Receptive field dimensionality increases from the auditory midbrain to cortex

Abstract

In the primary auditory cortex, spectrotemporal receptive fields (STRFs) are composed of multiple independent components that capture the processing of disparate stimulus aspects by any given neuron. The origin of these multidimensional stimulus filters in the central auditory system is unknown. To determine whether multicomponent STRFs emerge prior to the forebrain, we recorded from single neurons in the main obligatory station of the auditory midbrain, the inferior colliculus. By comparing results of different spike-triggered techniques, we found that the neural responses in the inferior colliculus can be accounted for by a single stimulus filter. This was observed for all temporal response patterns, from strongly phasic to tonic. Our results reveal that spectrotemporal stimulus encoding undergoes a fundamental transformation along the auditory neuraxis, with the emergence of multidimensional receptive fields beyond the auditory midbrain.

Fig. 1.

Response patterns and spike-triggered averages (STAs) for midbrain neurons. Each row (A–G) represents 1 neuron. Left: spike raster for pure tones. Raster responses are shown for the sound level at which the ripple stimulus was presented. Gray boxes indicate the tone stimulus duration. Middle: poststimulus time histogram (PSTH) across all pure-tone presentation sound levels. Right: STA.

Fig. 2.

STAs, maximally informative dimensions (MIDs), and nonlinearities. Each row represents the filters and nonlinearities for 1 neuron. A and C: STAs and MID1s have similar filter structure. E: MID2s are less structured. B and D: STA and MID1 nonlinearities are similar. Dashed horizontal lines indicate mean firing rate. F: MID2 nonlinearities show little deviation from the mean firing rate. For nonlinearities, increasing similarity values indicate increasing correlation between filter and stimulus.

Fig. 3.

Comparison between STA and MID1 filters and nonlinearities. A: latencies for STA and MID1 were highly similar (r = 0.995, P < 0.001, t-test). B: best frequencies (BF) for STA and MID1 were highly similar (r = 0.999, P < 0.001, t-test). C: excitatory bandwidths (BW) of the filters were highly similar (r = 0.982, P < 0.001, t-test). D: spectral tuning (Q = BW/BF) was highly similar (r = 0.975, P < 0.001, t-test). E: correlations between STA and MID1 filters were high [median = 0.948, median absolute deviation (m.a.d.) = 0.016]. F: correlations between STA and MID1 nonlinearities were also high (median = 0.905, m.a.d. = 0.087). In A–D, diagonal lines represent the unity relationship.

Fig. 4.

STA and MID information analysis in midbrain and primary auditory cortex. A: STA and MID1 information were highly similar in midbrain and moderately similar in cortex. B: STA sufficiency [100·Info(STA)/Info(MID1)] was greater in the midbrain than in the cortex (P < 0.001, rank sum test). C: MID1 information was similar to the joint MID1 and MID2 information for all neurons in midbrain and less similar in cortex. D: MID1 contribution [100·Info(MID1)/Info(MID1,MID2)] was greater in the midbrain, indicating that the MID1 information approximated the 2-filter information to a much greater degree than in cortex (P < 0.001, rank sum test). In A and C, diagonal lines represent the unity relationship.

Fig. 5.

Phasic-tonic index (PTI), response precision index (RPI), and firing rate of midbrain neurons. A: firing rate distribution for inferior colliculus (ICC) neurons (median = 12.5, m.a.d. = 9.4). B: RPI distribution (median = 0.114, m.a.d. = 0.049). C: PTI distribution (median = 0.426, m.a.d. = 0.048). D: RPI vs. firing rate (r = −0.428, P = 0.0005, t-test). E: PTI vs. firing rate (r = 0.326, P = 0.0097, t-test). Dashed line indicates the PTI value for an ideally tonic neuron (equal numbers of spikes throughout the stimulus duration). F: PTI vs. RPI (r = −0.135, P = 0.295, t-test).

Fig. 6.

Comparison between response metrics and information in the midbrain. A: STA sufficiency compared with firing rate for midbrain cells. C: STA sufficiency vs. RPI. E: STA sufficiency vs. PTI. For all metrics the STA sufficiency was high. B: MID1 contribution vs. firing rate. D: MID1 contribution vs. RPI. F: MID1 contribution vs. PTI. For all metrics, MID1 information was highly similar to the joint MID1 and MID2 information.

Fig. 7.

Midbrain and primary auditory cortex (AI) MID summary. A: cumulative distribution for STA sufficiency. Compared with cortical granular (Gran) and nongranular (NonGran) layers, midbrain STA information more closely approximated the MID1 information [granular layers vs. midbrain: P < 0.001, Kolmogorov-Smirnov (KS) test; nongranular layers vs. midbrain: P < 0.001, KS test]. B: cumulative distribution for MID1 contribution. Compared with cortex, in the midbrain MID1 accounted for a greater percentage of the joint MID1 and MID2 information (granular layers vs. midbrain: P < 0.001, KS test; nongranular layers vs. midbrain: P < 0.001, KS test). C: median STA sufficiency values. Midbrain values were higher than cortical values (***P < 0.001, rank sum test). D: median MID1 contribution values. Midbrain values were higher than those in cortex (***P < 0.001, rank sum test).