Vowel categorization induces departure of M100 latency from acoustic prediction

MEG studies have shown that the timing (latency) of the evoked response that peaks approximately 100 ms post-stimulus onset (M100) decreases as frequency increases for sinusoidal tones. We investigated M100 latency using a continuum of synthesized vowel stimuli in which the dominant formant frequency increases from 250 Hz (perceived /u/) to 750 Hz (perceived /a/) in 50 Hz steps. While M100 latency did vary inversely with formant frequency overall, frequency modulation was flattened within each vowel category. However, for mid-continuum ambiguous tokens (i.e. those with increased reaction time/decreased accuracy in the concurrent behavioral identification task), M100 reverted to formant frequency differences, agreeing with previous findings of frequency-dependence. A theory is proposed in which phonological categorization emerges from specific spatial distribution of frequency-tuned neurons.