Covert systematicity in a distributionally complex system

Abstract

Current thinking on inflection classes views them as organized networks rather than random assemblages of allomorphs (Carstairs-McCarthy 1994, Malouf & Ackerman 2010, Müller 2007), but we still find systems which appear to lack any visible implicative structure. A particularly striking example comes from Võro (a variety of South Estonian). Its system of verbal inflectional suffixes is formally simple but distributionally complex: although there are never more than three allomorphs in competition, nearly two dozen inflectional patterns emerge through rampant crossclassification of the allomorphs. Allomorph choice in one part of the paradigm thus fails to constrain allomorph choice in the rest, so it looks as if the paradigms would have to be memorized en masse. The key to these patterns lies outside the system of suffixation itself, in the more conventional formal complexity of stem alternations and their paradigmatic patterning. The computationally implemented analysis presented here provides a model of inflection in which the implicational network of phonological, morphophonological and morphological conditions on formal realization are unified in a single representation.

Morphosyntactic complexity: A typology of lexical splits

“…4 For these morphologists, there is still the issue of determining which of the problematic mappings are a matter of morphology and which are not (see Baerman 2014 for an instance of complex interactions of phonological, morphophonological, and morphological conditions). However, given that in this view there are clear instances of intraparadigmatic structuring (and hence paradigms are an essential component of the theoretical apparatus, rather than just a handy means of presenting and discussing data), there is no imperative to explain away difficult mappings at any cost.…”

A phonological account of Tlapanec (Mè’phàà) tonal alternation

“…In studying a language's inflectional paradigms, it is important to separate phonological processes and phonologically motivated alternations from morphologically motivated alternations (Baerman 2014(Baerman , 2015Corbett 2015: 147-148). However, this is not always an easy task especially in the case of Otomanguean languages, particularly due to the complex interactions of phonology and morphology.…”

Modeling morphological learning, typology, and change: What can the neural sequence-to-sequence framework contribute?

“…Importantly, the issue is not just that morphological analogy can overestimate the difficulty of the PCFP, but that the overestimation problem is likely to be larger for some languages than for others, so that the overly simplistic implementation based on pairs of exponents give an unrealistic description of the typological space. Baerman (2014) suggests that there is a typologically interesting class of languages in which information beyond what is captured by this narrow notion of morphological analogy contributes heavily to determining exponence; such a conjecture is difficult to test at a large scale without a model which is capable of exploiting these regularities as it learns to predict inflectional forms.…”

“…It is thus necessary to attend to the syntagmatic dimension when modeling the predictability of inflectional exponence. In fact, Baerman (2014) argues that in Võro, a variety of Estonian, inflectional exponents are predictable predominantly from stem shape (specifically, how stem alternants are distributed in the paradigm), and that the exponents of other inflected forms of the same lexeme are uninformative. Morphological stem shape can also matter: the inflection class that a lexeme belongs to may be predictable from its possible in some theoretical frameworks and we see no good motivation for it in this case.…”