Figure 6.

Vocoded but not token-based VT stimuli are encoded similarly to auditory spoken words in the mid-STG following VT speech training. Linear mixed-effects analysis revealed a significant two-way interaction between training phase and algorithm (β = 0.240, t_(31.1) = 2.679, p = 0.012). To investigate this interaction, we created interaction effects plots. A, Opaque lines indicate the mean Fisher-transformed Pearson correlation between neural and model RDMs estimated from the mixed-effects model for the vocoded group. For the VT-vocoded group, post hoc tests show a significant difference between pretraining and post-training in the right (t_(31.1) = 3.380, p = 0.008 Sidak-adjusted) but not the left STG (t_(31.1) = 1.781, p = 0.298 Sidak-adjusted). B, Same as in A, but for the token-based group. Post hoc tests show no significant difference in the right (t_(31.1) = −0.408, p = 0.990 Sidak-adjusted) or left STG (t_(31.1) = 0.250, p = 0.999 Sidak-adjusted). Values above each violin reflect the uncorrected p value from a one-sample t test against 0. Semitransparent lines indicate raw individual subject correlations from either the left (teal) or right (orange) STG. Horizontal lines in the violin plots indicate the median. Green asterisk represents significant (p ≤ 0.05) differences after multiple comparisons correction.