Categorizing human vocal signals depends on an integrated auditory-frontal cortical network

Abstract

Voice signals are relevant for auditory communication and suggested to be processed in dedicated auditory cortex (AC) regions. While recent reports highlighted an additional role of the inferior frontal cortex (IFC), a detailed description of the integrated functioning of the AC-IFC network and its task relevance for voice processing is missing. Using neuroimaging, we tested sound categorization while human participants either focused on the higher-order vocal-sound dimension (voice task) or feature-based intensity dimension (loudness task) while listening to the same sound material. We found differential involvements of the AC and IFC depending on the task performed and whether the voice dimension was of task relevance or not. First, when comparing neural vocal-sound processing of our task-based with previously reported passive listening designs we observed highly similar cortical activations in the AC and IFC. Second, during task-based vocal-sound processing we observed voice-sensitive responses in the AC and IFC whereas intensity processing was restricted to distinct AC regions. Third, the IFC flexibly adapted to the vocal-sounds' task relevance, being only active when the voice dimension was task relevant. Forth and finally, connectivity modeling revealed that vocal signals independent of their task relevance provided significant input to bilateral AC. However, only when attention was on the voice dimension, we found significant modulations of auditory-frontal connections. Our findings suggest an integrated auditory-frontal network to be essential for behaviorally relevant vocal-sounds processing. The IFC seems to be an important hub of the extended voice network when representing higher-order vocal objects and guiding goal-directed behavior.

Keywords: DCM; auditory-frontal network; decision-making; fMRI; voice.

FIGURE 1

Behavioral performance during the voice and loudness task. Reaction times (blue) and accuracy rate (red) for the voice task (voc = vocal, nvc = nonvocal) and the loudness task (low = low intensity, high = high intensity)

FIGURE 2

Neural activation for task‐based and passive vocal‐sound processing. (a) The general task‐based voice‐sensitive cortex (temporal voice area [tTVA]) revealed by the contrast vocal versus nonvocal sound trials during the voice and loudness task (red dashed line indicates the TVA defined by our task‐based design) in the temporal lobe covers large parts of primary (Te1.0–1.2), secondary, and higher‐level auditory cortex (AC) (Te3) as indicated by the white dashed line. (b) T‐map of the TVA as reported by Pernet et al. (2015) resulting from passive listening to vocal and nonvocal sounds. Red dashed line indicates the tTVA

FIGURE 3

Neural activation for vocal and nonvocal sound trials during the two tasks. (a) Contrasting vocal against nonvocal sound trials separately for task‐relevant (voice task) and task‐irrelevant (loudness task) voice processing revealed extended bilateral activity in the auditory cortex (AC) with peaks in superior temporal cortex (mSTC) and posterior STC (pSTC). Additional activity was found in the inferior frontal cortex (IFC) for vocal trials in the voice task (upper panel) located in bilateral IFGorb and IFGtri. An interaction contrast revealed specific activity in left IFGorb for vocal versus nonvocal sound trials in the voice compared to the loudness task. (b) Functional activity for high versus low‐intensity trials during loudness task (upper panel), as well as the interaction contrasts defining specific activity to high against low‐intensity trials during the loudness task (mid panel) and the vocal against nonvocal trials during the voice task (lower panel). (c) Signal estimates in regions of interest in 3 mm sphere around peak coordinates. The left IFGorb showed an interaction effect for vocal‐sound trials during the voice task, while a similar pattern of activity was found in the right IFGorb, this activation did not survive the cluster threshold. All activations thresholded at voxel p = .005 and a cluster extent of k = 65 (corresponds to p = .05 correct at the cluster level)

FIGURE 4

Neural activation including reactions time as covariate. (a) Task‐based voice‐sensitive auditory cortical regions by contrasting vocal against nonvocal sound trials across both tasks. (b) Functional activations for comparing vocal against nonvocal sound trials within the voice task (upper panel) and the loudness task (lower panel). All activations thresholded at voxel p = .005 and a cluster extent of k = 65 (corresponds to p = .05 correct at the cluster level)

FIGURE 5

Dynamic causal modeling (DCM) of brain activity during task‐based vocal‐sound processing. (a) Two‐step procedure for the DCM modeling, including a first estimation of the most likely input model (upper panel) based on permuting through the model space (left panel), determining the wining model (mid panel), and determining significant driving inputs (right panel). In the second step, using the significant driving inputs as fixed parameters, we permuted through all possible modulations of connections including three major families (bidirect, forward, and backward). The generic model takes voice trials during the voice and loudness task (all voice) and the task‐relevant voice trials during the voice task (voice task). (b) The bidirectional family achieved the highest posterior probability for the generic DCM models (upper panel) and the control DCM models (lower panel). The control models take voice trials during voice and loudness task (all voice) and the voice trials during task‐irrelevant sound processing during the loudness task. The right panel shows the posterior probability (black) and the relative log‐evidence (blue) of all models. (c) Bayesian model averaging revealed significant modulation of connections by all voice trails during the voice and loudness task (red) and by the voice trials during the voice task (blue) between bilateral auditory cortex (AC) and inferior frontal cortex (IFC) for the generic DCM models (upper panel); no such significant modulation was found for the control DCM models (lower panel). Light gray lines indicate significant intrinsic connections with positive connections as full and negative connections as dashed lines