Small syntactic terminal nodes, large Vocabulary Items: a spanning approach to irregular Romance verb inflection

In a series of papers within the framework of Distributed Morphology (DM), Calabrese (2012 et seq.) has convincingly argued that there is a link between the (a)thematicity and the (ir)regularity of verbal forms: the presence of a Theme Vowel (ThV) has a direct effect on the regularity of the respective verbal forms whereas its absence may cause allomorphy. However, Calabrese posits numerous postsyntactic processes and idiosyncratic rules, which – in our opinion – lack cognitive plausibility. In this paper, we will show that spanning (Svenonius 2012; Merchant 2015) is an economical and adequate way to implement verbal allomorphy in Romance. We argue that morpho-phonologically realized ThVs function as a kind of intermediate domain delimiter and we show, following the DM-based Vocabulary Insertion-Only Model (Haugen & Siddiqi 2016), that many of the context-specific rules and processes proposed in other works can be reduced to Vocabulary Insertion. Our analysis keeps the elements in syntax as small as possible, but allows spanning Vocabulary Items (VIs), i.e. VIs that realize more than one syntactic terminal node at once. We will illustrate our approach through analyzes of grammatically determined athematicity, athematic conjugation classes (CC), and inherited athematicity in irregular Romance verb inflection.