Karlos Arregi, University of Chicago

Fission of person and number in agreement and pronominal paradigms is well-attested cross-linguistically (e.g. Noyer 1992 for Arabic, Halle and Marantz 1993 for Georgian, Halle 1997 for Warlpiri 1997, Arregi and Nevins 2012 for Basque). In many of these languages, fission only affects second and third person (splitting their exponence of person and number), but not first (where person and number are realized by a single exponent). The opposite pattern, where fission only affects first person, is not attested.  Based on parallelisms found with diphthongization in the phonology of several languages (Calabrese 1995, 2005), Arregi and Nevins 2012 propose a markedness-based analysis of fission and explore its consequences for the expression of number in the pronominal clitics of several Basque dialects. In this talk, I provide a more comprehensive picture of dialectal variation in Basque pronominal clitics (based in part on Zuazo 2008), and argue that the markedness-based theory of fission affords an elegant account of this variation in terms of generalization/simplification of the feature specification of plural clitic vocabulary entries.

The analysis is based on two components. First, the asymmetry between first and second/third person noted above is accounted for in terms of the following markedness constraint that specifically targets non-first person ([-author]) terminal nodes:

(1) *[-author, -singular]

This constraint against the coexponence of certain features triggers fission, i.e. splitting of a terminal node specified for these features into two separate terminals. This defines [-author] and [-singular] as 'antagonistic' features; other features in the affected terminal are 'orthogonal'. The second component of the analysis has to do with the specific operation that splits the terminal node: the two output nodes differ in the specification of the antagonistic features, but share all the orthogonal features. For instance, fission of second plural ergative ([ergative, +participant, -author, -singular]) results in the output nodes [ergative, +participant, -author] and [ergative, +participant, -singular].

Under this analysis, the so-called plural morpheme resulting from fission is actually specified for case and (partially) for person, in addition to number. In most Basque dialects, this plural morpheme is realized by the allomorphs -e and -te. I show that the distribution of the allomorph -te can only be accounted for in a constrained theory of allomorphy by making reference to the orthogonal (case and person) features present in the output plural morpheme. In addition, dialectal variation in the distribution of -te is the result of the generalization of its featural specification, such that in the more restrictive dialects it is richly specified for orthogonal features, and in the progressively less restrictive dialects it has a gradually poorer specification for these features. This provides evidence for the markedness-based theory of fission sketched above.