Figure 5.

Task-dependent classification of Speakers and Vowels. A, Group averaged (Mean (SE)) classification accuracies for speakers and vowels in the superior temporal cortex (STC ROI), during the speaker and vowel task. LH= left hemisphere, RH= right hemisphere. Statistical significance was determined with respect to empirical (permutation-based) chance level (dotted lines, range: 0.331–0.336, mean: 0.333, i.e., corresponding to theoretical chance level). B, Discriminative maps are illustrated for speaker classification during the speaker task and for vowel classification during the vowel task. These binary maps show for how many subjects a voxel was among the 25% most discriminative voxels and are visualized on inflated and aligned group-averaged representations of the temporal cortex. C, Self-organizing map (SOM) displays illustrating brain-based representation of speakers during the speaker task and of vowels during the vowel task. The maps are based on the 15 most discriminative speaker/vowel STC voxels across the 10 participants. The colors (speakers) and symbols (vowels) show which stimulus condition was assigned to a unit. The size of the unit indicates how often the stimulus condition was assigned to that unit (unit selectivity, see Materials and Methods).