Task difficulty and performance induce diverse adaptive patterns in gain and shape of primary auditory cortical receptive fields

Abstract

Attention is essential for navigating complex acoustic scenes, when the listener seeks to extract a foreground source while suppressing background acoustic clutter. This study explored the neural correlates of this perceptual ability by measuring rapid changes of spectrotemporal receptive fields (STRFs) in primary auditory cortex during detection of a target tone embedded in noise. Compared with responses in the passive state, STRF gain decreased during task performance in most cells. By contrast, STRF shape changes were excitatory and specific, and were strongest in cells with best frequencies near the target tone. The net effect of these adaptations was to accentuate the representation of the target tone relative to the noise by enhancing responses of near-target cells to the tone during high-signal-to-noise ratio (SNR) tasks while suppressing responses of far-from-target cells to the masking noise in low-SNR tasks. These adaptive STRF changes were largest in high-performance sessions, confirming a close correlation with behavior.

Figure 1

Schematic of the experimental stimuli and data analysis. A series of spectrotemporal modulated noise bursts (TORC1,TORC2, …) served as reference sounds that ended with an embedded target tone whose level relative to the noise was adjusted in different tasks to range from high SNR (>=0 dB), mid SNR (−5 dB >= to <0 dB), low SNR (< −5 dB). Animals licked water from a spout during the reference bursts (blue region), and were trained to withhold licking upon hearing the target tone (pink region). The response latency to the target was shorter in high SNR (top panel) than in low SNR tasks (bottom panel). STRF measurements were made only from responses during presentation of the reference TORCs, not from target responses.

Figure 2. Detection of tone in noise in different SNR tasks

(A) Lick-rate during reference TORCs was relatively flat (top panel) and provided a baseline against which cessation of licking during target tone presentation was measured. Dashed horizontal line indicates the lick-rate at half of this baseline. (Bottom panel) Lick rate began to drop at ~200 ms following onset of the tone. Lick rate decreased more rapidly in higher SNR tasks, continued to diminish gradually during the tone (shaded interval of 1 sec duration), and ceased rapidly after the onset of the shock period (250 ms after the end of the tone). (B) Response latency measured in three animals as the lick withdraw time (LWT) from onset of target tone to the point at which lick rate decreased by 50% of average lick rate during the reference period.

Figure 3

STRF changes in three units during high SNR tasks. (A) STRF measured pre-, during, and post-task performance. Target tone was set at 4.8 kHz, near the BF of the cell that was approximately at 4 kHz. During the task, the STRF sharpened its outlines considerably and became enhanced relative to the pre-task STRF. The STRF reverted to its original shape after the task. The starting time of each STRF measurement (relative to the beginning of the series) was noted at the bottom right corner of each panel. (B) STRF changed when the target tone (2 kHz) was placed far from the BF (9 kHz) of a cell. During the task, STRF became suppressed and partially recovered afterwards. (C) STRF changes in a sequence of two tasks. In the first, the target tone was placed near the BF (4.9 kHz) causing the STRF to become enhanced. When the target tone was placed far from the BF (8 kHz), the STRF became suppressed during the task, and partially recovered afterwards. All STRFs are shown relative to the same color scale.

Figure 4

STRF changes in three units during low SNR tasks. (A) STRF measured pre-, during, and post-task performance. Target tone was set at 3.2 kHz, near the BF of the cell (4.7 kHz). During the task, the STRF became suppressed relative to the pre-task STRF. The STRF partially recovered its original shape after the task. (B) STRF changed when the target tone (7 kHz) was placed far from the BF (3 kHz) of a cell. During the task, STRF became strongly suppressed, and recovered partially afterwards. (C) STRF changes in a sequence of two tasks. In the first, target tone (4 kHz), the BF (6.3 kHz) and the STRF became suppressed. When the target tone was placed near the BF (6.3 kHz), suppression during the task was weaker than during the first task and the STRF recovered afterwards. All plots are shown relative to the same color scale with red (blue) denoting increase (decrease) relative to the green baseline.

Figure 5

STRF amplitude changes as a function of target SNR. (A) Histogram of STRF changes at BF (Δ_BF) in all cells tested during high, mid, and low SNR tasks showing progressively deeper suppression with an average Δ_BF of −1.1%., −2.2% and 13.1% in high, mid, and low SNR tasks, respectively (B) Average STRF differences (STRF_diff) between the active and passive states. Net suppression increased with increasing task difficulty (lower SNRs). Note that the time axis here is not the same as in Figs.3 and 4, but instead is normalized relative to when the Δ_BF occurred in each cell. (C) STRF changes for each SNR divided into near and far groups. In all tasks, changes in the far cells were more suppressive than in near. Note that near cells showed an enhancement (positive Δ_BF) in the high SNR task, which weakened with increasing task difficulty. All plots are shown relative to the same color scale

Figure 6

STRF changes during tasks with best performance. All data were measured and presented exactly as in Figure 5, except in tasks with the best performance (see text). (A) Histogram of Δ_BF changes in high, mid, and low SNR tasks with average Δ_BF of 14.3%, −2.5% and −17.3% for high, mid and low SNR respectively. Changes in the high and low SNR cases are significant (p < .05). (B) Average STRF_diff between active and passive states. (C) STRF changes at each SNR divided into near and far groups. All plots are shown relative to the same color scale. All other details are as in Fig.5 panels.

Figure 7

Contributions of gain and shape changes to TORC responses and STRF plasticity. (A) Dependence of PSTH reference responses on task difficulty and BF separation from target tone. PSTH curves were computed from all cells selected for Figure 6 during the passive state (faint gray curves), and during the high SNR and low SNR tasks (left and right panels, respectively). In each panel, responses of near (red) and far (blue) cells are shown. All responses were normalized relative to their corresponding passive responses (after subtracting out the spontaneous activity). In general, far cells were more suppressed relative to the near cells. For both near and far cells, increasing task difficulty (from high to low SNR) caused comparable suppression of about 30%. (B) Distribution of STRF gain changes during behavior relative to the pre-behavioral state. These distributions depended weakly on task conditions and target distance from BF as shown in bottom and right panels. Specifically, these pure gain changes resulted in increased suppression of far STRFs (bottom panels) and of STRFs during low SNR relative to high SNR tasks (first and third distributions from the top). (C) Contribution of STRF shape changes to plasticity was comparable in strength but almost completely positive and focused when the target tone was in a high SNR and near the BF of the cell. General trends remained the same as in Figs.5 and 6, indicating diminished shape changes in more difficult tasks (low SNR) and in far STRFs. All plots are shown relative to the same color scale for all panels in this figure.

Figure 8

Average STRF differences (STRF_diff) between the active and passive states taking into account the superposition of both gain and shape changes. All details of the averaging are as in previous plots (Figs.5–7). The pattern of changes is analogous to that due to shape only (Fig.7C), except for an overall depression (blue) added to all panels. A net positive STRF change (red) is seen in the near cells during high SNR tasks. By contrast, a net deep suppression is seen in far cells during low SNR tasks. The two panels at the top display the dependence on task difficulty independent of distance from target. All panels are shown relative to the same color scale.