Figure 5.

Color-coded cortical discriminative maps for vowels or consonants and for actual or imagined speech, respectively. The color-coded (see color bar) cortical patterns show the locations with the best decoding performance and were superimposed for all left-hemisphere subjects (seven for actual speech and six for imagined speech, respectively). Color gives z-scores indicating how much better accuracy at the respective location was compared to chance (p-value: 0.0023 at a z-score of 2). A and B: Discriminative maps for decoding vowels and consonants during overt word repetition. C and D: Discriminative maps for decoding vowels and consonants during covert word repetition.