Fig. 6: Phonemic to semantic processing.

To explore the cortical phonemic versus semantic processing, a joint Phonemic-Semantic VM consisting of single phoneme, diphone, triphone and semantic feature spaces was constructed. Variance partitioning was then used to obtain the cortical representation of phonemic and semantic models. The flat maps show the prediction performance for these models for one subject. a Shows that the Phonemic-Semantic VM  produces accurate predictions in LTC, VTC, LPC, MPC, SPFC and IPFC. b Shows that the Phonemic VM  produces accurate predictions in LTC, LPC, MPC, SPFC and IPFC. d Shows that the Semantic VM produces accurate predictions in VTC, LPC, MPC, SPFC and IPFC. c Show the mean and spread of the raw prediction performance across all subjects in ROIs in the temporal cortex, Broca’s area and the entire cerebral cortex. It reveals that the Phonemic VM produces significantly higher prediction accuracy than the Semantic VM in STG and STS. In contrast, the Semantic VM produces significantly higher prediction accuracy than the Phonemic VM in other regions of cortex, Broca’s area and LTC (*p < 0.05, **p < 0.01 and ***p < 0.001). The statistics are derived from the performance of significant Phonemic-Semantic voxels for each ROI (n = 2884 to 380,229 voxels) across 11 independent subjects. All tests are two sided and corrected for multiple comparisons. The exact p-values can be found in the Phonemic versus semantic cortical representations section. The box plots are defined the same way as in Fig. 4. Cortical regions referred are: PAC primary auditory cortex, STG superior temporal gyrus, STS superior temporal sulcus, LTC lateral temporal cortex, SPFC superior prefrontal cortex, IPFC inferior prefrontal cortex, MPC medial parietal cortex, LPC lateral parietal cortex, VTC ventral temporal cortex, MTC medial temporal cortex, VC visual cortex, sPMv ventral speech premotor area. Data used to generate this figure has been provided in source data.