An onset is an onset: evidence from abstraction of newly-learned phonotactic constraints

Phonotactic constraints involve language-specific patterns for sequences of speech sounds. Traditionally, phonotactic constraints are characterized by listing which sounds or sound sequences can occur at certain syllable positions (e.g. /h/ must begin syllables in English). However, most studies of these patterns have used monosyllabic stimuli (e.g., 'pef') and, thus, they do not tell us whether the patterns concern syllable or word positions. In a series of experiments we investigated the learning of new phonotactic constraints in an experimental setting that is informative as to whether these constraints truly concern syllable positions. In a continuous recognition-memory task, participants heard training nonwords which restricted particular consonants to either the syllable onset or coda position, mixed in with novel test nonwords that either followed (legal) or violated (illegal) the training constraints. Participants more often falsely recognized legal than illegal test nonwords, whether or not they matched the training nonwords in word structure or position of the restricted consonants. For example, having learned that /f/ is an onset and /p/ is a coda, participants generalized from one- to two-syllable items, and from word-edge to word-medial positions (or the reverse). These results suggest abstract representations of newly-learned phonotactic constraints in which the syllable, rather than the word, governs how those constraints are represented: an onset is an onset, a coda is a coda, regardless of word structure and word position