Introductory gestures before songbird vocal displays are shaped by learning and biological predispositions

Introductory gestures are present at the beginning of many animal displays. For example, lizards start their head-bobbing displays with introductory push-ups and animal vocal displays begin with introductory notes. Songbirds also begin their vocal displays by repeating introductory notes (INs) before producing their learned song and these INs are thought to reflect motor preparation. Between individuals of a given species, the acoustic structure of INs and the number of times INs are repeated before song varies considerably. While similar variation in songs between individuals is known to be a result of learning, whether INs are also learned remains poorly understood. Here, using natural and experimental tutoring with male zebra finches, we show that mean IN number and IN acoustic structure are learned from a tutor, independent of song learning. We also reveal biological predispositions in IN production; birds artificially tutored with songs lacking INs still repeated a short-duration syllable, thrice on average, before their songs. Overall, these results show that INs, just like elements in song, are shaped both by learning and biological predispositions and suggest multiple, independent, learning processes underlying the acquisition of complex vocal displays.


INTRODUCTION 23
Animal produce various complex displays to communicate with their conspecifics [1]. Many of these 24 communicative displays begin with the repetition of introductory gestures. For example, Anolis lizards 25 produce a few introductory "push-ups" or "tail-flicks" before starting their head bobbing display [2,3], 26 frogs and toadfish produce introductory vocalizations before their advertisement calls [4,5] and a 27 number of songbirds also produce introductory vocalizations before the start of their complex songs [6-28 17]. Many different functions have been attributed to these introductory gestures, including, an 29 "alerting" function [5,18,19], a species-specific signal that aids learning of song [20], a "local-dialect" 30 signal [21] and more recently, a role in motor "preparation" [22,23]. Given that these introductory 31 gestures are a part of both learned and unlearned displays, the extent to which they are learned, in the 32 context of learned displays like bird song, remains poorly understood. 33 Bird song is a well-studied example of a complex vocal display that is learned by imitation from a tutor 34 [24][25][26]. Many different species of songbirds, including the zebra finch, begin their displays with the 35 repetition of a short, simple, vocalization called an introductory note (IN) [6][7][8][9][10][11][12]14,16] before 36 producing their more complex song. Among individuals of a given species, both the repetition and 37 acoustic structure of INs vary considerably [23,[27][28][29]. What is the source of this variation? Variation 38 in elements of song between individuals is a consequence of learning [8,10,24,25]. Similarly, variation 39 in IN number and IN acoustic structure could also be learned from a tutor. Alternatively, as predicted 40 by the motor preparation hypothesis [22,23], variation in IN number and structure across birds could be 41 correlated with variation in their respective songs. Finally, variation in IN number and structure across 42 birds could also be a result of biological predispositions similar to the biological predispositions in the 43 production of elements of song [30]. Here, we examine these different predictions in the zebra finch, a 44 well-studied songbird [14]. 45 Adult male zebra finches also begin their vocal displays by repeating an IN before their song (Fig. 1A) 46 [8,10,14]. Both mean IN number and IN acoustic structure vary considerably between birds (Fig. 1A In order to prevent song learning from their father, juvenile zebra finches  81 were separated from their father around phd 10 (range phd 6-16) and reared with their mother and 82 siblings until phd 35 (range: phd . Previous studies have shown that exposure to the father's song 83 before phd 25 does not significantly impact learning [32], so our birds are unlikely to have learned their 84 father's song during this period. Once juveniles had reached nutritional independence they were 85 separated and housed individually in small cages (except for 2 birds who were housed together). These 86 birds were then used for the different groups as outlined below. 87

Playback tutored birds 88
Using active tutoring methods [33,34], birds were tutored with synthesized zebra finch songs that were 89 played back through a speaker ( Fig. 5A were part of a different study where song (but not IN) learning was described [35]. For these birds, INs 98 of the tutors was not a consideration during the choice of tutor. Therefore, to investigate IN learning, 99 only a subset of these tutored birds were analyzed here. Specifically, only socially tutored birds in 100 which tutors produced a different number of INs than the biological father (mean difference in IN 101 number between father and tutor -2.06; range: 0.5-4.78) were included in these analyses. 102 The tutoring phase lasted ~1.5 months (phd 34-40 to phd 91-97) for birds tutored at IISER Pune and for 103 5 days for birds tutored at McGill. Significant song learning is observed for socially tutored birds even 104 after just 5 days of tutoring [35]. 105 Isolate birds 106 5 lab bred male zebra finches (from two different nests) were reared in isolation. After birds had 107 reached nutritional independence, they were separated and kept in visual but not acoustic isolation from 108 other birds. Earlier studies have shown that visual isolation is sufficient to prevent song learning from 109 other birds [36]. 110

Data analysis 111
All analyses were conducted using custom written scripts in Matlab (Mathworks). 112

Song segmentation and syllable categorization 113
Song files were segmented and labeled as described previously [22,23] (see Supplementary Methods 114 for further details). The motif was identified based on the most common sequence of syllables across 115 all bouts. Bouts with one or more motif syllables were considered "song bouts". Importantly, 116 vocalizations that preceded the first motif of a bout were considered introductory notes (INs). 117 Vocalizations that were also produced outside of song bouts were considered as calls and not as 118 introductory notes. There was variation in the sample size of birds in each nest. To check whether this difference in sample 126 size could influence the correlation in IN number between fathers and normally reared birds, we used a 127 random reassignment procedure to assess significance of the observed correlation. For this procedure, 128 all sons were randomly reassigned to different fathers, while maintaining the actual number of birds per 129 nest, and then the correlation between fathers and sons was calculated. This procedure was repeated 130 10000 times and the 95% confidence intervals were estimated. This same randomization procedure was 131 also used to assess the significance of the correlation between socially-tutored birds and their tutors. In 132 this case, the total number of socially tutored birds was maintained constant during the random 133 reassignment procedure. 134 While classifying syllables as INs or motif syllables, we were not blind to the relationship between 135 birds or the experimental group. To control for any potential biases that this might have introduced in 136 our classification of INs and motif syllables, we also used a script based categorization of INs and 7 motif syllables. All song bouts were considered and syllable sequences within a bout with ≤ 500 ms 138 inter-syllable interval were considered. Across all these syllable sequences, any syllable that was 139 repeated (self-transition probability > 0) and occurred as the first syllable of a bout in > 2% of song 140 bouts, was considered an IN. Syllables that were present in 90% of bouts and did not occur as the first 141 syllable of a bout were considered motif syllables. The remaining syllables that did not satisfy either of 142 these criteria were classified as calls. 143  Fig. 1C, Fig. S1). INs (and song) of fathers were acoustically more similar to INs (and song) of 165 their sons as compared to INs (and song) of unrelated birds ( Fig. 1D; p = 0.041 for INs and p < 0.001 166 for song, Wilcoxon rank sum test). Mean IN number before song was also significantly correlated 167 between fathers and their sons (Fig. 1E). This correlation in mean IN number was not influenced by 168 removal of individual nests from the analysis (Fig. S2A) and was significantly different from the 169 correlations obtained by randomly re-assigning individual birds to different nests (Fig. S2B). Further, 170 this correlation in mean IN number was significant even when syllables in individual birds were 171 categorized into INs or song syllables based on pre-set rules (Fig. S2C,  obtained by randomly re-assigning birds to different social tutors (Fig. S3A) and was significant even 186 when syllables were categorized based on pre-set rules (Fig. S3B). These two sets of results indicate 187 that the number and acoustic structure of INs are learned from a tutor (father for normally reared birds 188 and social tutor for socially-tutored birds). 189

Accurate learning of INs is independent of accurate learning of song 190
Previous studies have suggested that INs represent motor preparation before song [22,23] However, across all normally reared and socially tutored birds, the degree of song similarity between 194 birds and their tutors was not correlated with similarity in IN number (Fig. 3A, Fig. S4A) or similarity 195 in IN acoustic structure (Fig. 3B, Fig. S4B) between birds and their tutors (father or social tutor). 196 Further, across all of these birds, mean IN number was not correlated with differences in various 197 temporal (Fig. S5A -duration of song; Fig. S5B -duration of first song syllable) and spectral aspects of 198 song (Fig. S5C -mean frequency of first motif syllable, Fig. S5D -complexity of first motif syllable). 199 These results showed that the number of times an IN was repeated was independent of the song that 200 followed and suggest two independent processes involved in IN and song learning. 201

Biological predispositions in IN production and learning 202
Our results suggest that INs are learned from a tutor just like birds learn their song motifs from their 203 tutor. Previous studies have demonstrated the presence of biological predispositions in zebra finch song 204 learning [30,39]. For example, juvenile zebra finches that are individually tutored with random 205 sequences of syllables converge on similar motif sequences [30]. To identify biological predispositions, 206 if any, in IN production, we experimentally tutored juvenile zebra finches with songs that lacked INs 207 ( Fig. 4A, n=22, see Methods for details of tutoring) [30,33,34]. Despite being tutored without INs, 208 juveniles tutored in this manner produced INs before their songs (Fig. 4B for an example). The number 209 of these INs was not correlated those of the father (Fig. 4C). The acoustic similarity of these INs to 210 those of the father was comparable to the similarity with unrelated birds, showing that these INs were 211 acoustically different from those of the father. (Fig. 4D, p = 0.82 for INs and p = 0.71 for motif, 212 Wilcoxon rank sum test). These results suggest more general, species-specific, predispositions in IN 213 production rather than direct genetic components from the father. 214 Another way to reveal biological predispositions in IN production is to analyze INs in the songs of 215 untutored birds. As observed in a previous study [10], we found that untutored birds produce INs before 216 their songs (n=5, see Methods for details of untutored birds). Interestingly, the mean number of such 217 INs produced before song ( Fig. 4E; p = 0.57, one-way ANOVA) and the duration of these INs ( Fig. 4F; 218 p = 0.3662, one-way ANOVA) was similar across all categories of birds, irrespective of tutoring 219 experience. Other features of INs differed between birds experimentally tutored without INs and birds 220 normally reared or socially tutored with INs (Fig. S6A, S6B, S6C) highlighting the role of learning. 221 Thus, on average, normally reared, socially tutored, operantly tutored and untutored birds produced 222 three, 60ms long, INs before starting their songs (Fig. 4E, 4F). This bias to produce ~3INs before song 223 can also be observed in the data for birds tutored with songs that contained INs. Juveniles tutored by 224 adults that produced, on average, < 3 INs tended to produce more INs than their tutor (Fig. 1E, 2E). On 225 the other hand, juveniles tutored by adults that produce, on average, > 3 INs tended to produce fewer 226 INs than their tutor (Fig. 1E, 2E). 227

DISCUSSION 228
The complex vocal displays of many songbirds begin with the repetition of simple, introductory notes 229

Functional significance of INs and other such introductory gestures 247
What is the functional role of INs in zebra finch song? Many different roles have been proposed for 248 introductory gestures including an "alerting" function [5,18,19], a species-specific signal that aids 249 learning of song [20], a "local-dialect" signal [21] and a reflection of motor "preparation" [22,23]. 250 While the degree to which INs in birdsong serve an alerting or identification function remain largely 251 unknown, our data showing learning of IN number and IN acoustic structure suggests that these two 252 aspects may not reflect motor preparation. Rather, learning of these two aspects suggests possible 253 communicative functions of INs, such as in individual, regional or species identification. 254 However, our data also reveals a biological predisposition in IN production. Specifically, we observed 255 that zebra finches are biased to produce short-duration vocalizations as INs, approximately three times 256 before their songs, regardless of their tutoring experiences during development. Interestingly, INs are 257 also seen before the start of song in suboscine birds [13,15,17] where song is not learned and before 258 advertisement calls in other vertebrates that produce unlearned vocalizations including frogs [4] and 259 toadfish [5]. These data support our findings showing the presence of biological predispositions in IN 260 production even before learned vocalizations, like birdsong. 261 Overall, irrespective of mechanism and function, our results show that the zebra finch can be an 262