Phonological features and phonotactic constraints in speech production

Languages are subject to phonotactic constraints - Restrictions on sound sequences. An implicit learning paradigm examined whether participants could acquire constraints at two levels of representation through exposure to a set of syllables. Participants were exposed to categorical segment-level constraints (e.g., /f/ only in onset, /s/ only in coda) as well as gradient featural-level constraints (e.g., labiodental fricatives /v/ and /f/ occurred in onset position 75% of the time, coda 25%). Speech errors revealed that participants encoded constraints at both levels of representation. By biasing errors towards a single syllable position, segmental constraints strengthened the tendency of errors to preserve target syllable position (e.g., virtually no /s/ errors occurred in onset). In contrast, since the featural constraint allowed errors to occur in both syllable positions, encoding it weakened the tendency to preserve target syllable position (e.g., /f/ errors, influenced by featural as well as segmental constraints, surfaced in coda more often than /s/ errors surfaced in onset). Finally, participants in a second study failed to learn featural constraints for dorsal stop consonants. The implications of these results for the representation and processing of features are discussed.