Word-types, not word-tokens, facilitate extraction of phonotactic sequences by adults

Phonotactics-the permissibility of sound sequences within a word-correspond to lexical statistics, but controversy persists over which statistics are being tracked. In this study, lexical type and token counts were compared as they contributed to phonotactic extraction from an artificial lexicon. Young-adult participants were familiarized with a set of CVCCVC nonwords contextualized as a lexicon of Martian animal names. The type and token frequencies of word-medial consonant sequences within those names were varied systematically. Participants then rated new nonwords, containing the same medial sequences, on a 7-point scale for similarity to the Martian animal names. Higher ratings only followed high type frequency familiarization conditions, suggesting that word-types drove phonotactic extraction. Additionally, participants reversed the typical preference for high frequency English sequences, likely because they rated nonwords according to their membership to an unknown language. This finding suggests cognitively separable tracking of artificial language statistics and preexisting representations.