Photo by Leszek Leszczynski
Primatologists distinguish between three types of vocal learning: production, usage, and comprehension. Comprehension learning entails forming an association between other individuals’ usage of a call and an appropriate behavioral response (e.g., learning the meaning of a predator-specific alarm call). Conversely, usage learning is defined as experience-dependent changes in the contexts in which a vocalization occurs (e.g., learning which alarm call to utter for which predator). Finally, production learning is defined as experience-dependent changes in the acoustic structure of vocalizations. Put differently, vocal usage refers to when and where a vocalization is produced, while vocal production refers to how it is produced; vocal production learning, then, means that an animal learns how to produce a vocalization.
Human language requires a huge amount of learning at all three of these levels, as I can attest firsthand having watched my fiance’s nephew develop into a talking machine over the last several months. This real-time MRI video of the vocal tract during speech, posted by the Max Planck Institute just a few days ago, illustrates just how intricate the muscle movements we acquire through vocal production learning really are. Songbirds (as well as parrots, hummingbirds, elephants, and some marine mammals) also undergo vocal production learning: like human infants, young birds learn by listening to the adults around them. Certain species show geographic variation in song, similar to human accents or dialects. The strongest evidence for production learning in songbirds comes from lab-rearing experiments (many of which were conducted by Peter Marler, my academic grandfather): birds reared in the absence of singing adults (and hence without the opportunity to learn how to sing properly) grow up with impaired, simplistic songs. Remember this experimental paradigm—I’ll be coming back to it.
Nonhuman primates, on the other hand, are generally thought to lack vocal production learning, at least early in development. They certainly undergo vocal usage learning: an infant vervet monkey startled by a falling leaf might inappropriately utter a predator alarm call, before eventually learning when and where that call should be used. Adults usually ignore alarm calls coming from infants, due to their unreliable nature—one of many examples of vocal comprehension learning. But unlike songbirds (and humans) monkeys have been shown to acquire normal-sounding vocalizations despite social isolation, cross-fostering, and even deafness. Furthermore, the neural connections known to play a key role in vocal flexibility in humans and songbirds alike are absent from the (admittedly few) monkey species whose brains have been studied in depth. These findings have led most primatologists to the view that nonhuman primates are born with the production of species-typical vocalizations already hard-wired.
Enter the common marmoset. Marmosets (along with their cousins, the tamarins) are a group of New World primate species that have evolved a diminutive body size and subsist on insects, and sap which they extract while clinging to trees with tiny claws (an evolutionary reversal from the nails shared by most primates). You can listen to some of their vocalizations here. Due in part to marmosets’ small size, which makes them hard to study in the field, marmoset vocal communication has received relatively little scientific attention, until recently. Within the last several years, there has been a string of publications from researchers studying marmoset vocal behavior in the lab, using methods similar to those traditionally utilized to investigate bird song.
Earlier this month, Yasemin B. Gultekin and Steffen R. Hage at the Werner Reichardt Centre for Integrative Neuroscience, University of Tübingen, Germany, published the results of the latest in this string of studies, in the journal Science Advances. With the goal of examining whether young marmosets undergo vocal production learning, they emulated the experimental paradigm I described earlier: they analyzed vocalizations from three marmosets who had received only limited parental interaction, and compared them to the normal vocalizations of two parent-reared individuals. The limited-interaction marmosets grew up to produce “phee” vocalizations that were “immature,” sharing acoustic similarities with infant cries. This pattern persisted into the subadult stage of development, unlike in previous studies where vocal deficits due to limited parental interaction were temporary. The vocal differences are unlikely to be the result of general developmental stunting in the limited-interaction marmosets, because in other respects (body weight, and pitch of vocalizations), they seemed more mature than the parent-reared group (appropriately so—they were older).
These findings demonstrate that, like humans and songbirds, marmosets require social interactions with adults for the development of normal vocal production. This study provides some of the first evidence that vocal learning in nonhuman primates can go beyond usage and comprehension, to include production learning. The researchers suggest that marmosets may not be special among nonhuman primates in this respect. Rather, they argue, the widespread view of primates as lacking vocal production learning may be the result of insufficient acoustic and statistical analytical tools through much of the history of this research.
A couple of caveats. First, this study examined only five marmosets—two parent-reared and three limited-interaction individuals—which is an incredibly small sample. It’s difficult to know whether the results would hold true for other common marmosets, let alone any other species. Second, while the acoustic structure of the limited-interaction marmosets’ “phee” calls was off, they were still recognizably “phees.” This contrasts with the songs of socially isolated songbirds, which can be radically simpler than species-typical songs. Furthermore, other vocalizations (“twitters” and “trills”) did not sound significantly different in the limited-interaction vs. parent-reared marmosets. The degree of vocal production learning demonstrated in common marmosets in this study is far less dramatic than that which allows humans and some other animals to learn a huge variety of different sounds. Nevertheless, this study’s findings pose a serious challenge to the long-held view of primates as entirely lacking vocal production learning in early development.
Finally, those who care about animal welfare, as I do, may be concerned about the psychological consequences of separating young primates from their parents—which we know can be traumatic, thanks to the controversial and (I’ll say it) cruel maternal deprivation experiments conducted by Harry Harlow in the mid-20th century. Gultekin and Hage report that the three limited-interaction marmosets were not separated from their parents for the purposes of this study per se, but rather for practical reasons. One of them was rejected by his parents as a newborn, and after the other two had finished weaning, they were introduced to the rejected individual to form a new social group. My own research with howler monkeys suggests that facilitating social interaction among orphaned primates can lead to beneficial outcomes. Regardless, as long as this facility was going to limit these marmosets’ social interaction anyway, even the staunchest animal welfare activist would have a hard time objecting to these researchers taking advantage of the opportunity to investigate vocal development.