Online contributions of auditory feedback to neural activity in avian song control circuitry

Abstract

Birdsong, like human speech, relies critically on auditory feedback to provide information about the quality of vocalizations. Although the importance of auditory feedback to vocal learning is well established, whether and how feedback signals influence vocal premotor circuitry has remained obscure. Previous studies in singing birds have not detected changes to vocal premotor activity after perturbations of auditory feedback, leading to the hypothesis that contributions of feedback to vocal plasticity might rely on"offline" processing. Here, we recorded single and multiunit activity in the premotor nucleus HVC (proper name) of singing Bengalese finches in response to feedback perturbations that are known to drive plastic changes in song. We found that transient feedback perturbation caused reliable decreases in HVC activity at short latencies (20-80 ms). Similar changes to HVC activity occurred in awake, nonsinging finches when the bird's own song was played back with auditory perturbations that simulated those experienced by singing birds. These data indicate that neurons in avian vocal premotor circuitry are rapidly influenced by perturbations of auditory feedback and support the possibility that feedback information in HVC contributes "online" to the production and plasticity of vocalizations.

Figure 1.

Schematic of the avian song system. HVC is a vocal premotor nucleus that provides major input to the rest of the song system, including the anterior forebrain pathway [Area X, medial portion of the dorsal lateral nucleus of the anterior thalamus (DLM), and lateral magnocellular nucleus of the anterior nidopallium (LMAN)], which is implicated in song plasticity, and RA (Doupe and Konishi, 1991; Vicario and Yohay, 1993). HVC receives inputs from auditory areas that include the NIf and nucleus Uva (Bottjer et al., 1989; Foster and Bottjer, 1998; Cardin and Schmidt, 2004; Coleman and Mooney, 2004; Cardin et al., 2005; Coleman et al., 2007).

Figure 2.

Singing-related activity of HVC neurons in the Bengalese finch. a, Multiunit activity during singing at one recording site. Plotted from top to bottom are the spectrogram of the song, an oscillogram of the song, and a raw trace of HVC activity during song production. Above the spectrogram are labels indicating recurring sequences of syllables (e.g., “abcd,” “efgg…”). b, Summary of singing-related activity during the fixed sequence “efgg” (same site as in a). Plotted from top to bottom are the spectrogram of the sequence, raster plots of firing during 15 iterations of the sequence and a PSTH summarizing firing rate (mean ± 1 SEM) during production of the sequence. The dashed line represents the mean firing rate at this site when the bird was quiescent. Neural data are aligned by the onset of the syllable “e.” c, Well isolated unit activity (putative interneuron) at a recording site in another bird. Same organization as a. Above the spectrogram are labels indicating a recurring sequence of syllables (“sabb…”). d, Summary of singing-related activity during the fixed sequence “sabb” (same neuron as in c). Plotted from top to bottom are the spectrogram of the sequence, raster plots of firing during 15 iterations of the sequence and a PSTH summarizing firing rate (mean ± 1 SEM) during production of the sequence. Neural data are aligned by the onset of the syllable “a,” and the dashed line represents the mean firing rate when the bird was quiescent. In both cases, HVC singing-related activity increased before the onset of song, remained elevated throughout song and exhibited a consistent pattern of modulation across multiple renditions of the sequence. On average activity was 12.6 times higher during singing than when birds were quiescent [16.9 times higher for multiunit sites (n = 7) and 6.6 times higher for single unit sites (n = 5)].

Figure 3.

Experimental design for altering auditory feedback. We selected a target syllable and created spectral templates to that syllable. After detection during ongoing song, birds experienced either altered feedback (feedback trials) or normal feedback (control trials) with equal probability. During feedback trials, a prerecorded sound (feedback element: syllable from the male's repertoire) was played back at a short and fixed latency via a free-field speaker located above the bird so that the singing bird experienced a superposition of the extraneous syllable on his own normal feedback. Feedback and control trials were randomly interleaved, which allowed us to directly compare HVC activity at single recording sites under altered and normal feedback conditions.

Figure 4.

Effect of auditory feedback perturbation on multiunit and single unit HVC activity during experiments in which acute changes to song production were not observed (sensory-only). a, Representative example of the effect of feedback perturbation on HVC multiunit activity. Spectrogram is plotted at top with a PSTH (mean ± 1 SEM) of spiking activity plotted below. Data are aligned by the onset of the feedback element (0 ms). A significant, localized decrease in ongoing singing-related activity was observed beginning 54 ms after the onset of feedback perturbation. Red trace corresponds to feedback trials (n = 85) and blue trace to control trials in which the same sequence of syllables was produced but feedback remained normal (n = 68). The asterisk and bar indicate the period (54–63 ms) during which HVC activity was significantly reduced after altered feedback. i, The same data are plotted on an expanded time base for the first 100 ms after feedback perturbation. ii, Raster plot of spiking activity 0–100 ms after the onset of perturbative feedback for the first 30 control and feedback trials. The shaded region corresponds to the period of time in which HVC activity was significantly reduced during feedback trials relative to control trials. iii, Histogram summarizing neural activity during control (blue) and feedback (red) trials when d′ values were maximal (59 ms after feedback onset). Triangles indicate means for control (blue) and feedback (red) trials. iv, Distribution of d′ values derived from a randomization test under the assumption that there was no difference between control and feedback trials at this time point (see Materials and Methods). The measured peak value of d′ (0.91) exceeded the 99th percentile of this distribution (dotted line), indicating a significant effect of feedback perturbation. The magnitude of the effect at this site was near the median value for all experiments (median peak d′ = 1.03). b, Example illustrating one of the largest observed effects of feedback perturbation on HVC activity (feedback, n = 24; control, n = 23). The d′ had a peak value of 1.62 at 43 ms after feedback onset. Layout same as in a. c, Example illustrating the effect of feedback perturbation for a well isolated neuron (putative interneuron) in HVC (same neuron as in Fig. 2c; feedback, n = 17; control, n = 24). The d′ had a peak value of 1.47 at 53 ms after feedback onset. Layout same as previous panels.

Figure 5.

Changes to HVC activity in feedback experiments without (sensory-only) and with (sensory-motor) changes to the temporal structure of song. a, A sensory-only experiment in which feedback perturbation did not significantly affect song output (top) but led to a transient decrease in HVC activity beginning ∼40 ms after feedback onset, indicated by bars above PSTHs (bottom). The red and blue traces correspond to data from feedback and control trials, respectively. The dark bars above the PSTH correspond to the periods in which HVC activity was significantly decreased during feedback trials relative to control trials. Below the plot of HVC activity are bars indicating the duration of three time windows used in our analysis: “pre window” (−150–0 ms), early window (20–80 ms after feedback onset), and late window (80–140 ms after feedback onset). b, Percentage changes in HVC activity caused by feedback alteration for each sensory-only (n = 25) experiment for the early and late windows. Gray circles represent multiunit experiments, and white circles represent single unit experiments. The dashed line indicates the mean (±1 SEM) difference in HVC activity during the pre window, a control period during which vocal motor output and sensory feedback were matched between control and feedback trials. The change in activity caused by feedback perturbation was significant only for the early window (13.7 ± 1.9% decrease; t test, H₀, mean = 0; p < 0.0001). The gray bar represents the mean percentage change for each time window, with the black lines corresponding to one SEM. p < 0.05. c, Peak (signed) d′ values during the early window for sensory-only experiments. Negative values indicate that HVC activity was reduced during feedback trials relative to control trials. The mean d′ value was significantly less than zero (−0.98 ± 0.15; t test; p < 0.0001). Gray and white bars indicate multiunit and single unit experiments, respectively. d, A sensory-motor experiment in which the superposition of an extraneous syllable caused a localized slowing of song (top), indicated by a rightward shift in the sound trace during feedback trials (red) relative to control trials (blue). Correspondingly, there was a shift in neural activity during feedback trials relative to control trials (bottom). This persistent shift resulted in repeated periods during which HVC activity was significantly different between feedback and control trials, indicated by bars above PSTHs. The dark and white bars above the PSTH correspond to the periods during which HVC activity was significantly decreased and increased, respectively, on feedback trials. e, Percentage changes in HVC activity caused by feedback alteration for each sensory-motor (n = 23) experiment for the early and late windows. Gray circles represent multiunit experiments, and white circles represent single unit experiments. The dashed line indicates the mean (±1 SEM) difference in HVC activity during the pre window. The change in activity caused by feedback perturbation was significant for the early (17.9 ± 3.4% decrease; t test; p < 0.005) and late (11.3 ± 3.3%; t test; p = 0.002) windows. p < 0.05. f, Peak (signed) d′ values during the early window for sensory-motor experiments. The mean d′ was significantly less than zero, and similar to that observed in sensory-only experiments (−0.98 ± 0.10; t test; p < 0.0001). Gray and white bars indicate multiunit and single unit experiments, respectively.

Figure 6.

HVC neurons of awake Bengalese finches were selectively responsive to playback of BOS relative to other complex stimuli. a, Example of multiunit auditory responses in HVC (same site as depicted in Fig. 2a). In the left column are data for playbacks of BOS, and to the right are data for playbacks of rBOS. i, Spectrogram of the BOS. ii, Oscillograms of BOS and rBOS. iii, Raw trace of neural activity elicited in response to one playback. Calibration: 0.1 mV (y-axis), 1 s (x-axis). iv, Raster plots showing the timing of spikes for 20 stimulus presentations. v, PSTH of the mean firing rate during stimulus presentation. The RS for BOS and rBOS for this site were 19.1 and 2.2, and the d′_BOS-rBOS was 3.53, indicating strong selectivity for BOS. b, Auditory responses at a recording site with well isolated unit activity (putative interneuron) in an awake, passively listening Bengalese finch (same neuron as depicted in Fig. 2c). Organization is the same as for a. The RS for BOS and rBOS for this neuron were 8.4 and 0.3, respectively, and the d′_BOS-rBOS was 1.36. c, Summary of auditory responses for 55 experiments in 11 birds in which stimuli (BOS, rBOS, or CON) were played back to awake but quiescent birds. The two top panels compare the RS of BOS to rBOS (left) and to CON (right). Data obtained when lights were turned off 1–5 min before playbacks are plotted as filled squares (n = 35); those obtained when lights remained on are plotted as empty circles (n = 20). The bottom panel shows a histogram of d′_BOS-rBOS values from these 55 experiments (lights off, dark bars; lights on, empty bars). Both RS and d′_BOS-rBOS scores were not significantly different when data were obtained with the lights on vs off. Responses of HVC neurons were selective for BOS (d′_BOS-rBOS >0.5) for 53 of the 55 experiments, with an average d′ of 3.01.

Figure 7.

Responses of HVC neurons to auditory stimuli that simulated those experienced while birds sang under conditions of experimentally perturbed feedback. a, Example of an experiment in which a feedback syllable was superimposed on playback of BOS. Top, Spectrogram of auditory stimulus. Bottom, PSTH of HVC activity during normal (blue) and altered (red) renditions of BOS. Playback of BOS selectively and robustly increased HVC activity at this site. HVC activity decreased ∼50 ms after the onset of the superimposed feedback element. b, Plot comparing mean (±SEM) multiunit HVC activity during a 60 ms window (20–80 ms after the onset of perturbative stimulus) for trials with normal and altered BOS across 16 experiments. For 13 of these experiments, HVC activity during this window was significantly lower during altered BOS trials than during normal BOS trials, and for two experiments, HVC activity was significantly higher during altered BOS trials. Overall, mean HVC activity was significantly lower when BOS was altered (paired t test; p = 0.0111). These data indicate that the response of HVC neurons to localized deviations from the sound of the bird's own song during playback is qualitatively similar to the response of HVC neurons to the same perturbation of feedback during singing.