A native speaker knows how to produce an unlimited number of words and possible words in their language, which are related to each other by various semantic, phonological, and morphological properties. The production of an intended word is not independent of these relations: similar or associated words may influence speech processing and articulation in the short-term, and the cumulative effects may change the phonological forms of words in the long-term. Here, I present three studies investigating how such relations affect the acoustic durations of words, and how that interacts with phonological representation.
It is well-known that if a word can be easily predicted based on the surrounding words, it is likely to be produced in a shortened form. In the first study, I show that some words almost always occur in contexts where they can be easily predicted, and are thus almost always produced in a shortened form. Through two corpus experiments, I demonstrate that these lexical patterns affect the phonological forms of words over the long-term: words that are typically predictable on the basis of nearby words become permanently shortened, and are produced with a shortened form even when they are not predictable.
In the second study, I examine how competition with phonologically-similar words affects hyperarticulation strategies. In a web-based experiment, I find that when speakers need to communicate a word clearly, they may actually reduce the duration of parts of that word when doing so would increase the acoustic contrast with a contextually-relevant minimal-pair competitor. The result provides evidence about how speakers use implicit knowledge of lexical contrasts in hyperarticulation.
In the final study, I investigate a mechanism through which paradigms of morphologically-related words might interact with each other in speech production. Phonetic paradigm uniformity proposes that the articulation of a word is influenced by the articulatory plans of morphologically-related words. In a laboratory experiment, I demonstrate that there is evidence for this influence based on the durations of morphologically-distinct homophones like FREES and FREEZE. Over the long-term, these patterns may help explain categorical changes in phonological structure.