at SciVerse ScienceDirect

Animal Behaviour 83 (2012) e6ee9

Contents lists available

Animal Behaviour

journal homepage: www.elsevier .com/locate/anbehav

Teaching can teach us a lot

Alex Thornton a,*, Katherine McAuliffe b

aDepartment of Experimental Psychology, University of Cambridge, Cambridge, U.K.bHuman Evolutionary Biology, Harvard University, Cambridge, MA, U.S.A.

a r t i c l e i n f o

Article history:Received 12 November 2011Initial acceptance 7 December 2011Final acceptance 5 January 2012Available online 15 February 2012MS. number: 11-00912

Keywords:cognitive mechanismcooperationcultureevolutionfunctional definitionsocial learningteaching

* Correspondence: A. Thornton, Department oUniversity of Cambridge, Downing Street, Cambridge

E-mail address: jant2@cam.ac.uk (A. Thornton).

0003-3472/$38.00 2012 The Association for the Studydoi:10.1016/j.anbehav.2012.01.029

In a recent Commentary, Byrne & Rapaport (2011; henceforth B& R) question the value of the functional perspective on teaching innonhuman animals in understanding the basis of teaching inhumans. They argue that the established operational definition ofteaching by Caro & Hauser (1992; henceforth C & H) is overlyrestrictive, misses instances where teaching serves to correctindividual failings in slow learners, and inhibits progress in ourunderstanding of the cognitive underpinnings of human teaching.While we welcome increased focus on the cognitive foundations ofteaching, this need not come at the costs of reducing rigour in thisnascent field. Here, we mount a defence of the C & H definition andargue that it can be applied at both population and individuallevels. We suggest that the development of the field will best beserved by considering both whether teaching occurs and, if so, howit is achieved.

IS IT TEACHING?

A rigorous definition is necessary to distinguish teaching fromother forms of behaviour. From an evolutionary perspective, much

f Experimental Psychology,CB2 3EB, U.K.

of Animal Behaviour. Published b

of the interest in teaching arises because it is a specialized form ofcooperation, whereby knowledgeable individuals invest in helpingothers to learn (Thornton & Raihani 2008). Caro & Hauser’s (1992)operational definition is critical in allowing us to discriminate suchactive assistance from cases of inadvertent social learning, whereindividuals acquire information from others who are simply goingabout their usual business and attempting to maximize their ownimmediate benefits. This parallels the distinction between coop-erative behaviour that involves initial investments and is selectedfor its beneficial effects on recipients and by-product mutualism,where the self-serving actions of one individual inadvertentlybenefit another (Clutton-Brock 2002; West et al. 2007). The C & Hdefinition has been enormously productive, spurring on a previ-ously moribund area of research and setting the scene for empirical(Hoppitt et al. 2008; Thornton & Raihani 2008) and theoreticalanalyses (Riboli-Sasco et al. 2008; Fogarty et al. 2011), with impactsfelt in evolutionary biology, anthropology and psychology (e.g.Tehrani & Riede 2008; Thornton & Raihani 2008; Hrdy 2009;Shettleworth 2010a). Nevertheless, B & R argue that strict adher-ence to the definition risks neglecting certain cases and, critically,that many instances of human teaching would be excluded.

In fact, we argue that, by placing the emphasis on measurableparameters, the C & H definition does rather a good job of capturinga diverse range of teaching interactions. The definition is dividedinto three main criteria: a knowledgeable individual (1) changes its

y Elsevier Ltd.

A. Thornton, K. McAuliffe / Animal Behaviour 83 (2012) e6ee9 e7

behaviour in the presence of a naïve companion in a way that (2)provides no immediate benefit, or imposes short-term costs, but (3)causes the companion to learn. In fact, the two cases B & R presentas examples of the definition’s failure to capture ‘unambiguouscases’ of human teaching fit the criteria perfectly. In the firstexample, an Aché father (1) modifies his usual behaviour whencrafting a bow by calling his son over, choosing a wide seat toaccommodate the son and shifting his position periodically to allowthe son a better view. These subtle behavioural changes (2) provideno current benefits to the father relative to his usual bow makingwhen alone, and may well slow down the process. Finally, (3) theson learns a skill he may not have learned otherwise. In the secondexample, a school teacher (1) writes cursive on the blackboard ina classroom full of children and moulds the hand of a particularlypoor pupil. This behaviour (2) is time consuming and provides noimmediate benefits to the teacher, but (3) the children all learnsomething they would not have done otherwise, with the poorpupil perhaps showing particular improvements owing to the extraattention he receives. B & R are particularly concerned that in somecases the costs of teachingmay beminimal and have nomeasurableimpact on energy budgets. However, this issue is not as damagingto the C & H approach as they imply. The critical point is that thereis no current benefit: the teachers’ changes to their usual routineserve no purpose other than to help others to learn. Thus, the C & Hdefinition may not be as restrictive as B & R assume.

Nevertheless, there will clearly be genuine cases of teaching forwhich it proves difficult to generate unequivocal evidence. Indeed,we have previously made the point that, because of these diffi-culties, we currently underestimate the prevalence of teaching(Thornton & McAuliffe 2006; Thornton & Raihani 2008), and wehave provided numerous suggestions as to methodologies thatmight improve data collection (Thornton & Raihani 2010). With thefield as it currently stands, it is important to consider how we maytreat cases where the evidence is suggestive but inconclusive. B & Rprovide no alternative standards by which to judge whether oneindividual goes out of its way to help another to learn and, ina curious reversal of Occam’s razor, suggest that it may beproductive to grant ambiguous cases the benefit of the doubt untilteaching is disproven. We feel that moving away from a clearlydelineated and testable definition risks creating confusion anderoding standards of evidence in this nascent field. Moreover, thereis a serious risk of inadvertently setting up a double standard,whereby special dispensation of weak evidence is granted tocertain species simply by virtue of their large brains, presumedcognitive sophistication or phylogenetic proximity to humans. Forinstance, given inconclusive evidence for teaching from, say, a fishand a primate, there may be a danger that the latter evidence maybe treated more generously, and that the underlying mechanismsmay be assumed by default to be more complex. Similar concernshave been raised by other comparative researchers (e.g. Laland &Hoppitt 2003; Chittka & Niven 2009) and may be partly assuagedby adhering to carefully defined, measurable criteria. Rather thanabandoning rigour, it is more productive to present inconclusiveevidence for teaching with due caution and consideration foralternative explanations. Such cases can undoubtedly contribute toour growing understanding of teaching and will help to spurfurther research. However, there is little to be gained fromaccepting equivocal evidence for teaching uncritically if simplerexplanations that do not require the active participation ofknowledgeable individuals cannot be ruled out.

SENSITIVE TEACHERS

As B & R rightly point out, population-level analyses, such asthose employed in accepted cases of teaching in tandem-running

ants, Temnothorax albipennis, pied babblers, Turdoides bicolor, andmeerkats, Suricata suricatta (Franks & Richardson 2006; Raihani &Ridley 2008; Thornton & McAuliffe 2006, respectively), may notcapture instances where particularly sensitive teachers act to bringslow learners up to speed. This is not, however, a reason to abandonthe operational definition, but instead should lead us to considerhow best to test it, and to determine the level at which to conductour analyses. The most direct route is to withhold or increaseteaching experimentally for certain pupils andmeasure the impacts(Thornton & Raihani 2010). Experiments may also serve not only toestablish the occurrence of teaching, but also to document theextent to which teachers are sensitive to their pupils’ needs. Forinstance, by interrupting bouts of teaching and experimentallyinhibiting the capacity of pupils to learn, researchers have shownthat knowledgeable tandem-running ants evaluate the progress oftheir pupils and adjust their movements accordingly (Richardsonet al. 2007; Franklin et al. 2011). Where experiments are unfea-sible, careful analyses of observational data can also be highlyinformative. In population-level studies, the potential impacts ofindividual variability may often be accounted through the use ofmixed-effects models controlling for repeated measures of indi-viduals. Where there is strong reason to suspect that teachers showsensitivity to individual pupils’ competence, within-subjects anal-yses would allow us to detect improvements in individual pupils’skills relative to teachers’ investments. Further analyses coulddetermine the impact of putative teachers by comparing learningby poor pupils that do or do not receive extra assistance. Thus, in B& R’s schoolteacher example, population-level analyses of the C & Hcriteriawould reveal overall improvements in the children’s cursiveskills in relation to their attendance of cursive lessons, and certainlyrelative to children that did not attend lessons. Similarly, at anindividual level, we could compare the poor pupil’s skills beforeand after lessons and measure his learning trajectory relative tounschooled pupils of the same age.

COGNITIVE MECHANISMS OF TEACHING

One of B & R’s principal concerns is that the focus on the func-tion of teaching detracts from our understanding of its cognitivefoundations. However, adherence to a rigorous operational defini-tion based on evolutionary function does not preclude in any wayanalysis of cognitive mechanisms. Indeed, it is very difficult tounderstand themechanisms bywhich one individual helps anotherto learn if, in fact, we do not know that that is indeed what it isdoing.

From a psychological perspective, much of the interest inteaching stems from the fact that much (although not all) of humanteaching involves recognizing that a given individual lacks knowl-edge or skills, allowing flexible, targeted teaching across contexts.However, it is important to note that skill monitoring may beachieved through a number of different mechanisms. In manycases, responses to physical or behavioural cues may suffice forteaching to be targeted appropriately. Adult meerkats, for instance,respond to age-related changes in pups’ begging calls, presentingdead or disabled prey to young pups and live prey to old pups. Theyalso monitor individual pups handling prey and will intervene byrecapturing or modifying prey if a pup is having difficulty(Thornton & McAuliffe 2006). Similarly, tandem-running ants usetactile cues to ensure that teachers and pupils respond sensitivelyto one another’s movements, thus facilitating learning (Franks &Richardson 2006; Richardson et al. 2007; Franklin et al. 2011).These findings provide a powerful illustration of the power ofsimple mechanisms in generating seemingly complex behaviour.Such low-level mechanisms are increasingly recognized as havinggreat importance in both human and nonhuman behaviour

A. Thornton, K. McAuliffe / Animal Behaviour 83 (2012) e6ee9e8

(Shettleworth 2010b) and their role in human teaching ought not tobe underestimated. This is particularly true for cases such as scaf-folding (parental support and encouragement of infant motor skilldevelopment) and motherese (simplified, exaggerated parentalspeech to facilitate infant language learning), both of which, like allknown examples of nonhuman teaching, are adaptations topromote learning in specific contexts (Thornton & Raihani 2008).

Of course, it would be fascinating to know whether anynonhuman animal goes beyond stimuluseresponse mechanismsand, like humans, is also motivated by a strong desire to shareintentions and equipped with the ability to attribute knowledgeand recognize ignorance (Tomasello et al. 2005). However, dilutingthe definition of teaching will not take us any closer to under-standing such motivational and cognitive traits. B & R suggest thata focus on rare cases where teaching is reserved for specialinstances to assist slow learners will be particularly informativeabout the evolution of human targeted instruction. We havepreviously argued that if selection favours teaching because it isnecessary to promote learning of critical skills, it should becommonwithin populations (Thornton & Raihani 2008). Indeed, allhuman children need to be taught, even if slow learners receivespecial instruction. Nevertheless, we concede that a speciesequipped with the ability to identify slow learners and target themfor special instruction might be seen to teach relatively rarely. Weagree that such cases of rare teaching may be of great value inunderstanding the cognitive foundations of human teaching, butonly if (1) we can determine that teaching is occurring and (2) wespecifically investigate the cognitive mechanisms involved. If wefind that an animal seems to teach, albeit infrequently, this tells uslittle about whether this is a case of ‘intentional’ or ‘targeted’instruction, rather than a response to behavioural cues. Forexample, B & R cite a recent study showing that adult femaleelephants, Loxodonta africana, occasionally simulate oestrus,apparently in an attempt to demonstrate to young relatives expe-riencing their first oestrus how to behave towards males (Bateset al. 2010). The jury must remain out on whether this is a case ofteaching (eight of the 19 cases of false oestrus occurred when therewere no young, first-oestrus females around to act as potentialpupils) but even if we accept this explanation, the mechanisms bywhich matriarchs may or may not recognize ‘those few youngsterswho really needed the help’ (B & R, page 1208) are unknown. Forinstance, mature females may respond to behavioural indicators ofthe youngsters’ incompetence (as in meerkats) or false oestruscould simply be a nonadaptive by-product of hormonal changes, ashas been found in chimpanzees, Pan troglodytes (Elder & Yerkes1936), orang-utans, Pongo pygmaeus (Schultz 1938) and gibbons,Hylobates lar (Barelli et al. 2007). We argue it may be moreproductive to determine whether elephants teach and then todesign experiments to test how they do it.

While we share B & R’s desire to further our understanding ofthe evolution of human teaching, we do not agree that a reliance onthe C & H definition to detect cases of teaching in nature riskshindering this enterprise. Instead, we suggest that the quest tounderstand the origins of human teaching will best be servedthrough two parallel lines of research. First, we can use the C & Hdefinition to determine whether an animal teaches and thenexamine how it does so. Second, as many comparative researchersare already doing, we can examine the domain-general mecha-nisms employed in some forms of human teaching.

The C & H criteria allow us to detect cases of teaching in nature,paving the way for experimental studies to investigate the mecha-nisms employed. Comparing the similarities and differences in themechanismsusedbyhumanandnonhuman teacherswillbecrucial inunderstanding theevolutionofdistinctlyhumanattributes.Moreover,adherence to the C & H definition when searching for animals that

teach does not prevent us from examining whether the domain-general mechanisms employed in some forms of human teachingalso occur in other species. Many of the mechanisms involved inhuman teaching are also deployed in a range of other situations.Consequently, comparative studies seeking to understand the evolu-tionary origins of these mechanisms do not necessarily have to occurin the context of teaching, or use species that are known to teach. Forexample, experimental studies of joint attention, theory of mind,imitation, causal reasoning and mental time travel in great apes,corvids and other animals for which there is no evidence of teachingare nevertheless crucial in helping us understand the evolution ofcognitive traits that humans incorporate into much of their teaching.As a parallel, monkeys cannot talk, but studies examining their abilityto represent hierarchically structured social relations (Seyfarth et al.2005) or form semantic combinations (Arnold & Zuberbühler 2006)may provide useful insights into precursors of language.

CONCLUSIONS

The C & H definition provides us with the means to documentteaching across a range of taxa and to unravel the selective pres-sures that may lead individuals, including humans, to invest inhelping others to learn. We full heartedly agree with B & R thata focus on the cognitive mechanisms of teaching, with analyses ofindividuals as well as populations, should be a priority for futureresearch. However, far from stifling cognitive research, the C & Hdefinition allows us to determine whether teaching occurs ina given species and, if so, to examine the underlying mechanisms.Moreover, productive experimental studies of the mechanismsinvolved in human teaching can proceed in animals that are notknown to teach. Eroding standards of evidence in studies ofteaching will take us no closer to understanding mechanisms, butmay muddy the waters and hinder future progress.

We benefited greatly from stimulating discussions with NicholaRaihani and Fred Ispanze and from comments by two anonymousreferees. Richard Byrne and Lisa Rapaport provided helpful andgracious comments on the paper. A.T. is funded by a David PhillipsResearch Fellowship from the BBSRC.

References

Arnold, K. & Zuberbühler, K. 2006. Semantic combinations in primate calls. Nature,441, 303.

Barelli, C., Heistermann, M., Boesch, C. & Reichard, U. H. 2007. Sexual swellings inwild white-handed gibbon females (Hylobates lar) indicate the probability ofovulation. Hormones and Behavior, 51, 221e230.

Bates, L. A., Handford, R., Lee, P. C., Njiraini, N., Poole, J. H., Sayialel, K., Sayialel, S.,Moss, C. J. & Byrne, R. W. 2010. Why do African elephants (Loxodonta africana)simulate oestrus? An analysis of longitudinal data. PloS One, 5, e10052.

Byrne, R. W. & Rapaport, L. G. 2011. What are we learning from teaching? AnimalBehaviour, 82, 1207e1211.

Caro, T. M. & Hauser, M. D. 1992. Is there teaching in nonhuman animals? QuarterlyReview of Biology, 67, 151e174.

Chittka, L. & Niven, J. 2009. Are bigger brains better? Current Biology, 19,R995eR1008.

Clutton-Brock, T. 2002. Breeding together: kin selection and mutualism in coop-erative vertebrates. Science, 296, 69e72.

Elder, J. H. & Yerkes, R. M. 1936. The sexual cycle of the chimpanzee.The Anatomical Record, 67, 119e143.

Fogarty, L., Strimling, P. & Laland, K. N. 2011. The evolution of teaching. Evolution,65, 2760e2770.

Franks, N. R. & Richardson, T. 2006. Teaching in tandem-running ants. Nature, 439,153.

Franklin, E. L., Richardson, T. O., Sendova-Franks, A. B., Robinson, E. J. H. &Franks, N. R. 2011. Blinkered teaching: tandem running by visually impairedants. Behavioral Ecology & Sociobiology, 65, 569e579.

Hoppitt, W. J. E., Brown, G. R., Kendal, R., Rendell, L., Thornton, A.,Webster, M. M. & Laland, K. N. 2008. Lessons from animal teaching. Trends inEcology & Evolution, 23, 486e493.

Hrdy, S. B. 2009. Mothers and Others: the Evolutionary Origins of Mutual Under-standing. Cambridge, Massachusetts: Harvard University Press.

A. Thornton, K. McAuliffe / Animal Behaviour 83 (2012) e6ee9 e9

Laland, K. N. & Hoppitt, W. J. E. 2003. Do animals have culture? EvolutionaryAnthropology, 12, 150e159.

Raihani, N. J. & Ridley, A. R. 2008. Experimental evidence for teaching in wild piedbabblers. Animal Behaviour, 75, 3e11.

Riboli-Sasco, L., Brown, S. P. & Taddei, F. 2008. Why teach? The evolutionaryorigins and ecological consequences of costly information transfer. In: Socio-biology of Communication: an Interdisciplinary Perspective (Ed. by P. d’Etorre &D. P. Hughes), pp. 265e274. Oxford: Oxford University Press.

Richardson, T. O., Sleeman, P. A., McNamara, J. M., Houston, A. I. & Franks, N. R.2007. Teaching with evaluation in ants. Current Biology, 17, 1520e1526.

Schultz, A. H. 1938. Genital swelling in the female orang-utan. Journal ofMammalogy, 19, 363e366.

Seyfarth, R. M., Cheney, D. L. & Bergman, T. J. 2005. Primate social cognition andthe origins of language. Trends in Cognitive Sciences, 9, 264e266.

Shettleworth, S. J. 2010a. Cognition, Evolution and Behaviour. 2nd edn. Oxford:Oxford University Press.

Shettleworth, S. J. 2010b. Clever animals and killjoy explanations in comparativepsychology. Trends in Cognitive Sciences, 14, 477e481.

Tehrani, J. J. & Riede, F. 2008. Towards an archaeology of pedagogy: learning,teaching and the generation of material culture traditions. World Archaeology,40, 316e331.

Thornton, A.&McAuliffe, K.2006. Teaching inwildmeerkats. Science,313, 227e229.Thornton, A. & Raihani, N. J. 2008. The evolution of teaching. Animal Behaviour, 75,

1823e1836.Thornton, A. & Raihani, N. J. 2010. Identifying teaching in wild animals. Learning &

Behavior, 38, 297e309.Tomasello, M., Carpenter, M., Call, J., Behne, T. & Moll, H. 2005. Understanding

and sharing intentions: the origins of shared cognition. Behavioral and BrainSciences, 28, 675e735.

West, S. A., Griffin, A. S. & Gardner, A. 2007. Social semantics: altruism, coopera-tion, mutualism, strong reciprocity and group selection. Journal of EvolutionaryBiology, 20, 415e432.