Figure 3

Relationship between the preference for father’s song and imitation of that song. (A) The white line shows the average trajectory of the preference for father’s song. The black area indicates the 95% confidence interval. The horizontal dashed line indicates no preference for either song. Vertical dashed lines indicate boundaries between developmental phases (see Fig. 1A). Concentric blue lines represent the 2D density kernel estimation of the daily preference scores. Ticks near the bottom of the plot indicate the age at which individual birds reached their peak preference for father’s song prior to the crystallization phase (for individual trajectories, see Fig. S2). Given that crystallization may begin as early as 70 dph²⁶, we chose the maximum preference reached before that age (see Fig. S5 for an alternative analysis using the peak preference over the entire learning trajectory, including crystallization). Ticks are drawn with an offset to prevent overlap. Preferences peaking during the phase known as auditory learning (see Fig. 1) are marked in orange, and those peaking during the plastic song phase are marked in teal. (B) Peak preference for father’s song was significantly correlated with mean similarity to that song. The curve is the fit of a beta regression. Vertical bars show the standard error of the mean similarity score. Data points are colored according to the phase during which the bird reached maximum preference for father’s song, as indicated in (A). (C) Spectrograms of the songs of three pupil exemplars are shown next to the spectrograms of the songs of their respective father and neighbor. The spectrograms indicate that these three pupils imitated father song and not neighbor song. All other pupils also imitated father song or, in the case of one individual, did so most of the time (see text). Spectrograms were generated in R⁷³ using Hanning windows containing 512 samples each (11.61 ms for recordings sampled at 44,100 kHz), with 50% overlap between windows.