This could, in turn, inform a time-dependent model of gain control (e.g., Model 7 in Table S2), though we did not cross-validate such a model. Reliable estimates of time constants were obtained for both the switch from low- to high-contrast context (τL→H) and the switch from high- to low-contrast context (τH→L) for 18 units. Adaptation to high-contrast context occurred with a median τL→H of 86 ms, compared with a slower adaptation to low-contrast context with a median τH→L of 157 ms. This difference was significant (p < 0.001, sign-rank) and evident for 14/18 of the individual

units ( Figure 6F). Thus, the time courses for increases and decreases in neural gain are asymmetric. To explore the mechanism for gain control, we asked whether gain is modulated

by the contrast within MAPK Inhibitor Library a local region of frequency space or whether it is a function of the global statistics of the input. To address this, we varied the contrast of the DRC stimuli within two separate frequency regions. One region was denoted the “test,” centered around a chosen unit’s BF and spanning 0.5, 0.67, or 1.2 octaves. The remaining frequency bands were denoted BMN 673 solubility dmso the “mask” (Figure 7A). We aimed to situate the test stimulus over the “responsive frequency range” (ΦRFΦRF; see Experimental Procedures), the frequencies to which a given neuron (linearly) responded. However, since we recorded multiple units simultaneously (usually bilaterally), we actually sampled a range of conditions where the test stimulus covered the neuron’s responsive frequency range, overlapped it, or lay entirely outside it. This enabled us to measure how contrast gain depended on the amount of overlap between the test stimulus and ΦRFΦRF. We presented nine separate DRCs, where the contrasts in the test (σtest  ) and mask (σmask  ) were independently chosen from σL   = 2.9 dB, 5.8 dB, or 8.7 dB (c =   33%, 64%, or 92%).

Rebamipide We found that the gain of each neuron was most strongly modulated by contrast within the responsive frequency range. Thus, varying σtest   had the strongest effect on gain when the test stimulus completely covered ΦRFΦRF ( Figure 7B). Similarly, varying σmask   had the strongest effect when the mask completely covered ΦRFΦRF ( Figure 7C). However, contrast away from the responsive frequency range also had an impact on gain. For example, even when the test stimulus completely covered ΦRFΦRF, decreasing σmask   still resulted in an increase in gain ( Figure 7C). There were also interactions between contrast within and outside ΦRFΦRF (compare Figure 7B with 7D and Figure 7C with 7E). This is summarized in Figure 7F for 24 units where the test completely covered ΦRFΦRF.