Invited presentation at plenary round table “Formal and functional approaches to grammatical variation”, 55th Annual meeting of the Societas Linguistica Europaea (SLE 2022), Bucharest, 24–27 August 2022.
I offer two general suggestions: we should be more ambitious in what we try to explain; and we sh... more I offer two general suggestions: we should be more ambitious in what we try to explain; and we should devote more effort to careful measuring of the linguistic data, when we evaluate claimed explanations. Moreover, while explaining variation is a valid goal, I note that the shape of variation in one phenomenon, P, can often be used to test proposed explanations for another phenomenon, Q. Illustrating these points is made difficult by the fact that formalists and functionalists tend to tackle different issues. One area where interests overlap is agreement. Hence I consider three types of variation in agreement: (i) the dramatic differences between languages according to how pervasive agreement is; (ii) the extreme variation of agreement targets (probes), including within single parts of speech (syntactic categories); (iii) fine-grained variation among alternative agreements within languages and within constructions.
Given (i), the variation in how pervasive agreement is (phenomenon P), we seek to explain why languages can have agreement at all (phenomenon Q). A suggested explanation for Q is that agreement facilitates reference tracking; to evaluate this explanation, the variation within agreement targets (ii), that is, phenomenon P, proves a useful way to measure and thus test the proposed explanation for Q. The suggested explanation is found wanting (Nichols 2018, Fedden in press, Feist 2020).
Turning to explanations of variation itself, fine-grained variation in agreement (iii) has been tackled by formalists from differing perspectives (some relying on different syntactic structures, e.g. Polinsky 2016, others giving morphosyntactic features an important role, e.g. Borsley 2016, Landau 2016, Sadler 2016). Much of this variation involves Agreement Hierarchy effects (Corbett in press); here Wechsler & Zlatić (2003, and subsequent papers) make a noteworthy contribution, and see more recently Shen (2019), Smith (2021) among many others. We need explanations for (i) why alternatives are possible, and (ii) the factors which favour one or other outcome. This is an area where there are rich data to measure, and canonical approaches play an important role in specifying valid baselines and criteria (Round & Corbett 2020).
Ambitious explanation should be based on solid evidence, the result of careful measuring. The existence of agreement, and the variation we find, are deeply puzzling. But this is an environment where linguists of different persuasions can contribute, where careful measurement is being undertaken, and where variation can be used as tool of explanation as well as a problem which itself is to be explained.
References
Bond, Oliver, Greville G. Corbett, Marina Chumakina & Dunstan Brown (2016), Archi: Complexities of agreement in cross-theoretical perspective. Oxford: Oxford University Press.
Borsley, Robert D. (2016), HPSG and the nature of agreement in Archi. In Bond et al., 118-149.
Corbett, Greville G. (In press), The Agreement Hierarchy revisited: the typology of controllers. To appear in Word Structure (special issue The many facets of agreement, ed. by Tania Paciaroni, Alice Idone and Michele Loporcaro).
Fedden, Sebastian (In press), Agreement and argument realization in Mian discourse. To appear in Word Structure (special issue The many facets of agreement, ed. by Tania Paciaroni, Alice Idone and Michele Loporcaro).
Feist, Timothy (2020), Nominal classification: Does it play a role in referent disambiguation? Studies in Language 44.199–239.
Landau, Idan (2016), DP-internal semantic agreement: A configurational analysis. Natural Language and Linguistic Theory 34. 975-1020. doi:10.1007/s11049-015-9319-3.
Nichols, Johanna (2018), Agreement with overt and null arguments in Ingush. Linguistics 56.845-863.
Polinsky, Maria (2016), Agreement in Archi from a minimalist perspective. In Bond et al., 184-232.
Round, Erich R. & Greville G. Corbett (2020), Comparability and measurement in typological science: the bright future for linguistics. Linguistic Typology 24.489-525.
Sadler, Louisa, (2016), Agreement in Archi: An LFG perspective. In Bond et al., 150-183.
Shen, Zheng (2019), The multi-valuation Agreement Hierarchy. Glossa: A Journal of General Linguistics 4(1), 46. doi: 10.5334/gjgl.585.
Smith, Peter W. (2021), Morphology-semantics mismatches and the nature of grammatical features. Berlin: De Gruyter Mouton. doi: 10.1515/9781501511127
Wechsler, Stephen & Larisa Zlatić (2003), The many faces of agreement. Stanford: CSLI.
Bookmarks Related papers MentionsView impact
Uploads
Papers
inflectional forms of the same lexical item are not related phonologically, has a special
place in morphology. Part of its importance is that it sets one of the outer bounds for the
notion ‘possible word’ in a human language. It provokes questions about how such
forms are to be treated in our theories, and how they are stored (Carstairs-McCarthy
1994). There has been considerable work on suppletion, particularly from Osthoff
(1899) onwards. Current interest in the topic is shown by the recent appearance of two
dissertations (Veselinova 2003 and Veselinovič 2003). While the body of research is
extensive, the range of languages investigated is rather restricted in many publications.
In order to stimulate further progress, we have constructed and made available a
database (Brown, Chumakina, Corbett and Hippisley 2004). We hope this will help to
put future research on a broader empirical base An annotated bibliography is now
available (Chumakina 2004); it contains over seventy entries on works written in five
different languages (English, French, German, Italian and Russian) and this will give
the reader a view of the literature.
inflectional forms of the same lexical item are not related phonologically, has a special
place in morphology. Part of its importance is that it sets one of the outer bounds for the
notion ‘possible word’ in a human language. It provokes questions about how such
forms are to be treated in our theories, and how they are stored (Carstairs-McCarthy
1994). There has been considerable work on suppletion, particularly from Osthoff
(1899) onwards. Current interest in the topic is shown by the recent appearance of two
dissertations (Veselinova 2003 and Veselinovič 2003). While the body of research is
extensive, the range of languages investigated is rather restricted in many publications.
In order to stimulate further progress, we have constructed and made available a
database (Brown, Chumakina, Corbett and Hippisley 2004). We hope this will help to
put future research on a broader empirical base An annotated bibliography is now
available (Chumakina 2004); it contains over seventy entries on works written in five
different languages (English, French, German, Italian and Russian) and this will give
the reader a view of the literature.
In this volume, senior scholars and junior researchers highlight novel perspectives on conceptualizing morphological complexity, and offer concrete means for measuring, quantifying and analysing it. Examples are drawn from a wide range of languages, including those of North America, New Guinea, Australia, and Asia, alongside a number of European languages. The book will be a valuable resource for all those studying complexity phenomena in morphology, and for theoretical linguists more generally, from graduate level upwards
Part I takes four key languages from diverse families - Nakh-Daghestanian, Gunwinyguan (Australian), Uralic and Indo-European - as examples of languages in which periphrasis poses particular problems for current linguistic theories. Part II views periphrasis in different contexts, determining its place within the morphological and syntactic systems of the languages it is found in, its relations to other linguistic phenomena, and the typological variation represented by periphrastic constructions. Treating periphrasis as a morphological and syntactic phenomenon at the same time and applying the criteria worked out within the Canonical Typology approach allows linguists to view periphrasis as a family of phenomena within a typological space of syntactic constructions used to fulfil grammatical functions.
Although such gaps have been known to us since the days of Classical grammarians, they remain poorly understood. Defectiveness contradicts basic assumptions about the way inflectional rules operate, because it seems to require that speakers know that for certain words, not only should one not employ the expected rule, one should not employ any rule at all. This is a serious problem, since it is probably safe to say that all reigning models of grammar were designed as if defectiveness did not exist, and would lose a considerable amount of their elegance if it were properly factored in.
This volume addressed these issues from a number of analytical approaches - historical, statistical and theoretical - and by using studies from a range of languages.
features and establishes new bases for their use in the study and
understanding of language. Features are fundamental components
of linguistic description: they include gender (feminine, masculine,
neuter); number (singular, plural, dual); person (1st, 2nd, 3rd); tense
(present, past, future); and case (nominative, accusative, genitive,
ergative). Despite their ubiquity and centrality in linguistic description, much remains to be discovered about them: there is, for example, no readily available inventory showing which features are found in which of the world's languages; there is no consensus about how they operate across different components of language; and there is no certainty about how they interact. This book seeks both to highlight and to tackle these problems. It brings together perspectives from phonology to formal syntax and semantics, expounding the use of linguistic features in typology, computer applications, and logic. Linguists representing different standpoints spell out clearly the assumptions they bring to different kinds of features and describe how they use them. Their contrasting contributions highlight the areas of difference and the common ground between their perspectives. The book brings together
original work by leading international scholars. It will appeal to linguists of all theoretical persuasions.
The grammar of a language can also force us to classify. When we use a pronoun in English we have to choose between ‘he’ for males, ‘she’ for females and ‘it’ for inanimates. This type of categorization runs along the lines of biological sex. In a language with a gender system all nouns are treated as either masculine or feminine — even those nouns whose meanings have nothing to do with biological sex.
Quite a different approach is taken by languages with a classifier system. Here categorization is based on fine-grained meaning, involving shape, function, arrangement, place or time interval. One such language is Kilivila (an Oceanic language spoken on the Trobriand Islands in Papua New Guinea), which has at least 177 distinct classifiers.
Mostly a language will have only one system or the other, gender or classifiers, but in a few interesting cases we find both systems together. A key language for this project is Mian, a Papuan language spoken by 1,700 people in Papua New Guinea. Mian has both a gender system and a system of classifiers in the form of prefixes on verbs of object handling or movement (e.g. give, take, put, lift, throw, fall).
This database documents this almost unresearched phenomenon: how it works, how it varies, what it does, and where it comes from. Since the key languages investigated are highly endangered, the project was timely, and as a by-product, resulted in some partial primary documentation work. In most Oceanic languages of the North West Solomonic subgroup (spoken in Bougainville and the western Solomon Islands), some use is made of apparent possessive morphology to index subject and encode aspect on verbs.
The bibliography comprises collections and special issues devoted to agreement (section A), monograph-length studies of agreement , mainly studies of agreement in particular languages (section B), articles and book chapters devoted to agreement (section C). There is a good deal of material on agreement in the Slavonic languages which is given separately (section D).
This bibliography does not in general include works which may refer to agreement morphology in connection with language acquisition, language reconstruction or sign language.
comparing the relationship between
grammatical ambiguity (syncretism) in nouns,
as represented in a default inheritance
hierarchy, with textual frequency distributions.
In order to do this we consider a language with
a reasonable number of grammatical
distinctions and where syncretism occurs in
different morphological classes. We
investigated this relationship for Russian
nouns. Our results suggest that there is an
intricate relationship between textual
frequency and inflectional syncretism.
Uninflectedness is nameworthy because the phenomenon is unexpected and significant. We should ask, then, why we expect inflectedness, and why its lack is significant. This leads us to distinguish it from related phenomena, including syncretism and defectiveness. And while full uninflectedness has a history of discussion, we should not treat it as an absolute: rather, there is an interesting canonical scale from fully inflected to uninflected. Items may be uninflected for a part of their paradigm (thus Polish muzeum ‘museum’ is uninflected in the singular only, an unusual type of heteroclisis), while some Macedonian adjectives show a featural split, being uninflected for gender though inflected for number. And when items move towards being inflected, the change may affect particular specific uninflected cells of the paradigm.
Since uninflectedness is an unexpected phenomenon within inflectional morphology, we might assume it would have no consequences outside inflection. And indeed, derivation may remain unaffected. Thus Upper Sorbian abbé ‘priest’ does not inflect, but it derives the possessive abbéowy ‘priest’s’. In syntax, however, while uninflected items often fit smoothly into their expected syntactic slot, this is not always the case. Wechsler & Zlatić (2013: 115-169) argue that uninflected nouns in Serbo-Croat are restricted in the contexts in which they can occur. They cannot occur in a nominal phrase assigned dative or instrumental, unless the case value is morphologically realized by some other element in the phrase.
Thus uninflectedness varies along a range of criteria, to be carefully defined. These criteria will be exemplified from two main sources. First, Slavonic languages, since these show dramatic variation, and have attracted considerable interest. The second main source will be Dagestanian languages, since these can have substantial numbers (even majorities) of uninflecting items, within parts of speech which can inflect. Given this, we need to refine our definitions.
Uninflected items are indeed surprising. They are also more varied than most accounts allow for, and it is only when we map out the typological possibilities that we can appreciate their significance.
References: • Wechsler, Stephen & Zlatić, Larisa. 2003. The Many Faces of Agreement. Stanford: CSLI.Wechsler & Zlatić. 2013.
Given (i), the variation in how pervasive agreement is (phenomenon P), we seek to explain why languages can have agreement at all (phenomenon Q). A suggested explanation for Q is that agreement facilitates reference tracking; to evaluate this explanation, the variation within agreement targets (ii), that is, phenomenon P, proves a useful way to measure and thus test the proposed explanation for Q. The suggested explanation is found wanting (Nichols 2018, Fedden in press, Feist 2020).
Turning to explanations of variation itself, fine-grained variation in agreement (iii) has been tackled by formalists from differing perspectives (some relying on different syntactic structures, e.g. Polinsky 2016, others giving morphosyntactic features an important role, e.g. Borsley 2016, Landau 2016, Sadler 2016). Much of this variation involves Agreement Hierarchy effects (Corbett in press); here Wechsler & Zlatić (2003, and subsequent papers) make a noteworthy contribution, and see more recently Shen (2019), Smith (2021) among many others. We need explanations for (i) why alternatives are possible, and (ii) the factors which favour one or other outcome. This is an area where there are rich data to measure, and canonical approaches play an important role in specifying valid baselines and criteria (Round & Corbett 2020).
Ambitious explanation should be based on solid evidence, the result of careful measuring. The existence of agreement, and the variation we find, are deeply puzzling. But this is an environment where linguists of different persuasions can contribute, where careful measurement is being undertaken, and where variation can be used as tool of explanation as well as a problem which itself is to be explained.
References
Bond, Oliver, Greville G. Corbett, Marina Chumakina & Dunstan Brown (2016), Archi: Complexities of agreement in cross-theoretical perspective. Oxford: Oxford University Press.
Borsley, Robert D. (2016), HPSG and the nature of agreement in Archi. In Bond et al., 118-149.
Corbett, Greville G. (In press), The Agreement Hierarchy revisited: the typology of controllers. To appear in Word Structure (special issue The many facets of agreement, ed. by Tania Paciaroni, Alice Idone and Michele Loporcaro).
Fedden, Sebastian (In press), Agreement and argument realization in Mian discourse. To appear in Word Structure (special issue The many facets of agreement, ed. by Tania Paciaroni, Alice Idone and Michele Loporcaro).
Feist, Timothy (2020), Nominal classification: Does it play a role in referent disambiguation? Studies in Language 44.199–239.
Landau, Idan (2016), DP-internal semantic agreement: A configurational analysis. Natural Language and Linguistic Theory 34. 975-1020. doi:10.1007/s11049-015-9319-3.
Nichols, Johanna (2018), Agreement with overt and null arguments in Ingush. Linguistics 56.845-863.
Polinsky, Maria (2016), Agreement in Archi from a minimalist perspective. In Bond et al., 184-232.
Round, Erich R. & Greville G. Corbett (2020), Comparability and measurement in typological science: the bright future for linguistics. Linguistic Typology 24.489-525.
Sadler, Louisa, (2016), Agreement in Archi: An LFG perspective. In Bond et al., 150-183.
Shen, Zheng (2019), The multi-valuation Agreement Hierarchy. Glossa: A Journal of General Linguistics 4(1), 46. doi: 10.5334/gjgl.585.
Smith, Peter W. (2021), Morphology-semantics mismatches and the nature of grammatical features. Berlin: De Gruyter Mouton. doi: 10.1515/9781501511127
Wechsler, Stephen & Larisa Zlatić (2003), The many faces of agreement. Stanford: CSLI.
Sebastian Fedden, Matías Guzmán Naranjo & Greville G. Corbett
Keywords: assignment rules, data mining, gender, German, morphosyntax, typology
In recent years linguistic typology has increasingly profited from computational methods; the hope is to discover patterns in large data sets more quickly and more accurately than would be possible for a human researcher. This is commonly known as ‘data mining’. A linguistic system which could benefit from such an approach is German gender.
The German gender system is a gem among the assignment systems found in the world, for the complexity of its interacting semantic, morphological and phonological assignment principles. As fast as it offers partial results it raises new questions. This is the more remarkable since there are just three gender values: masculine, feminine, and neuter. Furthermore, the basic semantic assignment rules are relatively straightforward. Much more challenging are (i) phonological assignment (investigated by Köpcke 1982, Köpcke & Zubin 1983, among others), and (ii) the relation between gender and inflection class (see Pavlov 1995, Bittner 1999, and Kürschner & Nübling 2011). And yet, despite the progress which has been made, and the great typological interest of German gender, no attempt has been made to analyse the system as a whole.
In a system as complex as German there are at least three pitfalls:
1. cherry picking: observations of alleged regularity are sometimes based on few examples and the overall applicability of these regularities is left unexplored;
2. generalizations without a baseline: thus a prediction of a particular gender value for, say, 35% of the nouns is hardly remarkable if 35% of the nouns overall are of that gender;
3. not allowing for overlapping factors: given that phonological, morphological and semantic properties may make the same gender value more probable, making a claim for a particular generalization (e.g. phonological) requires us also to eliminate the possible effects of morphology and semantics.
To avoid these pitfalls and make progress towards a holistic analysis of the German gender system, we mine a database of more than 30,000 German nouns from WebCELEX (Baayen et al. 1995), coded for gender, frequency, phonological shape, inflection class, and derived/compounded status, which we have cleaned and to which we added semantic information (human, animal, object, abstract, mass) and frequency (based on the COW corpus, Schäfer 2015). We then built a series of analogical models using machine learning algorithms (similar to Guzmán Naranjo 2020), including different combinations of predictors (morphology, semantics, phonology, inflection class).
The overall accuracy results show clearly that the system is anything but arbitrary. The combined factors reach a predictive success of over 96%. Individual
factors are also strong predictors, most notably phonological shape and inflection class. The German gender assignment system – while complex and unusual – represents a typologically well-known type: a combination of semantic and formal (morphological/phonological) assignment principles (Corbett 1991). Our conclusions relate to German gender, but we also make a larger point by showing how typologists can benefit from data mining. And we hope to reduce the ill-informed comments still made about German gender, sometimes even by linguists.
Acknowledgments: This work was partly funded by the grant “Optimal categorisation: the origin and nature of gender from a psycholinguistic perspective” (ESRC UK Grant RN0362A) and the public grant overseen by the French National Research Agency (ANR) as part of the “Investissements d’Avenir” program (reference: ANR10LABX 0083).
References
Baayen, R. Harald, Richard Piepenbrock & Leon Gulikers (1995), The CELEX Lexical Database (CD-ROM), Linguistic Data Consortium, University of Pennsylvania, Philadelphia.
Bittner, Dagmar (1999), Gender classification and the inflectional system of German nouns, in Barbara Unterbeck (ed), (1999), Gender in Grammar and Cognition, Part 1: Approaches to Gender, Berlin: Mouton de Gruyter, 1–23.
Corbett, Greville G. (1991), Gender, Cambridge: Cambridge University Press.
Guzmán Naranjo, Matías (2020), Analogy, complexity and predictability in the Russian nominal inflection system, Morphology 30(3), 219–262.
Köpcke, Klaus-Michael (1982), Untersuchungen zum Genussystem der deutschen Gegenwartssprache, Tübingen: Niemeyer.
Köpcke, Klaus-Michael & David A. Zubin (1983), Die kognitive Organisation der Genuszuweisung zu den einsilbigen Nomen der deutschen Gegenwartssprache, Zeitschrift für germanistische Linguistik 11, 166–182.
Kürschner, Sebastian & Damaris Nübling (2011), The interaction of gender and declension in Germanic languages, Folia Linguistica 45(2), 355–388.
Pavlov, Vladimir (1995), Die Deklination der deutschen Substantive. Synchronie und Diachronie, Frankfurt a. M.: Peter Lang.
Schäfer, Roland (2015), Processing and Querying Large Web Corpora with the COW14 Architecture, in Proceedings of Challenges in the Management of Large Corpora (CMLC-3), 28–34.
As an example, if we pull apart the possible criteria for reported speech, we find a wide range of possibilities, as Evans (2013) shows. Interestingly, natural languages vary considerably, but each avails itself of few of the possibilities. A large part of the theoretical space is not occupied. Conversely, the extreme instance of inflection classes, as defined by the combination of the extreme values of the canonical criteria, would appear to be highly unlikely on functional grounds. And yet, it is indeed found, in Burmeso (Donohue 2001, discussed in Corbett 2009).
Once we measure carefully, we are no longer confined to labelling label items as just “hot” or “cold” (as if the world were that simple); rather we can explore their finer-grained nature. Variability and empirical uncertainty become easier to characterize, and can now be incorporated into our analyses, rather than being factored out of them (see Round forthcoming for helpful discussion). As we measure more carefully, additional tools become available, and we can enter mainstream (social) science (Bickel 2015). We do not need exceptional devices for linguistics. Rather we measure the variability of the length of vowels, or the range of the genitive case value, using carefully defined criteria (Corbett 2012); we do so similarly when comparing the idiolects of two Russian speakers, when comparing older/younger speakers, Russian speakers with Polish speakers, and Polish speakers with Archi or Tamil speakers.
It is hard to imagine science without consistent measurement. We should follow Fahrenheit’s lead.
References
Bickel, Balthasar. 2015. Distributional typology: statistical inquiries into the dynamics of linguistic diversity. In: Bernd Heine & Heiko Narrog (eds) The Oxford Handbook of Linguistic Analysis, 2nd edition, 901-923. Oxford: Oxford University Press.
Corbett, Greville G. 2009. Canonical inflectional classes. In: Fabio Montermini, Gilles Boyé and Jesse Tseng (eds) Selected Proceedings of the 6th Décembrettes: Morphology in Bordeaux, 1-11. Somerville, MA: Cascadilla Proceedings Project. Available at: http://www.lingref.com/cpp/decemb/6/abstract2231.html.
Corbett, Greville G. 2012. Features. Cambridge: Cambridge University Press.
Donohue, Mark. 2001. Animacy, class and gender in Burmeso. In: Andrew Pawley, Malcolm Ross & Darrell Tryon (eds) The Boy from Bundaberg: Studies in Melanesian Linguistics in Honour of Tom Dutton (Pacific Linguistics 514), 97–115. Canberra: Pacific Linguistics.
Evans, Nicholas. 2013. Some problems in the typology of quotation: a canonical approach. In: Dunstan Brown, Marina Chumakina & Greville G. Corbett (eds) Canonical Morphology and Syntax, 66-98. Oxford: Oxford University Press.
Nikolaeva, Irina. 2013. Unpacking finiteness. In: Dunstan Brown, Marina Chumakina & Greville G. Corbett (eds) Canonical Morphology and Syntax, 99-122. Oxford: Oxford University Press.
Round, Erich R. forthcoming. Review of Matthew K. Gordon, Phonological Typology, Oxford University Press 2016. To appear in Folia Linguistica.
Greville G. Corbett
Surrey Morphology Group
University of Surrey
In trying to understand natural language, we need to consider what is a ‘possible word’ (lexeme). We find simple lexemes that are internally homogeneous and externally consistent. On the other hand, there are others with splits in their internal structure and inconsistencies in their external behaviour. I first explore the characteristics of the most straightforward lexemes, in order to establish a point in the theoretical space from which we can calibrate the real examples we find. I then schematize the interesting phenomena which deviate from this idealization: these deviations include suppletion, syncretism, deponency and defectiveness. Next I analyse the different ways in which lexemes are split into two or more segments by such phenomena. I set out a typology of possible splits, along four dimensions: (i) splits based on the composition/feature signature of the paradigm versus those based solely on morphological form; (ii) motivated (following a boundary motivated from outside the paradigm) versus purely morphology-internal (‘morphomic’); (iii) regular (extending across the lexicon) versus irregular (lexically specified); (iv) externally relevant versus irrelevant. I identify instances of these four dimensions separately: they are orthogonal to each other. Their interaction gives a substantial typology, and it proves to be surprisingly complete: the possibilities specified are all attested. The typology also allows for the unexpected patterns of behaviour to overlap in particular lexemes, producing some remarkable examples. Such examples show that the notion ‘possible word’ is more challenging than has generally been realized.
The following conventions, beginning with the more superficial and progressing to those with greater analytical significance, all deserve discussion:
• we conventionally represent different features by different dimensions (as in a case X number layout rather than a simple list of forms); this is difficult when we need more than two dimensions
• portrait view is favoured over landscape view, leading to specific choices of paradigm layout: for instance, person values in rows and number values in columns
• we follow traditional ordering of feature values (absolutive before elative)
• we split or combine cells according to unspoken conventions about majority distributions within and across lexemes; for instance, Russian nouns are represented with six case values, though an additional four values occur in different combinations on subsets of nouns
• we “know” that some conditions on paradigms belong in the representation while others are textual notes; for example, mass nouns have no separate paradigm, rather we state somewhere that the plural is available only for nouns of particular semantic types
• we appreciate elegance (witness the original minimal-augmented analyses of number systems)
• we (at least some of us) believe that syntax is morphology-free; hence we include non-autonomous values (such as the Romanian neuter gender) in paradigms. This approach avoids invoking strange rules of agreement or government, but requires additional paradigm cells which are systematically syncretic
• we represent morphosyntactic patterns rather than morphomic patterns
All of these conventions individually have merit. Good practice requires us to be fully explicit about our use of them and our departures from them, particularly where there come into conflict.
Conclusions:
• the substance matters more than the representation; conventions should help make clear what the analyst intends, so that the reader is able to agree or disagree with the actual intention
• the representation has enormous potential: it can clarify our understanding of our material (so that claims are made with full awareness) or it can mislead the unwary
• we are cleaning up our act with regard to morphosyntactic glossing. It is time to begin being more explicit about how we represent the forms of lexemes. Our largely unspoken conventions form a good basis.
into segments. Splits may be ‘motivated’, that is they may correspond to morphosemantic, morphosyntactic, or phonological specifications. Alternatively the split may lack such motivation, in which case we have a morphomic split, one which arguably increases the complexity of the system with no obvious corresponding return. We shall focus on the difference between these two types, so that we can recognise
morphomic splits. There are some properties which the two types of split share: for instance, both motivated and morphomic splits can be viewed in terms of Wurzel’s Paradigm Structure Conditions, that is, there can be predictive relations within the segments; and both types can persist over long periods of time.2 But they
are also interestingly different, which makes drawing the distinction valuable. It bears on the important notion that syntax is morphology-free. Our main question, then, is ‘How do morphomic splits differ from motivated splits?’
systems that are more informative or more rigid appear to be more efficient. The study implies that the influence of language on cognition may vary across languages, and that not all nominal classification systems employ this optimal trade-off between simplicity and informativeness. These novel data provide a new perspective on the origin and nature of nominal classification.