The lexical representation of coda /t/ by 14-month-olds

We evaluated 14-month-olds’ lexical representation of coda /t/. Pronunciation variation of /t/, one of the most frequent segments in English, is well documented. Crucially, in American English infant-directed speech, /t/s are most often produced as glottal stops. This is particularly the case in word-final position, where words like cat and night are produced as [kæʔ] and [naiʔ] around 42% of the time, but produced with word-final canonical stop consonants (e.g., [kæt] or [nait]) only 25% of the time. In Experiment 1, we found that infants preferred to listen to familiar monosyllabic words ending in glottal stops (like [kæʔ]) when compared to nonce words ([kɪp]). This suggests that infants encode the most frequently occurring variant in their lexical entries. We are now testing whether the preference for familiar words produced with a glottal stop in Experiment 1 is the result of coda underspecification or encoding of glottal stop. If codas are underspecified in infant’s lexical representations, we expect no preference between lists of familiar words with glottal stops ([kæʔ]), compared to minimal pairs with coda mispronunciations ([kæk]). We will discuss implications of these findings for infants’ developing lexical representations and knowledge of allophony, focusing on the contribution of input frequency.