Exceptional Faithfulness and Exceptional Alternation: A Case Study of Polish Vowel-zero Alternations as Deletion and Epenthesis

The present thesis explores the issue of lexical exceptionality by examining vowel-zero alternations (also known as yers/jers and e-ruchome) in Polish nouns. The phenomenon refers to stem alternations such as sweter~swetry ‘sweater, sg.~pl’, whereby a vowel is present between the final two consonants of a stem when the inflectional suffix is null, whereas all other inflectional suffixes occur with a stem form that is missing this vowel. The difficulty of analyzing vowel-zero lies in the fact that it is not found categorically across all lexemes: it is possible to find counterexamples to an epenthesis analysis in paradigms that maintain the consonant cluster (e.g. metr~metry ‘meter, sg.~pl.’); and counterexamples to a deletion analysis in paradigms that maintain the vowel (e.g. seter~setery ‘setter dog, sg.~pl.’). Traditional analyses contended that the presence of counterexamples means that vowel-zero is unpredictable and thus every instance must be lexically encoded. However, this thesis shows that phonological and morphological patterns in the distribution of vowel-zero exist. Via three major contributions to the body of knowledge on vowel-zero—a statistical analysis of the lexicon, a wug test, and a grammaticality judgment study—it becomes clear that the number of exceptions to vowel-zero is quite small, and that vowel-zero is a productive process among Polish speakers but one that is not tolerated in all environments. Results thus indicate that regular rules governing vowel-zero exist in the grammar and, unlike other analyses of Polish vowel-zero, this thesis argues that a conspiracy of both epenthesis and deletion rules can best explain the vowel-zero patterns. Additionally, three main models proposed by linguists to encode exceptions are evaluated in order to account for remaining exceptions. The thesis argues that exceptionality is best explained through a combination of two models: morpheme-indexing to faithfulness best explains exceptional faithfulness (i.e. faithfulness where alternation is expected), while multiple listed allomorphs best explain exceptional alternation (i.e. alternation where faithfulness is expected). A sketch grammar of vowel-zero is presented to show that these two models can work together to explain the patterns and highlights the need to distinguish the two types of exceptions.Ph.D