Extended Data Fig. 4: Neural similarity in budgerigar AAC is not driven by vocal reuse of elemental components.

a, Sonogram of an example zebra finch song recorded using the same piezoelectric microphone as used for budgerigar recordings. b, Schematic of the methodology for quantifying correlations between unique and repeated vocal segments. Each zebra finch syllable was decomposed into 20-ms segments using a sliding window with a 10-ms step size. Correlations for repeated segments (vocal reuse) were calculated between corresponding segments across 10 renditions (2 renditions shown for illustration) of each syllable. Correlations for unique segments (vocal similarity) were calculated between segments from different syllables. c, Distributions of correlation values between vocal similarity (orange) and reuse (blue), with 0.6 set as a threshold to distinguish between the two conditions. d, Distribution of correlation values across all vocal segments analyzed in Fig. 2 for each budgerigar, with percentage of correlations above 0.6 indicated on the right, representing potential vocal reuse. e, Correlation between spectral similarity and neural similarity matrices for an example budgerigar (left) and accompanying population data (right), excluding all cases with spectral correlation values exceeding 0.6. ρ denotes Spearman’s correlation value. P-values displayed within each plot are from two-sided permutation tests (n = 216,848, 1,113,918 pairs of vocal segments for each plot, respectively).