of change.
In the well-known passages in the chapter on "Drift" and elsewhere in the
book Language, Sapir established that "the linguistic drift has direction"
(1921:155), and he did not hesitate to claim that the explanation for this
observable—the explanation both for the drift and for its direction—is to be
found in the language itself: "language moves down a current of its own
making" thanks to the "structural genius" that provides the "deep controlling
impulse to form that dominates [its] drift" (144, 170).
Sapir's reference to "language" here is not to some object existing outside
or independently of the speakers of the language, as in some contemporary
theorizing; the reference is squarely to speakers' grammars, which are assumed
to be substantially similar among members of a community. It is this substantial
similarity of speakers' grammars that makes it possible to understand that "the
drift of a language is constituted by the unconscious selection on the part of its
speakers of those individual variations that are cumulative in a certain direction"
(155).
246 HENNING ANDERSEN

Sapir was well aware that this account begs the question of how speakers
can know to make just those selections that are cumulative in a specific direction.
This is indeed a cardinal explanandum for any realistic theory of change: how do
new speakers of a language know or discern which way it is changing, whither it
is drifting. Sapir was convinced that the explanation for this is in the very
formation of the linguistic patterns. He saw that "we shall not advance seriously
until we study the intuitional bases of speech" and asked rhetorically, "How can
we understand the nature of the drift that frays and reforms ... [linguistic]
patterns when we have never thought of studying ... patterning as such and the
'weights' and psychic [= cognitive] relations of the single elements ... in these
patterns?" (183)
Elsewhere in this volume it has been suggested that the intuitional bases of
speech include an inomissible, a priori 'weighting' of all the single elements that
enter into the cognitive relations that constitute the patterns of a speaker's
grammar (Andersen 2001:45-47). In this paper I have tried to show how in
Evolutive change such synchronic asymmetries—through the unconscious,
collective choices made by members of speech communities—come to be
transformed into the changes and drifts we can observe in traditions of speaking.

REFERENCES
The American Heritage Dictionary of the English Language ed. by Anne H.
Soukharov et al. Third edition. Boston & New York & London: Houghton
Mifflin Co., 1992.
Andersen, Henning. 1973. "Diphthongization". Language 48.11-50.
Andersen, Henning. 1973. "Abductive and deductive change". Language
49.567-595.
Andersen, Henning. 1974. "Towards a typology of change: bifurcating changes
and binary relations". Historical Linguistics. Proceedings of the First
International Conference on Historical Linguistics, I—II ed. by John M.
Anderson & Charles Jones, vol. II, 17-60. Amsterdam: North-Holland.
Andersen, Henning. 1980a. "Morphological change: towards a typology".
Historical Morphology ed. by Jacek Fisiak, 1-50. (Trends in Linguistics.
Studies and Monographs, 17.) The Hague: Mouton.
Andersen, Henning. 1980b. "Russian conjugation: acquisition and evolutive
change", Papers from the Fourth International Conference on Historical
Linguistics ed. by Elizabeth Closs Traugott et al., 285-301. Amsterdam:
Benjamins.
Andersen, Henning. 1989. "Understanding linguistic innovations". Language
Change. Contributions to the Study of Its Causes ed. by Leiv Egil Breivik &
Ernst Håkon Jahr, 5-28. Berlin & New York: Mouton de Gruyter.
Andersen, Henning. 1990. "The structure of drift". Historical Linguistics 1987.
Papers from the 8th International Conference on Historical Linguistics ed.
 ACTUALIZATION AND UNIDIRECTIONALITY 247

by Henning Andersen & Konrad Koerner, 1-20. (Current Issues in

Linguistic Theory, 66.) Amsterdam & Philadelphia: John Benjamins.
Andersen, Henning. 2001. "Markedness and the theory of change". In this
volume, 21-57.
Anttila, Raimo. 1989. Historical and Comparative Liunguistics. Second edition.
Amsterdam & Philadelphia: John Benjamins.
Bloomfield, Leonard. 1935. Language. London: George Allen & Unwin.
Bubenik, Vit. 2001. "On the actualization of the active-to-ergative shift in Pre-
Islamic India". In this volume, 95-118.
Busse, Ulrich. 2001. "Markedness and the use of address pronouns in Early
Modern English". In this volume, 119-142.
Coseriu, Eugenio. 1962. "Sistema, norma y habla". In his Teoría del lenguaje y
lingüística general, 11-113. Madrid: Editorial Gredos.
Coseriu, Eugenio. [1965] 1968. "Sincronía, diacronía y tipología". Actos del XI
Congreso Internacional de Lingüistica i Filología Románicas, Vol. 1,
269-283. Madrid, 1965. Reprinted as "Synchronie, Diachronie und
Typologie" in his Sprache. Strukturen und Funktionen. XII Aufsätze zur
allgemeinen und romanischen Sprach-wissenschaft, 91-108. (Tübinger
Beiträge zur Linguistik, 2.) Tübingen: Gunter Narr.
Croft, William. 2000. Explaining Language Change. An Evolutionary Approach.
Harlow, England: Longman.
Dressier, Wolfgang U. 1980. "A semiotic model of diachronic process
phonology". Wiener Linguïstische Gazette 22/23.31-94.
Garcia, Erika  1990. "Reanalysing actualization and actualizing reanalysis".
Historical Linguistics 1987. Papers from the 8th International Conference
on Historical Linguistics ed. by Henning Andersen & Konrad Koerner,
141-160. (Current Issues in Linguistic Theory, 66.) Amsterdam &
Philadelphia: Benjamins.
Garcia, Erica  1997. "Syntactic diffusion and the irreversibility of linguistic
change: personal a in Old Spanish". Sprachwandel und Sprachgeschichte.
Festschrift für Helmut Lüdtke zum 65. Geburtstag hrsg. Jürgen Schmidt-
Radefeldt und Andreas Harder, 32-48. Tübingen: Gunter Narr Verlag.
Halle, Morris. 1962. "Phonology in Generative Grammar". Word 18.54-72.
Reprinted in The Structure of Language: Readings in the Philosophy of
Language ed. by Jerry A. Fodor & Jerrold J. Katz. Englewood Cliffs, N.J.:
Prentice-Hall, Inc.
Hamming, Richard W. 1980. Coding and Information Theory. Englewood
Cliffs, N.J.: Prentice-Hall Inc..
Harris, Alice & Lyle Campbell. 1995. Historical syntax in cross-linguistic
perspective. (Cambridge Studies in Linguistics, 74.) Cambridge: University
Press.
Heine, Bernd & Tania Kuteva. MS. World lexicon of grammaticalization.
Cambridge: Cambridge University Press.
Kaufman, Terrence & Sarah Grey Thomason. 1988. Language Contact,
Creolization, and Genetic Linguistics. Berkeley & Los Angeles & London:
University of California Press.
248 HENNING ANDERSEN

Kurylowicz, Jerzy. [1947] 1979. "Sur les procès dits 'analogiques'. Acta
Linguistica Hafniensia 5.15-37.
Lightfoot, David W. 1991. How to set parameters: arguments from language
change. Cambridge, Mass.: The MIT Press.
Lightfoot, David. 1999. The Development of Language: Acquisition, Change,
and Evolution. Malden, Mass.: Blackwell.
Milroy, James. 1992. Linguistic Variation and Change. Oxford: Blackwell.
Milroy, James. 1993. "On the social origins of language change". Historical
Linguistics. Problems and Perspectives ed. by Charles Jones, 215-236.
London & New York: Longman.
Newmeyer, Frederick J. 1998. Language Form and Language Function.
Cambridge, Mass.: The MIT Press.
Rittel, Teodozja. 1975. Szyk czlonów w obrebie form czasu przeszlego i trybu
przypuszczajqcego. Wroclaw & Warsaw & Cracow & Gdańsk: Ossolineum.
Sapir, Edward. [1921] 1949. Language. An Introduction to the Study of Speech.
New York: Harcourt, Brace & World.
Shapiro, Michael. 2001. "Markedness, causation, and linguistic change: a
semiotic perspective". In this volume, 187-202.
Stern, Gustaf. 1931. Meaning and Change of Meaning. With special reference to
the English language. Bloomington: Indiana University Press.
Tabor, Whitney & Elizabeth Closs Traugott. 1998. "Structural scope expansion
and grammaticalization". The Limits of Grammaticalization, ed. by Anna
Giacalone Ramat & Paul J. Hopper, 227-270. Amsterdam & Philadelphia:
Benjamins.
Tauli, Valter. 1958. The structural tendencies of languages. (Annales
Academiae Scientiarum Fennicae, ser. , torn. 115,1.) Helsinki: Suoma-
lainen Tiedaakatemia.
Timberlake, Alan. 1981. "Dual reflexes of *dj and a morphological constraint on
sound change". International Journal of Slavic Linguistics and Poetics
23.25-54.
Tabor, Whitney & Elizabeth Closs Traugott. 1995. Studies in scope expansion
and grammaticalization". The Limits of Grammaticalization, ed. by Anna
Giacalone Ramat & Paul J. Hopper, 229-272. (Typological Studies in
Language, 37.) Amsterdam & Philadelphia: Benjamins
Traugott, Elizabeth Closs. MS. "From etymology to historical pragmatics". Paper
presented at the conference "Studies in the History of English Language",
UCLA, May 26-28, 2000.
Vennemann, Theo. 1988. Preference laws for syllable structure and the
explanation of sound change. Berlin & New York & Amsterdam: Mouton
de Gruyter.
Weinreich, Uriel & William Labov & Marvin I. Herzog 1968. "Empirical
foundations for a theory of language change". Directions for Historical
Linguistics. A Symposium ed. by Winfred P. Lehmann & Yakov Malkiel,
95-188. Austin & London: University of Texas Press.
Zipf, George. 1935. The Psychobiology of Language. New York: Houghton
Mifflin.
 GENERAL INDEX

A diglossia 9, 17, 95, 99-102, 106, 109,

actualization 3, 10, 14, 17, 31-34, 52, 61, 112-14, 116-117, 170, 175-80, 226, 231
71-72, 79, 82, 92, 95, 99, 101, 104, drift 4, 9, 32, 37, 52, 101, 115-17, 127-28,
109-11, 115, 117, 143-44, 156, 162, 166, 225, 228, 245-46
169-70, 175-76, 182-83, 203-4, 207, 214,
217-18,225-48 E
actuation 4, 79, 82, 91,92, embedding problem 138, 143, 226
ambiguity 5, 95, 103, 104, 106, 111-16, 145, evolutive change 9, 15, 101, 116, 143-66,
166, 200, 213, 233-35 169-83, 203-18, 233-40, 242, 246
analogy 36, 69, 71-72, 204, exclusive opposition 39-50
appropriateness
see norms F
asymmetrical relation figure and ground 45, 50, 128
see markedness frequency 4-8, 11-12, 16-17, 21-22, 25,
32-33, 47, 49-52, 63-64, 66-67, 74-75,
 82, 90, 96, 120, 123, 126-28, 131, 137-39,
base grammar 17, 231-32, 236-37, 239 165-66, 206, 210-11, 216, 232, 237-39,
243-44
 functionality, functionalism 7, 96, 99, 107,
causality 113, 115, 127, 134, 147-50, 153, 158,
see motivation, teleology 170-71,173, 176-77, 181, 195, 203-18
change from above 4, 14, 34, 84, 239
change from below 9, 14, 35, 235 G
change, lexical 61, 63-75, 119-39, 229-31 gradation, gradient
morphosyntactic 6, 7, 31, 32, 81-90, see hierarchy
103-11,229-31,234 grammaticalization 5-6, 10, 13, 85, 144, 156,
phonological 15, 31, 33, 59, 63, 229-31, 164-67, 179, 181-82, 203, 241-42
234
syntactic 6, 9, 35, 37, 203-18, 226-27, H
229-31 hierarchy 7, 10-13, 33, 38, 42, 82, 115-16,
see also contact change; evolutive change; 171-76, 179, 181-82, 198, 204-7, 237
motivation, pragmatic hypercorrection 20, 60, 71, 236
clarification changes 14, 15, 239
contact change 3, 8-9, 15, 33, 35, 61, 69-71, I
74-75, 95, 99, 101, 115-16, 120, 230-31, inclusive opposition
239 see markedness
contrast indeterminacy
see opposition see ambiguity
convention 17, 28, 51, 135-38, 147, 167, 188 innovation
see change
250 INDEX

innovation 1-2, 9-10, 14-16, 31-35, 45, 79, productivity 1-2, 33, 59-61, 64, 67-68, 71,
84, 91, 96, 99, 108, 111, 127-30, 145, 167, 73, 95, 166, 239
172, 178-79, 181-82, 200, 204, 225-26,
228-33, 237-39 R
rank, ranking 3-4, 11, 81, 234
L realism 9, 193-94
lexical diffusion 8, 63, 67, 69, 71 reanalysis 5, 15-18, 51, 86, 95-96, 98-99,
104, 112, 115-16, 145, 152, 156, 159, 166
M 169, 207, 225-27, 231-35, 237-38, 240-
markedness agreement 3-4, 6-9, 13-14, 17, 41,244
27-31, 36-37, 49-50, 52, 79-80, 82, 84- reinterpretation
85,89-90,92, 109, 111, 116,237-38 see reanalysis
markedness shift 9, 16, 51, 96-97, 115, 135,
212, 217, 238-39, S
markedness 1-8, 10-11, 13-18,21-57, 165, S-curve 5, 10, 16, 96, 120, 226
169-85, 187, 199-200, 203-18, 225, 237- simplicity
39, 243-46 see speech processing
motivation 9, 15, 26, 33, 65, 69, 89-90, 92, speech processing
95, 120, 144, 165, 229-30, 238-39, processing effort 7, 49-50, 81, 84, 129,
242-43 165, 199-200, 206-7, 212-14
internal: see evolutive change processing strategies 7, 10, 115-16
external: see contact change
pragmatic 79, 84, 91, 108, 111, 175-76 Τ
teleology 9, 17, 18,68,71
Ν typology 11, 90-92, 98-99, 103, 115, 117,
naturalness 4, 12, 49, 67, 79, 84, 89, 143, 225, 240-41
193-94, 204-5, 212, 245
nominalism 9, 25, 193-94 U
norms 15-17, 33, 84, 95, 102, 121-23, usage rules 1, 18, 33-34, 51, 232, 236-39
125-26, 133, 137-38, 232, 234-36
V
 variation, variants 8-18, 34-36, 44, 49-51,
obscuration 65, 67, 69-70, 75, 87-88, 119-21, 123,
see clarification 127-28, 130, 139, 164, 169, 175, 180-81,
opposition 5-6, 11-14, 21, 27, 34, 37-52, 60, 195-96, 199-200, 203, 206-7, 212, 216,
89-90, 96-98, 103, 113, 128-29, 173, 182, 225, 228-30, 232, 234, 236-40, 243-45
191,205-7 lexical 2, 8, 15,34-35
optimality 193, 199, 205-6, 212 morphosyntactic 2, 30-32, 44, 81-90,
100-104, 169-76
Ρ phonological 2, 22, 31,49
perceptual factors pragmatic 6-7, 9, 16, 105-10, 120-39,
see speech processing 179-82
polarity 34-35, 41,49 syntactic, 96, 100, 105, 107-11, 176-79
process
see change