CHAPTER 1

Introduction

The study of social learning sits at the interface of a truly astonishing number ofacademic disciplines. How many other fields could boast being central to bothsocial anthropology and human evolution; core material for both experimentalpsychologists and theoretically minded economists; or emerging influences inthe fields of both cognitive neuroscience and artificial intelligence?

The observation that many animals, including humans, acquire valuable lifeskills and knowledge through copying others has been the focus of attention ofanimal behaviorists dating back to Darwin. Likewise, social learning, the diffusionof innovations, conformity, and social influences on child development havebeen key concepts within the social sciences for over a century. However, in recentdecades, the field of social learning has received such unprecedented attentionspread across the sciences, social sciences, and humanities, and experiencedsuch growth, that researchers within the field have referred to an "explosion ofinterest" in the topic (Galef and Giraldeau 2001; Shettleworth 2001). Long goneare the days when research on imitation could (ungenerously) be characterizedas the esoteric province of an obscure branch of comparative psychology. Now,social learning is a rapidly growing subfield of animal cognition research; whilebiological anthropologists and archaeologists are constructing models of culturalevolution, economists are frequently talking norms and herding behavior, neuroscientistsare mapping circuitry associated with social influences on decisionmaking, and engineers are building imitating robots.

The diverse backgrounds of the researchers studying social learning contributeto the field's controversies. East African but not West African chimpanzees usestalks to fish for termites, while western but not eastern chimps crack open nutswith stone hammers. Capuchin monkeys in Costa Rica exhibit extraordinarysocial conventions, such as sniffing each other's hands and placing fingers in eachother's mouths. Humpback whales and chaffinches sing different songs than theirfellow conspecifics living in different regions. If these, and other, animal traditionsare acquired through social learning, are behavioral scientists justified inspeaking of animals possessing "culture"? Or are anthropologists correct to assertthat human cultures are so imbued with meaning, so permeated with symbolism,and so reliant on uniquely human aspects of cognition, that to liken them to thebehavioral traditions of animals is, frankly, ridiculous? Is human social learningshaped by evolved structure in the mind then biased to acquire content—fromchoosing sugar rich foods to admiring specific body shapes—that proved adaptiveamong our Pleistocene ancestors, as suggested by many evolutionary psychologists?Or is human learning dominated by general rules (e.g., copying thehighest payoff behavior or conforming to the local norm) that are for the mostpart acquired independently of their content, as claimed by cultural evolutionists?Is imitation critically dependent on the ability to take another individual'sperspective, to understand their goals, or on complex cognition? Are sociallylearned traditions constrained in order to be adaptive? Can cultural processessupport a viable form of group selection? And so on, and so forth. The controversiesare multiple, ripe, and engaging, enriched by the varied standpoints thatcharacterize adjacent disciplines struggling to understand a common topic.

The disparate backgrounds of those drawn into the field have also contributedto the newly emerging methods that are appearing to address these challenges.Until recently, experimental studies of social learning were restricted to behavioralinvestigations, typically conducted in the laboratories of comparative or developmentalpsychologists, and focused on very specific questions, such as: "Cananimals imitate?" or "Do children acquire violent dispositions from others?"Similarly, with the exception of some early experiments on birdsong learning,biologists' interest in social learning was pursued almost entirely through observationsand recordings of the natural behavior of animals, largely by ethologistsand primatologists.

In recent years, however, new methods have become available, considerablyexpanding the social learning researcher's toolbox. These include (i) experimentaland statistical methods that allow researchers to categorize cases of sociallearning according to their underlying psychological processes and learningmechanisms; (ii) neuroscientiic methods for identifying the brain structures,neural circuitry, and physiological processes underlying both social learning andsocial influences on decision making; (iii) mathematical and statistical methodsfor identifying social learning when it occurs in natural populations (or in naturalisticcontexts in captivity); (iv) experimental and statistical methods to predictand explain when humans and other animals copy, from whom they learn("social learning strategies"), and to also detect the strategies deployed; and (v)mathematical methods for predicting the pattern of diffusion of novel learned innovations,and for modeling cultural evolution and gene-culture coevolution. Finally,numerous and diverse tools and procedures are available for applying sociallearning methods outside of academia. These range from commercial hatcheriestraining hatchery-reared fish to recognize predators and thereby enhance restockingefforts, to predicting the pattern of spread of technological innovationsand identifying likely targets for uptake.

This book is designed to be a complete and accessible practical guide for thesocial learning researcher and their students, as well as for others whose interestin social learning is less central. As it currently stands, there is no single sourcethat reviews the aforementioned conceptual and methodological developments,and the field's new theory and tools are to be found in a diverse collection ofarticles in academic journals. This book is first and foremost a monograph onsocial learning concepts and methods; it seeks to summarize and extend newdevelopments in the field, rendering the new tools available to a broader constituencyand heightening awareness of this emerging research topic. However, such isthe growth of interest in social learning that we envisage the material containedin this book to be of interest to many individuals, academic and nonacademic,who are not directly connected to the immediate field. Social learning and imitation,including their mechanisms, methods, and models, have become centralto a broad range of disciplines, and significant progress in their scientific understandingis potentially of widespread interest, both within and outside academia.

In order to discuss the conceptual foundations of social learning some definitionswill be necessary from the outset. Certain terms, including "social learning,""imitation," "innovation," "tradition," and "culture," will appear repeatedlythroughout this book. In almost all cases, there exists no universally accepteddefinition, and in many instances the terms are the focus of considerable debate.Accordingly, at the outset we provide initial definitions for these labels, thoughlater in the book we will suggest refinements. A summary of these definitions isgiven in box 1.1, with a justification presented below.

1.1 What Is Social Learning?

The most commonly used definition of social learning is "learning that is influencedby observation of, or interaction with, another animal (typically a conspecific)or its products" (Heyes 1994); this draws on a similar earlier generic definitionby Box (1984). This definition is extremely broad, and indeed might be regardedas problematically so. The concern here is that the phrase "influenced by" lacksspecificity. To take matters to the extreme, the presence of another individual mighteven impede learning, yet would still meet the above definition, because the other'sactivities constitute an influence on learning. It is then perhaps not surprising thatothers have used the term "social learning" to refer to a more specific concept. Forexample, Lonsdorf and Bonnie (2010) restrict social learning to cases in which seeinganother individual performing a behavior pattern causes the observer to learnthe same pattern; this approximates our use of the term "social transmission"

This ambiguity in the meaning of the term "social learning" has the potentialto lead researchers into the logical error of equivocation, which occurs whenthe meaning of a term is changed during a line of reasoning. For example, thosedefinitions of "culture" (see below) that emphasize geographic differences in behaviorcaused by social learning (e.g., Laland and Hoppitt 2003) typically have inmind a narrower conception of social learning, more akin to the Lonsdorf andBonnie definition than the broader Heyes definition. There could be a danger thatresearchers might prematurely make the claim of culture if variant forms of sociallearning in the broader sense were identified for a given species. This is particularlyproblematic if different researchers interpret the term "social learning" withvariant levels of breadth, unaware of discrepancies in usage.

On the face of it, a solution might be to adopt a narrower definition of sociallearning, with another term for the broader category; for example, "socially biasedlearning" could be used (e.g., Humle et al. 2009). However, there are twoconcerns here. First, the broad definition is widely used and accepted, and henceit may lead to further confusion if we to advocate a change in usage. Second andmore importantly, narrower definitions could miss out on some phenomena ofinterest to those who would consider themselves social learning researchers. Forexample, a definition that requires matching behavior would not include caseswhere an observer learns what not to do, perhaps after watching another individualmake a mistake. Consequently, on balance, we maintain that a broad andgeneric term still has some currency and propose the following definition, modifiedfrom Heyes (1994):

Social learning is learning that is facilitated by observation of, or interactionwith, another individual (or its products).

Our use of the term "facilitated" restricts forms of social interaction thatgenerate social learning to those that have a positive influence on the observerlearning. Instances where social interaction impedes learning are better capturedby a different term, such as "social inhibition" (Brown and Laland 2002). Evenwith this refinement, the definition of social learning remains broad, meaning aresearcher's primary task is to investigate the manner in which learning is social,rather than to answer whether learning is social learning or not. This kind ofreasoning justifies researchers' tendency to focus on general social processesthought to be widely important in promoting social learning, potentially acrossdifferent contexts and modalities, rather than on isolated or specialized instances.

Throughout this book we will refer to the individual who learns socially as the"observer" and the individual that they learn from as the "demonstrator." Whilewe acknowledge drawbacks to both of these terms, there are no obviously superioralternatives.

1.2 Social Transmission

Notwithstanding the utility of a generic term, researchers clearly require morespecific terms to describe different types of social learning. In recent years therehas been a concerted effort to develop methods to "detect social learning" in naturalpopulations (see chapters 5-7). Such methods are not really aimed at detectingsocial learning in the broad sense, but rather to detect what we term "socialtransmission." This term was initially coined by Galef (1976), and subsequentlydescribed as "cases of social learning that result in increased homogeneity of behaviorof interactants that extends beyond the period of their interaction" (Galef1988, 13). Familiar examples include the spread of sweet potato washing in Japanesemonkeys (Kawai 1965) and hybrid-corn use by American farmers (Ryanand Gross 1943). Our definition of social transmission is very much in the samespirit as Galef's, but is designed to confer greater precision:

Social transmission occurs when the prior acquisition of a behavioral traitT by one individual A, when expressed either directly in the performanceof T or in some other behavior associated with T, exerts a lasting positivecausal influence on the rate at which another individual B acquires and/orperforms T.

Here a behavioral trait is a target behavior pattern, which might be specific to aparticular context (e.g., pigeons pecking at a lever in response to a green light;Dorrance and Zentall 2002). Inclusion of the phrase, "expressed either directly inthe performance of T or in some other behavior associated with T" is recognitionof the fact that acquired information and behavior can be transmitted with orwithout performance of the trait. Examples of the latter are not uncommon. In thecase of humans, for example, we can provide written or spoken instructions onhow to perform the trait, instead of demonstrating it directly. For example, socialtransmission of a recipe between chefs might occur when one chef passes a recipeto another. The number of times a chef cooks the recipe might have no causalinfluence on the acquisition of the recipe by others, but we would still considerthis social transmission (see fig. 1.1c). In the case of other animals, signals suchas food calls (e.g., the vocalizations of callitrichid monkeys that recruit infants toa desirable or novel food source [Rapaport and Brown 2008], or mother hens thatwarn their chicks from consuming a toxic foods [Nicol and Pope 1996]) or theopportunity teaching exhibited by several carnivores (e.g., when meerkat helpersprovision pups with disabled prey; Thornton and McAuliffe 2006) can facilitatesocial transmission without direct performance of the trait by a demonstrator.

We also recognize two types of social transmission, which we call social transmissionof trait acquisition (where the effect of social transmission is on the rateat which B acquires T) and social transmission of trait performance (where theeffect is on the rate at which B performs T). In practice, many cases are likelyto involve both the social transmission of trait acquisition and of trait performance—whereindividuals who observe a demonstrator both acquire the traitsooner and perform it more frequently once they acquire it, than individuals whodo not observe a demonstrator. Nonetheless, we see the two as logically distinct.Observers might acquire T sooner, but may not perform T any more frequentlythan individuals who have acquired it asocially.

It might seem strange to some readers to discuss a rate at which acquisition ofa trait occurs, when each individual can only acquire a trait once (except whenan individual has forgotten a trait and then does learn it again). Here our use ofthe term "rate" is in the sense of a "hazard rate" in time to event or in "survival"models—comparable to the rate of death or the rate at which a machine componentfails—which therefore refers to the pattern of learning spread across asample of individuals. For example, if a large sample of individuals all had alearning rate of 0.5 per hour, we would expect the mean time to acquisition to be1/0.5 = 2 hours.

Our definition also covers the case in which there are two or more traits thatare alternative variants for the same function, and (1) the variant possessed by Acauses B to be more likely to acquire the same variant, and/or (2) the preferencesof A for one variant causes B to develop a similar preference. Definitions in termsof rates also have the advantage that they allow researchers to model trait acquisitionor performance as a stochastic process, and thereby infer social transmissionfrom relatively complex patterns of acquisition or performance (see chapter 5).As indicated above, we interpret the phrase "performance" broadly to encompasstransmission mediated by an animal's products, including, in the case of humans,signs, signals, and linguistic information.

There are clearly a number of causal routes by which a trait T could be sociallytransmitted between individuals A and B. Figure 1.1 illustrates three such cases.Perhaps the most obvious is social transmission through observation: B observingtrait T being performed by A, causes B to acquire T (fig. 1.1a). It is this causalpathway that most traditional observer-demonstrator experiments aim to detect.There are also a number of documented cases where the physical products of theperformance of T (excluding visual and auditory cues) can play a role in socialtransmission (fig. 1.1b). For example, Heyes and Dawson (1990) found that inrats the trait of pushing a joystick in a particular direction for a food reward wassocially transmitted from trained demonstrators to observers, and that olfactorycues left on the joystick were sufficient for this effect to occur (Mitchell et al. 1999;see also Galef and Beck 1985; Laland and Plotkin 1991). Some researchers arguethat social transmission mediated through physical traces left by animals in theirenvironment is likely to be highly important in natural populations (Fragaszy,2012b). We discuss several such examples in chapter 3.