Gaze durations during speech reflect word selection and phonological encoding.

Speakers produced the sentence frame The A and the B are above the C to describe three pictured objects while their eye movements were monitored. Object B or C varied in codability (the number of alternative names for it) and in the frequency of its dominant name. Codability is known to affect speed of word selection, and word frequency, speed to retrieve a word's pronunciation (phonological encoding). Speakers gazed longer at lower codability and lower frequency objects before naming them. However, the codability and frequency of B and C did not affect when speakers began naming A, even when utterances were perfectly fluent. The results indicate that speakers began "The A..." once they had a name prepared for A, before selecting names for B and C. Similar gaze patterns during less constrained scene description tasks in other studies suggest that speakers often incrementally select and phonologically encode nouns in fluent utterances.