Modelling the interaction of regularity and morphological structure: the case of Russian verb inflection

The main focus of this paper is to investigate how aspects of morphologicalregularity may have an impact on early stages of word processing,prior to full lexical access. Here I explore the interaction of regularity andmorphological structure by using a computational simulation of the processof learning Russian verb forms, without any morpho-syntactic or morphosemanticadditional information. With a recurrent variant of self-organisingmemories, namely a Temporal Self-Organising Map, or TSOM, experimentalresults allow an investigation of the impact of incremental learning and onlineprocessing principles on paradigm organisation, by assessing the differentialimpact of several aspects of regularity, ranging from formal transparency andpredictability to allomorphy, on the processing/learning behaviour in a connectionistframework. The proposed analysis suggests a performance-orientedaccount of inflectional regularity in morphology, whereby perception of morphologicalstructure is not the by-product of the design of the human wordprocessor, with rules separated from exceptions, but rather an emergent propertyof the dynamic self-organisation of stored lexical representations, dependenton the adaptive processing history of inflected word forms, intrinsicallygraded and probabilistic.