The temporal lobes differentiate between the voices of famous and unknown people: an event-related fMRI study on speaker recognition

Abstract

It is widely accepted that the perception of human voices is supported by neural structures located along the superior temporal sulci. However, there is an ongoing discussion to what extent the activations found in fMRI studies are evoked by the vocal features themselves or are the result of phonetic processing. To show that the temporal lobes are indeed engaged in voice processing, short utterances spoken by famous and unknown people were presented to healthy young participants whose task it was to identify the familiar speakers. In two event-related fMRI experiments, the temporal lobes were found to differentiate between familiar and unfamiliar voices such that named voices elicited higher BOLD signal intensities than unfamiliar voices. Yet, the temporal cortices did not only discriminate between familiar and unfamiliar voices. Experiment 2, which required overtly spoken responses and allowed to distinguish between four familiarity grades, revealed that there was a fine-grained differentiation between all of these familiarity levels with higher familiarity being associated with larger BOLD signal amplitudes. Finally, we observed a gradual response change such that the BOLD signal differences between unfamiliar and highly familiar voices increased with the distance of an area from the transverse temporal gyri, especially towards the anterior temporal cortex and the middle temporal gyri. Therefore, the results suggest that (the anterior and non-superior portions of) the temporal lobes participate in voice-specific processing independent from phonetic components also involved in spoken speech material.

Figure 1. Experiment 2– Response times to indicate voice familiarity by overt speaking.

Each box gives the median with the 1st and 3rd quartile, the whiskers show the minimum and maximum values.

Figure 2. Examples of activation maps.

Shown for the contrast Voices > Rest at the significance level of (Experiment 1, A) or (Experiment 2, B) in different participants shown in sagittal plane at .

Figure 3. Experiment 1– Averaged BOLD signal time courses shown for all regions-of-interest in the left (A) and right (B) hemisphere.

Time t is given as the duration to scan one volume (2 s). Blue =  unfamiliar voices, green =  familiar voices, red =  named voices. Regions: a, anterior; ma, mid-anterior; mp, mid-posterior; iSTG, inferior part of the STG; LH, left hemisphere; p, posterior; RH, right hemisphere; sMTG, superior part of the MTG; sSTG, superior part of the STG.

Figure 4. Experiment 2– Averaged BOLD signal time courses shown for all regions-of-interest in the left (A) and right (B) hemisphere.

Blue =  unfamiliar voices, green =  familiar voices, yellow =  associated voices, red =  named voices. For abbreviations and further information see Figure 3.

Figure 5. Example for the position of the regions of interest.

(A) The sagittal view of a single subject’s brain shows the position of the ROIs along the superior temporal sulcus at (right hemisphere). The upper row of ROIs covers the superior part of the STG, the middle row the inferior part of the STG, and the lower row the superior part of the MTG. Red ROIs are located in the anterior temporal lobe, green ROIs in the mid-anterior part, blue ROIs in the mid-posterior part, and purple ROIs in the posterior temporal lobe. The mean Talairach coordinates are given in Table 5. (B) Sagittal view () of the first functional EPI volume showing those brain regions that produced an MR signal.