The Syntax-Morphology Interface: A Study of Syncretism

by Matthew Baerman, Greville G. Corbett, Dunstan Brown

Overview

Syncretism—where a single form serves two or more morphosyntactic functions—is a persistent problem at the syntax-morphology interface. It results from a 'mismatch', whereby the syntax of a language makes a particular distinction, but the morphology does not. This pioneering book provides the first full-length study of inflectional syncretism, presenting a typology of its occurrence across a wide range of languages. It will be welcomed by linguists interested in the relation between words and the larger units of which they are a part.

Product Details

ISBN-13:	9780521102759
Publisher:	Cambridge University Press
Publication date:	03/12/2009
Series:	Cambridge Studies in Linguistics , #109
Pages:	304
Product dimensions:	5.98(w) x 9.02(h) x 0.67(d)

About the Author

Matthew Baerman is Research Fellow in Linguistics at the University of Surrey.

Dunstan Brown investigates autonomous morphology, morphology-syntax interaction, and typology. His recent work has focused on describing and understanding different aspects of morphological complexity. After graduating with a BA in Modern Languages and a Master of Linguistics from the University of Manchester, he completed a PhD in Linguistics at the University of Surrey and worked there for many years before taking up a 50th Anniversary Chair at the University of York in 2012.

Greville G. Corbett is Distinguished Professor of Linguistics at the University of Surrey.

Read an Excerpt

Cambridge University Press
0521821819 - The Syntax-Morphology Interface - A Study of Syncretism - by Matthew Baerman, Dunstan Brown and Greville G. Corbett
Excerpt

1 Introduction

We might expect that a language's sentence-structure and word-structure would mesh rather straightforwardly. For instance, if the syntax of a particular language distinguishes different arguments of the verb, and the morphology distinguishes different cases, it seems natural to assume that the two systems will line up. In real languages the situation is often more complex. One of the most persistent and interesting problems at this syntax-morphology interface is syncretism. As a first informal characterization, syncretism is the situation where the morphology 'lets down' the syntax. To make that more concrete, let us take some Russian examples. (Normally we shall give detailed glosses but here, so as not to build our assumptions into the examples, we shall give only the basics.)

(1)	Maša čitaet knigu
	Masha reads book
	'Masha reads a book.'
(2)	Na stole ležit kniga
	on table lies book
	'A book lies on the table', 'There's a book on the table.'

These sentences are representative, in that they show what they seem to show. Russian distinguishes subject from object in its syntax, by a variety of means (for example, the verb agrees with its subject but not with its object). This appears to be reflected in the inflectional morphology. In (1) we have knigu 'book', in the accusative case, as opposed to kniga in (2) when it is in the nominative. (Similarly Maša in (1) is in the nominative.) As any reasonable non-linguist would expect, the two systems work hand in hand to distinguish subject and object and to facilitate the task of the hearer.

But now compare:

(3)	Maša čitaet pis ′ mo
	Masha reads letter
	'Masha reads a letter.'
(4)	Na stole ležit pis ′ mo
	on table lies letter
	'A letter lies on the table', 'There's a letter on the table.'

Here the syntactic structures are as in the earlier examples, but the noun fails to show the expected morphological distinction. We say that pis′mo 'letter' in (3) and (4) shows syncretism of nominative and accusative case.

We understand syncretism therefore as a mismatch between syntax and morphology. We know that Russian syntax requires reference to subject and object. This is reflected in the morphosyntactic category of case, which distinguishes nominative and accusative (as in (1) and (2) above). However, the morphology of pis′mo in (3) and (4) fails to make this distinction. The key components of the definition are:

1. a morphological distinction which is syntactically relevant (i.e. it is an inflectional distinction)
2. a failure to make this distinction under particular (morphological) conditions
3. a resulting mismatch between syntax and morphology.

Thus syncretism is the failure to make a morphosyntactically relevant distinction.

A good way to look at it is to say that examples (1) and (2) set up the expectation that there will be two forms of pis′mo in (3) and (4). Syncretism is the breaking of that expectation; the nominative singular and the accusative singular of pis′mo are identical. As Spencer (1991: 45) puts it 'a single inflected form may correspond to more than one morphosyntactic description.' A similar definition is: 'Identity in form between two grammatically different inflections' (Trask 1997: 215).

There are various questions to be asked about our example pis′mo 'letter'. For example, is it an odd exception, going against the general trend? No, there are thousands of nouns in Russian which behave similarly; there are also many thousands of the kniga 'book' type. And pis′mo does not fail to draw certain other inflectional distinctions. It has a distinct locative, as shown by: v pis′me 'in the letter'. We might think that we could simply divide Russian nouns into those which distinguish nominative and accusative and those like pis′mo which do not. But here we find that kniga 'book', which marks the distinction in the singular, fails to do so in the plural (both forms are knigi).

We shall investigate which distinctions can fail to be drawn. We have seen an instance where case is involved, but there are several other inflectional distinctions which may be treated similarly. We shall ask when this occurs. It may be that our expectation is based on some lexemes (like kniga) and is not met by others (like pis′mo). Or it may be that other features provide the circumstances (kniga 'should' distinguish nominative and accusative, as the singular shows, but it fails to do so in the plural). It is not self-evident that all the phenomena which fall under the broad umbrella of syncretism should be modelled in the same way, and so we shall look carefully at the arguments for particular ways of treating different instances of syncretism.

While we shall be inclusive in our coverage we shall ensure that we are indeed dealing with inflection. That is, the expectation of a difference in form must arise from the syntax and morphology of the given language. Languages may have a distinct locative case, as in Russian, but there is no language-internal evidence for such a morphological case in languages like English. We shall not, therefore, treat English book as syncretic between nominative and locative. Nor shall we be concerned with derivational morphology. There are, of course, interesting coincidences of form in derivational morphology, but we do not have the same expections of consistency and completeness for derivational morphology as for inflection. Syncretism is also distinct from lexical homonymy, where there is a single form with distinct meanings (as in bank (of river) and bank (financial institution); this could be characterised as the lexicon letting down the semantics.

1.1 History of the notion

The term 'syncretism' ultimately descends from the Greek συγκρητισμóς 'union, federation of Cretan communities' (Liddell and Scott 1996), referring to the practice that the continually feuding Cretan communities had of laying aside their differences and banding together in the face of a common enemy. In post-classical times Erasmus of Rotterdam reintroduced the term, using it to designate 'the coherence of dissenters in spite of their difference of opinions, especially with reference to theological divisions' (Herbermann, Pace, Pallen, Shahan and Wynne 1907-18). At some point, through learned folk etymology, the term was confounded with συγκɛράννυμι 'to mix' and its derivatives (e.g. σύγκρατoς 'mixed'), so acquiring a more general meaning of a mixture of originally different elements or viewpoints, typically with respect to religion (Grimm and Grimm 2004). In the nineteenth century, use of the term became fashionable in other realms as well. Pott (1836) is generally credited with introducing the term into linguistics (Curtius 1863: 160; Wackernagel 1924: 32; Hjelmslev 1935-7: 59),¹ where it is used to mean the diachronic collapse of originally distinct inflectional forms, either through merger of the forms, or through the merger of their underlying functions. Pott himself uses the term only in passing (p. 638),² but it is more generously applied in the index (p. 792) - written not by Pott but by Heinrich Bindseil - where it refers the reader to sections discussing the merger (typically through sound change) both of case marking in nouns and person marking in verbs. In spite of this early use of the term, it did not enter into general use until the 1890s (Meiser 1992: 212, fn. 2). In the twentieth century, this diachronic approach to syncretism was recast in synchronic terms within the framework of structuralist linguistics (Hjelmslev 1935-7: 60; Jakobson 1936 [1971]: 67). On this approach, syncretism involves the contrast not between an earlier and a later stage of a language, but between an underlying system and its concrete realization. This is how we understand syncretism in the present work, though of course we shall not ignore the diachronic processes that may have led to such a state of affairs.

1.2 Delimiting the notion of syncretism

Some authors have suggested that the term 'syncretism' should be reserved for the products of certain diachronic developments. Roughly speaking, the diachronic merger of forms within an inflectional paradigm can have two sources, either as the result of blind phonological change, or the result of a more complex morphosyntactic readjustment. Blind phonological change can be illustrated by the merger of nominative and accusative singular in first declension nouns in Vulgar Latin (5), a result of the regular loss of word-final -m. Note that in other declension classes the two forms remained distinct, as their difference was not solely due to the presence or absence of -m.

(5) Nominative/accusative singular in Vulgar Latin (Coleman 1976: 50-4)

Image not available in HTML version

Merger resulting from morphosyntactic change is illustrated by the development of the Proto-Indo-European dative and locative singular, which were combined in Ancient Greek into the case traditionally known as the dative (6). By the laws of sound change that applied between Proto-Indo-European and Ancient Greek, these two forms should be distinct in o-stem

(6) Dative/locative singular in Ancient Greek (Buck 1933: 180-5)

		Proto-Indo-European	>	Greek
o-stem	DAT SG	-ōi		-ōi
	LOC SG	-oi, -ei
consonant stem	DAT SG	-ei		-i
	LOC SG	-i

and consonant-stem nouns, so their merger cannot be attributed to phonological change. Nor can it be attributed to the outright loss of one of the cases, since both the dative and locative have contributed to the syncretic forms:

• In the o-stems, the Greek form descends from the Proto-Indo-European dative singular. The Proto-Indo-European locative singular is marginally preserved in some adverbialized forms, such as oikoi 'at home', originally from oikos 'house'.
• In consonant stems, the Greek form descends from the Proto-Indo-European locative singular. The original dative form has been lost completely.³

Thus, the Vulgar Latin example represents a superficial inflectional homophony, the result of blind sound change. The Ancient Greek dative singular results from a blending of both the forms and the functions of the Proto-Indo-European dative and locative, the result of some fundamental reanalysis of the system of morphosyntactic oppositions. Meiser (1992: 190) and Luraghi (2000) both suggest that the term 'syncretism' be restricted to the second kind of change. Luraghi suggests that the product of phonological change should be called 'homophony' while Meiser (p. 190) proposes reviving the term 'synemptosis', used by ancient Greek grammarians for situations where a morphological expression belonged to several grammatical categories.

However, though this distinction is one of undoubted theoretical significance, in practice it is often difficult to draw the line. Consider the merger of ablative and dative singular in second declension nouns in Latin (7). This seems to have been the result of two independent sound changes. On the one hand, final consonants were lost after a long vowel, so that ablative

(7) Ablative and dative singular in Latin (Buck 1933: 176, 181)

Image not available in HTML version

singular -ōd became -ō (Buck 1933: 157). On the other hand, long diphthongs were monophthongized, losing their second element, so that dative singular -ōi became -ō (Buck 1933: 90). The first change was quite regular and is also found in the first declension (thus, ablative singular -ād became -ā). However, the development of the original long diphthongs in Latin turns out to have been erratic. Sometimes they were monophthongized, but in other contexts they were shortened; the conditions which determined which change took place remain obscure (Leumann 1977: 271-2).⁴ While the dative singular of the second declension underwent the first change, the dative singular of the first declension underwent the second change, with -āi becoming -ai (orthographically -ae), though there is evidence from early inscriptions that monophthongization took place sporadically as well (Buck 1933: 176). Consequently, there is no syncretism between ablative and dative singular in the first declension.

Although the merger of ablative and dative singular can be portrayed as the result of sound change, a key element remains unaccounted for, namely why did the long diphthong develop one way in one declension class and another way in the other? For some reason syncretism was favoured in the second declension but not in the first, in a way that does not obviously follow from phonological developments. Such examples show that there is not always a clear distinction between phonological and morphological change, much less a way to classify phenomena whose history remains unknown. It seems useful, then, to retain 'syncretism' as a cover term that will apply to all instances of inflectional homophony, regardless of their origin or interpretation; indeed, this is how the term was first used by Pott (and Bindseil) in 1836.

1.3 Scope of the investigation

1.3.1 Typological methodology

The core of the present book is a cross-linguistic typological investigation of syncretism, with two complementary goals. On the one hand, we explore the logical space of syncretism: what features may be involved, and what sort of patterns do these describe? On the other hand, we have aimed for a diverse sampling of the world's languages. In particular, we have brought the evidence of non-Indo-European languages to bear, since these remain relatively under-represented, a legacy of the fact that the notion of syncretism was born in comparative Indo-European studies. To ensure genetic breadth we followed, in part, the selection of languages used for the World Atlas of Language Structures (Haspelmath, Dryer, Gil and Comrie 2005). While genetic breadth can be achieved by the application of apriori criteria, typological breadth - in order to see how many of the logical possibilities are in fact attested - can be attained only by sifting through masses of information, unconstrained by prior notions of what one may find. This is a task which will never and can never be completed, but the present study represents at least an introduction.

1.3.2 Selection of forms

In this book we focus our attention on syncretism between inflected whole word forms. In principle, one could speak of syncretism between the individual components of inflected words, e.g. by comparing prefixes to prefixes, stems to stems and suffixes to suffixes, regardless of whether there is homophony at the level of the word form as a whole. For example, in the Nilo-Saharan language Mursi (8), the verb stem is identical in the first and second singular of the indefinite aspect (baaγio), and in the first and third singular of the indefinite aspect (baaka). However, in neither aspect is there syncretism at the level of the whole word, because 1SG is distinguished by the prefix ka-.

Stump (2001: 217) terms this block syncretism, because the pattern obtains

(8) Singular indicative forms of the verb 'eat' in Mursi (Turton 1981: 341-2)

	indefinite aspect		definite aspect
1SG	ka-	baaγio	wa-	ka-	baaka
2SG		baaγio	wa-		baaku
3SG		baato	wa-		baaku

within a given block of rules. So long as a word form is transparently segmentable, such an approach has the advantage of widening the field of investigation. But the question of segmentation into components (rule blocks, morphemes or formants) is not always easy to resolve. For example, take Pike's (1965) analysis of the six present tense forms and the infinitive of the German verb sein 'to be' (9). At the level of the whole word, these show 1PL/3PL syncretism alone.

Within these forms, however, eight distinct patterns of identity can be

(9) German sein 'to be'

Image not available in HTML version

isolated, as shown in (10): the elements b-, s-, -ei-, -ist-, -in-, -i-, -n- and -t (note that orthographic d in (10) is indistinguishable from t in word-final

(10) Patterns of identity in German sein (adapted from Pike 1965: 198)

Image not available in HTML version

position) each combine different person-number values.

Whatever the merits of such an analysis, it is not one which is compatible with most morphological models. The drawback to such an approach is that the more a word is broken up into components, the more the resulting parts are peculiar to a specific analysis. This is not desirable in the context of a large-scale typological investigation such as ours. On the other hand, the status of the whole word, while hardly self-evident, is nevertheless more uncontroversial than that of such elements as morphemes or formants. By limiting the investigation to whole word forms, we aim to keep the typological part of the investigation theory-neutral.

1.4 Accidental versus systematic homophony

The focus of this book is inflectional morphology: what interests us are instances of inflectional homophony that might be seen as systematic, that is somehow represented in morphological structure. However, it is undoubtedly the case that not all instances of homophony within inflectional paradigms are morphologically encoded as such. Consider the Russian forms in (11).

For the noun mesto 'place', the genitive singular form is identical to the

(11) a. stem-stress 'place' b. end-stress 'wine'

	orthographic	phonetic	orthographic	phonetic
NOM/ACC SG	mesto	'mje.stə	vino	vji.'no
GEN SG	mesta	'mje.stə	vina	vji.'na

nominative/accusative singular, while for vino 'wine', the genitive singular is distinct. However, the difference between the two nouns is easily explained in phonological terms. Russian has a general rule whereby /a/ and /o/ are distinguished only under stress. In this declension class, the nominative/accusative singular ending is -o and the genitive -a, as reflected in the orthography. For vino, these endings are stressed, and remain distinct, while for mesto they are unstressed, and hence homophonous. The collapse of the genitive with the nominative/accusative is a superficial by-product of phonology and need not be reflected in a morphological analysis.

In general we have excluded such obvious examples of accidental homophony from consideration. However, it should be borne in mind that the question is seldom so clear-cut. On the one hand, a pattern of syncretism may be restricted to a particular phonological environment without there being any generally applicable phonological rule that would account for it. On the other hand, there is evidence that originally accidental homophony may be ascribed by speakers to a morphological rule, a reanalysis which remains covert until revealed by diachronic processes (see Chapter 4: §4.5.1). Therefore, if we have erred in our selection of material for presention, it has been on the side of inclusiveness.

1.5 Using this book

1.5.1 Supporting materials

The text is supported by a range of additional materials which have been made available to the reader. There is an annotated bibliography of syncretism (Baerman 2002a), which contains details of 100 items. This is freely available online. Since this bibliography is available, we can restrict the references section in the book to those items which are discussed in the text.

It is important that our account of syncretism is based on a wide range of languages rather than on the few usual suspects. Our work is grounded on an investigation of syncretism in a sample of genetically diverse languages. These data are available in the Surrey Syncretisms Database (Baerman, Brown and Corbett 2002a), which records all instances of syncretism in thirty genetically diverse languages, comprising 1,256 separate entries. This can be searched online. Its rationale is explained in Brown (2001). The provision of this database means that the reader can frame hypotheses about syncretism and investigate them on-line. The essential information for using the database is provided in readme files at the web address given. There is a second database covering a more limited phenomenon over a much broader range of languages,

---

© Cambridge University Press

Table of Contents

Preface; List of abbreviations and symbols; 1. Introduction; 2. Characteristics of syncretism; 3. Cross-linguistic typology of features; 4. Formal representation; 5. Formal framework and case studies; 6. Conclusion; References; Indexes.

Editorial Reviews

'This book is a milestone in the understanding of syncretism, and will interest both specialists and non-specialists … the existing literature on syncretism is widely documented and discussed at length before the Network Morphology model is proposed, so that the theoretical inquiry matches the quality of the empirical data. Cercles

From the Publisher