Universal underpinnings of language-specific categories

The goal of this paper is to argue that the assumption that there are universal underpinnings for the construction of language specific categories is a useful, if not necessary assumption for the discovery and comparison of categories. Specifically, I will explore three empirical domains: grammatical categories of the familiar kind (e.g., tense, voice, demonstrative, etc.); categories associated with the language of interaction (e.g., sentence final tags, response particles, interjections, etc.), and categories that express emotions (e.g., ideophones, certain types of intonational tunes, expressives, etc.) The argument will be developed as follows. I start by introducing the framework for the analysis of grammatical categories I have developed in Wiltschko 2014 . This approach seeks to reconcile the tension between the two opposing views which this volume addresses: typologists observe that languages differ in their categorial inventories but some linguists (especially of the generative tradition) assume that there is a core which all languages share, including a set of universal categories. The key to reconciling this tension, I argue, is to assume that the categories we observe are always constructed on a language-specific basis, but that there are some universal building blocks involved in their construction, namely the universal spine , a hierarchically organized set of functions which is at the core of constructing sentential meanings. The spine has to be associated with units of language (I use the term unit of language as opposed to morpheme or word because I include – among other things - features as well and intonational tunes in the set of elements that can associate with the spine). Familiar grammatical categories are constructed via this association: that is, units of language per se do not form grammatical categories, they do so only in interaction with the spine. It follows that grammatical categories will always be language-specific, since the units of language are language-specific (for traditional morphemes this follows from the Saussurian assumption that the relation between form and meaning is arbitrary – hence must be conventionalized on a language specific basis). What this assumption allows us to do is to compare language-specific categories via a third element (Humboldt’s tertium comparationis ), namely the spine. Comparing language-specific categories directly to each other is typically meaning-based, but categories of similar meaning do not always have the same distribution and hence cannot be classified as universal categories (assuming that the hallmark of units of language of the same category is that they display the same distributional patterns). I then proceed to show that the same framework can be used for the discovery and comparison of categories which are not typically assumed to be part of grammar proper: interactive and emotive categories. I first show that they, too, display the patterns of grammatical categories: we find classes of UoLs which enter into syntagmatic and paradigmatic relations; and they display patterns of contrast and patterns of multi-functionality.