Available online at www.sciencedirect.com ScienceDirect The evolution of vocal learning 1 2 Stephen Nowicki and William A Searcy Vocal learning, in which animals modify their vocalizations to across multiple lineages, as we attempt here in a qualita- imitate those of others, has evolved independently in scattered tive fashion. lineages of birds and mammals. Comparative evidence supports two hypotheses for the selective advantages leading When we speak of vocal learning, we are referring specifi- to the origin of vocal learning. The sexual selection hypothesis cally to vocal production learning, in which signalers alter proposes that vocal learning evolves to allow expansion of the form of their vocalizations by imitating the vocaliza- vocal repertoires in response to mating preferences for more tions of others [4]. Additional types of vocal learning complex vocalizations. The information-sharing hypothesis occur, such as usage learning, in which signalers learn also proposes that vocal learning evolves to allow expansion of the context in which to produce specific vocalizations, and vocal repertoires, but in this case in response to kin selection comprehension learning, in which receivers learn to favoring sharing of information among relatives. modify their response to vocalizations produced by Addresses others [4]. These additional forms of vocal learning have 1 Department of Biology, Duke University, Durham, NC 27708, USA a wider taxonomic occurrence [1] and perhaps can be 2 Department of Biology, University of Miami, Coral Gables, FL 33124, accomplished without the considerable neural apparatus USA that undergirds vocal production learning [2], and con- sequently their evolution seems less puzzling. Corresponding author: Nowicki, Stephen ([email protected]) Because vocal production learning requires complex Current Opinion in Neurobiology 2014, 28:48–53 neural adaptations that are presumably costly, this type of learning is not expected to evolve except in response to This review comes from a themed issue on Communication and language some compensating fitness advantage. Some of the hypothesized advantages of vocal learning are experi- Edited by Michael Brainard and Tecumseh Fitch enced by an individual only when most or all of the other For a complete overview see the Issue and the Editorial signalers in its population are vocal learners and receiver Available online 15th July 2014 response has evolved to adjust to this circumstance. http://dx.doi.org/10.1016/j.conb.2014.06.007 Advantages such as these are not relevant to the evol- 0959-4388/# 2014 Elsevier Ltd. All rights reserved. utionary origin of vocal learning because they only would accrue when the trait has already evolved, although they may be important to its evolutionary maintenance. Our focus here is principally on the origins of vocal production learning, and thus on the benefits that accrue to the first individuals that begin to modify their vocalizations based Introduction on the sounds they hear from others. Although vocal learning has evolved multiple times in vertebrates, relatively few lineages possess this ability Advantages of song learning in songbirds [1,2]. Vocal learning is especially rare among mammals, Many of the proposals on the selective advantages of and among primates is well developed only in humans [3]. vocal learning have emerged from the study of song The rarity of vocal learning may be explained in part by learning in songbirds. The songs of songbirds typically the complexity of the neurological underpinnings that have dual functions in mate attraction and territory this ability seems to require [2]. Given the complexity of defense [5]. Song development in songbirds appears to vocal learning as an adaptation, and its centrality to the be universally influenced by imitation learning, but as life histories of some species, it is not surprising that there are over 4000 species of songbirds (suborder Pas- attention has been given to explaining its evolution. seri), there is tremendous scope for variation in patterns of Because the evolution of vocal learning in any one lin- learning [6]. One typical pattern is for young males to eage, including our own, is typically a one-off affair, memorize songs or parts of songs in the first few months of progress in testing ideas on the evolution of vocal learning life [7] in the neighborhood where they will later establish requires a comparative approach. This approach is chal- their own territories [8]. The memorized sounds later lenging because the lineages that have evolved vocal guide song development when the young males begin to learning are few and diverse, hampering the application sing as they approach their first breeding season [9]. of formal comparative methods, which require a much larger number of taxa for statistical comparison. None- In an early and still influential paper, Nottebohm [10] theless, progress is only possible by synthesizing patterns suggested that one of the two main advantages of vocal Current Opinion in Neurobiology 2014, 28:48–53 www.sciencedirect.com Evolution of vocal learning Nowicki and Searcy 49 learning in songbirds is the production of vocal dialects, does not happen [12], at least not at a relevant geographic which he argued would be advantageous in promoting scale. If geographic variation in song does not occur unless genetic adaptation to local environmental conditions. song learning is already established, then the vocal dialect Nottebohm assumed that males learn the local dialect hypothesis can explain the maintenance of vocal learning where they are born, and that females learn preferences for but not its origin. their own local dialects and then choose mates that sing those dialects. Song learning would thus promote assorta- The second main advantage of vocal learning proposed by tive mating between males and females that have been Nottebohm is in meeting female preferences for complex selected for adaptation to local conditions (Figure 1; Hy- sounds [10] (Figure 1; Hypothesis 2). Nottebohm assumed pothesis 1). Various aspects of this scenario have since that learned vocalizations can achieve greater complexity been challenged on empirical grounds [11,12]: whether than innate vocalizations, an assumption that still seems males actually learn songs where they are born rather than reasonable. He proposed that females generally prefer to after they disperse, whether females actually prefer to mate with males having more elaborate songs. Nottebohm mate with males singing their own natal dialects, and could adduce little evidence for this assumption at the whether dialect populations actually show genetic differ- time, but since then laboratory experiments and field ences indicative of local adaptation. Beyond these empiri- studies both have shown that female songbirds prefer more cal problems, the vocal dialect hypothesis also has some complex song in many species [13,14] though not all [15]. logical difficulties. First, a hypothesis to explain the evo- lution of vocal learning must provide a selective advantage In some songbirds, female preferences for complex song to the male that learns, not an advantage to the female that may currently be adaptive because of associations be- chooses to mate with him or to the population to which he tween song complexity and male quality [16 ]. The belongs. The vocal dialect hypothesis can be stated in ‘developmental stress hypothesis’ suggests that brain terms of an individual level advantage to the learner, structures underlying song learning develop during a though often it is stated otherwise. Second, and more period in which other aspects of the phenotype are critically, the vocal dialect hypothesis provides a selective developing, and in which the young bird is particularly advantage to song learning only if songs already vary susceptible to developmental stress. Those individuals geographically and females already prefer local variants. that escape stress, or whose genotypes are resistant to Although it seems logically possible for geographic vari- stress, develop more complex and more accurately imi- ation in song to occur without song learning, in practice this tated songs as well as more robust phenotypes overall Figure 1 evolution of vocal learning local repertoire fit of signals individual dialects expansion to environment distinctiveness assortative mate information improved individual mating choice sharing transmission recognition local genetic increased enhanced more effective enhanced social adaptation mating success fitness of kin signaling Interactions Hypothesis 1 Hypothesis 2 Hypothesis 3 Hypothesis 4 Hypothesis 5 Current Opinion in Neurobiology Five hypotheses on the selective benefits leading to the evolution of vocal learning. For each hypothesis, vocal learning has a proximate effect (in green) that translates by some mechanism (in blue) to a selective benefit (in red). The hypotheses are the (1) vocal dialect hypothesis, (2) sexual selection hypothesis, (3) information sharing hypothesis, (4) environmental adaptation hypothesis, and (5) individual recognition hypothesis. www.sciencedirect.com Current Opinion in Neurobiology 2014, 28:48–53 50 Communication and language [17,18,19 ]. The developmental stress hypothesis pre- been considered, abilities other than vocal learning have dicts associations between male quality and vocal com- drawn more attention, such as the capacity to use and plexity specifically for learned songs. Unlearned understand recursion [33] or the ability to use syntax to vocalizations