Distinct timescales for the neuronal encoding of vocal signals in a high-order auditory area

Abstract

The ability of the auditory system to selectively recognize natural sound categories while maintaining a certain degree of tolerance towards variations within these categories, which may have functional roles, is thought to be crucial for vocal communication. To date, it is still largely unknown how the balance between tolerance and sensitivity to variations in acoustic signals is coded at a neuronal level. Here, we investigate whether neurons in a high-order auditory area in zebra finches, a songbird species, are sensitive to natural variations in vocal signals by recording their responses to repeated exposures to identical and variant sound sequences. We used the songs of male birds which tend to be highly repetitive with only subtle variations between renditions. When playing these songs to both anesthetized and awake birds, we found that variations between songs did not affect the neuron firing rate but the temporal reliability of responses. This suggests that auditory processing operates on a range of distinct timescales, namely a short one to detect variations in vocal signals, and longer ones that allow the birds to tolerate variations in vocal signal structure and to encode the global context.

Figure 1

A single sequence or sequences with natural variations found in individual’s songs were used to build two series types: ABAB-Same and ABAB-Var series. (a) Schematic diagram of the structure of ABAB-Same (top) and ABAB-Var (bottom) series. A and B depict two syllable types used to form ABAB sequences. The ABAB-Same series consisted of 60 repetitions of a single ABAB sequence while the ABAB-Var series consisted of 60 distinct renditions of a given ABAB sequence. These renditions called sequence “variants” were labelled as A_nB_nA_nB_n (n varying from 1 to 60). A_n and B_n were distinct exemplars of a single syllable type that were extracted from the song’s repertoire of a given individual. Each sequence was presented at a rate of one per second. (b) Example spectrograms of two consecutive sequences within an ABAB-Same (i, no variants) and ABAB-Var (ii, variants) series. Note the subtle changes between A₁B₁A₁B₁ and A₂B₂A₂B₂ sequences of the ABAB-Var serie (e.g. power at ~ 5 kHz on syllable B. Underneath each spectrogram are the accuracy scores (%) computed with SAP 2011 (see main text for further details) between A and B syllables across the two successive example renditions of the ABAB-Same and ABAB-Var sequences. (c) Mean (+ /- STD) of the accuracy scores computed between A and B syllables across the 60 renditions of all the ABAB-Same (top) and ABAB-Var (bottom) sequences. *** p < 0.001.

Figure 2

Auditory responses to 60 repetitions of a single sequence (ABAB-Same series) and to 60 sequence variants (ABAB-Var series) in awake birds. Responses of a representative unit to the ABAB-Same (a) and the ABAB-Var (b) series used as auditory stimuli. Neuronal responses that are time-aligned with sequence spectrograms (ai and bi: the sequence repeated 60 times for the ABAB-Same example series and one sequence variant for the ABAB-Var example series) are shown as raster plots (aii and bii, 60 iterations, dark colors for the 10 first and 10 last trials) and peristimulus time histograms (aiii and biii; 10 ms bin width; for the 10 first (top) and the 10 last (bottom) trials). (c) Modulation of responses over the six successive blocks of ten trials (each block for the ABAB-Var series includes 10 variants of the auditory sequence). The RS values estimated the strength of the responses driven by the series used as auditory stimulus. Thick line indicates mean responses for the population of recording sites (n = 56). Hatched area represents SEM. (d) Adaptation rate (mean ± SEM) of responses computed over the 10 first trials did not significantly differ between the two series.

Figure 3

Auditory responses in anesthetized birds. From rendition to rendition, spike timing greatly changed when sequence variants (ABAB-Var series) were played back. No such changes were observed when the same sequence was repeated (ABAB-Same series). (a,b) Neuronal responses of representative putative narrow spike (a) and broad-spike (b) cells to playback of ABAB-Same (left panels) and ABAB-Var (right panels) series (spectrogram on top, ai: the sequence repeated 60 times for the ABAB-Same example series and one sequence variant for the ABAB-Var example series) are shown as raster plots (aii and bii, 60 iterations, dark colors for the 10 first and 10 last trials) and peristimulus time histograms (aiii and biii; 10 ms bin width; for the 10 first (top) and the 10 last (bottom) trials) that are time-aligned with sequence spectrograms. (c) ABAB-Same series evoked higher responses (RS values) than ABAB-Same series at the population level (left) and for the sub-population of narrow spike cells (right), but not for broad spike cells (middle). Thick line indicates mean values and shaded area represents SEM. (d) Despite of the difference in the response strength, adaptation rate computed over the first ten stimuli presentations, thus indicated that response strength similarly changed with repeated exposure to sequences. Thick line indicates mean values; shaded area represents SEM. Significant difference: *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 4

Reliability of spike trains in awake and anesthetized birds. (a) To compute the reliability of the spike trains (CorrCoef), a convolution of spike times with a Gaussian window of 10 ms was performed. Pearson correlation coefficients were computed on convoluted spike trains between two renditions. CorrCoef values within a block, e.g. block 1 vs 1 (1|1), correspond to the mean of the Pearson correlation coefficients between renditions one to ten. CorrCoef values between blocks, e.g. block 1 vs 2 (1|2), correspond to the mean of all the Pearson correlation coefficients between renditions one to ten and eleven to twenty (i.e. rendition 1 vs 11 to 20, rendition 2 vs 11 to 20, etc.). Reliability of spike trains in awake (b) and anesthetized (e) birds illustrated by heatmaps (right: ABAB-Same series; left: ABAB-Var series). Spike-trains reliability, quantified by the CorrCoef index, was lower when sequence variants were presented. Blue color indicates low CorrCoef values. At the population level in awake (c,d) and anesthetized (f,g) birds, differences in spike-timing reliability and in its time course between the two series. CorrCoef values were computed from spike trains evoked by the first ten trials and those evoked by the ten trials of the six blocks (block 1 to 6, c, f) and within each block (d,g). In both awake and anesthetized birds, CorrCoef (mean ± SEM) changed with stimulus exposure when the same sequence (blue line) was repeated while it remained similar when sequence variants (red line) were played back. CorrCoef values were higher than those of spike trains in which spike timing was randomly permutated (yellow line). (h) Varying the Gaussian window width used to compute the convolution of spike trains from 1 to 200 ms affects CorrCoef values. In the present study, a 10 ms Gaussian window width to compute CorrCoef values (vertical dashed line) and CorrCoef values differed between the two series. No difference between ABAB-Same and ABAB-Var was observed when the time window exceeds 98 ms. CorrCoef values were also computed on spike trains after a random permutation of the spike timing. Significant difference: *p < 0.05, **p < 0.01, ***p < 0.001 (see main text for statistics details).

Figure 5

Responses to the two AB pairs that form ABAB sequences reflects sensitivity to the context in awake birds. (a) For ABAB-Same series, the same syllables A and B were used to build pairs of syllables AB that were repeated twice within a sequence, so two consecutive sequence renditions (n) include two repetitions of the same pairs of syllable AB. For ABAB-Var series, variants of the syllables A and B were used to build pairs of syllables AB that were repeated twice within a sequence so two consecutive sequence renditions include variants of the pairs of syllable AB repeated twice. (b) Strength of responses (RS values) changed from the first AB pair to the second one. The exposure to the first pair of syllables AB impacts the responses to the second pair of syllables AB within a stimulus rendition in awake birds (b–d). Evoked auditory responses (b) and CorrCoef (c) were overall higher for ABAB-Same than for ABAB-Var sequences and were lower for the second pair of syllables AB than for the first pair. Yet, Pearson correlation coefficient measured on each individual spike train between the first and second pair of syllables AB was lower for ABAB-Var than ABAB-Same sequences (d). *, ** and ***, p < 0.05, 0.01 and 0.001, respectively (see main text for statistics details).