Slide Slam O7
Predicting words and syllables: neural oscillations to “koguka” in adults and children
Valerie Shafer1, Margaret Kamowski-Shakibai1, Yan Yu2; 1The Graduate Center, CUNY, 2St. Johns University
Perception of speech makes use of inferences shaped by experience with the native-language phonology. These inferences are developed early in life via experience with input. The current study examines this process using neural oscillations to syllables sequences to elucidate the development of these prediction skills. The electroencephalogram (EEG) was recorded to sequences of 10 3-syllable words, where the second and third syllable were exchanged in two of the 10 words (koguka -koguka -koguka- kokagu, koguka, etc.). The words (.55 s duration) occurred at a rate of .95 s within the sequence of 10, with an inter-sequence interval of 1.9 s. Fifty percent of the sequences had the stimulus change (deviant) in the 4th and 8th positions, 25% had the deviants in the 5th and 10th and 25% had deviants in the 6th and 10th positions. The EEG was recorded from 12 adults and eight one- to four-year old children for a total of 400 sequences from 65 scalp sites using a geodesic net. The trains were segmented into 12.56 seconds, beginning 2 sec before the 1st stimulus of the sequence. Results focus on the sequences with the change in the 4th and 8th because we had sufficient trials from the children. Analyses on Fz (average-referenced) revealed peaks of power at 1.035 Hz (.966 period) for adults and 1.11 Hz (.897 period) for the children, roughly corresponding to the word rate of speech. The adult and children also showed a clear peak at 2.07 Hz (.483 s period). Only the adults showed a clear peak at 5.25 Hz (.19 s period), with children showing multiple peaks between 6 and 8 Hz. The Continuous Wavelet Transform (CWT – using Morlet wavelet) on the single trials revealed increased power between 2 and 5 Hz at the onset of the sequence of 10 stimuli for adults, but not for children. Adults, but not children showed increased power in the 1-Hz range after the 5th stimulus, perhaps related to predicting the .95 s rate of words; adults also showed suppression of 2-Hz power following the onset to the deviant (kokagu) word, but this suppression occurred closer in time to the onset when in the 8th compared to 4th word position. Children showed increased power between 2 and 5 Hz at the onset of the deviant in 8th position. These findings suggest that adults were tracking and predicting timing of sequence onsets, word rate, word duration, syllable duration, and the deviant in the 8th position. Children showed evidence of predicting word rate, word duration and the deviant in the 8th position. The different pattern to the deviant, specifically, power suppression for adults and increased power for children may reflect different types of mismatches responses, with adults showing a negative mismatch response and very young children showing a positive mismatch response. These findings are consistent with other studies indicating that young children compute statistical probabilities for speech, but also provide insight on the brain correlates associated with this process.