Gradient Perceptual Facilitation from Phonotactic Knowledge

1 Introduction In classical generative phonology, phonotactic constraints define the set of possible sound patterns in a given language by restricting the distribution of sounds, e.g. in certain structural positions or next to certain other sounds. Speakers' knowledge of the phonotactic constraints in their language is manifested in various linguistic tasks they perform. Speakers will judge a sound pattern to be grammatical if it is phonotactically legal and ungrammatical if it is phonotactically illegal. Their knowledge is also manifested in their performance in on-line speech processing tasks, such as speech perception and production. For example, Brown and Hildum (1956) showed that adult English speakers perceive non-words beginning with phonotactically legal onset clusters (e.g. /pr/) more accurately than non-words beginning with illegal onset clusters (e.g. /zdr/). One obvious source of their sensitivity to phonotactic legality of sound patterns is the distribution of sound patterns in the language. They have encountered one or more examples of phonotactically legal sound patterns but they have not encountered any examples of phonotactically illegal sound patterns in the language. Under the classical view, the focus was on the categorical distinction between phonotactically legal vs. illegal sound patterns. More recent studies make a finer distinction and focus on the gradience of phonotactic constraints. One commonly made argument is that a better description of phonotactic constraints should include not only what are legal and illegal sound patterns in the language but also the frequency or probability with which legal sound patterns occur in the language. Evidence in support of the argument comes from studies that relate gradience in speakers' task performance with lexical statistics. Speakers judge non-words consisting of more frequent sound patterns to be more acceptable than non-words consisting of less frequent sound patterns (Coleman and Pierrehumbert, 1997; Treiman et al., 2000). Speakers perceive non-words with high-frequency syllables more quickly than non-words with low frequency syllables (Vitevitch et al., 1997). Similar to the idea that presence vs. absence of sound patterns in the language results in speakers' sensitivity to the phonotactic legality of sound patterns, the gradience in performance is often attributed to difference in the amount of statistical evidence for different sound patterns. That is, speakers become more familiar with the phonotactic constraint governing a