(2002). Experiments 1 and 2 tested the hypothesis that variability along the contrastive

dimension of voicing helps infants define the phonological categories for the words, while simultaneously eliminating noncontrastive variation that might be expected to impede processing. CFTR activator If true, it might suggest that further development of the internal statistical structure of VOT distributions is necessary for phonological categories to be engaged in this case. We used the same words as Rost and McMurray (2009): /buk/ and /puk/. These differ in voicing, for which VOT is the dominant cue. In the present study, the effects of variability in VOT alone were investigated by training and testing infants using auditory stimuli from a single speaker, but with a VOT distribution as shown in Figure 1c that mirrored distributions in the child’s language as well as the distribution found in the original Rost and McMurray study. This is an important contrast with the work of Maye et al. (2002, 2008), in that our continua spanned a dimension that infants had significant familiarity

with, and used asymetrical (although more natural) distributions. Given Ceritinib order the purpose of augmenting their natural categories (to explain our prior results), this seemed a better test. If variation in VOT is sufficient to drive learning, then we should observe good word learning using a set of exemplars with this distribution of VOT, and no variation in any of the additional cues present

in multitalker input (e.g., pitch, vowel quality, prosody or timbre). Infants between 13 and 15 months old were recruited from county birth records. Infants were eligible if they were monolingual English learning, with no history of developmental disorder or recurrent ear infection. Twenty-six infants Vorinostat in vivo participated; data from 10 were excluded due to their failure to habituate (5), experimenter error (2), fussiness (2), and ear infection (1). Sixteen infants (9 boys; M age = 14 months 4 days, range=13 months 5 days to 14 months 22 days) were included in the final analysis. A female native speaker of the local dialect produced a series of /buk/ and /puk/ tokens in an infant-directed register. In order to create a continuum with sufficient variation we included prevoicing (so that /b/ could be more variable while still being distinct from /p/). Praat (Boersma, 2001) was used for all stimulus manipulation. One /buk/ token was chosen by five adults as being the “best” exemplar, and it was modified to have a VOT of close to 0 msec by cutting the prevoicing. One /puk/ token was chosen as having the most natural aspiration which was longer than 100 msec. From these we constructed a 29-point VOT continuum ranging from −40 to +100 in steps of approximately 5 msec (limited by the availability of splice-points) using the following procedure.