1). Local and temporary administration of the GABA-A receptor antagonist gabazine was performed simultaneously with electrophysiology using a carbon electrode coupled to a three-barrel pipette (Carbostar). Two pipettes were filled with 0.9% saline and one pipette was filled with 2.7 mM gabazine diluted in 0.9% saline. An injection current of 30 nA was used to deliver both drug and vehicle, and a retention current of −30 nA was used at all other times. A variable Verteporfin current was passed through the second saline barrel to balance the net current at the tip of the electrode. Physiology experiments during gabazine administration were started 2–5 min after beginning iontophoresis,
check details which was continued throughout the drug phase. Immediately following gabazine administration, saline was administered for 5 min before and continuously throughout the wash-out phase. To simulate the activity of a primary AC neuron, we convolved the STRF of a primary AC neuron with the spectrograms of songs, chorus, and auditory scenes. By rectifying the resultant with an exponential, we generated a simulated PSTH that was highly similar to the PSTH recorded in vivo (r > 0.60). We generated spike trains by sampling each PSTH with a Poisson spike generator and we simulated 10 trials of every stimulus. The kernel defining the the
BS temporal filter was a mixture of excitatory and inhibitory Gaussians with different delays and variances, representing excitation from the primary AC and delayed inhibition from NS neurons, and was constant for every simulated BS neuron. We simulated multiple BS neurons, each of which had the same temporal filter but received input from a different primary AC neuron. In this way, each BS neuron inherited a spectrotemporal filter from the primary AC, onto which was applied a temporal kernel. The width of the excitatory Gaussian corresponded to the duration of a typical BS spiking event (∼15 ms) and the width of the
inhibitory Gaussian corresponded to the duration over which contextual suppression was observed in vivo (∼100 ms). Because a single primary AC neuron provided input to the BS and NS neuron, the excitation and inhibition that each BS neuron received were cotuned. To simulate BS spiking activity, we convolved a primary AC PSTH with the BS temporal kernel shown in Figure 5A. We added an offset to the resultant of this convolution, rectified the outcome with an exponential filter, and generated spiking activity with a Poisson spike generator. We quantified simulated primary AC and BS spike trains with the same methods described above for recorded spike trains. For statistical analysis, the nonparametric Kruskal-Wallis and Wilcoxon rank-sum tests were used. We thank J. Moore, J.