The Syntax-Morphology Interface
A Study of Syncretism
Matthew Baerman, Dunstan Brown & Greville G. Corbett
Cambridge : University Press, 2005
Cambridge Studies in Linguistics, 109
Hardback. xix+281 pages. £66.00. ISBN-13: 978-0521821810*
Reviewed by Laure Gardelle
École Normale Supérieure de Lyon
In this study, the authors investigate cases in which a single morphological form serves several syntactic functions – a phenomenon known as syncretism. For example, the Greek adjective for “wise”, when it takes neuter agreement, is realised as soph-on both in the nominative singular and accusative singular. Consequently, there is no longer a one-to-one correspondence between morphological and syntactic features.
Syncretism only concerns inflections, which excludes homonyms (such as bank in the sense of financial institution vs. the bank of a river) as well as cases in which forms are identical but are not expected to show inflectional difference (e.g. the form book in English does not reflect syncretism of the nominative / accusative /... cases because there is no morphological marking of those cases in the language). The authors take a broad approach to the phenomenon: they include both cases that result from blind phonological change and those that originate in a more complex morphosyntactic readjustment.
The word “syncretism” itself is not originally a linguistic concept. It was first used in the religious domain, to refer to a mixture of different elements or viewpoints, and might even be traced back to a Greek word meaning “union”, which referred to the practise that feuding Cretan communities had of laying aside their differences to unite against a common enemy. It was extended to linguistics by Pott, in 1836, to refer specifically to the diachronic collapse of originally distinct inflectional forms, as a result of the merger of either the forms or their underlying functions. Use of the word in linguistics did not become widespread until the 1890s, however, and it was only from the 20th century, with structuralist theory (Hjelmslev, Jakobson, ...), that it was applied to synchrony.
This study is ground-breaking in that it adopts a truly cross-linguistic approach. It is not based mainly on Indo-European languages, unlike much of the existing literature, but on thirty genetically diverse languages. More specifically, it uses the data of the Surrey Syncretisms Database (compiled by the authors), which records all instances of syncretism in those thirty languages and provides a total 1,256 entries. Moreover, the analyses lead the authors to propose a new theoretical model for syncretism, based on the Network Morphology framework.
The structure of the book follows an inductive approach, from facts to theory. Following the introduction (Chapter 1), Chapter 2 lays out the typological parameters and defines basic concepts. Chapter 3 offers an analysis of the empirical data, which then enables theoretical modelling, from an investigation of existing theories and their shortcomings (Chapter 4) to the model proposed by the authors (Chapter 5). Untypically, readers are invited to test the theory for themselves, thanks to a range of supporting materials which complement the bulk of the book. At the end of the study are six appendices with samples, fuller versions of which are available online at: http://www.surrey.ac.uk/LIS/SMG/Syncretism. Readers can thus check that the examples chosen by the authors in the course of their analyses are truly representative of the data and not just “convenient” instances. In addition to this valuable resource, the Surrey Syncretisms Database can be searched online at: http://www.smg.surrey.ac.uk (readme files explain how to browse it). At the same address, readers can also find a database covering no less than 111 languages (instead of 30), which addresses a more specific phenomenon: syncretism of subject person marking in intransitive verbs.
A typology of syncretism reveals that it exhibits different types. With simple syncretism, two or more values are merged into a single morphological form. This can be compounded across different environments, resulting in nested syncretism. For instance, in the West Slavonic language Upper Sorbian, all nominals syncretise the dative, the locative and the instrumental in the dual, but a-stem nouns also syncretise the dative and the locative in the singular; the syncretic pattern of the singular can therefore be said to be nested within the larger syncretic pattern of the dual. A third type is contrary syncretism, in which pairings in each paradigm are mutually exclusive. In the Nilo-Saharan language Nuer, for example, in the singular, the noun for “egret” displays syncretism of the genitive with the locative, whereas in the noun for “girl”, the genitive is syncretic with the nominative. The existence of three different types of syncretism raises the question of the relationship between the feature values involved; according to the type, they stand in a hierarchical or a non-hierarchical structure.
Syncretism is also varied in its distribution: it can affect verbal affixes (conjugations) and features (voice, mood), as well as nominals (case, gender, person, number) and even negation when realised in the morphology. For instance, in Dongola Nubian, the verb for “drink”, in a negative context (“not drink”), shows syncretism of all persons except the third person plural (nimunun), whereas in a positive context, there is only syncretism of the second and third person singular on the one hand (nin), and of the first and second person plural on the other (ninu).
Another source of variety is that syncretic patterns may have different morphological realisations. Three criteria help describe them: regularity (whether the pattern is repeated across multiple exponents), directionality (whether one of the values of the form is its source and another its extension), and unmarkedness (whether the merging of the morphological forms results in use of the originally unmarked form).
Finally, instances of syncretism may be interpreted in three different ways. In some cases, they are the result of neutralisation. For instance, adjectives in Russian only inflect for gender in the singular, so that the syncretism of gender in the plural can be regarded as syntactic neutralisation. In other words, the lack of formal distinction in the plural reflects the irrelevance of the feature for syntax. A second interpretation equates syncretism with uninflectedness. For example, in Russian, pal’to (coat) is one of a small group of nouns that are undeclinable and so have no distinct number or case forms. Unlike cases of neutralisation, here differentiation would be syntactically relevant, but morphology does not respond to that need. For both neutralisation and uninflectedness, the underlying principle for syncretism is underspecification. The third interpretation, however, termed “canonical syncretism”, is more problematic: it involves the partial (as opposed to total) collapse of syntactically relevant feature distinctions. For instance, in Slovene, the l-participle marks masculine, feminine and neuter gender in the singular and the plural, but in the dual, the feminine and neuter forms are syncretised (the masculine form remains distinctive). Underspecification is insufficient to account for such syncretism, hence the need to search for a better theoretical model.
This model, proposed by the authors in the last chapter, is built from an in-depth, cross-linguistic analysis of various features: case, person, gender, number, tense, aspect, mood and polarity effects (a rare phenomenon, found for instance in the Somali definite article, for which -ta is the form for the masculine singular and the feminine plural, and -ka the form for the femine singular and the masculine plural). To give one example here, the authors show that for case, outside Indo-European languages, most occurrences of syncretism result in each of the marked core cases (the accusative and the ergative) taking the morphological form of a different peripheral case. These patterns show a sensitivity to basic syntactic relations, perhaps motivated by the need to differentiate between objects and agents. More generally, analysis of the data also shows that some of the features interact in syncretism. In the nominal domain, case and gender are more likely to syncretise than number; for example, gender syncretism is very common in the plural. In the verbal domain, agreement features on verbs are also more prone to syncretism than the tense, aspect or mood features.
Modelisation of syncretism is particularly difficult, because it must explain why some patterns are very common (these are therefore termed “natural” patterns) while others (termed “uncommon”) are rare; it must also remain open-ended, so as to allow for isolated, language-specific patterns that do not permit generalisations. The model proposed by the authors meets these requirements. It is built on the Network Morphology framework. This is an inferential (as opposed to a lexical) model, in that the morphological markings are not regarded as elements of the lexicon, but as relations between the lexical stems and their paradigm of word forms. The choice of a word form, therefore, is a matter of inference, that is, of rules. Network Morphology also considers that inflections do not necessarily increase the information given by a word; in that sense, it is a realisational (as opposed to an incremental) approach.
Another key concept in its analysis of morphology is that of default inheritance. Networks consist of nodes and connections between them; default inheritance means that information is automatically inherited from higher nodes unless it is specifically overridden. For instance, past participles in English are typically realised with an -ed ending, sharing the same form as the past; this morphological information is inscribed in the node VERB. But one subset of verbs shows -en (such as sown or been). For these, there is an intermediate node between VERB and SEW or BE: EN_VERB. That node registers a morphological realisation of the past participle as having an -en form; the rule overrides the general -ed rule, but all the other information from the VERB node is still valid, i.e. inherited by default. The default inheritance principle accounts for the varying degrees of morphological regularity: the higher the syncretism is located in the inheritance hierarchy, the more regularity there is.
The authors’ Network Morphology analysis of syncretism leads them to conclude that syncretism is not a minor morphological phenomenon, i.e. a small set of exceptions in the one-to-one correspondence between syntactic function and form. Instead, it shows that morphology, at least to some degree, exists independently of meaning, as confirmed by the fact that no case of syncretism is determined solely by semantics. Rather, syncretism is “morphology failing syntax”, and this can occur in three ways. One, as stated above, is underspecification; in this case, syncretism is due not to semantic naturalness, but to a tendency for the given morphological feature to syncretize. The other two ways are indexing, which involves an autonomous structure that cuts across morphosyntactic natural classes, and referrals, in which a default form is associated with one default primary function and is also used in another subset of lexical items or morphosyntactic contexts.
This book is a milestone in the understanding of syncretism, and will interest both specialists and non-specialists. First, the impressive amount of data allows for the truly cross-linguistic approach sought by the authors; in that respect, it lies in the wake of other invaluable Cambridge University Press publications, such as Person, Gender or Number (the last two written by G. Corbett, co-author of the present book). Secondly, the authors manage the feat of combining numerous, in-depth analyses with theoretical considerations, while remaining extremely clear and progressive. Only the theoretical presentation of the Network Morphology framework makes rather difficult reading for the non-specialist; but the authors go into more detail later on in the chapter. In addition, the three indexes at the end enable specific searches by author, language or subject – in the subject index, a few concepts could have been usefully added, especially referral, which is part of the theoretical framework, but it is otherwise very useful. Finally, 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.
* Paperback reissue, 2009. £23.99. ISBN-13: 978-0521102759.
Cercles © 2011