Figure 3.

High-gamma single-trial rasters across the reading task from four manually selected electrodes in Subject 1. Trials, plotted along the Y axis, were sorted first by the place of articulation for the consonant, then by consonant voicing. Trials were aligned with response-onset time set to 0 s, denoted by a black vertical line at the center of each raster. Color denotes the high-gamma feature z-score normalized to a pooled pre-trial baseline period. Activity in electrode (A) appears to represent a bilabial place of articulation, whereas activity in electrode (B) appears to indicate an alveolar place of articulation. Timing differences of high-gamma activity relative to the voice onset time encoded the voicing of bilabial and alveolar consonants in these areas. Electrode (C) exhibited consistent high-gamma amplitude and timing for all utterances; informing neural VAD but less useful for keyword discrimination. Electrode (D) appeared to encode consonant voicing across all places of articulation. No clear patterns emerged if the trials were sorted by vowel height (/a/ vs /i/) for any electrodes in Subject 1.