Avoidance learning facilitates temporal processing in the primary auditory cortex

Abstract

The primary auditory cortex is now known to be involved in learning and memory, as well as auditory perception. For example, spectral tuning often shifts toward or to the frequency of the conditioned stimulus during associative learning. As previous research has focused on tonal frequency, less is known about how learning might alter temporal parameters of response in the auditory cortex. This study addressed the effects of learning on the fidelity of temporal processing. Adult male rats were trained to avoid shock that was signaled by an 8.0 kHz tone. A novel control group received non-contingent tone and shock with shock probability decreasing over days to match the reduced number of shocks received by the avoidance group as they mastered the task. An untrained (nai ve) group served as a baseline. Following training, neuronal responses to white noise and a broad spectrum of tones were determined across the primary auditory cortex in a terminal experiment with subjects under general anesthesia. Avoidance conditioning significantly improved the precision of spike-timing: the coefficient of variation of 1st spike latency was significantly reduced in avoidance animals compared to controls and nai ves, both for tones and for noise. Additionally, avoidance learning was accompanied by a reduction of the latency of peak response, by 2.0-2.5 ms relative to nai ves and approximately 1.0 ms relative to controls. The shock-matched controls also exhibited significantly shorter peak latency of response than nai ves, demonstrating the importance of this non-avoidance control. Plasticity of temporal processing showed no evidence of frequency specificity and developed independently of the non-temporal parameters magnitude of response, frequency tuning and neural threshold, none of which were facilitated. The facilitation of temporal processing suggests that avoidance learning may increase synaptic strength either within the auditory cortex, in the subcortical auditory system, or both.

Figure 1

(A) Auditory Avoidance Task. Adult male rats avoided shock if they moved to the opposite side of a shuttlebox within 5 s after onset of a tone (8 kHz, 70 dB), or escaped shock (maximum duration = 10 s) if their response latencies were > 5 s. Responses terminated the tone on successful avoidance trials and terminated both tone and shock on escape trials. Daily training sessions consisted of 50 tone presentations. (B) Control Task. Control rats received the same density of tone and shock, which were presented pseudo-randomly (minimum inter-stimulus intervals = 20 s). Across sessions, the number of pre-programmed foot shocks administered to control rats was gradually decreased to match the reduction in foot shocks experienced by avoidance rats as their performance improved. Responses terminated the tone and provided escape from shock.

Figure 2

Behavioral Measurements. (A) Group Performance. Avoidance rats demonstrated rapid improvement across experimental sessions, attaining ~90% avoidance responses. Due to the nature of the task, control rats were unable to avoid shock. Control rats showed no such learning function. (B) Shock Density. The average number of shocks received by groups across sessions. The programmed number of shocks presented to control rats was not significantly different from the avoidance group. (C) Escape Latency. Both groups of rats showed good escape behavior. However, across sessions the escape latencies of the avoidance group were consistently shorter than control group latencies. This difference may reflect escape responses when avoidance rats were in the act of crossing the shuttle box just before the shock was presented.

Figure 3

Neural Population Responses to Pure Tones. The graph shows the average neural responses (bin-width = 3 ms) to tones determined at 10 dB (A), and 20 dB (B) above CF threshold. Responses for each cluster were converted to z-scores using the mean and standard deviation of spike rate during the first 60 ms after tone onset. The plots demonstrate a reduced peak latency in the avoidance group (AV = solid), relative to both naïve (NV = gray) and control (CN = dashed) responses, and a reduced peak magnitude in the control group. These between-group differences were confirmed on the basis of individual cluster analyses, and across a broad range of stimulus levels (see text).

Figure 4

The Effects of Behavioral Training on Peak Responses to Pure Tones. (A) Peak Latencies vs. Stimulus Level. Peak latencies for each cluster were determined from average spike rates (bin-width = 3 ms) measured for the first 60 ms after tone onset. Neural clusters recorded from avoidance rats (AV = solid) showed significantly reduced peak latencies relative to both naïve (NV = gray) and control (CN = dashed) groups (p < 10⁻⁴ for both comparisons). Non-contingent presentation of tone and shock had a smaller effect on peak latencies; however compared to group NV, the reduction was still statistically significant (p < 10⁻⁶). (B) Peak Magnitudes vs. Stimulus Level. Avoidance training had no effect on the magnitude of the peak response. Peak magnitudes were obtained from responses that were first converted to z-scores using the mean and standard deviation of spike rates during the first 60 ms after tone onset. Neural clusters recorded from CN rats showed markedly reduced peak magnitudes relative to groups AV and NV (p < 10⁻⁷ for both comparisons).

Figure 5

The Effects of Behavioral Training on 1^st Spikes Following Pure Tones. (A) 1^st Spike Latencies vs. Stimulus Level. Avoidance training had no effect on the latency of the 1^st spike. A small but statistically significant difference was detected for 1^st spike latencies of group CN, such that latencies were slightly shorter at the lowest stimulus levels and slightly longer at the highest stimulus levels (p < 0.05 for both comparisons). (B–D) 1^st Spike Latency Variability vs. Stimulus Level. For all three groups the standard deviation of 1^st spike latencies declined as the stimulus level was increased (B). At all stimulus levels, the standard deviation (hence, variability) of latencies was lowest for the AV group. This was evident even when variability was expressed in terms of the coefficient of variation (C). When within-group CVs were pooled across stimulus level, those of group AV were found to be significantly lower than groups CN and NV (p < 0.05 for both comparisons; Bonferroni test). CVs for groups CN and NV did not differ significantly (D). AV = solid, CN = dashed, NV = gray for all graphs.

Figure 6

The Effects of Behavioral Training on Peak Responses to Noise Bursts. (A) Peak Latencies vs. Stimulus Level. As for pure tones, peak latencies for noise bursts were determined from average spike rates (bin-width = 3 ms) measured for the first 60 ms after burst onset. Again, neural clusters recorded from avoidance rats (AV = solid) showed significantly reduced peak latencies relative to both naïve (NV = gray) and control (CN = dashed) groups (AV vs. NV: p < 10⁻⁶, F-test; AV vs. CN: p = 0.033, binomial test). Likewise, peak latencies for group CN were also significantly shorter than group NV (p < 0.002). (B) Peak Magnitudes vs. Stimulus Level. As for pure tones avoidance training had no effect on the magnitude of the peak response to noise bursts. However, group CN exhibited significantly smaller peak magnitudes relative to groups AV and NV (p < 10⁻³ for both comparisons). Peak magnitudes for noise were normalized by the same procedure applied to pure tone data (see Figure 4 legend). AV = solid, CN = dashed, NV = gray.

Figure 7

The Effects of Behavioral Training on 1^st Spikes Following Noise Bursts. (A) 1^st Spike Latencies vs. Stimulus Level. Across stimulus levels avoidance training appeared to increase 1^st spike latencies by ~1 ms, relative to both CN and NV groups (p < 10⁻⁵ for both comparisons). Latencies for CN and NV were statistically indistinguishable. (B–D) 1^st Spike Latency Variability vs. Stimulus Level. The variability of 1^st spike latencies to noise bursts is represented here as it was for pure tone data (see Figure 5B–D). Once again, CVs pooled across stimulus levels were significantly reduced for group AV, relative to both CN and NV groups (p < 0.05 for both comparisons; Bonferroni test). Again, CVs for groups CN and NV did not differ significantly (D). AV = solid, CN = dashed, NV = gray for all graphs.