Article Figures & Data
Figures
- Figure 1.
Auditory neurons are characterized by their receptive fields. a–c, Sound pressure waveforms of example stimuli are shown in the top left. Spectrograms, in which sound intensity is indicated by color, are in the middle. Red is high and blue is low. Spike trains evoked by 10 presentations of the sound shown above are at the bottom. a, Responses to pure tones were used to measure a neuron's classical receptive field (CRF, right). Color indicates response strength, which is the driven firing rate minus the spontaneous firing rate at each frequency–intensity combination. Red regions show higher firing rates compared with spontaneous firing rates, and blue regions show lower firing rates (maximum = 145 spikes/s). As with most neurons we recorded, this neuron was not inhibited by any single frequency–intensity combination and therefore the CRF has no blue regions. b, c, Spectrotemporal receptive fields were calculated independently from responses to song (b) and noise (c). STRFs are on the right. Red regions of the STRF show frequencies correlated with increased firing rates, and blue regions show frequencies correlated with decreased firing rates. Measurements of STRF parameters are indicated in b and c.
- Figure 2.
STRF excitatory bandwidth is stimulus-dependent. a, Song STRFs (left) and noise STRFs (right) for three representative MLd neurons. Song and noise STRFs of the neuron in the top row are highly similar. The neurons in the middle and bottom rows have song and noise STRFs that differ in their spectral and temporal tuning. b, Song and noise STRFs differ at the single neuron level. Each bar shows the degree (mean + SD) of differences between song and noise STRFs for single neurons, normalized by the range of each parameter observed across all neurons. EI index is the normalized ratio of excitation and inhibition and Sum is the sum of the absolute values of all pixel values in the STRF. Other abbreviations are as described in Figure 1. c, Song and noise STRFs were used to predict neural responses to novel within-class and between-class stimuli. Across the population of neurons, STRFs were significantly better at predicting neural responses to within-class stimuli compared with between-class stimuli (p = 3 × 10−10). d, The degree to which STRFs differentially predicted responses to within-class and between-class stimuli can be accounted for by differences in the measured parameters of song and noise STRFs. Together, the STRF parameters accounted for 72% of the difference in STRF predictions between sound classes (Δr). Each bar shows the fraction of this variance accounted for by the STRF parameters used independently. Differences in BW alone accounted for more than one-third of the explainable variance (*p < 0.05; **p < 0.01). e, The distribution of bandwidth differences across the population of recorded neurons shows that many neurons respond to a different range of frequencies during the processing of song compared with noise and that the distribution is biased toward broader song STRFs (mean = 225.58 Hz, p < 0.005).
- Figure 3.
Correlated stimuli could modulate neural responses by recruiting subthreshold inputs. a, Song and noise stimuli have different correlations. Spectrograms of 2 s samples of song (top) and noise (bottom) are on the left and 60 ms samples of song and noise are on the right. b, The average spectral profiles of 20 ms samples of song (solid line) and noise (dashed line) aligned at their peaks (i.e., the spectral autocorrelation). Song energy is more correlated across multiple contiguous frequency bands than is noise energy. c, Histogram showing the average baseline firing rate recorded during the 1 s of silence preceding stimulus presentation. d, Histogram showing the difference in excitatory bandwidth measured from pure tones at 90 dB SPL and at 20 dB above each neuron's threshold. Bandwidth expands substantially as sound intensity increases. e, A model of a neuron with V-shaped excitatory classical (solid lines) and extra-classical (dashed lines) receptive fields. The colored dots represent isointensity tone stimuli at four different frequencies. Tones that fall within the solid V evoke spiking responses. Tones that fall within the dashed V evoke subthreshold changes in membrane potential, but not spikes. f, The top shows a spectrogram of the stimuli shown in e. Dots at the top of the spectrogram show the location of each stimulus in the neuron's receptive field. A diagram of changes in membrane potential and spiking activity (vertical lines) during each stimulus is shown below.
- Figure 4.
Paired tones reveal extra-classical excitation and inhibition. a, b, The top panel shows the spectrogram of three successive stimuli: a pure tone at the neuron's BF, a side tone (ST) near the BF but outside of the CRF, and the two tones presented simultaneously. The black PSTHs below show the average neural response to each stimulus. The bar graphs at the right show the firing rates in response to multiple presentations of tone pairs (RBF+ST, red; mean + SD), and the sum of firing rates in response to the two tones (BF and ST) played independently (RBF + RST, blue). In a, the tone pair produced significantly higher firing rates than predicted by the two tones played independently, indicating extra-classical excitation. In b, the tone pair produced significantly lower firing rates, indicating extra-classical inhibition.
- Figure 5.
Extra-classical excitation measured with tone pairs. a, c, e, g, The song STRF, noise STRF, and CRF for four representative single neurons. The intensity that is enclosed by a dashed box within the CRF shows the intensity at which tones were presented for these experiments (60, 70, 70 and 50 dB, respectively). The darkest red pixel in each CRF corresponds to the maximum response strengths, which are 59.2 Hz (a), 17.5 Hz (c), 65 Hz (e), and 31 Hz (g). Green pixels show response strength of zero. b, d, f, h, The top row shows the spectral profiles of song and noise STRFs. The middle row shows the strength of the neural responses (firing rate, mean + SD) to isointensity pure tones ranging in frequency between 0.5 and 8 kHz. Asterisks indicate frequencies that drove responses significantly above baseline (p < 0.05) and were therefore in the CRF. The bold bar marks the frequency that was used as the BF in the tone pair experiments (1, 2.5, 2.5,and 2.5 kHz, respectively). The bottom row shows the response to tone pairs (red) comprised of the BF played simultaneously with tones ranging from 0.5 to 8 kHz and the sum of the two tones played independently (b, d, f) or the BF played alone (h) (blue). Significant differences between the blue and red bars at frequencies outside of the CRF show the eCRF. Asterisks (*) indicate frequencies that interacted significantly with the BF (p < 0.05). In b, d, f these interactions were excitatory. In h, these interactions were inhibitory. The top, middle, and bottom panels share the same frequency axis.
- Figure 6.
Extra-classical tuning predicts stimulus-dependent STRFs. a, The fraction of neurons with inhibitory (blue), excitatory (red), or no eCRFs (black) is at the top. The differences between the song and noise STRF BWs for the neurons in each group are given below. The center line of each box is the median of the distribution, and the outer box edges show the 25th and 75th percentiles. Bar ends mark the upper and lower bounds of each distribution. The distributions were all significantly different (p < 0.05, Wilcoxon rank-sum test). The STRF differences in the excitatory group were significantly greater than zero (**p < 0.005, Wilcoxon signed-rank test). b, Plot showing the relationship between stimulus-dependent spectral tuning and eCRFs for individual neurons. Each line shows the extent (Hz) and valence [excitatory (Exc) or inhibitory (Inh)] of extra-classical tuning for a single neuron. Red lines extending to the right show neurons with extra-classical excitation. Blue lines extending to the left show neurons with extra-classical inhibition. Black circles indicate neurons with no extra-classical tuning. The location of the line on the ordinate shows the difference between the song STRF and noise STRF BWs. The histograms show the distributions of extra-classical bandwidths (top) and STRF bandwidth differences (right). c, d, For neurons with excitatory eCRFs, the extra-classical excitation was typically asymmetric with respect to BF (either in lower frequencies, or in higher frequencies). Similarly, differences in the song and noise STRF BWs were also asymmetric with respect to BF. c, The fraction of neurons for which excitatory eCRFs had the same (matched) frequency asymmetry as STRFs. Neurons with mixed asymmetries include those with excitatory eCRFs but no differences in STRF BW, and those with differences in STRF BW but no eCRF. Neurons with mismatched asymmetries had eCRFs and STRF BW differences in frequency channels on opposite sides of the BF. d, The fraction of neurons for which inhibitory eCRFs had matched, mixed, or mismatched asymmetries.
- Figure 7.
Simulated neurons with eCRFs and a spiking threshold exhibit stimulus-dependent STRFs. a, The neural response to isointensity tones for a simulated neuron with subthreshold excitation is on the left. Frequencies that cause the neural response to exceed spike threshold (Vth) lead to an increased spiking probability and are in the CRF. Frequencies that cause the neural response to deviate from the resting potential (Vr) but not cross Vth do not evoke spikes and are in the eCRF. A STRF with the same frequency tuning as the spectral profile is on the right. b, Spike trains were simulated using the STRF in a as a generative model. The top raster shows 10 spike trains in response to a single song, and the bottom raster shows 10 spike trains in response to a single noise sample. c, Separate song and noise STRFs were computed from the simulated spike trains to 20 songs and 10 noise stimuli, generated using the STRF in a. d, The spectral profile of the original STRF used in the generative model (pink), along with the spectral profiles of the resulting song STRF (solid gray) and the resulting noise STRF (dashed black). The left shows the spectral profiles of the neuron described in a–c, which had extra-classical excitation. The middle shows the spectral profiles for a simulated neuron with extra-classical (sideband) inhibition. The right shows the spectral profiles for a neuron with no extra-classical tuning. e, Histograms showing the difference in excitatory bandwidth between the song and noise STRFs of simulated neurons with extra-classical excitation (top), extra-classical inhibition (middle), and no extra-classical tuning (bottom). For neurons with extra-classical excitation and inhibition, the difference between song and noise STRF BWs was significantly different than 0 (p < 0.002, both comparisons) and highly similar to the BW differences that were measured experimentally.
