Atomic Linguistic Elements of Phi (ALEPH)

ALEPH seeks to integrate and advance understanding of foundational issues in morphology, semantics, and syntax, and to link these to linguistic typology and non-linguistic cognition.

The first, emerging from recent deep typological studies (Corbett 2000, 2006, Cysouw 2003, Siewierska 2004), concerns the person/number systems that recur across languages. Language exhibits four persons (first (inclusive/exclusive), second, third) and eleven numbers (general, singular, dual, trial, minimal, (unit) augmented, (lesser) paucal, (greater) plural). The term 'person/number system' refers to the total set of such distinctions that a language makes: e.g., in English, the person system distinguishes first/second/third (but not whether 'we' in/excludes the hearer), and the number system, singular/plural (but not dual, etc.) . Naively, there are 2^4=16 person systems, 2^11=2048 number systems. Yet typologists have shown that far fewer are attested.

The other insight, stemming from Alexander M. Bell (1867) and Jakobson, Karcevsky & Trubetskoy (1928), is that traditional categories of descriptive linguistics are not the atomic grammatical units that the human mind manipulates. Rather, sounds like S/Z, and cases like nominative/dative, are composites of more basic units, known, since 1928, as 'features'. Seminal work by Hale (1973) and Silverstein (1976) extended featural analysis to person categories, e.g., first person in/excluding the hearer, and number categories, e.g., singular/dual/plural. They showed, for example, that the featural composition of the dual shares features with both singular and plural.

Such research has received sustained attention since the 1990's and continues to achieve important results (e.g., Noyer 1992, Harley & Ritter 2002, Neeleman & Szendroi 2007). However, it has yet to achieve descriptive adequacy: even recent proposals undergenerate for number but require constraints to avoid overgeneration for person. (For example, Harley & Ritter treat dual/paucal as different uses of the same features, which wrongly precludes their cooccurrence in, e.g., Yimas.)

Thus, two complementary challenges arise in the attempt to unify feature-theoretic insights into person/number with full typological variation: the feature inventory proposed must possess sufficient expressive power to yield analyses of all person/number categories; but it must be constrained enough not to generate unattested categories/systems. Only with such a perfect fit between features and categories/systems can linguists claim to have an adequate account.

The proposed research methodology combines analytic techniques from diverse theoretical domains. Specifically, we will use techniques from morphology (decomposition, allomorphic/syncretic analysis) to identify the features behind different person/number systems, from syntax (discontinuous/split/partial/defective agreement) to identify the structures that features form, and from semantics (lattice theory, lambda calculus) to ensure precise interpretation of individual features and the structures in which they cooccur.

An account of person/number is essential to theoretical linguistics. However, its significance is much broader. (1) Interest in these features amongst syntacticians, semanticists and morphologists has recently grown (e.g., Harbour et al eds 2008). A project uniting research diverse methods/concerns is timely and important. (2) Feature families are not frequently investigated in depth (even feature-centered theories, like HPSG, focus on feature structures, not e.g. person/number). A model of how to identify and define features is a significant result that could guide investigation of, e.g., tense/aspect/location. (3) The cognitive primitives of groups/quantities/persons are of import beyond linguistics (e.g., grouphood/agency in political science, infant/animal number cognition) and linguistics has very real potential to inform such research (e.g., Gold & Harbour 2008).