Words and Gaps
A dissertation in linguistics by Kyle Gorman
University of Pennsylvania

Supervisor: Charles Yang (Linguistics, Psychology, and Computer & Information Science)
Committee: Stephen Anderson (Yale University, Linguistics, Psychology, and Cognitive Science), Eugene Buckley, Mark Liberman (Linguistics and Computer & Information Science), Rolf Noyer

This dissertation considers two kinds of unattested linguistic structures, and in both cases, shows that current generative models provide a natural account for the facts. 

The first part of the dissertation considers apparent constraints on underlying representations in a number of languages. It is proposed that Universal Grammar countenances no language-specific constraints on the sequences that make up underlying forms: any apparent counterexample is either a historical accident or a reflection of a constraint which is in force for domains larger or smaller than the morph. A number of studies are presented in favor of this hypothesis. A corpus study reveals that a number of constraints used to describe the inventory of medial syllable contact clusters in English monomorphs are in fact accidental gaps. A study of Turkish wordlikeness judgements reveals that speakers are sensitive to morpheme structure constraints which are derived by phonological alternations, and thus apply to a domain larger than the morph, but not those which are only static generalizations. A meta-analysis of English wordlikenss judgements reveals that modern gradient phonotactic models make unreliable predictions once the presence or absence of gross phonotactic violations are controlled for. 

The second part investigates "inflectional gaps": ineffable inflectional forms. It is proposed that all cases of inflectional gaps are morpholexical in nature, and cases from Hungarian and Swedish are used to illustrate. A case from Kinande is used to demonstrate that a lexically-specific constraint learning model will fail to converge on such grammars. A case from Spanish is used to show that inflectional gaps cannot be attributed solely to lexical storage. An analysis of inflectional gaps as reflexes of unproductive generalizations is put forth and used to analyze cases from English, Spanish, Greek, and Russian.