Predicting surface forms in complex inflectional paradigms through phonological constraints

Abstract

This paper describes a new approach to the 'cell-filling problem' in inflectional paradigms (Ackerman and Malouf 2013) that builds on an account in Rosen (2019) in the framework of Gradient Symbolic Computation ('GSC') (Smolensky and Goldrick 2016). Exponents in a paradigm are derived through (a) Faithfulness to gradiently-weighted input blends of phonological material that occur both as part of a lexeme and on a morphosyntactic feature combination and (b) a gradiently weighted constraint that rewards a candidate for identity with a principal part form (Finkel and Stump 2007) in the paradigm. The analysis is applied to complex paradigms in Ngiti and Kwerba (Finkel and Stump 2007), both of which the previous account was unable to capture, and is presented as evidence in favour of the partially-activated features and weighted constrains in the GSC model.