Figure 1.
Repetition rate transfer functions for VS and FR.
(A) Poststimulus time histograms for 20 repetition rates for an AAF site. Response strength was normalized to the maximum response at 1 Hz. Maximum height of the FR ordinate: 15 spikes. Information values for 111L-S98: VS info: 0.62 bits/stimulus; FR info: 0.23 bits/stimulus; ISI (1 ms) info: 0.51 bits/stimulus; ISI (10 ms) info: 1.95 bits/stimulus. (B) Corresponding RRTFs for VS (magenta line) and FR (blue line). Data points are fit by a polynomial cubic spline. Filled circles are significant VS values (Rayleigh test, p<0.001). Gray background illustrates the repetition rate range at the focus in this study. (C) ISI histogram for 6 repetition rates. Multiple ISI peaks correspond to integer multiples of stimulus periods. Information values for 111L-S93: VS info: 0.09 bits/stimulus; FR info: 0.11 bits/stimulus; ISI (1 ms) info: 1.20 bits/stimulus; ISI (10 ms): 2.28 bits/stimulus. (D) Population distribution of the coefficient of variation (CV) of ISIs for hemisphere 111L (n = 130). CV of ISI that estimates the variability of ISIs was computed by dividing the standard deviation of ISIs by the mean. Histograms of CV distributions, smoothed by a polynomial cubic spline, are illustrated for six different repetition-rates.
Figure 2.
Mutual information (MI) contained in VS, FR, and ISI.
(A) Mean (± standard error of the mean) of MI values for VS, FR, ISI (1 ms) and ISI (10 ms) for all three hemispheres. The global mean is indicated as a dashed line across the three hemispheres. MI for ISI (10 ms) was based on intervals equal or larger than 10 ms, whereas MI for ISI (1 ms) contained all intervals equal or larger than 1 ms. Paired t tests adjusted by the sequential Bonferroni correction for multiple comparisons (p<0.001) were performed for the three global mean measures. A theoretical MI value for distinguishing six different repetition rate stimuli is 2.58 bits/stimulus ( = log2(6)). (B) Information captured for different combinations of a joint repetition rate code. Black/gray bars: additive combination of the two codes (Code(x) + Code(y)). The number of sites that resulted in valid joint information value was lower than the total number of sites for the individual information analysis. The summed information is based on recording sites that had a valid joint information. White bars: joint information values for two codes (Code(x) × Code(y)). Unpaired t tests for the comparison between additive and joint codes (p<0.001).
Figure 3.
Correlation between information and three temporal response measures.
(A) Positive correlation between VS max and ISI info (p<0.05). (B) Negative exponential correlation between ISI info and FR max (p<0.0001). (C) Negative correlation between ISI info and CV min (p<0.0001). (D) Weak positive correlation between VS info and VS max (p<0.05).
Table 1.
Principal component analysis.
Figure 4.
Spatial distribution of population response to different repetition rates.
(A) Tonotopic gradient smoothed by a weighted least-squares linear regression model is reconstructed on the cortical surface by Voronoi-Dirichlet tessellation. An approximate location of AAF is indicated by the suprasylvian sulcus (sss) and the anterior ectosylvian sulcus (aes; thick black lines). Hemisphere 111L; D: dorsal, A: anterior, scale bars: 1 mm. (B) Spatial representation of VS as a function of different repetition rate. Repetition rates are shown on the top. White polygons indicate sites not tested for the corresponding repetition rates, which also apply to (C, D). Raw VS values of the Voronoi-Dirichlet tessellation maps were smoothed by a weighted least-squares linear regression model. (C) Spatial distribution of FR as a function of repetition rate. FR magnitude was normalized to the peak rate for the corresponding repetition rate. Normalized FR magnitudes were smoothed by a weighted least-squares linear regression model. High activity sites with FR>0.75 in the smoothed maps were categorized and shown in the bottom panel as red polygons. Sites with FR<0.75 are illustrated by gray polygons. (D) Spatial distribution of CV of ISI as a function of repetition rate. Spatially smoothed maps are shown. D: dorsal, A: anterior, scale bars: 1 mm. The scales also apply to (B, C).
Figure 5.
Map similarity for repetition rate differences.
(A) Spatial cross-correlation values of raw VS values generated by repetition rates between 1 and 30Hz (see Fig. S4A) and plotted as a function of the logarithmic repetition rate difference for all three hemispheres. The solid line is a logarithmic fit. The gray area indicates non-significant correlation values (n = 16/45). (B) Cross-correlation values of raw FR generated by repetition rates between 1 and 30Hz (see Fig. S4B) and plotted as a function of the linear repetition rate difference. Non-significant correlations: n = 19/45. (C) Cross-correlation values of CV of ISI generated by repetition rates between 1 and 30Hz (see Fig. S4C) and plotted as a function of the logarithmic repetition rate difference. Non-significant correlations: n = 12/45.
Table 2.
Spatial clustering statistics of cortical maps.
Figure 6.
Spatial distributions of temporal response measures and mutual information.
(A) Spatial distributions of CV min of ISI, VS max, and FR max for hemisphere 111L. Minimum or maximum value of the measures for any of the repetition rates is shown. White dots indicate polygons with statistically similar values as their direct neighbors (compared to random re-distribution of all neighbors, see Materials and Methods). Gray polygons indicate sites not available due to the four tested repetition rates, which also apply to (B, C). (B) Spatial distributions of mutual information values of ISI, VS, and FR based on repetition rate discrimination. (C) Spatial distribution of three temporal factors emerging from a principal component analysis of CV min, FR max, VS max, as well as the three corresponding information measures. Both white and black dots indicate polygons with statistically similar values as their direct neighbors.
Figure 7.
Spatial distributions of temporal response factors.
(A) Spatial distribution of the magnitudes of three temporal factors based on a principal component analysis of CV min, FR max, VS max as well as the three corresponding information measures (hemisphere 073L). White and black dots indicate polygons with statistically similar values than their direct neighbors (compared to random re-distribution of all neighbors, see Materials and Methods). Significant local clustering: *: p<0.05; **: p<0.01. D: dorsal, A: anterior, scale bars: 1 mm. For the tonotopic gradient, see Figure S3. (B) Same as (A) for hemisphere 073R.
Figure 8.
Local versus global spatial organization in AAF.
A scatter plot of the mean proportion of polygons with high magnitude similarity to directly neighboring polygons (local spatial organization) versus a mean spatial autocorrelation measure (Geary's C; global spatial organization) for all three hemispheres (see Table 2). Error bars indicate standard error of the mean. Light gray shading indicates statistically non-significant regions for either measure. Dark gray area corresponds to value range that is not statistically significant for either local or global measures. A linear regression line is shown (r2 = 0.92, p<0.001). CF = characteristic frequency; Q40 = frequency tuning curve bandwidth (at 40 dB above threshold)/CF; Lat = minimum response latency at CF; F1(ISI) = strongest temporal factor comprising CV min, ISI info, and FR max; F2(VS) = second strongest temporal factor comprising VS max and VS info F5(FR) = third strongest temporal factor comprising FR info.