Non-Standard Allomorphy in Russian Prefixes: Corpus, Experimental, and Statistical Exploration

Abstract

This dissertation challenges the traditional idealized model of allomorphy by confronting it with comprehensive data on 15 Russian aspectual prefixes (RAZ-, RAS-, RAZO-, S-, SO-, PERE-, PRE-, VZ-, VOZ-, O-, OB-, OBO-, U-, VY-, IZ-) collected from corpus and linguistic experiments. The traditional definition narrows allomorphy down to a mere variation of form where the meaning remains constant and variants are distributed complementarily. My findings show that submorphemic semantic differences and distributional overlap are not uncommon properties of morpheme variants. I suggest that allomorphy is a broader phenomenon that goes beyond the axioms of complementary distribution and identical meaning. I examine non-trivial cases of prefix polysemy and multifactorial conditioning of prefix distribution that make it difficult to assess the traditional criteria for allomorphy. Moreover, I present studies of semantic dissimilation of allomorphs and overlap in distribution that violate the absolute criteria for allomorphic relationship. I take the perspective of Cognitive Linguistics and propose an alternative usage-based model of allomorphy that is flexible enough to capture both standard exemplars and non-standard deviations. This model offers detailed applications of several advanced statistical models that optimize the criteria of both semantic “sameness” and distributional complementarity. According to this model, allomorphy is a scalar relationship between morpheme variants – a relationship that can vary in terms of closeness and regularity. Statistical modeling turns the concept of allomorphy into a measurable and verifiable correspondence of form-meaning variation. This makes it possible to measure semantic similarity and divergence and distinguish robust patterns of distribution from random effects.