Recovery of pitch target requires practice.

(A) Three hypotheses on birds’ ability to recover a song target away from their current vocal output (green): 1) they could recall its motor program (M), or 2) its sensory representation (S) plus the mapping (left black arrow) required to hit it, or, 3) the sensory target (T) and the circuit for translating that into the original motor program (arrows). (B) WNm birds were first pitch-reinforced using white noise (WN), then muted, and subsequently unmuted. Pitch recovery from the reinforced (R) state towards the baseline (B) target is evaluated in early (E, no practice) and late (L, with practice) analysis windows (all windows are time-aligned to the first 2 h of songs after withdrawal of reinforcement, E) and compared to recovery in unmuted control birds (WNC). (C) Syllable pitches (dots, red=reinforced syllables) of an example bird that while muted recovered only about 27% of pitch difference to baseline despite three spontaneous unmuting events (arrows). (D) Same bird, spectrograms of example song motifs from 5 epochs: during baseline (B), reinforcement (R) with WN (green bar), spontaneous unmuting (spont. unmut), and during permanent unmuting (early – E and late - L). (E) Example syllables from same 5 epochs. (F) Stack plot of pitch traces (pitch indicated by color, see color scale) of the first 40 targeted syllables in each epoch (‘reinforced’: only traces without WN are shown). (G)Average pitch traces from F), revealing a pitch increase during the pitch-measurement window (dashed black lines) and pitch recovery late after unmuting. (H)WNm birds (blue lines, N = 8) showed a normalized residual pitch (NRP) far from zero several days after reinforcement (circles indicate unmuting events, arrow shows bird from C) unlike WNC birds (gray lines, N = 18). (I) Violin plots of same data restricted to early and late analysis windows (***p < 0.001, *p < 0.05, two-tailed t-test of NRP = 0).

Recovery of pitch target is impaired after deafening.

(A) WNd birds were first pitch-reinforced using white noise (WN) and then deafened by bilateral cochlea removal. Analysis windows (letters) as in Fig. 1. (B) Syllable pitches (dots, red=reinforced syllables) of example WNd bird that shifted pitch down by d’ = −2.7 during WN reinforcement and subsequently did not recover baseline pitch during the test period. (C) WNd birds (N = 10) do not recover baseline pitch without auditory feedback (circles=early window after deafening events, cross=late). (D) Violin plots of same data restricted to early and late analysis windows (***p < 0.001, *p < 0.05, two-tailed t-test of NRP = 0).

Deaf birds do not recover pitch target after light-induced mismatch.

(A) dLO birds were first deafened and then pitch-reinforced using a brief light-off (LO) stimulus. Analysis windows (letters) as in Fig. 1. (B) Syllable pitches (dots, blue=LO-reinforced syllables) of example dLO bird that shifted pitch up by d’ = 3.5 within a week, but showed no signs of pitch recovery during the test period. (C) dLO birds (N = 8) do not recover baseline pitch without auditory feedback. (D) Violin plots of same data restricted to the late analysis window (***p < 0.001, two-tailed t-test of NRP = 0).

Target mismatch experience is necessary for revertive pitch changes.

(A) WNd birds heard a target mismatch during reinforcement whereas dLO birds did not. dC birds were not pitch reinforced, their analysis windows matched those of manipulated birds in terms of time-since-deafening. (B, C) Pitch change between the last 2 h of reinforcement (R) and 2 h windows during the test period time-aligned to the first 2 h of song after withdrawal of reinforcement (E) in std for WNd (red, B), dLO (blue, C) birds. (D) WNd (red) perform both early and late pitch changes in the direction of the baseline target (by about one standard deviation, * p < 0.05, one-tailed t-test), similar to WNC (gray) and unlike dLO (blue) birds without mismatch experience. (E) Bootstrapped pitch differences between reinforced WNd (blue) and dLO (red) and 10’000 times randomly matched dC birds, shown for early (solid line) and late (dashed line) analysis windows. The stars indicate the bootstrapped probability of a zero average pitch difference between reinforced and dC birds (n.s. not significant, ** p < 0.01, *** p < 0.001).

Schematic illustrating the goal-directed planning of vocal changes.

The song variability from the last few hours (green arrows, black density) limits (blue dashed lines) the song targets (filled black dot in M) that birds can overtly plan without sensory feedback. Sensory experience (green circles under S) is a prerequisite for consolidating motor plans and for reaching (curved blue arrow) a target (T) beyond the planning range, e.g. a song variant produced several days ago.