Learning strategy trumps motivational level in determining learning-induced auditory cortical plasticity

Abstract

Associative memory for auditory-cued events involves specific plasticity in the primary auditory cortex (A1) that facilitates responses to tones which gain behavioral significance, by modifying representational parameters of sensory coding. Learning strategy, rather than the amount or content of learning, can determine this learning-induced cortical (high order) associative representational plasticity (HARP). Thus, tone-contingent learning with signaled errors can be accomplished either by (1) responding only during tone duration ("tone-duration" strategy, T-Dur), or (2) responding from tone onset until receiving an error signal for responses made immediately after tone offset ("tone-onset-to-error", TOTE). While rats using both strategies achieve the same high level of performance, only those using the TOTE strategy develop HARP, viz., frequency-specific decreased threshold (increased sensitivity) and decreased bandwidth (increased selectivity) (Berlau & Weinberger, 2008). The present study challenged the generality of learning strategy by determining if high motivation dominates in the formation of HARP. Two groups of adult male rats were trained to bar-press during a 5.0kHz (10s, 70dB) tone for a water reward under either high (HiMot) or moderate (ModMot) levels of motivation. The HiMot group achieved a higher level of correct performance. However, terminal mapping of A1 showed that only the ModMot group developed HARP, i.e., increased sensitivity and selectivity in the signal-frequency band. Behavioral analysis revealed that the ModMot group used the TOTE strategy while HiMot subjects used the T-Dur strategy. Thus, type of learning strategy, not level of learning or motivation, is dominant for the formation of cortical plasticity.

Figure 1

Training protocol includes a post-tone grace period (PTG, 2 s) to determine learning strategy. A, Each subject was required to bar-press (BP) to a 10 s, 5.0 kHz (70 dB SPL) pure tone to receive water reward. BPs made during inter-trial-intervals (ITI) were signaled as errors with a flashing overhead light for the duration of a time-out penalty period. BPs made during the PTG were neither rewarded nor penalized. B, The absence or presence of responses during the PTG reveal different learning strategies. If BPs are absent, both tone onset is used to initiate and tone offset to terminate responses. This comprises a “tone-duration” strategy (T-Dur). If BPs are present during the PTG, the only acoustic cue used is the tone onset to initiate reponses. The tone offset is ignored and instead the presentation of an error signal cues the termination of responding. This defines a “tone-onset-to-error” learning strategy (TOTE). Gray bars demarcate the distribution of BPs across the duration of a training trial in animals using either T-Dur or TOTE learning strategies. Note that animals using a TOTE strategy continue BPs until receiving an error signal.

Figure 2

Groups were trained in the same protocol under two different motivational levels. One group was highly motivated (HiMot, n=6) while the other group was moderately motivated (ModMot, n=8). A, Each group was water deprived to maintain body weight at ~70% of ad lib control rats for the HiMot group, or ~85% for the ModMot group. Values represent significantly different mean percent body weight for each group across all days of training (*p < 0.0001). B, An immediate effect of motivation level was evident in the latency to bar-press to tone presentations. Highly motivated animals were consistently faster to respond than moderately motivated animals beginning on the first day of training (p < 0.001). Only the first 11 sessions are shown to contrast the immediate difference between groups early in training.

Figure 3

Performance curves for ModMot and HiMot groups. A, Performance was calculated as the number of BPs during CS tones divided by the total number of tone or ITI BPs during the session ((# CS BPs / [# CS BPs + # ITI BPs])*100). Note that BPs made during the PTG period during which responses are neither rewarded nor penalized are not taken into account in the measure for performance. Initial learning was the same between groups across the first 6 days of training (p > 0.10), however asymptotic levels of performance were greater in the HiMot group (p < 0.001). The number of subjects in each group (HiMot/ModMot) by day (d) are as follows: d1(6/8), d2(6/8), d3(6/8), d4(6/8), d5(6/8), d6(6/8), d7(6/8), d8(6/8), d9(6/8), d10(6/8), d11(6/4), d12(5/0), d13(4/0), d14(4/0), d15(3/0), d16(3/0), d17(3/0), d18(3/0), d19(3/0), d20(3/0), d21(3/0), d22 (1/0, not shown in group mean). B, The difference in asymptotic performance level is accounted for by fewer errors of omission in the HiMot group than in the ModMot group (top, *p < 0.001). There was no difference in errors of commission between groups (bottom, p > 0.10). Values shown represent the mean number of errors during a session at asymptote.

Figure 4

Post-tone grace period (PTG) bar-press responses are diagnostic of learning strategy. Highly motivated subjects (HiMot) have fewer responses during the PTG than moderately motivated subjects (ModMot). PTG responses are normalized to the total number of BPs made during a session: Percent of PTG response = PTG BPs / [CS BPs + ITI BPs + PTG BPs]. HiMot responses decrease towards zero throughout training unlike ModMot responses which initially increase and later stabilize at an elevated level. Reduced numbers of PTG BPs indicate that HiMot group uses the tone offset to cue the termination of responses in a tone duration (T-Dur) learning strategy. Elevated PTG BPs in the ModMot group indicate that the tone offset is ignored and instead the presence of an error signal is used to cue response termination. Thus, ModMot subjects use a tone-onset-to-error (TOTE) strategy. Only the first 11 sessions are shown, however HiMot subjects that were trained for longer periods continued to show near-zero PTG BPs. Inset, PTG BPs at asymptote. HiMot subjects respond significantly less during the PTG than ModMot subjects who exhibit PTG BPs even at asymptote.

Figure 5

Frequency-specificity of bar-pressing behavior. Stimulus-frequency generalization tests were conducted twenty-four hours immediately after the last day of training. There were no significant differences between mean group frequency generalization gradients. Values represent the proportion of all “BP” vs. “no-BP” responses during the test session made in response to each test frequency. Both groups exhibited peak responses at (5.0 kHz) or near (7.5 kHz) the training frequency.

Figure 6

Learning does not induce A1 plasticity in highly motivated subjects. Instead, only subjects that use a TOTE learning strategy exhibit CS-specific plasticity in minimum threshold and tuning bandwidth. A, FRAs were constructed for each recording site in A1 to determine response threshold and tuning bandwidth (right), and pooled according to CF octave bands for each group. Each “V” represents the tip of the average group tuning curve in each CF octave showing response threshold (y-axis) and bandwidth 10 dB SPL above threshold (BW10, x-axis). Solid lines surrounded by shaded areas represent group means ± SE, respectively. Only the ModMot group has significant decreases in both threshold and BW10 that are specific to the signal-tone frequency band (CF threshold and BW10, ANOVA: p < 0.05; marked with asterix*). CF threshold is not significantly different from naives in any frequency band in the HiMot group. The decrease in threshold for the ModMot group in the octave band immediately above that of the CS-frequency is significant only before correction for multiple comparisons (before Bonferroni correction: CS Octave Band: F_(2,111)=2.39, p < 0.05) and the difference is only between the two trained groups (between groups posthoc: ModMot vs. naïve, p > 0.05; HiMot vs. naïve, p > 0.05; HiMot vs. ModMot, p < 0.05). B, Decreases in bandwidth 10 dB SPL above threshold (BW10) were only present in the ModMot group. For BW20, the decrease in tuning bandwidth was only significant prior to correction for multiple comparisons (before Bonferroni correction: CS Octave Band: F_(2,111)=2.39, p < 0.05) and the difference was only between the two trained groups (between groups posthoc: ModMot vs. naïve, p > 0.05; HiMot vs. naïve, p > 0.05; HiMot vs. ModMot, p < 0.05). The HiMot group is not significantly different from naives in any frequency band except in a non-specific decrease in BW10 the highest CF band (BW10, ANOVA: p < 0.05; marked with asterix*). See text for detailed statistics.

Figure 7

Learning does not change cortical area in either HiMot or ModMot groups. The amount of cortical area representing the CS and all other CF octave bands were equivalent among the trained and naïve groups. Therefore, learning did not induce areal reorganization in A1 either instead of (in the case of the HiMot group) or in addition to (for the ModMot group) FRA changes in threshold and bandwidth. Each bar represents the mean (+SE) percentage of the total area of A1 within each CF octave band.