Mechanisms underlying the social enhancement of vocal learning in songbirds

Abstract

Social processes profoundly influence speech and language acquisition. Despite the importance of social influences, little is known about how social interactions modulate vocal learning. Like humans, songbirds learn their vocalizations during development, and they provide an excellent opportunity to reveal mechanisms of social influences on vocal learning. Using yoked experimental designs, we demonstrate that social interactions with adult tutors for as little as 1 d significantly enhanced vocal learning. Social influences on attention to song seemed central to the social enhancement of learning because socially tutored birds were more attentive to the tutor's songs than passively tutored birds, and because variation in attentiveness and in the social modulation of attention significantly predicted variation in vocal learning. Attention to song was influenced by both the nature and amount of tutor song: Pupils paid more attention to songs that tutors directed at them and to tutors that produced fewer songs. Tutors altered their song structure when directing songs at pupils in a manner that resembled how humans alter their vocalizations when speaking to infants, that was distinct from how tutors changed their songs when singing to females, and that could influence attention and learning. Furthermore, social interactions that rapidly enhanced learning increased the activity of noradrenergic and dopaminergic midbrain neurons. These data highlight striking parallels between humans and songbirds in the social modulation of vocal learning and suggest that social influences on attention and midbrain circuitry could represent shared mechanisms underlying the social modulation of vocal learning.

Keywords: attention; birdsong; catecholamines; social influences; speech.

Fig. 1.

Social interactions with tutors for as little as 1 d enhance vocal learning. (A) Birds that were socially (n = 10) or passively (n = 9) tutored (SI Appendix, Fig. S1A) for 5 d produced songs as adults that were significantly more similar to tutor songs than the adult songs of birds that remained untutored throughout development (“untut. birds”) (SI Appendix). Additionally, socially tutored pupils produced songs as adults that were significantly more similar to their tutor’s song than the adult songs of passively tutored pupils. *P < 0.05. (B) In another experiment, pupils were allowed to visually and acoustically interact with the tutor (V-A pupil), to only acoustically interact with the tutor (A pupil), or to passively hear the tutor’s song (P pupil) for <1 d (SI Appendix, Fig. S1B). Overall, V-A pupils (n = 6) but not A (n = 5) or P (n = 6) pupils produced songs as adults that were significantly more similar to tutor songs than the songs of untutored birds (“untut. birds”; *P < 0.05). Similarity scores were not significantly different between V-A, A, and P pupils (P = 0.1394). Bars ± error bars and dashed lines ± shaded areas represent the mean ± SEM.

Fig. 2.

Attention to song is affected by interactions with tutors and is correlated with song learning. (A) Across socially tutored (●; n = 10) and passively tutored (□; n = 9) birds, there was a significant and positive relationship between attention to tutor’s songs and the similarity of the pupil’s adult song to his tutor’s song. (B) There was a significant and positive relationship between the magnitude of difference in attention to song and the magnitude of difference in song learning between socially and passively tutored pupils of a cohort. (C) Tutors produced songs that were directed at pupils [pupil-directed (PD) songs] or not directed at pupils [undirected (UD) songs], and, overall, pupils were more attentive to PD songs than UD songs. However, this difference was significant only for socially tutored birds (●; Tukey’s HSD; *P < 0.05). Relatedly, socially tutored birds paid significantly more attention than passively tutored birds (□) to PD songs (*P < 0.05) but not to UD songs.

Fig. 3.

Tutors changed their songs in distinct ways when directing songs to pupils than when directing songs to females. (A) An example of a single bout of zebra finch song. Above the spectrogram are labels that identify individual syllables as well as introductory notes that are repeated before song onset. (Scale bar: 500 ms.) (B) Relative to undirected songs, pupil-directed songs were preceded by more introductory notes, contained longer intervals between motifs, and consisted of syllables with lower spectral entropy and mean frequency and higher goodness of pitch (n = 10). Introductory notes and intermotif intervals were analyzed per bird whereas spectral features were analyzed per individual syllable using mixed effects models. Although raw data were analyzed, plotted is the percent change from undirected to pupil-directed song. *P < 0.05. (C) Relative to undirected songs, female-directed songs (n = 7) consisted of syllables that were significantly more stereotyped across renditions [i.e., lower coefficient of variation (CV) for fundamental frequency (FF)] and tended to be preceded by more introductory notes. Plotted is the percent change from undirected to female-directed song. *P < 0.05, ^#P < 0.10. For both B and C, the magnitude of change for the first three features (○) scales to the left y axis whereas the magnitude of change to the last three features (□) scales to the right y axis.

Fig. 4.

Tutor song rate affects pupil attention and learning. Pupil attention (A) and song learning (B) were inversely related to the rate at which tutors produced song [songs per hour; cube-root (cbrt)-transformed]. (C) The relationship between tutor song rate and pupil song learning was not significant after controlling for pupil attention. Plotted are the residuals from the analyses of how attention relates to tutor song rate and to song learning.

Fig. 5.

Social but not passive tutoring significantly increased EGR-1 expression in noradrenergic neurons in the LC and dopaminergic neurons in the VTA. (A) We compared the percentage of tyrosine hydroxylase (TH)-immunoreactive (TH-ir) neurons that expressed EGR-1 across socially tutored pupils, passively tutored pupils, and untutored control pupils. In the LC and VTA but not the SN, a significantly higher percentage of TH-ir neurons expressed EGR-1 in socially tutored birds than in passively tutored birds or untutored birds. Passive tutoring did not increase EGR-1 expression in TH-ir neurons. (Inset) Photomicrograph of midbrain neurons that express TH (green) and EGR-1 (red). Yellow arrow highlights a TH-ir neuron expressing EGR-1. (Scale bar: 40 μm.) (B) Social exposure to a tutor that produced song (“social-song”) but not social exposure to an adult tutor that did not produce song (“social-nosong”) increased the percentage of TH-ir neurons expressing EGR-1 in the LC and VTA. For both A and B, *P < 0.05, and lines indicate significant contrasts.