Shift of ITD tuning is observed with different methods of prediction.

Transcription

1 Supplementary Figure 1 Shift of ITD tuning is observed with different methods of prediction. (a) ritdfs and preditdfs corresponding to a positive and negative binaural beat (resp. ipsi/contra stimulus frequency 300 Hz/301 Hz and 300 Hz/299 Hz). Shift is unrelated to the sign of the binaural beat. This shows that the shift is dependent on ITD and excludes a dynamic artifact of the stimulus. Characteristic frequency = 4.6 khz. (b) PredITDf for different prediction threshold values. Black line is ritdf. Stimulus: 300/301 Hz 70 db (characteristic frequency = 1895 Hz). (c) Difference between BD of ritdf and preditdfs from b as a function of threshold. Predicted BD is not dependent on threshold. (d,e) ITD function is compared with the monaural prediction for two datasets. PredITDf is similar whether the complete monaural responses are summed and then thresholded (solid red line; as in Fig. 3), or the monaural EPSP period histograms are crosscorrelated (dashed red line). The latter method of prediction uses only the timing of the EPSP peaks and diminishes the weight of the amplitude differences between EPSPs, and ignores other aspects of the monaural input, such as hyperpolarizing events. Stimulus 400/401 Hz 70 db (d) and 400/401 Hz 80 db (e). Characteristic frequencies: 1741 Hz (d) and 1149 Hz (e). (f) Comparison of shift for both methods of prediction as in d and e at the population level. Only datasets with significant suprathreshold ITD tuning were included (Rayleigh test α <= 0.001). Symbols correspond to Figure 3b,d. Dashed line is linear fit (linear correlation; t(56) = 6.23; 58 datasets from 24 neurons).

2 Supplementary Figure 2 Individual traces preceding supraepsps and subepsps for ITDs with the same supraepsp rate. ITD 1 and ITD 2 as indicated in Figure 6a. Rows correspond to the datasets in Figure 6a-c. Data traces are aligned at the peak of the first derivative of the EPSP. In the datasets with a shift (top two rows), subepsps are preceded more frequently by smaller EPSPs (arrows), corresponding to the relative depolarization in Figure 6b.

3 Supplementary Figure 3 Leading depolarization hinders spiking in response to broadband noise. (a-h) SupraEPSPs (black) and subepsps (red) are stacked for eight MSO neurons during binaural stimulation with broadband noise. Events were aligned on the peak of the first derivative of the EPSP. Responses were pooled across binaural conditions (e.g. different ITDs, correlated and uncorrelated noise). Right panels show the averages for supraepsps (black) and subepsps (red) separately. As for stimulation with tones, preceding relative depolarizations are associated with spike failures. CF is indicated in the right panels when known. (i) Average V m preceding subepsps and supraepsps 1 to 0.5 ms before the maximum of the first derivative of the EPSP for the neurons in a-h. Grey lines connect individual values. Red and black lines are population averages. Significance was assessed using a one-tailed paired t-test (t(7) = 7.46). Characteristic frequency is indicated in panels a-h when known.

4

5 Supplementary Figure 5 Coincidence detection in mammalian sound localization is adaptive rather than instantaneous. (a) Left, scheme of instantaneous coincidence detection. Output results directly from events coinciding within a narrow time window (coincidence window, cw). Right, adaptive coincidence detection takes into account preceding EPSPs and is not simply predicted by coincidence of events in the coincidence window. (b) Top two rows: cycle histograms for symmetrical (first row) and asymmetrical (second row) inputs at different ITDs. Bottom: ITD functions corresponding to symmetrical inputs (black), instantaneous coincidence detection for asymmetrical inputs (dashed red line) and adaptive coincidence detection for asymmetrical inputs (solid red line). For the asymmetrical inputs, the contralateral ear evokes fewer early EPSPs. With instantaneous coincidence detection this shifts the ITD curve to the right, favoring sounds that arrive at the contralateral ear first. With adaptive coincidence detection, the small group of leading EPSPs in the contralateral input decrease spiking when sounds arrive at the contralateral ear first, thereby shifting the ITD curve to the left.