Fig. 3. State-Space Analysis of Motor Nuclei Activity in the Song-Motor Pathway.

a. Amplitude waveform and spectrogram of a recorded song bout from a male zebra finch. The bout features multiple motifs, comprising individual syllables ($S_1$ to $S_5$), color-coded from black to brown, with additional intra-motif notes ($S_6$, green) and introductory notes ($i$, pink) preceding the first motif. b. HVC neural trajectories during motif repetition, inferred by projecting HVC activity onto a latent manifold defined by the leading three GPFA factors (left). Trajectory subsegments are color-coded to match associated song syllables. The same trajectories are also shown projected onto each of the three-dimensional axes (GPFA factors), arranged in descending order by variance explained (right). c. Same as in (b) but for GPFA-inferred neural trajectories in RA. Comparing b and c highlights the larger degree of smoothness and temporal structure in the activity of RA across vocal repetitions compared to the greater variability observed in HVC during the same song bout.