Social factors in bird-song development: Learning to sing with friends and rivals

Abstract

Laboratory studies have revealed that social factors are key in bird-song learning. Nevertheless, little is known about how or why birds choose the songs they do learn from the many they will hear under natural conditions. We focus on various theories concerning social song learning that have been offered to date, with special attention paid to two axes of social factors. First, does song learning occur via direct interaction of the young bird with song tutors, or via social eavesdropping by the young bird on interacting singers (social modeling of song)? Social modeling, a hypothesis first proposed by Pepperberg (Zeitschrift für Tierpsychologie, 55(2), 139-160, 1981), and direct interaction are not mutually exclusive hypotheses, and the evidence we review suggests both play a role in song learning. Second, does song learning occur via interactions with rivals (territorial competitors) or with friends (mutually tolerant or even cooperative territorial neighbors). These are largely mutually exclusive hypotheses, and can really only be tested in the field. There is little evidence on this contrast to date. We review our recent study on song sparrows, which indicates that both the young bird and his primary tutor may benefit from song learning/tutoring. If this mutual benefit result is confirmed by further studies, we believe that song “tutoring” in these cases may be more than a term of convenience: that it may qualify as true teaching.

Introduction

Young songbirds learn their songs from adult conspecifics. The many similarities between bird-song learning and human language learning have long been noted (Marler, 1970a), but one striking difference is often overlooked: In humans the vocal tutors are typically parents, relatives, and other interested parties, whereas in most songbirds they are typically unrelated individuals who are future or present territorial neighbors, and thus potential competitors for space, food, and mating opportunities. Consequently, while human language learning is a cooperative, mutually beneficial process, the opposite may well be true of bird-song learning, at least in the common case where tutor and tutee are potential competitors. In this paper we focus on this puzzling difference between human and songbird vocal learning and ask whether song learning is a truly competitive process. We focus on our study species, the song sparrow (Melospiza melodia), but draw on data from many other species as well.

While it is generally assumed that young songbirds benefit from the particular songs they have learned, generally no such assumption is made about a benefit to the song “tutors.” Whereas in human parlance, a “tutor” is someone who benefits from his or her instructional efforts – as does a parent or another relative, or a paid tutor, or a mentor vis-à-vis an apprentice – in the bird song literature the term “song tutor” is used only as a term of convenience. Except for rare case where song tutor and tutee are related, we are unaware of any proposal that the older birds who provide the songs to be learned are actually tutoring in the usual sense of the word, such that tutoring a young bird might ultimately benefit them as well as the tutee. To the contrary, it is often assumed that the young bird learns its songs by eavesdropping on the older bird while it is singing for other reasons – to post its territory, interact with territorial neighbors, or attract a mate. We will refer to this model of song learning as the Simple Eavesdropping model. It has two key features. First, simple overhearing of song is sufficient for the young bird to learn the song. Second, the older bird’s song “tutoring” is incidental, not motivated, i.e., not actual tutoring. The tutor himself is likely unaware that he is “tutoring.” We will consider the possibility that the older bird may be purposefully tutoring the young bird below, following consideration of several older hypotheses.

The Simple Eavesdropping model is a straight-forward extrapolation from Peter Marler’s ground-breaking studies on song learning in white-crowned sparrows, Zonotrichia leucophrys (Marler, 1970b) . Marler showed that white-crowned sparrows learn their species-specific but dialect-variable song in an early sensitive period, and that mere overhearing of the song – from a loudspeaker in the bird’s chamber – is sufficient. The model for this process is similar to the classical ethological model of imprinting (Lorenz, 1935). Moreover, Marler found that a young bird would reject heterospecific song (even song that was highly similar, e.g., song-sparrow song) that he heard in the sensitive period, either outright if that is all he hears, or in favor of white-crown song, if he heard both. But he will learn a white-crown song from a different dialect area than the one he hatched in, demonstrating clearly that the song is learned. Marler postulated a sensorineural “template,” which allowed the young bird to discriminate conspecific from heterospecific song, with the template being generic enough that the white crown could learn any particular dialect of conspecific song that he was exposed to. That Marler was correct about this being a sensory constraint, rather than a motor constraint, was confirmed by later studies showing that white-crowned sparrows could learn heterospecific song under certain conditions (see below). Marler’s laboratory findings, taken together, suggested that under natural circumstances, on hearing adult songbirds singing nearby, a young white crown would home in, auditorily, on conspecific song, and learn that song.

The Simple Eavesdropping model was given a shock, however, when Baptista and Petrinovich (Baptista & Petrinovich, 1984, 1986) showed that live interactive adults were far more potent song “tutors” than was playback of recorded song. In their studies, also carried out with white-crowned sparrows, they showed that live tutors were effective beyond the end of the sensitive period that had been defined by the playback experiments, and that the young birds would even learn heterospecific song if the song tutor was a live bird. The greater effectiveness of live tutors has been shown in many other songbird species subsequently (see reviews in Beecher, 2017, and Casey & Baker, 1993). These studies point to the importance of social factors in song learning, and they have spawned a number of hypotheses to explain the special effectiveness of social tutors.

According to the Aggression hypothesis, young birds learn more from aggressive song tutors. Payne (1981), for example, found that young indigo buntings (Passerina cyanea) were more likely to learn from an adult with whom they could interact directly than from an adult they could only hear and see. During these interactions, aggression was observed but the study did not isolate aggression from other variables that relate to social contact. In another study, Clayton (1987) found that young male zebra finches (Taeniopygia guttata) that were housed with two adult males from 35 days post-hatching on learned preferentially from the adult that showed more aggression towards them. In contrast, Casey and Baker (1993) found that young white-crowned sparrows developed abnormal songs when aggressive adults were the only available song tutors. Finally, in a field study on song sparrows, our group tested the role in song learning of the aggressiveness of potential song tutors (Akçay, Campbell, Reed, & Beecher, 2014b). We assayed aggressive behaviors (such as approaching the speaker and flying around it) and signalling behaviors of potential tutors through repeated territorial intrusions simulated via playback in the field during the song-learning period of a cohort of young sparrows. We also recorded these young birds and traced their repertoires back to potential tutors. While the potential tutors showed individually consistent differences in aggressiveness, level of aggressive signalling, and level of normal singing (Akçay, Campbell, & Beecher, 2014a), we found no effect of any of these variables on a bird’s tutoring success. Thus, overall aggressiveness or responsiveness of tutors doesn’t seem to be an important factor in whether that bird is selected as a tutor, although this finding doesn’t rule out that direct interactions with that tutor may be important (see below).

According to the Visual Signal hypothesis, it is the live tutor’s visual presence that makes him more effective than the playback of recorded song. Several studies of zebra finches – and ideal model species for this purpose since young finches generally do not learn from song playback – have produced mixed results. In one study, zebra finch fledglings prevented from seeing by eye patches still learned from a tutor in the same cage (Adret, 2004). On the other hand, a recent neurobiological study showed that neurons in the midbrain periaqueductal grey of juvenile zebra finches are selectively excited by the sight and sound of a singing male finch, releasing dopamine in the cortical song nucleus high vocal center (HVC) and facilitating song learning (Tanaka, Sun, Li, & Mooney, 2018).

In song sparrows, several lab studies by our group found evidence against both the aggression and visual signal hypotheses. In Nordby and colleagues’ studies (Nordby, Campbell, & Beecher, 2001; Nordby, Campbell, Burt, & Beecher, 2000), four adult singing birds were given aviary “territories” on the roof of our laboratory building. Following an initial period in which the young birds were rotated to be physically close to and in visual contact with each of the four adults for an equal amount of time (simulating prospecting for territories in the first summer of the young bird), they were placed near one of these tutors. From this point on, they could see and hear this particular tutor, and interact with him to a limited degree (though they were physically separated), but could only hear at a distance the other three singing adults, each of which was similarly positioned near a young song learner. This second period simulated territorial settlement next to an adult male in the fall. In the first study (Nordby et al., 2000), we found that the young birds learned most (though not exclusively) from the closest adult in the second phase, as we had expected based on field data (Beecher, Campbell, & Stoddard, 1994; Nordby, Campbell, & Beecher, 1999). However, in the second study (Nordby et al., 2001), which was a replication study in most respects, except that two of the tutors in the first phase were replaced with two new tutors in the second phase, subjects all learned the most from one of the adults introduced in the second phase. This super tutor sang more than the other three tutors, and sang more interactively with the other tutors than any of them did. This tutor tutored more songs despite the handicap of entering the experiment on 1 October, long after the mid-summer sensitive period. The most relevant aspect of the findings here is that three-quarters of the tutees only heard this super tutor at a distance, i.e., with no visual presence, and no aggression. Thus, it seems that neither visual contact, nor aggressive interactions (at least to the extent they occurred between different aviaries) are necessary for a tutor to be selected by young birds.

One point to consider is that variables that are readily manipulated or observed in laboratory settings may not be the critical variables from the bird’s perspective. For example, although we can manipulate visual presence or absence of the song tutor in an experiment, the manipulation may be meaningless if the young bird infers the presence of the unseen bird from its vocal behavior and other acoustic cues such as noises made when flying between perches. That is, that the bird’s perception of the presence of a live interacting bird may be the critical variable, not whether or not this bird can be seen. Similarly, aggressive behavior seen in an artificial laboratory setting may have no counterpart in the field context, because young birds can easily distance themselves from an adult bird that might be aggressive or because they are generally tolerated by adult birds in that context (Templeton, Campbell, & Beecher, 2012a). Indeed, we often see young song sparrows approaching singing adult birds, even in the midst of an interaction between adults, or between an adult and a simulated intruding bird, i.e., a taxidermic mount plus song playback (Templeton et al., 2012a).

Is direct interaction between adult and young bird the key social ingredient in song learning?

The most obvious social candidate for a role in song learning is direct interaction between song tutor and tutee. This is the possibility that comes immediately to mind when we consider the human analogy of language learning, where parents (and other interested parties) directly teach infants and toddlers words and phrases, both modelling of the correct words and phrases and shaping the infant’s vocal attempts. We will refer to this as the Direct Interaction model. The only strong evidence for direct interaction effects in birds comes from studies of brown-headed cowbirds (Molothrus ater) and zebra finches (Carouso-Peck & Goldstein, 2019; West & King, 1988). Laboratory studies in both species have revealed that song in young males can be shaped by visual feedback from non-singing females during the song development phase. However, in contrast, as noted above, close exposure to males, at least in captive settings, tends to inhibit song learning in males (Beecher, Burt, O'Loghlen, Templeton, & Campbell, 2007; Casey & Baker, 1993; West & King, 1980). But both of these effects – positive shaping by adult females, inhibition by adult males – could be artefacts of unnatural laboratory conditions. We will argue below that the role of direct interaction, and particularly the possibility of positive shaping, can clearly be unraveled only by careful field studies.

Is social modelling the key ingredient in song learning?

Pepperberg (1981, 1985) was the first to propose that the key social component in vocal learning may be the observation by the learner of a vocal interaction between adults, or between an adult and another learner. Extrapolating from a model of social learning in humans – “social modelling” theory (Bandura, 1969) – she proposed that observation by the young bird of vocal interactions between individuals who have mastered the communication system might be critical for vocal learning. Although Pepperberg developed her model in the context of teaching language to African grey parrots, she thought the model should apply to natural song learning in songbirds. An intriguing part of this social modelling hypothesis is that the young bird learns not only particular songs but also singing usage rules, for example, the appropriate replies to particular songs in particular contexts.

Several laboratory and field studies from our group have pointed to the importance of eavesdropped or observed vocal interactions in song learning by song sparrows. In addition to the Nordby et al. (2001) study described above, another semi-natural study also indicated that more songs could be learned from an unseen singer than from a singer directly available. Beecher et al. (2007) exposed hand-raised song sparrows to four adult tutors in May and June (early). In January, February, and March (late) they were exposed to only two of those tutors, one in close visual contact, and the other overheard (but not seen) interacting with another young bird. This yoked design is shown schematically in Fig. 1. As expected from previous studies, the subjects learned more from the tutors they experienced late as well as early than they did from tutors they experienced only early. However, to our initial surprise, subjects learned more from the tutor they overheard interacting with another young bird than they did from the tutor they saw and interacted with vocally themselves. That is, social modelling was more potent than direct interaction, perhaps in part because the direct interaction was too intense, and thus inhibitory. Evidence for the inhibition interpretation is supported by the results of an additional manipulation: half the subjects had a blind placed between them and their interactive tutor, and those subjects learned relatively more from that tutor than did those subjects who had no visual separation from their interactive tutor. As cautioned earlier, laboratory contexts are always artificial, and may have no counterpart in nature, and the up-close and personal (and perhaps overly aggressive) interaction of tutor and tutee in laboratory settings may mislead us about the role of direct interactions between tutor and tutee in nature.

Fig. 1

Schematic diagram of tutoring arrangement in Beecher et al. (2007). Subject 1 interacted directly with (though physically separate from) Tutor A, and on alternate days overheard Tutor B and Subject 2 interacting. Subject 2 interacted directly with (though physically separate from) Tutor B, and on alternate days overheard Tutor A and Subject 1 interacting

In a second, similar semi-natural experiment, Burt et al. (2007) exposed hand-raised song sparrows sequentially to two pairs of live tutors in May and June (early), and in January, February, and March (late) to computer presentation of songs from two of the four original tutors: one of the two “virtual” tutors sang interactively with the young bird, while the other sang asynchronously. Subjects learned more from the interactive virtual tutor than they did from the non-interactive virtual tutor, but they also retained more songs from the early-only partner of the interactive tutor than they did from the late interactive tutor. These findings indicate that not only is the bird more likely to learn from an interactive tutor, but he is more likely to retain for his final repertoire songs that he heard early that had been used as replies to the interactive tutor. Note that this suggests that the bird learned more than particular songs, he learned something about song usage: he retained songs that his early experience had indicated were appropriate replies to the songs of his interactive tutor. Again, this result points to the important role of eavesdropping on singing interactions for song learning.

Templeton et al. (2010) carried out a field experiment that suggests an important role for overheard vocal interactions in song learning. The response of radio-tagged juvenile song sparrows to simulated adult counter-singing was compared with that to simulated solo singing during the mid-summer sensitive period. Juveniles approached the simulated interactive singing of two song sparrows but not the simulated solo singing of a single song sparrow, which they treated similarly to heterospecific singing (a control condition). This finding suggests that during the prime song-learning phase, juvenile song sparrows are attracted to singing interactions between adults but not to singing by a single bird, and suggest that observed or overheard singing interactions may be particularly powerful song-tutoring events. A recent field study with Savannah sparrows found that speakers simulating counter-singing birds (in the sense that the timing of the songs from two speakers was coordinated) did not yield to greater song learning from these speakers compared to speakers that did not coordinate timing of their songs, simulating solo singing (Mennill et al., 2019). This study, impressive as it is, may not have been entirely successful in simulating a truly interactive singing scenario, since simply manipulating the timing of songs may not have been perceived by young birds as counter-singing, given all the other songs sung by the real birds at the same time.

In another field experiment, Templeton et al. (2012a, b) examined how adult male song sparrows treat juvenile and adult intruders during different developmental periods – summer (the classical sensitive period), autumn, and spring (when most late learning occurs) – using taxidermic mounts and song playback. Territorial adults responded aggressively to adult intruders throughout the year, but were tolerant of juvenile intruders in the summer. In the autumn, territorial adults displayed somewhat reduced aggression toward juveniles, but by the spring they treated them as aggressively as adult intruders. In addition, in the summer trials, we noted wild juveniles approaching our simulated “interactions” between subjects and adult mounts; these wild juveniles were also tolerated, even at close distances. That juveniles can and do closely approach adults during the early sensitive phase of song learning suggests that direct interactions with adults are possible and might be important for song learning, although notably our observations occurred in the context of the young birds eavesdropping on singing interactions, albeit closely.

Simple eavesdropping, social modelling, or direct interaction?

The three hypotheses identifying different contexts in which song learning might occur are schematized in Fig. 2. Simple Eavesdropping may allow song learning in a socially impoverished situation such as in a laboratory isolation box, and perhaps even in natural settings in the field if the songs are presented with enough frequency (Mennill et al., 2018). But, in general, social interaction is a key component of song learning in birds, just as it is human language learning. The Social Modelling vs. Direct Interaction hypotheses are not mutually exclusive, however, and the studies reviewed above suggest that both processes may be involved in song learning. The negative evidence for direct interaction compared to social modelling (or indirect interaction) may be largely an artefact of the laboratory. For this reason, we concluded that the best window into song learning would be found in the field. The remaining studies considered have all been field studies.

Fig. 2

Schematic diagram distinguishing between models of song learning. Simple eavesdropping: the young bird merely overhears an adult conspecific solo singing. Social modelling: he hears (and perhaps witnesses) two adults counter-singing. Direct interaction: he interacts with adult, with adult singing and perhaps tutee singing as a part of the interaction

Tutor and tutee, competitive or cooperative?

It seems clear that social interactions – whether direct interaction or social modelling – are key to song learning, but it is unclear exactly how or why they are. An approach our group has taken is to focus on the why? – that is, to directly examine the function of song learning. We therefore return to the question posed at the beginning of this paper – is song learning in birds competitive or cooperative?

Human language learning has traditionally been viewed as a cooperative enterprise between “tutor” and “tutee,” e.g., between parent (or other teacher) and child, and in some cases bird-song learning may fit this model. For example, zebra finches appear to learn from their father (Goldstein, King, & West, 2003; Immelmann, 1969), as do at least two of the Darwin’s finches (medium ground finch and cactus finch; Grant & Grant, 1996). Presumably in such cases offspring benefit from learning and parents from tutoring, hence these would be considered cooperative interactions. A clear example where tutor and tutee both benefit occurs in superb fairy-wrens, where mothers teach their embryos a vocal password that chicks produce as a begging call after hatching to elicit parental food provisioning. Both parties benefit because parents use the begging call to discriminate between their own young and brood parasitic cuckoo chicks (Colombelli-Négrel et al., 2012; Kleindorfer, Evans, & Colombelli-Négrel, 2014). Some communal and cooperative breeders as well as lekking species may present additional cooperative examples (Brown & Farabaugh, 1997; Brown, Farabaugh, & Veltman, 1988; Payne & Payne, 1997; Price, 1998; Sharp, McGowan, Wood, & Hatchwell, 2005; Trainer, 1989). But most songbirds, including our study species the song sparrow, follow the much more common pathway of the young bird learning songs from older, unrelated birds, after the young bird has dispersed from the natal area (Beecher, 2008). In this circumstance, bird-song learning has generally been viewed as an essentially one-sided enterprise in which the young bird presumably benefits from song learning but no assumptions are made concerning the older “tutors.” Indeed, whether or not song tutoring benefits the older bird has rarely been considered, but given that the young bird and his tutor are potential future competitors, it is possible that tutoring might even be costly for the tutor. In this section, we will consider two hypotheses: the competition hypothesis that the tutee benefits but the tutor does not, and may even suffer a cost, and the mutual benefit hypothesis that both tutor and tutee benefit from the song learning/tutoring interaction (Beecher & Akçay, 2014).

The competition hypothesis

Territorial neighbors are competitors. They dispute their territory boundaries, especially when they are new neighbors, but also when, for example, one of their mates decides to put her nest in a particularly desirable location on the wrong side of the negotiated boundary. Thus, it is certainly plausible that an older bird would not welcome a young bird into or near his territory, that he might not benefit in the long run by having a new competitor to deal with. The period in which song sparrows learn their songs overlaps with the period in which they prospect for territories, therefore a young bird might attempt to learn songs from an older bird all or part of whose territory he covets. If sharing songs with the older bird helps him establish a territory, and if that territory establishment comes at the expense of the established bird, then this situation resembles a zero-sum game, and we might expect a negative correlation between the territorial success of young birds and their older tutor neighbors. In song sparrows, we have observed that a young bird rarely replaces his primary tutor (the older bird from whom he learned the most songs), which rules out a strong version of the competition hypothesis, but neighboring song sparrows nevertheless do compete. In addition to fights over territory boundaries, they also engage in extra-pair mating at the expense of their neighbors: on average, 25% of the eggs in a male’s nest are the result of extra-pair mating with neighbors (Hill, Akçay, Campbell, & Beecher, 2011; Sardell, Keller, Arcese, Bucher, & Reid, 2010). Moreover, during the female’s fertile phase, males guard their mates against extra-pair copulations with neighbors (Akçay et al., 2012) and the “Dear Enemy” relationship between neighboring males (discussed next) breaks down (Moser-Purdy, MacDougall-Shackleton, & Mennill, 2017).

The mutual benefit hypothesis

The mutual benefit hypothesis of song learning/tutoring follows fairly directly from the well-studied “Dear Enemy” relationship (Fisher, 1954). The Dear Enemy relationship occurs when territorial animals are more tolerant of established neighbors than they are of newly arrived neighbors, where they prefer the neighbor they know to the neighbor they don’t know. In relationships between established, familiar neighbors, negotiation of territory boundaries can be minimized and the costs of territorial defense reduced, for both individuals. A number of studies have shown that these Dear Enemy relationships may go beyond mutual tolerance and reduced aggression, with neighbors sometimes forming defensive coalitions against intruding strangers or engaging in cooperative defence against predators (see review in Beecher, 2017). Moreover, a number of studies have shown that familiarity with neighbors is linked to a number of fitness-related benefits, including faster growth rates, enhanced body condition and increased reproductive success (Beletsky & Orians, 1989; Grabowska-Zhang, Wilkin, & Sheldon, 2012; Liebgold & Cabe, 2008; Seppä, Laurila, Peuhkuri, Piironen, & Lower, 2001).

In our studies, we have shown that song sparrows display the Dear Enemy effect, responding less aggressively to established neighbors than to strangers, but differentiating among their neighbors as well, responding more aggressively to neighbors who have recently been simulated to intrude onto their own or their neighbor’s territory (Akçay, Reed, Campbell, Templeton, & Beecher, 2010; Akçay et al., 2009; Stoddard, Beecher, Horning, & Campbell, 1991). In eastern song sparrows, Hyman and Hughes (2006) have found that territorial males respond more strongly to aggressive neighbors than to non-aggressive neighbors.

The Dear Enemy context might form the basis for a mutually beneficial song-learning/tutoring strategy, where sharing songs with one older bird (or tutoring a young bird) is one component of a partially cooperative enterprise between a young bird and an older bird who will become territorial neighbors. The Dear Enemy relationship could begin early on, perhaps even in the natal summer when adults are quite tolerant of juveniles. There is evidence in favor of this hypothesis in song sparrows, including the greater survival of young birds sharing more songs with their first-year neighbors (Beecher, Campbell, & Nordby, 2000a; Wilson, Towner, & Vehrencamp, 2000), and the finding that song sparrows are less aggressive toward neighbors with whom they share more songs (Wilson & Vehrencamp, 2001).

Returning to the tutor-tutee context, the relationship might be a true alliance, or it could simply be one of mutual tolerance. Either way, the relationship gives the older bird a young neighbor who is less threatening in terms of both territorial integrity and paternity confidence, and the young bird the ability to establish and hold a territory more easily than he could otherwise. Although being allies does not necessarily require song sharing between the tutee and tutor, song sharing may facilitate this relationship, for example, by signaling to each other and to others that these neighbors are allies. Shared vocalizations have been shown to facilitate group cohesion, coordination, and affiliation in several group-living animals (Brown et al., 1988; King & McGregor, 2016; Trainer, 1989; Tyack, 2008). It is possible that in territorial species song sharing may facilitate cooperation between territorial allies and serve as a signal to third parties.

A test of the mutual benefit and competition hypotheses

These two functional hypotheses can best be tested in the field, where we can establish who learned which songs from whom, and either directly measure the level of competition or mutual tolerance/cooperation between the tutors and tutees, or measure some component of fitness to determine whether song tutoring has benefited or cost the older bird. In our first attempt to get at this question, we tried to assess responses of tutees to their primary tutors by simulating the latter singing within the territory of the tutee (Akçay, Campbell, & Beecher, 2017). We reasoned that if a tutee had a cooperative relationship with his tutor, he would respond less aggressively to the tutor songs compared to the song of a distant bird (a stranger). To our surprise, we found that the tutees responded more strongly to tutor songs than to stranger songs, and the strength of the response depended on song learning: the more the tutee learned from the tutor, the stronger his response to the tutor’s songs (Akçay et al., 2017). Although at the time we interpreted this finding as indicating a competitive relationship between the tutor and tutee, at least from the perspective of the tutee, we have realized that there are two significant problems with this interpretation. First, it was carried out a few years into the territory tenures of the tutees, by which time most of the primary tutors had disappeared from the study area (died). Therefore, we ended up simulating a long-gone tutor/neighbor returning and singing in the middle of his erstwhile tutee’s territory, which objectively might be a more severe threat than a stranger doing the same thing. Second, the former tutee’s strong response to the simulated intrusion by his former neighbor might be seen as strong retaliation to a former Dear Enemy returning and trying to claim his former ally’s territory. As noted above, song sparrows do retaliate against current neighbors that are perceived to be aggressive or expansionist (Akçay et al., 2010; Akçay et al., 2009; Hyman & Hughes, 2006).

A less ambiguous test of the functional song-learning hypotheses is to correlate song learning with a measure of fitness. To that end, we collected data on song learning in situ for several cohorts of young song sparrows at our study site in the northwestern USA (Akçay, Campbell, Reed, et al., 2014b; Beecher, Akçay, & Campbell, 2020; Nordby et al., 1999). We recently completed an analysis correlating song learning for three cohorts of young birds (2009, 2010, and 2011) with years of survival on territory (generally the major component of lifetime reproductive success in a territorial songbird) for those young birds and their primary tutors (Beecher et al., 2020). We summarize those results here.

In our sedentary population of song sparrows, the processes of song learning and territory establishment are linked (Beecher et al., 1994). A male song sparrow disperses from his natal area at about 1 month of age and acquires a repertoire of about nine songs (e.g., Fig. 3) during his first year of life, learning from unrelated adults in the area where he attempts to establish his territory (Akçay, Campbell, Reed, et al., 2014b; Beecher et al., 1994; Nordby et al., 1999). The bird crystallizes his song repertoire by the time he is fully territorial sometime between January and March of the next spring, and keeps his territory and his song repertoire (unchanged) for the rest of his life (average 3 years). Song sparrows in this population typically learn most of their songs from one older bird, the “primary” tutor. In the data set to be considered (Beecher et al., 2020), the mean percentage learned from the primary tutor was 55%’ (range 30–100%). The primary tutor-tutee relationship is central to our analysis, for it suggests a special relationship between that older bird and the young bird, particularly in those cases where sharing between the two is high. It is the basis for our prediction that the extent of the tutee’s learning of the primary tutor’s songs will reflect either the degree of tolerance or cooperation between the two (mutual benefit hypothesis) or the intensity of the competition between the two (competition hypothesis). To be clear, by “tolerance” we mean that the older bird and the young bird are less aggressive and competitive toward another than they are toward other neighbors. By “cooperative” we mean the two birds going beyond tolerance and engaging in cooperative behaviors such as together driving off an intruder or predator that has appeared on one of their territories, or perhaps not attempting an extra-pair copulation with the other’s mate.

Fig. 3

Six songs of a bird’s full repertoire of nine songs

We tested the mutual benefit and competition hypotheses by measuring the survival of a tutee and his primary song tutor as a function of how many of the tutee’s songs he had learned from the tutor (Beecher et al., 2020). We reasoned that if song learning is mutually beneficial for the song tutor and the young song learner, then the more of the tutor neighbor’s songs the young bird learns, the more years the two will coexist on their territories (i.e., their mutual survival will increase). This hypothesis assumes that the full benefit of song sharing between tutor and tutee occurs only so long as both survive on territory as neighbors. Hence the number of songs the tutee learns from the tutor should predict their mutual survival, but not necessarily the survival of tutor or tutee alone. In contrast, according to the competition hypothesis, the more of his songs the young bird learns from the older tutor, the fewer years the two will coexist on territory (their mutual survival). This hypothesis assumes that how much the young bird learns from this tutor reflects the degree of their territorial competition and that the presence of one of them makes the future survival of the other one less likely (Akçay et al., 2017).

Our results support the mutual benefit hypothesis (Fig. 4). The number of songs that the tutee learned from the primary tutor most strongly predicted the mutual survival of the tutor and tutee in the full data set. There was a weaker positive prediction of the primary tutor’s survival, but because song learning continues into the early part of the first spring, we excluded those tutors who did not survive into the first spring of the tutee, and in that case only mutual survival showed a significant positive relationship with the number of songs learned from the primary tutor. There were non-significant correlations of number of songs learned with tutor and tutee survival in the reduced data set. In a series of control tests, we ruled out several alternative explanations that did not invoke a mutualistic relationship between tutee and primary tutor. Strong primary tutors did not differ from others in terms of quality, condition or aggressiveness, or of territory quality, or of attractiveness of their particular songs (Beecher et al., 2020; see also Akçay et al., 2014a, b; Nordby et al., 1999).

Fig. 4

Years (breeding seasons) that primary tutor, tutee, or both survived a function of the number of songs learned from the primary tutor. From Beecher et al., in press

The finding that song learning is associated more strongly with mutual survival than it is with the survival of the tutee alone or the tutor alone suggests that the benefits of song sharing accrue only so long as both individuals are alive and present on territory. If the strength of the tutor-tutee relationship is indicated by the number of songs they share, then it appears there is a mutual advantage in a strong tutor–tutee relationship. Previous research on song sparrows has found support for only the tutee side of this benefit (Beecher, Campbell, & Nordby, 2000b; Wilson et al., 2000). Similar results have been found for the indigo bunting, a territorial songbird in which males sing a single, complex song (Payne, 1982; Payne, Payne, & Doehlert, 1988). Payne and colleagues found that young indigo buntings who shared their song with an older neighbor (the presumptive song tutor) in their first breeding season had greater breeding success than did young males who shared with a distant male (the presumptive song tutor) or with no one at all in the population. That is, breeding success was dependent on the coexistence of both the young bird and his presumptive tutor on adjacent territories, another parallel with Beecher et al. (2020).

In summary, our test of the mutual benefit and competition hypotheses suggests the following scenario for song sparrows in our population. In the normal course of events, the Dear Enemy relationship between a bird and his primary tutor will become stronger than that between the bird and his other neighbors who were weak tutors or non-tutors. As noted earlier, not all Dear Enemies are equally dear (Stoddard, 1996). For instance, in another resident western population of song sparrows, (Wilson & Vehrencamp, 2001), neighbors who shared more songs were less aggressive to playback of the neighbor’s song at their mutual territory boundary than were neighbors who shared fewer songs. Although the authors of that study did not identify tutor–tutee relationships, the high degree of sharing between the neighbors in their study suggests that at least half of these pairs were tutee and primary tutor. We note again that the fact that young song sparrows rarely replace their primary tutors (Akçay, Campbell, & Beecher, 2014a; Nordby et al., 1999) is inconsistent with the hypothesis that song learning from a particular older bird is part of a strategy to take over his territory or that it reflects particularly strong competition between the two birds.

Although our results suggest that learning many songs from a particular tutor-neighbor is beneficial for both parties in song sparrows, we need to gather direct evidence on the behaviors that could underlie the mutual benefit effect. These would include behaviors that might be considered teaching by the older birds, cases where proto-territorial young birds are tolerated by older birds, and actual examples of joint defence by putative tutor–tutee pairs. Radio-tracking studies of young birds in our population have revealed numerous cases where they are tolerated by older territorial birds, but we have as yet been unable to connect these cases to song learning (Templeton, Campbell, et al., 2012a; Templeton, Reed, Campbell, & Beecher, 2012b). Joint defence could be studied by playback experiments simulating intrusions near common territory boundaries (Elfström, 1997). We would predict more cooperative, effective defence carried out by strong primary tutor-tutee pairs compared to weak tutor-tutee pairs. In the absence of song-learning information, the extent of sharing between a young bird in his first breeding season and his older neighbor could be used as a proxy for song-tutoring.

Conclusion: Hypotheses on social factors on song learning

Table 1 summarizes the eight different hypotheses on bird-song learning, all of which address in one way or another the critical nature of social interactions. Note that the hypotheses refer to different levels of analysis (ultimate function vs. proximate mechanism), and so are not necessarily exclusive. And some merely describe the context in which song learning is thought to occur. Some of these hypotheses are mutually exclusive (the competition vs. mutual benefit hypotheses) and some contradict others if assumed to be exclusively true (simple eavesdropping vs. social modelling vs. direct interaction). But we view the evidence to date as supporting the idea that some combination of them will be needed to explain the song-learning process in a given species, and certainly no one hypothesis will be able to account for the very different song-learning programs that are found among the many species of songbirds (see review in Beecher & Brenowitz, 2005).

Table 1 Song-learning hypotheses. For the most part these hypotheses are not mutually exclusive (notable exceptions being the Mutual Benefit and Competition hypotheses). For instance, all the mechanistic hypotheses would be compatible with either Direct Interaction or Social Modelling contexts, although they would not apply in the Simple Eavesdropping context

Bird-song learning as an instance of teaching in non-human animals?

We conclude by noting that a great deal of the evidence discussed in this paper supports Pepperberg’s original suggestion that vocal learning in birds (and other animals) may often occur in the context of social eavesdropping via social modelling, and, indeed, in a context where one animal is explicitly or at least effectively teaching another. Taken with the long-term mutual benefit of song learning for tutor-tutee pairs found in Beecher et al. (2020), we suggest that it might be profitable to view bird-song tutoring as an instance of teaching in non-human animals (Caro & Hauser, 1992; Hoppitt et al., 2008; Thornton & Raihani, 2010). Accordingly, we add that as one of the hypotheses in Table 1. The commonly accepted criteria for demonstrating teaching in non-human animals are that (1) teachers should modify their behavior in the presence of the learner, (2) this change in behavior should result in no immediate benefit to the teacher, and (3) the learner should acquire a behavior quicker or better as a result (Caro & Hauser, 1992). Bird-song tutoring would fit all three criteria if, for example, the older bird reduces his usual aggression when a young bird appears on his territory (Templeton, Campbell, et al., 2012a), increases his counter-singing with the young bird, and the young bird prefers to learn the older bird’s songs over another neighbor’s songs. Alternatively, such “teaching” as the older bird does may be purely indirect or incidental, perhaps occurring because the primary tutor is a particularly interactive bird, giving the eavesdropping young bird the opportunity to observe many singing interactions between this older bird and other adults (i.e., social modelling without tutee-directed behaviors). We urge further research in songbirds designed to gather observational and experimental evidence on these points, so as to fully evaluate whether bird-song tutoring in song sparrows and perhaps other songbirds qualifies as an instance of teaching in non-human animals.

Open Practices Statement

The data and materials for the main experiment discussed in this review paper (Beecher, Akcay & Campbell, 2020) are available at https://doi.org/10.1016/j.anbehav.2020.05.015. That study was a field study containing no experiments, so nothing was preregistered.