Figure 1.
Experimental setup included wn-rearing and social isolation.
Experimental setup included normal acoustic and social rearing for control birds, and a continuous white-noise exposure and social isolation for wn-reared birds. The song output of one control male and one wn-reared male are shown in spectrogram form on the right: the control males sang a normal song, typical of normal/tutored zebra finches; and the wn-reared males sang a scratchy, perseverative song, typical of untutored/isolate song. Although we did not systematically study the differences in vocal output, we used song production as a means to verify the effectiveness of the white noise exposure and isolation since song development in males depends on both natural auditory spectro-temporal cues and social interactions.
Figure 2.
Power spectra and statistically-matched synthetic stimuli used in neural selectivity analyses.
(left panel) Spectra were estimated using all the sounds used in the experiments, approximately 40 s of sound for each stimulus type. Tones and Pips ensembles were designed to match the power spectrum of conspecific song (Con) and have very similar bell-shaped spectra. Discrepancies between the Pips and the Con ensembles are due to sampling errors. White Noise and Ripples stimuli have flat power spectra between 1 and 7 kHz. (right panels) Spectrographic representation (frequencies ranging from 500 to 8,000 Hz on the y-axis and time in seconds on the x-axis) of exemplars of matched synthetic stimulus types (Pips, Tones, Ripples) used in analyzing neural responsivity and selectivity in our study. Note that the sounds in these exemplars begin at 0.5 s and last about 2 s each.
Figure 3.
Example neural responses from a control bird and a wn-reared bird to a subset of the selectivity stimuli.
Spectrographic representation of exemplars of a subset of stimulus types (Con, Pips, Tones, WN) and corresponding neural responses for 2 recording sites, one from a control adult (top) recorded in L1 and the other from a wn-reared bird recorded in L3 (bottom). Note that sound begins at 0.5 s. For the neural response, both the spike raster for 10 trials (middle) and the PSTH (denoted by spikes/s on the bottom) are shown. These examples were chosen to reflect the characteristics in the average neural responses in control adults versus wn-reared adults: the recording site from the control adult shows robust responses to Con and WN, whereas the recording site from the wn-reared adult shows decreased responses to Con and enhanced responses to Tones. In this example, as in the average data, the wn-reared recording site was more variable (Con FF = 1.16 and Pip FF = 1.40) than the control recording site (Con FF = 0.91 and Pip FF = 0.91). While these responses were chosen as illustrative of the average neural responses, we also found a wide range of response properties, including neurons in wn-reared animals that showed strong and reliable responses to song (as shown below). The spontaneous background activity was variable across units in both control and wn-reared birds but similar in rate across the two rearing conditions. To conserve space, we omitted showing the response to Ripples.
Figure 4.
Neural responsivity and selectivity, as measured by z-scores and d′ values respectively.
A. Comparison of control and wn-reared mean z-scores for all the stimuli used in the selectivity analysis (Con, Pips, Tones, Ripples, and WN). Average z-scores for all stimulus-excited responsive units show that responses to the more complex stimuli of Con, Ripples, and WN are reduced in wn-reared birds compared to control birds. Error bars represent 2 SEs. B. Cumulative distribution functions (cdf) of d′ values for the Con-Pips, Con-Tones, Con-Ripples, and Con-WN comparisons for both controls and wn-reared adults. Selectivity analyses (and the resulting cumulative curves) show the greatest divergence in rearing conditions for the Con-Tones, followed by the Con-Pips and Con-Ripples comparisons. C. Working model of the development of the neural selectivity for natural sounds such as song (d′>0 for song as compared to synthetic sounds), as a function of development and natural rearing environments, in the auditory system of songbirds. The bars in this schematic summarize the distributions of d′ values obtained from our data from control adults, juveniles, and wn-reared and socially-isolated adults.
Figure 5.
Neural responsivity, selectivity, and reliability in subfields L1 (∼Layer III), L2 (∼Layer IV), and L3 (∼Layer V).
A. Comparison of control and wn-reared mean z-scores (pooled over all five stimuli: Con, Pips, Tones, Ripples, and WN) broken down by subdivisions in field L (as assessed by histological analysis). Average z-scores for all stimulus-excited responsive units show that responses to the more complex stimuli of Con, Ripples, and WN are reduced in L1 and L3 in wn-reared birds compared to control birds. Error bars represent 2 SEs. B. d′ values for the Con-Tones comparison for controls and wn-reared adults broken down by subdivisions in field L. Selectivity analysis shows noise-rearing had the greatest effect on subfields L1 and L3. Error bars represent 2 SEs. C. Fano factor for Pips (left) and Tones (right) for controls and wn-reared adults broken down by subdivisions in field L. Noise-rearing decreased the response reliability for neurons in subfields L1 and L3. Error bars represent 2 SEs.
Figure 6.
Single neuron information values greater for ML-Noise than song in wn-reared condition.
A. Pairwise difference in Gamma Information rates (bits/s) for Song vs. ML-Noise estimated for single neurons for control birds (left) and wn-reared birds (right). Neurons in the auditory forebrain of wn-reared birds encode for ML-Noise more optimally than Song, while auditory neurons in control birds encode both ML-Noise and Song equally well. Even upon adjusting for firing rate differences using a general linear model, wn-reared neurons have higher Gamma Information values than the control neurons for both Song and ML-Noise, with a larger effect for ML-Noise (see Results). B. To control for potentially small differences in the types of receptive fields found in wn-reared birds compared to control birds, we also estimated the Gamma Information (bits/s) for single neurons in control birds that had similar (‘matched’) SIs to the receptive fields found in wn-reared birds. The pairwise difference in Gamma Information for Song vs. ML-Noise is still negligible in the matched control case, while the neurons in wn-reared birds encode for ML-Noise more optimally than Song. Even upon adjusting for firing rate differences using a general linear model, neurons from the wn-reared condition still have higher Gamma Information values than the matched control neurons for both Song and ML-Noise, and once again the effect size is larger for the ML-Noise stimuli than for the Song stimuli (see Results).
Figure 7.
Ensemble neural coding for song more redundant in wn-reared birds.
A. Mutual Information for neuronal ensembles of two to ten neurons for the control and the wn-reared cases in response to Song (on the left panel, in bits/s and on the middle panel, in bits/spike). Error bars represent one standard error obtained by randomly sampling neurons from our dataset. Redundancy in information transmitted as a function of the number of neurons for the control and the wn-reared cases in response to Song (on the right panel). Errors bars represent one standard error obtained by randomly sampling neurons from our dataset. B. Same as in A but for ML-Noise stimuli.