Bernard, A. & Onishi, K. H. .
McGill
Languages display sound-position regularities (e.g., in English NG ends but cannot begin syllables). Adults are sensitive to these phonotactic patterns and can rapidly learn novel ones. We investigated the level(s) at which phonotactics are represented. Adults heard repeating words like baFPek and kiZDev in which constraints on the medial consonants could be represented as constraints on syllable-position (F, Z as codas; P, D as onsets) or co-occurrences (F co-occurring with P; Z with D). In test, false recognition was higher for novel words that maintained than violated syllable-position constraints, whether or not local co-occurrences were maintained (more false alarms to kiFPeb than kiPFeb, and to kiFDeb than kiDFeb; Experiments 1-2). False recognition was also higher for novel words that maintained than violated local co-occurrences, but only when syllable-position constraints were maintained (more false alarms to kiFPeb than kiFDeb, but similar rates to kiPFeb and kiDFeb; Experiments 3-4). Thus, across participant groups, syllable-level representations were available, and, if syllable-level constraints were maintained, co-occurrence level information was available. When test items from Experiments 1-4 were presented in a single experiment (Experiment 5), results were similar, suggesting the simultaneous availability of hierarchical phonotactic representations in which co-occurrence information is nested within syllable-position information.