Voice selectivity in the temporal voice area despite matched low-level acoustic cues

Abstract

In human listeners, the temporal voice areas (TVAs) are regions of the superior temporal gyrus and sulcus that respond more to vocal sounds than a range of nonvocal control sounds, including scrambled voices, environmental noises, and animal cries. One interpretation of the TVA's selectivity is based on low-level acoustic cues: compared to control sounds, vocal sounds may have stronger harmonic content or greater spectrotemporal complexity. Here, we show that the right TVA remains selective to the human voice even when accounting for a variety of acoustical cues. Using fMRI, single vowel stimuli were contrasted with single notes of musical instruments with balanced harmonic-to-noise ratios and pitches. We also used "auditory chimeras", which preserved subsets of acoustical features of the vocal sounds. The right TVA was preferentially activated only for the natural human voice. In particular, the TVA did not respond more to artificial chimeras preserving the exact spectral profile of voices. Additional acoustic measures, including temporal modulations and spectral complexity, could not account for the increased activation. These observations rule out simple acoustical cues as a basis for voice selectivity in the TVAs.

Figure 1

(A) The left and right TVAs, highlighted in yellow on a 3D rendering of a template brain, were identified using a“voice localizer” and the group-level contrast of vocal vs. nonvocal sounds (p < 0.05, FWE) (B) Auditory spectrograms of exemplars of the voice and instrument stimuli in Experiment 1, with amplitude envelopes over time (top panels) and auditory spectra (side panels). (C) Mean performance on the one-back task. (D) Mean and s.e.m. of parameter estimates in response to Voice and Instrument conditions in the left TVA. (E) As for panel D but for the right TVA, which shows greater activity to the voice stimuli.

Figure 2

(A) As for panel B, but for exemplars of the stimuli used in Experiment 2, organized in a 2 × 2 factorial design with temporal structure (T: Voice; Instrument) and auditory spectrum (S: Voice; Instrument) as factors. The top left (Instrument) and bottom right (Voice) stimuli correspond to natural categories, while the other two stimuli correspond to “chimeras”. (B) Average behavioral categorization as voice of the four stimulus categories. (C) Mean and s.e.m. of parameter estimates in response to the four stimulus conditions in the left TVA. (D) As for panel C, but for the right TVA, in which the response to Voice is significantly greater than for the other three stimulus categories. *p < 0.05.