The Effects of Pitch Shifts on Delay-Induced Changes in Vocal Sequencing in a Songbird

Experimental design. A, Auditory feedback manipulated using miniaturized headphones. A microphone transmits a bird’s vocalizations to online sound processing hardware, which is used to introduce a 175-ms delay (with or without a ±3 semitone pitch shift). B, Schedule of experimental conditions. Each bird was exposed to the delayed feedback alone (DAF) and delayed feedback plus a pitch shift (DAF+PS). The order of these conditions, as well as the direction of the pitch shift, was randomized across subjects. Before each DAF or DAF+PS epoch, birds sang in a null epoch free of delays or pitch shifts. The four experimental epochs lasted 5 d each.

Effects of auditory feedback manipulations on branch point probabilities. A, One branch point (in which syllable b can be followed by either syllable a or c), of nine total branch points in our dataset. Spectrogram plotting conventions as in Fig. 1A. B, In this experiment, after a null period of unmanipulated auditory feedback, the bird experienced DAF followed by another period of unmanipulated feedback, followed by DAF+PS. Green and orange traces show the probability of the b-to-c and b-to-a transitions, respectively.

Effects of auditory feedback manipulations on branch points (group data). A, Probability of the primary (i.e., most common; see Methods) transition in the null versus DAF conditions. Filled circles indicate that the difference between each probability in the null and DAF condition was statistically significant (p < 0.05, z-test for proportions). Across all cases, transition probabilities significantly decreased as a result of DAF (p < 0.01, one-sided Wilcoxon signed rank test). B, Transition probabilities in the null versus DAF+PS condition. Other plotting conventions as in A. A, Comparison of the change in transition probability induced by the DAF and DAF+PS conditions. As hypothesized, the change in probability was significantly smaller in the DAF+PS condition (p < 0.05, one-sided Wilcoxon signed rank test). In all plots, diamond symbols indicate data from the example shown in Fig. 3, and triangle symbols indicate data from the example shown in Fig. 5.

Additional example of effects of feedback manipulations on branch point probabilities. Shown is data from one branch point (of nine total branch points in our dataset). Plotting conventions as in Fig. 3. Data are from the branch point also shown in Fig. 1.

Effects of auditory feedback manipulations on repeat lengths. A, Spectrogram shows excerpt of a song that contains three repeated syllables (e, f, and g). In this excerpt, syllable g is repeated 10 times. Spectrogram plotting conventions as in Fig. 1A. B, Distribution of repeat lengths of syllable g in the null condition (black solid line) and DAF condition (blue solid line). Dashed lines show the mean of each distribution.