ways of thinking: from crows to children and back again · 2017-01-09 · of the mammalian...

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=pqje20

Download by: [Michigan State University] Date: 03 January 2017, At: 07:36

The Quarterly Journal of Experimental Psychology

ISSN: 1747-0218 (Print) 1747-0226 (Online) Journal homepage: http://www.tandfonline.com/loi/pqje20

Ways of thinking: From crows to children and backagain

Nicola S. Clayton

To cite this article: Nicola S. Clayton (2015) Ways of thinking: From crows to childrenand back again, The Quarterly Journal of Experimental Psychology, 68:2, 209-241, DOI:10.1080/17470218.2014.943673

To link to this article: http://dx.doi.org/10.1080/17470218.2014.943673

Published online: 10 Sep 2014.

Submit your article to this journal

Article views: 511

View related articles

View Crossmark data

http://www.tandfonline.com/action/journalInformation?journalCode=pqje20

http://www.tandfonline.com/loi/pqje20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/17470218.2014.943673

http://dx.doi.org/10.1080/17470218.2014.943673

http://www.tandfonline.com/action/authorSubmission?journalCode=pqje20&show=instructions

http://www.tandfonline.com/action/authorSubmission?journalCode=pqje20&show=instructions

http://www.tandfonline.com/doi/mlt/10.1080/17470218.2014.943673

http://www.tandfonline.com/doi/mlt/10.1080/17470218.2014.943673

http://crossmark.crossref.org/dialog/?doi=10.1080/17470218.2014.943673&domain=pdf&date_stamp=2014-09-10

http://crossmark.crossref.org/dialog/?doi=10.1080/17470218.2014.943673&domain=pdf&date_stamp=2014-09-10

EPS Mid-Career Award 2013

Ways of thinking: From crows to children and back again

Nicola S. Clayton

Department of Psychology, University of Cambridge, UK

This article reviews some of the recent work on the remarkable cognitive capacities of food-cachingcorvids. The focus will be on their ability to think about other minds and other times, and tool-using tests of physical problem solving. Research on developmental cognition suggests that young chil-dren do not pass similar tests until they are at least four years of age in the case of the social cognitionexperiments, and eight years of age in the case of the tasks that tap into physical cognition. This devel-opmental trajectory seems surprising. Intuitively, one might have thought that the social and planningtasks required more complex forms of cognitive process, namely Mental Time Travel and Theory ofMind. Perhaps the fact that children pass these tasks earlier than the physical problem-solving tasksis a reflection of cultural influences. Future research will hope to identify these cognitive milestonesby starting to develop tasks that might go some way towards understanding the mechanisms underlyingthese abilities in both children and corvids, to explore similarities and differences in their ways ofthinking.

Keywords: Corvid cognition; Cognitive development; Mental time travel; Mental attribution; Tool-use.

The last few decades have seen a shift in focuswithin the field of comparative cognition.Formerly the province of the primates, there isnow considerable evidence that a number of otheranimal groups also possess these cognitive abilities,including dolphins (Marino, 2002), hyaenas(Holekamp, Sakai, & Lundrigan, 2007), andcanids (Hare & Tomasello, 2005; Miklosi, Topal,& Csyani, 2004) and consequently there is increas-ing support for the hypothesis that cognition hasevolved convergently in these and other distantlyrelated groups that face similar physical and socialchallenges (Seed, Emery, & Clayton, 2009; vanHorik, Clayton, & Emery, 2012). Perhaps the

most striking case is that of corvids, members ofthe crow family that includes ravens, jays andmagpies (Emery & Clayton, 2004), because thesebirds are even more distantly related to primatesthan the mammalian examples listed above,sharing a common ancestor with the primatesover 300 million years ago. These birds have beenshown to be capable of feats that were thoughtuntil recently to be uniquely human, including theability to recall specific what-where-and-whenmemories about the past (Clayton & Dickinson,1998) and use such information to plan for thefuture (Cheke & Clayton, 2012; Correia,Dickinson, & Clayton, 2007; Raby, Alexis,

Correspondence should be addressed to Nicola S. Clayton, Department of Psychology, University of Cambridge, Downing Street,

Cambridge, CB2 3EB, UK. E-mail: [email protected]

I am deeply indebted to the Experimental Psychology Society for the Mid-Career Award. Financial support was provided by the

University of Cambridge, the Royal Society and Clare College, and research grants from the BBSRC and the ERC. None of this

research would have been possible without the help and hard work of the members of the Comparative Cognition lab, to whom I

am so grateful. Finally I thank my collaborators for their vital contributions to the work, in particular Nathan Emery and Tony

Dickinson for their support throughout my career, and Nathan for his wonderful artwork.

© 2014 The Experimental Psychology Society 209

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2015

Vol. 68, No. 2, 209–241, http://dx.doi.org/10.1080/17470218.2014.943673

mailto:[email protected]

Dickinson, & Clayton, 2007a). In addition tomental time travel (Clayton, Bussey, &Dickinson, 2003), these birds have been shown todisplay sophisticated social and physical cognition,such as perspective taking (Dally, Emery, &Clayton, 2006; Legg & Clayton, in press; Ostojic,́Shaw, Cheke, & Clayton, 2013) and experienceprojection (Emery & Clayton, 2001), cooperativeproblem solving (Seed, Clayton, & Emery, 2008)and creating novel tools to solve problems (Bird& Emery, 2009a, 2009b; Weir, Chappell, &Kacelnik, 2002). Furthermore, research on devel-opmental cognition in human children using com-parable non-verbal tests of cognition (e.g., Beck,Appeley, Chappell, Guthrie, & Cutting, 2011;Cheke, Loissel, & Clayton, 2012; Teufel,Clayton, & Russell, 2013) suggests that these abil-ities are not trivial, emerging relatively late in ahuman child’s cognitive development.

These findings raise a number of questionsabout what exactly it is that has evolved conver-gently. Are corvids really using human-like reason-ing to carry out these behaviours and solve thesetasks, and if not, what cognitive processes allowfor their impressive flexibility? How do thecorvids differ in their decision-making processesfrom those made by children? Are there a suite ofsimilarities and differences in the ways of thinkingused by corvids and by humans, and what can thistell us about the evolution of both the human andthe corvid mind?

In order to address these questions, this articlereviews some of the recent work on the remarkablecognitive capacities of food-caching corvids. Threesources of evidence will be discussed: social cogni-tion, mental time travel and physical problemsolving. In addition, the application of this workto research on developmental cognition in younghumans will be considered, for both comparativeand developmental cognition require the use oftasks that are, respectively, entirely or largely,non-verbal. The purpose is not to ask questionsabout whether or not corvids show cognitiveperformance equivalent to children of a certainage: adult corvids differ in many ways from youngchildren that make such questions meaningless.

The aim of these comparisons between corvidsand children is in fact twofold. First, to investigatetwo rather different kinds of mind in order tocompare and contrast the mechanisms they useto solve the tasks. Of interest here is the factthat the bird brain has a strikingly differentneural architecture from that of the mammalianbrain with a nuclear arrangement, bereft as it isof the mammalian six-layered neuroarchitecture(Jarvis et al., 2005). This raises interesting ques-tions about how information processing isachieved in these two kinds of mind, and specifi-cally how the information is passed betweengroupings in the avian brain as opposed tobetween layers in the mammalian cortex. Itshould be noted that, despite differences in thegross structure of the brain, there are in fact anumber of similarities in the way they are orga-nized, suggesting convergence between birds andmammals in the neural correlates of cognition(Güntürkün, 2012). For example, we now knowthat birds have a nidopallium caudolaterale,which is thought to be equivalent to the mamma-lian pre-frontal cortex (Güntürkün, 2005).Furthermore, it has recently been demonstratedthat the bird and mammalian brain share similarpatterns of large-scale network organization(Shanahan, Bingman, Shimizu, Wild, &Güntürkün, 2013).

Second, the comparison between corvids andchildren allows us to combine the strengths ofboth research areas. The field of comparative cog-nition has tended to focus on two questions,namely “do animals have cognitive processessimilar to humans?” and “how does cognitionevolve?”, whereas developmental psychologistsassume the eventual presence of human cognitionin young children and ask “when does itdevelop?” and “how do the reasoning processes ofyoung children differ from those of adults?” Anintegration of both fields may advance these ques-tions by enabling us to ask “what is it that evolvesand develops?”, focusing on the similarities anddifferences in performance on various experimentalinterventions to establish what mechanisms areused, when, and by whom.

210 THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2015, 68 (2)

CLAYTON

1. SOCIAL COGNITION

1.1. Corvid cache-protection tactics

Much of the work on social cognition in corvidsrevolves around the tactics that these birds use toprotect their caches of hidden food from beingstolen by other individuals, both conspecific (samespecies) and heterospecific (different species) pil-ferers (Vander Wall & Jenkins, 2003). Many ofthese species tend to live in large social groups,where they are likely to compete for food, andwhere the likelihood is that they can be seen byonlookers both when protecting their own cachesand when pilfering the caches of others. Indeed,pilferage rates can be quite high, ranging from 2to 30% per day, at least for artificial caches(reviewed in Vander Wall & Jenkins, 2003). Incorvids, the risk of pilferage by conspecifics isespecially great because these birds do not justrely on random search but possess observationalspatial memory and can therefore remember thelocations of caches that they have seen other indi-viduals make and return at a later date to steal thecontents of those caches when the cachers are notpresent to defend the cache sites (Bednekoff &Balda, 1996a, 1996b; Bugnyar & Kotrschal,2002a; Scheid & Bugnyar, 2008; Watanabe &Clayton, 2007).

It has been suggested that this observationalspatial memory may have acted as a catalyst in anevolutionary arms race between cacher and pilferer,resulting in the development of a suite of sophisti-cated strategies for cache-protection and pilfering(Bugnyar & Kotrschal, 2002b; Grodzinski &Clayton, 2010). As Bugnyar and Kotrschal(2002a) were the first to point out, observationalspatial memory markedly increases pilferingsuccess compared to random search, for example,and this has the knock-on effect of making itmore advantageous for cachers to evolve a protec-tive countermeasure to minimize the risk of cacheloss. Furthermore, once pilferers use observationalmemory, then the risk of pilfering becomescache-specific, and depends upon the pilferer’sknowledge of the cache, which in turn dependson whether or not the pilferer perceived the

caching event, and if so how accurately theyattended and remembered. This creates the possi-bility for a variety of cache-protection strategies toevolve that are concerned with altering a pilferer’sknowledge, e.g., hiding caches in sites that arehard for the potential pilferer to see, and minimiz-ing auditory information when the potential pil-ferer can hear but cannot see. For the purposes ofthis argument, “knowledge” simply refers to theinformation that is accessible to the cache-protectorand/or pilferer, without any necessary implicationabout the mechanisms with which they might attri-bute such information, through behaviour-readingor mind-reading for example. Indeed it remainsan open and much-debated question as towhether or not corvids (and non-human animalsin general) attribute knowledge or ignorance toothers, namely whether or not they possess a“theory of mind” (see Emery & Clayton, 2009a;Heyes, 1998; Lurz, 2009; Penn & Povinelli,2007). This is a point we shall return to when dis-cussing behaviourist and mentalist accounts ofsocial cognition in young children.

An added complexity is that any one individualcan play the role of both the protector of their owncaches and the pilferer of other birds’ caches. Thisrole-playing may have enabled shortcuts that accel-erated the evolution of more complex cognitivestrategies for pilfering and cache protection(Emery & Clayton, 2008). Consider the effect ofpilfering experience on a western scrub-jay’s pro-pensity to re-cache in new locations in private,when the original caching event had been observed(Emery &Clayton, 2001). The striking finding wasthat it is only those birds who themselves have beenthieves in the past that re-cache the food once thepotential pilferers have left the scene. Naïve jaysthat had not been given the opportunity of stealinganother bird’s caches did not do so. A capacity topredict high pilfering risk from the presence of anobserver by using their own experience as a pilferersuggests that the jays may be able to imagineanother individual’s point of view (Emery &Clayton, 2008). Such experience projection wouldcreate the opportunity for using many cache-pro-tection and pilfering strategies without learningthrough trial and error first, or the need for innate

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 2015, 68 (2) 211

COGNITION IN CHILDREN AND CROWS

strategies and counter-strategies to evolve. Ravenscan differentiate between knowledgeable andignorant competitors both as storers and as pilferers(Bugnyar & Heinrich, 2005). It would be fascinat-ing to determine whether they can also use theirexperience in one role to make decisions in theother.

Corvids use a suite of cache-protection strat-egies, ones which limit opportunities for potentialpilferers to witness caching events (Clayton,Dally, & Emery, 2007). In a series of experiments,we have investigated the caching strategies used bywestern scrub-jays to protect their caches whenbeing observed by a potential pilferer, by comparingthe behaviour of birds that are allowed to cache inprivate with those that are observed by a conspecificin a neighbouring cage. The key finding is that thebirds limit opportunities for conspecifics to witnessthese caching events: they preferentially cachebehind barriers when others are watching (Dally,Emery, & Clayton, 2005a), and use both distance(Dally et al., 2005a) and shade (Dally, Emery, &Clayton, 2004) to degrade the visual informationavailable to observers, preferences they do notshow when others cannot see what and wherethey cache. Their decisions are not limited to thevisual domain, however, for when the potential pil-ferers can hear but cannot see they conceal auditoryinformation by caching in a substrate in which theact of burying the food makes little noise, whereaswhen alone or when others can see and hear theyprefer to cache in noisy substrates (Stulp, Emery,Verhulst, & Clayton, 2009). This may also serveas a cache protection strategy, allowing the cacherto detect a potential cache-raid should the protectorbe within earshot of the pilferer but unable to see.

The jays also keep track of which particular indi-vidual bird watched them cache and when (Dallyet al., 2006) and take protective action accordingly,moving the high-risk caches only when the knowl-edgeable potential pilferer is no longer present(Dally et al., 2006; Emery & Clayton, 2001). Ifthe pilferer is present at the time of recovery,however, they then use “confusion tactics”, inwhich the cacher repeatedly moves the items fromone place to another, thereby sometimes makingfake caches and re-caches (Dally et al., 2006),

reminiscent of the Cups and Balls Effect andThree Shells Effect that magicians use to misdirectthe visual information that the onlooker sees. Indeciding which cache protection tactics to use thebirds take into account the dominance status ofthe potential pilferer in relation to their own dom-inance status, employing different strategies if theyare dominant to the onlooker than if they are sub-ordinate (Dally, Emery, & Clayton, 2005b, 2006).The variety of these cache protection strategies arenot exclusive to western scrub-jays: similar tacticshave been observed in ravens (e.g., Bugnyar &Heinrich, 2005; Bugnyar & Kotrschal, 2002a,2002b, 2004; Bugnyar, Stöwe, & Heinrich, 2007;Heinrich & Pepper, 1998), Clark’s nutcrackers(Clary & Kelley, 2011), Florida scrub-jays(Kulahci & Bowman, 2011) and Eurasian jays(Legg & Clayton, in press; Shaw & Clayton,2012, 2013, in press).

1.2. Experience projection by corvids

Perhaps the most striking finding, however, andone that was discussed earlier when consideringcomplex social cognitive skills including theory ofmind, is that it is only those birds who have beenthieves themselves in the past that engage in there-caching strategies of moving food to newplaces once the potential pilferer has left the scene(Emery & Clayton, 2001). Naïve birds that havenot stolen other birds’ caches do not do so, rulingout the possibility that such cache protection strat-egies are hardwired. Furthermore, the re-cachingbehaviour is not simply a response to the stress ofbeing observed and subsequent memory failure(Thom & Clayton, 2013), as has been suggestedrecently by computational modelling work (vander Vaart, Verbrugge, & Hemelrijk, 2012). It isimportant to note that in these experiments thejays were neither rewarded nor punished forre-caching, and in fact they were not given theopportunity to recover their re-caches and thus dis-cover whether or not re-caching was successful. Wecan therefore also rule out the possibility that thebirds could have learned about the benefits of re-caching. Emery and Clayton (2001) have arguedthat the inference is that the experienced pilferers


CLAYTON

engage in experience projection, predicting byanalogy of their own behaviour as a pilferer, antici-pating that the onlooker might do what they woulddo in such circumstances, and thus move theircaches to new places that the potential pilfererdoes not know about. Moreover, the responses ofthese experienced birds are highly flexible. Thebirds only re-cache the food when a potential pil-ferer has observed them cache—they do not movethe food to new places if the onlooker was theirmate, with whom they share their caches, or ifthe potential thief did not witness the cachingevent (Dally et al., 2005b; Emery & Clayton,2001).

In short, this finding suggests that the old adage“it takes a thief to know one” applies not only tohumans, but also to these food-caching jays. Thisis far from a trivial capacity. Experiments we haveconducted recently with young children suggestthat they have great difficulty projecting their ownexperience to infer what others have and have notseen, and do and do not know (Teufel et al., 2013).

1.3. Experience projection by children

In the experiment reported by Teufel et al. (2013),we tested two-year-old toddlers on a game in whichthe children had to put stickers of particularanimals onto the appropriate location on a gameboard. The task was based on a procedure devel-oped by O’Neill (1996) that was designed to testyoung children’s understanding of the relationshipbetween visual perception and knowledge for-mation, namely that seeing leads to knowing andnot seeing leads to not knowing. In the crucialpart of O’Neill’s experiment, the child sat at atable with the experimenter and a parent. Twoidentical opaque boxes were placed at the far endof the table beyond the child’s reach. In each trial,the experimenter dropped a sticker in one of thetwo boxes, and the child had to request help fromthe parent to get the sticker. In the “seeing” con-dition, when Mummy’s eyes were open and uncov-ered during the hiding event, the child shouldsimply ask for the sticker. By contrast, in the“unseeing” condition, in which Mummy hadcovered her eyes with her hands during the

hiding event, then the child would need to indicatewhich of the boxes contained the sticker onceMummy had opened her eyes again. In the originalexperiment by O’Neill (1996), young childrenspontaneously pointed to the correct box whenMummy’s eyes had been covered by her handsduring the hiding event, whereas if Mummy’seyes had been uncovered they would typically say“go get it” or words to that effect. This findingcould be taken to suggest that the toddlers havesome understanding of the relationship betweenvisual perception and knowledge formation, thatMummy’s lack of perceptual access to a certainfact leads to ignorance about that fact and that apointing gesture is an appropriate way of informingher about which box contains the sticker.

O’Neill (1996) argued that the results of herstudy did not necessarily indicate that seeing leadsto knowing, and not-seeing leads to not-knowing.Perhaps the toddlers simply realized that Mummywas disengaged from the hiding event and wantedto update her about this event. This simpler expla-nation could be of the form “tell other people aboutsignificant happenings they did not take part inwith me” O’Neill (1996, p. 674). Support for this“disengagement plus updating” account comesfrom a study by Dunham, Dunham, and O’Keefe(2000), who found that when Mummy firstcovered her eyes but then reopened them, eitherbefore or after the sticker was hidden, theyounger children pointed equally as often in bothconditions. Dunham et al. (2000) argued that theyounger children based their judgement ofwhether or not Mummy was disengaged on thebasis of global contextual cues of the hidinggame, while for the older children, disengagementwas associated with more specific behaviouralcues, namely whether or not Mummy’s eyes werevisible during the critical hiding event. Otherauthors have suggested simpler behaviouralaccounts of the young children’s behaviour (e.g.,Doherty & Anderson, 1999; Perner & Ruffman,2005). The children could be using a simple ruleabout the presence or absence of eyes, forexample, in order to decide whether or not topoint. They could also use their personal experienceof playing peekaboo and therefore learning about



the effects of covering and uncovering their eyes,without having to understand the causal connectionbetween perceptual access and knowledge transfer.That is why it was so important to contrast theexperienced pilferers with the naïve jays in the orig-inal experience projection experiment on the scrub-jays—because the birds had been hand raised, it waspossible to know what experiences these birds hadalready acquired (Emery & Clayton, 2001).

At issue, then, is whether, and to what extent,the children’s success in this experiment was cogni-tive or behavioural. To explore this further, weadded two artificial “seeing” and “unseeing” con-ditions, in which Mummy wore a pair of glassesthat meant that her eyes could no longer be seen(Teufel et al., 2013).1 From the perspective of theperson wearing the glasses, one pair was transparentand the other was completely opaque. It was notpossible to determine which pair of glasses wastransparent and which one was opaque by lookingat the glasses from the outside, however, becauseboth pairs had mirrored lenses. The two pairs ofglasses could be distinguished visually but not inways that predicted whether or not they were trans-parent or opaque. One pair was cased within a bluemask, and the other within a yellow one, such thatfor half of the tests yellow glasses were transparentand for half of the tests the blue ones weretransparent.

In Experiment 1, the children were given first-hand experience of wearing the glasses and there-fore of learning that they could see through thetransparent pair but not through the opaque pair,but importantly, they did not experience howanother person’s behaviour might change whenwearing the glasses. On test, in the two NaturalConditions, Mummy simply had her eyes open orcovered her eyes with her hands, as in the originalexperiment by O’Neill (1996). In the twoArtificial Conditions, Mummy put on a pair ofthe glasses and the children were asked whetheror not Mummy could see. In order to inferwhether or not Mummy could see the hiding ofthe sticker, the children had to use their ownfirst-hand visual experience of wearing the glasses.

If the toddlers understand that other people havevisual experiences that are similar to their ownthen they should be able to use their first-handexperience with the glasses to infer whether ornot Mummy can see. Furthermore, if the two-year-olds can understand the causal connectionbetween perceptual access and knowledge for-mation, they should be more likely to use pointinggestures in the condition in which, during thehiding of the sticker, Mummy was wearing theopaque glasses compared to the condition inwhich Mummy wore the transparent glasses, mir-roring the naturalistic “seen” (eyes open) and“unseen” (eyes covered) conditions.

In Experiment 2, the toddlers were given thereverse experiences. They were not given theopportunity to wear the glasses and therefore theycould not use their own first-hand experience toinfer whether or not Mummy could or could notsee. Instead, they received third-hand experience,namely by observing how Mummy’s behaviourchanged when she was wearing the glasses. Inthis way, they could learn that a particular pair ofglasses predicted a specific change in Mummy’s be-haviour, for example, that Mummy becomesclumsy or disengaged when she wears the opaqueglasses but not if she wears transparent glasses. Ifthe toddlers are using such behaviour-readingrather than an understanding of another’s mentalstates, then they should use pointing gestures dif-ferentially in response to a knowledgeable versusan ignorant parent only in Experiment 2.

We replicated the original findings of O’Neill(1996) in the Natural Seen and UnseenConditions, namely that when Mummy’s eyeswere open she could see and therefore she couldlocate the correct box. Whereas, when Mummy’seyes had been covered when the box had beenbaited, then she could not see and they needed topoint to the box containing the out-of-sightsticker. By contrast, we found some striking dis-sociations between the “Can Mummy see?” ques-tion and the pointing task in the children’sperformance in the Artificial Conditions (Teufelet al., 2013).

1The original idea was first suggested by Heyes (1998).


CLAYTON

In Experiment 1, the toddlers were able to cor-rectly infer whether or not Mummy could see byusing their first-hand experience of wearing theglasses and thus concluded that Mummy couldnot see when she wore the opaque glasses, butthat she could see when she wore the transparentglasses. However, they were unable to use thisinformation about whether or not Mummy couldsee to point specifically in those instances inwhich Mummy was ignorant of the correct boxlocation, namely when she had not witnessed thesticker being hidden because she had beenwearing the opaque glasses. In the absence ofseeing Mummy’s eyes open, the children weremuch more likely to point and there was no signifi-cant difference between the two artificial conditionson the child’s tendency to point. This result showsthat there is a dissociation between the seeing testand the pointing test, namely that the toddlerscan use their first-hand experience to pass the“Can Mummy see?” questions but not the pointingtest in these Artificial Seen and Unseen Conditions.So unlike the western scrub-jays in Emery andClayton’s (2001) experience projection experiment,the toddlers cannot use their experience to inferwhat another individual may or may not know.

It is not that the toddlers cannot use social infor-mation and behavioural cues to decide whether ornot they should point, as shown by the results ofExperiment 2. Here the opposite dissociation isfound in the two artificial conditions. The toddlersfail the “Can Mummy see?” question when theyonly have third-person experience of how Mummybehaves when she is wearing the glasses asopposed to first-hand experience of seeing throughthe glasses themselves. Yet they pass the pointingtask, only pointing when Mummy had worn theopaque glasses during the hiding of the sticker.

The results of the two experiments show adouble dissociation between the “Can Mummysee?” question and the pointing task. In each exper-iment, the toddlers pass one of the two conditionsbut not the other in the Artificial Seen and Unseenconditions, in which they cannot use the presenceor absence of eyes, or their personal past experienceof eyes being covered and uncovered by their hands.First-hand experience of wearing the mirrored

glasses can be used to infer whether or notMummy can see when she is wearing the glasses.However, the young children cannot use this infor-mation to infer whether or not they need to point.By contrast, the young children can use their third-hand experience of watching Mummy’s behaviourwhen she wears the mirrored glasses to inferwhether or not they need to point, perhaps basedon whether or not Mummy appears engaged or dis-engaged, but they fail to use this information toinfer whether Mummy can or cannot see. Takentogether, these findings suggest that the two-year-old child’s understanding of the “seeing leads toknowing” relationship is rather limited (Teufelet al., 2013).

1.4. Cooperative interactions

So far we have considered the social cognition ofcorvids in terms of competitive interactionsbetween caching and pilfering. However, theirsocial cognition is not restricted to such competitiveinteractions. Rooks can solve a cooperative task inwhich two birds must pull the string together inorder to reach food that would otherwise be outof beak reach (Seed et al., 2008). Even moreimpressive is the ability of the male Eurasian jayto understand what his female partner wants toeat, and share that food with her even when it isin conflict with his own desires (Ostojic ́ et al.,2013, 2014).

1.5. Corvid food-sharing

During the breeding season corvids will activelyprovision their partners with food. In the case ofEurasian jays, it is easy to observe the amountand type of food that the birds share, as each itemis transferred individually from beak to beak. Thisallowed us to investigate another aspect of Theoryof Mind, namely desire state attribution. Wetested whether the males could attribute changesin the desire state to their female partner duringcourtship feeding (Ostojic ́ et al., 2013, 2014).

Desire states can be manipulated by satiatingindividual jays with a particular type of food.Such satiety is specific because it reduces the



desire for that particular food without necessarilyimpacting on the desire for other food types. Toinduce specific satiety we gave the birds eitherwax worms (W) or mealworms (M) to eat duringthe pre-feeding stage and then a choice betweenthese two foods in the test phase (see Figure 1A;Ostojic ́ et al., 2013). All 13 jays showed specificsatiety, consuming a lower proportion of the W ifthey had been pre-fed W than if they had beenpre-fed M. As there was a large degree of variationbetween individuals in the amount of food eatenand in their preferences for one food overanother, Ostojic ́ et al. (2013) compared the birds’choices on the test trials with their eating patternin a baseline condition where the birds had beenpre-fed on their maintenance diet (MD), whichdoes not include W and M.

In terms of desire state attribution, the keyquestion is whether the male will take his partner’sspecific satiety into account in deciding whichfoods to offer her during courtship feeding. Amale that does take his partner’s specific satietyinto account should adjust his food-sharing behav-iour after observing her being pre-fed on a particu-lar food and share less of that food relative to abaseline in which the food was not devalued forthe female. In order to test this hypothesis, inthe Food-Sharing Experiment the desire state ofthe female was manipulated by pre-feeding hereither W or M or the baseline MD. The maleswere always fed the baseline MD to keep theirdesire constant over trials. There were two con-ditions. In the Seen Condition, the maleswatched their partner being pre-fed for 15 min inan adjacent compartment through a wire meshwindow that was covered by a transparent screen(Figure 1Bi). In the Unseen Condition, themales could not see what the females ate duringthe pre-feeding phase because an opaque screencovered the mesh window (Figure 1Bii), but themales could continue to hear the female. At theend of the pre-feeding phase of both conditions,the screen was removed and each male was given20 opportunities to choose a single W or M thathe could give to his partner through the wiremesh window.

If the male’s choice of what to share was influ-enced by what he had seen his partner eat, thenhis courtship feeding behaviour should differbetween these two conditions. The UnseenCondition is an important control to establishwhether the male’s choice is guided by what hehas seen her eat and thus knows about her desirestate, or whether he is simply behaviour-reading,that is responding to differences in her behaviourduring the food-sharing test phase. The femalescould, for example, have indicated their preferencefor a particular food type during the test phase bybegging more intensely for one of the foods, orimmediately after a sharing bout, the female couldhave indicated either acceptance or rejection ofthe food he had just offered her and this couldthen influence his subsequent choices of the twofoods.

The pattern of food-sharing did differ betweenthe Seen and Unseen conditions. It was crucial tocompare the sharing patterns in the test trials inwhich females were pre-fed either W or M tothat shown in the baseline when females werepre-fed MD because of inter-individual variationin the amount of sharing bouts and in the prefer-ence to share one food over another, just as wedid for the proportion of W eaten in the Food-Sharing Experiment.

In the Seen Condition, the males responded tothe current desire state of their partners bysharing a lower proportion of W when thefemales had been pre-fed W than when they hadbeen pre-fed M. By contrast, in the UnseenCondition there was no difference in the pro-portion of W given when the female was pre-fedW or M. This pattern of results suggests that themales needed to observe what the females hadeaten during the pre-feeding phase—simplyobserving her behaviour during the test phase didnot provide them with sufficient information toascertain which food she desired most. It is alsoimportant to note that the males never had visualaccess to the females during the test phase of thespecific satiety experiment and consequently thedecrease in the males’ sharing of the pre-fed foodin the Seen Condition could not be explained by


CLAYTON

Figure 1. The schematic of the experiments in Ostojic ́ et al. (2013).Note: Drawing by Nathan Emery.



them having had any prior experience of thefemales’ response to pre-feeding in this context.

One possible explanation for the differencebetween the Seen and the Unseen conditions isthat males developed specific satiety themselvesfrom watching the females being pre-fed on a par-ticular food and that their subsequent sharing be-haviour reflected their own decreased desire forthe pre-fed food, rather than an understanding ofthe female’s decreased desire as a consequence ofeating that food. To test this possibility, we ran athird experiment, which we termed theObservational Specific Satiety experiment(Figure 1C). The pre-feeding phase was identicalto the pre-feeding phase in the Seen Condition ofthe Food-Sharing experiment. The female wasthen released back into the aviary after the pre-

feeding stage and then the males were given theopportunity to eat W and M just as they were inthe original Specific Satiety Experiment.

If specific satiety can be induced simply bywatching the female eat a particular food asopposed to actually consuming the food them-selves, then the males’ eating behaviour in theObservational Specific Satiety experiment shouldshow the same pattern of results as in the originalSpecific Satiety experiment, namely a decrease inpreference for W if they had either been pre-fedW as opposed to M or watched the female beingpre-fed W as opposed to M. This was not thecase. The fact that observing the female beingpre-fed did not induce a decrease in the male’sown desire for that food rules out the possibilitythat males developed observational specific satiety

Figure 1. Continued.


CLAYTON

through watching the female being pre-fed on thetest foods during the Food-Sharing experiment.

Taken together, the results of these three exper-iments suggest that the jays can flexibly cater fortheir partner’s desire state provided they have seenwhat she has consumed beforehand during thepre-feeding stage, even when that desire state isin conflict with the male’s own motivational state(Ostojic ́ et al., 2013). Furthermore, a subsequentexperiment tested the extent to which the males’decisions are influenced by their own currentdesire (Ostojic ́ et al., 2014). We found that whenthe male’s desire matched his partner’s needs,then the males correctly shared the food that wasdesired by both parties. Even when the female’sdesire differed from the male’s own desire, thenhis decision of what to share was not driven entirelyby his own desire, suggesting that the males alsotook their partner’s desire into account. In short,male Eurasian jays can use this information toknow what their partners have been sated on andthus what they will want now. In common withthe research on experience projection by jays, thisfinding is particularly intriguing in the light ofsome experiments we have conducted recently ondesire state attribution in young children (Legg,Ostojic,́ Hughes, & Clayton, 2014).

1.6. Food sharing by children

A number of experiments have been conducted toinvestigate when young children develop theability to respond appropriately to another individ-ual who directly indicates a different desire to thechild’s own desire (conflicting desires tasks). Theimportance of the conflicting desire tasks thatinvolve inference lies in the possibility that theymight be tests of whether or not children are trulyrepresenting another individual’s desires as mentalstates. When the protagonist’s desires concur withtheir own, then young children can simply usetheir own desires without having to differentiatebetween their own mental states and those ofothers. By contrast, children who cannot differen-tiate other individuals’ mental states from theirown will make errors on the conflicting desirestasks when they have to infer what the protagonist

wants because they fail to understand that someonecould want something different to that which theywant. They do not need to do so, of course, whenthey are told explicitly about the other individual’sdesires because they can use simple behaviour-reading cues such as verbal and facial expressions.In the case where desires must be inferred,however, in order to pass the conflicting desirestask the children need to understand that mentalstates are subjective, that they can differ betweenpeople and even within the same individual atdifferent times. The power of perspective takingis that it allows you to understand other mindsand other times.

At first sight the results of the studies themselvesappear conflicting: in some studies, toddlers ofabout 18 months of age can pass such tasks(Repacholi & Gopnik, 1997), whereas otherstudies suggest that children under five years oldstruggle to inhibit their own desire when predictinganother individual’s desire (Moore et al., 1995).One difference between the two studies waswhether or not the children received explicit infor-mation about what their parents wanted. InRepacholi and Gopnik’s (1997) study the toddlersreceived direct and explicit information aboutwhat their parents wanted in the form of a verbaland facial expression. By contrast, in the exper-iment by Moore et al. (1995) the young childrenwere provided with rather limited information,namely a story about the protagonist being frigh-tened by an animal, and from that they had toinfer what the other individual desired.

Cassidy et al. (2005) directly tested whether ornot performance on conflicting desire tasks is influ-enced by how the children are informed of the otherindividual’s desires. In their study, three-year-oldchildren were asked to predict a protagonist’saction in a range of scenarios in which the chil-dren’s own desires either concurred or conflictedwith the protagonists’ desires. In some cases, theprotagonist’s desire was explicitly mentioned andin other cases it had to be inferred, for exampleby watching them eat and thus become sated on aparticular food. Cassidy et al. (2005) found thatthe three-year-olds were able to infer the otherindividual’s desires. They could also attribute



conflicting desires. However, when the taskrequired the children to do both things, namelyto infer what the protagonist wanted when thatdesire was in conflict with their own, then thethree-year-olds performed at chance. Takentogether, these findings might be taken to suggestthat although the very young children are able topass the task under some circumstances, they doso without truly representing the protagonist’sdesires. In line with studies on young children’sunderstanding of the relationship between visualperception and knowledge formation, a behav-iour-reading account seems more likely than a cog-nitive one. The children may have relied on rules ofthumb and past experience just as they may havedone in the “seeing leads to knowing” studies. Forexample, the items the children chose to sharewith the protagonist may have been informed bytheir past experience of contingencies between theprotagonist’s positive and negative expressionsand their future actions.

It should be noted, however, that failure in theconflicting desires task does not distinguishbetween children who lack an understanding ofthe subjectivity of mental states and those with aninability to inhibit control by their own mentalstate (Hughes & Russell, 1993; Perner, Lang, &Kloo, 2002). We have just begun to investigatechildren’s ability to comprehend the subjectivityof desires in a scenario where there was no needto inhibit their own desires (Legg et al., 2014).Conflicting desires tasks require children to rep-resent that their own desire is different fromanother individual’s. By contrast, in our currentstudy (Legg et al., 2014), the children were requiredto understand that another individual’s desire maychange over time. Consequently, subjectivity inthis task was expressed through changes to a thirdperson’s desires.

We asked children of four to six years of age topack a lunch box with food to give to another indi-vidual after observing them eating different foods tosatiety once per day across three consecutive days(Legg et al., 2014). The protagonist’s desires weremanipulated by pre-feeding them one of twofoods, just as we did for the jays. Over the threedays of testing, the children watched each of three

4-minute-long video clips depicting the same pro-tagonist (male for the boys, female for the girls)becoming sated in three different conditions: ontheir dinner (Eats Baseline); on crackers (EatsCrackers) or on slices of apple (Eats Apples). TheEats Baseline acted to assess the children’s baselinepreference for sharing apples and crackers in thesame way as the pre-fed MD baseline for the jays.Consequently, each child was always shown theEats Baseline video first, and then the order inwhich the children saw the Eats Apples and EatsCrackers videos was counterbalanced. Each 4-minute-long video was followed immediately by a10-second clip in which the protagonist sat infront of a bowl filled with crackers and a bowlfilled with slices of apples. In the Eats Baselinetrial the protagonist remained silent and made noindication of their preference for apples or crackers,whereas in the Eats Apples and the Eats Crackerstrials the protagonist indicated their preference forone food over the other (see below). The food-sharing phase followed immediately afterwards.The children were given 50 slices of apple and 50crackers and two lunch boxes, a blue-labelled con-tainer to give to the protagonist and a transparentcontainer for themselves. We introduced theadditional condition (Choose for Self) because pre-vious research has found that giving children theopportunity to choose some for themselvesimproves their ability to respond to others’ desirestates by reducing the salience of their own desire(Atance, Bernstein, & Meltzoff, 2010). The chil-dren were asked to choose food for the protagonistfirst before choosing their own food.

If children understand that another individual’sdesires can change over time, then the type of foodthey choose to share with that other person shoulddepend upon which food the protagonist is cur-rently sated on, just as the male jays’ decision ofwhat to feed his partner depended on what hehad seen her eat. In the first experiment, wetested whether or not children are able to changetheir choice of which food to give to the protago-nist in response to what that person directly indi-cated that they would like to eat. In the secondexperiment, children were tested using the sameparadigm but this time they saw either the


CLAYTON

protagonist indicating which food they were “fullof” or the children had to infer which food theprotagonist was sated on based on which foodthey had seen the protagonist eat during thepre-feeding phase. In this way, we couldcompare the children’s changes in sharing behav-iour when the protagonist’s different desires weredirectly indicated through verbal comments(Experiment 1: “I’d like some of that”;Experiment 2: “I’m full of that”) and when thesedifferent desires had to be inferred from observingwhich food the protagonist had eaten to satiety(Experiment 2: Infer group). The latter is analo-gous to the Seen Condition in the Food-Sharingexperiment of Ostojic ́ et al. 2013.

We found that the four- and five-year-old chil-dren responded correctly when they were directlyinformed of the protagonist’s desires (Legg et al.,2014), in line with the previous studies (e.g.,Cassidy et al., 2005). Unlike the jays in the Seencondition of Ostojic ́ et al. (2013), however, thefour- and five-year-old children failed to respondcorrectly when they had to infer the protagonist’sdesires based on what they had seen them eat. Itwas only the six-year-olds who could respondappropriately regardless of how the protagonist’sdesire was indicated. The results of our desiretask confirm previous studies by showing thatyoung children struggle to infer desires that conflictwith their own based on what they have seen themeat, but do not struggle when the protagonist hasdirectly indicated their desire in the conflictingdesire situation. Our findings add to the previouswork by demonstrating that the children’s difficultywith inferring desires is not limited to cases wheretheir own desire is in conflict with a protagonist’sbut is also present when a single individualchanges their desire over time. In short, wesuggest that young children below the age of sixstruggle to infer desires whenever they have to rep-resent the subjective nature of a desire state. Thisrequires them to engage in perspective taking, tounderstand that other minds may differ fromtheir own, but also that these minds, own andother, may change over time and thus differ fromthe present self. It is this topic of temporal perspec-tive taking to which we shall now turn.

2. MENTAL TIME TRAVEL

Although physical time travel remains within therealm of fiction, mental time travel is somethingwhich adult humans engage in much of the time,travelling backwards in the mind’s eye to rememberthe past (episodic memory) and forwards to plan forthe future (episodic future thinking). Temporalperspective taking allows us humans to possess anever present awareness of the passage of time, toacknowledge that what was once the future willrapidly become the past and that what happenedin the past can be recreated in the ever-changingyet persistent present to question which version iscorrect—just one, or all, or none. As Mark Twainassiduously remarked, “I have been through someterrible things in my life, some of which actuallyhappened”.

According to the Prospective Brain hypothesisproposed by Schacter and colleagues, the brain’scrucial function is to use our past experiences thathave been acquired by the episodic memorysystem in order to ponder, plan and predictfuture possibilities (Schacter, Addis, & Buchner,2007; Schacter et al., 2012). In other words, thefunction of episodic memory lies primarily in itsconstructive rather than its reconstructive ability—the system evolved in order to mentally simulatemultiple future scenarios by flexibly recombiningdetails of the events that have happened to us inthe past without having to physically engage inthe actual behaviour. There is a trade-off betweenflexibility and stability, however, and therefore thecost of using the episodic memory system is thatit is much more vulnerable than the semanticknowledge system to errors such as misattributionand false recognition because of its creativenature. This is the reason why our episodic mem-ories are surprisingly inaccurate compared to oursemantic knowledge-based memories, and whythey are more fragile and more labile—for eachtime we revisit these memories using our episodicsystem we recreate them. This does not happenwith semantic system, devoid as it is of a creativetime-travelling self (see Schacter, Addis, &Buckner, 2008 for an excellent and relativelyrecent review).



There has been considerable debate as towhether mental time travel is a uniquely humanability (e.g., Suddendorf & Corballis, 1997;Tulving, 2005), or whether we share this cognitivefeat with some other animals (e.g., Clayton, Bussey,& Dickinson, 2003; Corballis, 2013). Part of thedifficulty is that mental time travel has typicallybeen characterized in terms of its phenomenologi-cal consciousness. This episodic system has twokey features that the semantic knowledge systemdoes not. The first is an awareness of the subjectivesense of time, of re-experiencing now in the mind’seye an event that happened in the past (chronesthe-sia; Tulving, 2002). The second is an awareness ofbeing the owner and author of these memories andforethoughts (autonoesis; Wheeler, 2000). AsWilliam James so aptly wrote, “[m]emory requiresmore than the mere dating of a fact in the past. Itmust be dated in my past” (James, 1890, p. 6509).

In the absence of any agreed behaviouralmarkers of consciousness in non-linguisticanimals and preverbal children it is not possibleto evaluate empirically whether or not mentaltime travel is unique to humans, nor preciselywhen it develops in young humans (Griffiths,Dickinson, & Clayton, 1999). This dilemma canbe resolved to some degree, however, by differen-tiating between the phenomenological and behav-ioural criteria for episodic cognition. The latter isreferred to as episodic-like cognition in acknowl-edgement that these criteria explicitly ignore theinvolvement of phenomenological consciousness.

My colleagues and I have argued that episodic-like memory, the retrospective component ofmental time travel in animals, needs to fulfilthree criteria to meet the behavioural propertiesof episodic memory as defined for humans:content, structure and flexibility (Clayton, Bussey,& Dickinson, 2003). In terms of the content ofan episodic-like memory, the subject must remem-ber what happened, where and when on the basisof a single past experience, and in a way thatcannot be explained in terms of relative familiarity.Second, the what-where-and-when componentsform an integrated structure which allow thesubject to discriminate between similar episodesthat occurred at different times and places.

Finally the information must be capable of flexibledeployment, and thus be updated and generalizedacross situations.

2.1. Episodic-like memory in corvids

We tested whether western scrub-jays can episodi-cally recall specific past caching episodes in terms ofwhat happened where and when (Clayton &Dickinson, 1998). To do so we capitalized on thefact that these birds will readily cache perishableitems such as worms that degrade over time aswell as non-perishable nuts, and as they do noteat rotten worms, recovering the perishable foodcaches is only valuable when the food is still fresh.To investigate whether or not jays can rememberwhich foods they have hidden where and howlong ago, they were given the opportunity tocache worms and peanuts and then recover thesecaches of food either after a short delay of 4 hoursor a long delay of 124 hours (i.e., 5 days plus 4hours). This procedure ensured that the birdswere tested at the same time of day in both con-ditions so that they could not use a circadianrhythm to discriminate between short and longdelays but instead had to remember how long agothe caching event had occurred.

Although the jays had no cue to predict whetheror not the worms had perished other than thepassage of time between caching and recovery, thebirds rapidly learned that the highly preferredworms were still fresh when recovered 4 hoursafter caching, whereas they were rotten and tastedunpleasant after 124 hours. Consequently, thejays avoided the wax worm caches if they hadbeen cached 124 hours earlier, and instead onlyrecovered and ate the peanuts, which do notperish. After the birds had received this experienceof caching and recovering worms and peanuts afterthe short and long delays, we introduced test trialsin which the food was removed prior to recovery toensure that the birds relied on memory rather thanolfactory cues emanating directly from the food(Clayton & Dickinson, 1998). Subsequent testshave revealed that the jays also remember whichperishable foods they have hidden where and howlong ago, irrespective of whether the foods


CLAYTON

decayed or ripened (Clayton, Yu, & Dickinson,2001; de Kort, Dickinson, & Clayton, 2005).The jays also discriminate between similar episodesthat occurred at different times and places, demon-strating that they form integrated what-where-and-when components (Clayton et al., 2001).

We also found that the jays could update andgeneralize across situations, and therefore thatthese episodic-like memories also met the flexibledeployment criterion (Clayton, Yu, & Dickinson,2003). We tested this by allowing the birds tocache and recover perishable and non-perishablefood items using an interleaved procedure inwhich the birds cached in different trays on threesubsequent days, and only once they had completedthe caching trials did they get the opportunity torecover their caches from each tray. We reasonedthat, if the birds were capable of flexible deploy-ment, then they should be able to update theirknowledge about the rate of perishability of thefood and change their search behaviour at recoveryaccordingly, even though the episodic informationabout what they cached where and when wasencoded prior to the acquisition of the new knowl-edge about the decay rates, i.e., prior to any oppor-tunity to recover the caches. When the jays cachedperishable and non-perishable items in differentlocations in one tray and then subsequently discov-ered that the perishable items from another tray haddegraded more quickly than they expected, thebirds immediately switched their search preferencein favour of the non-perishable nuts when given theoriginal tray back. The birds continued to search forthe perishable food if it had been cached recently,thereby showing that they had not simply devel-oped a general aversion to searching for food thatmight rot. To our knowledge, this is the only pub-lished demonstration of the declarative flexibilitywith which animals can update their informationafter the time of encoding (Clayton, Yu, &Dickinson, 2003).

Since our initial studies, a number of other lab-oratories have investigated whether or not animalshave episodic-like memory using paradigms analo-gous to those employed with the jays. There is nowgood evidence that a diverse range of animals canremember the what-where-and-when of past

events, including rats (Babb & Crystal, 2006a,2006b), mice (Dere, Huston, & De Souza Silva,2005), magpies (Zinkivskay, Nazir, & Smilders,2009), chickadees (Feeney, Roberts, & Sherry,2009), chimpanzees (Martin-Ordas, Haun,Colmenares, & Call, 2010) and cuttlefish (Jozet-Alves, Bertin, & Clayton, 2013). The extent towhich these what-where-and-when memories ofanimals involve mental time travel remains anopen question, given that there is no way ofknowing whether or not animals possess autonoesisand chronesthesia. What we can investigate,however, is the developmental trajectories of thesecognitive abilities by determining whether or notchildren pass what-where-and-when tasks ataround the same age as they pass other tests ofmental time travel. Much of this work hasfocused on the prospective component, namely epi-sodic future thinking, rather than the mnemonicretrospective component.

2.2. Episodic-like memory and cognition bychildren

In the tests of episodic future thinking, the typicaldevelopmental trajectory is as follows: young chil-dren fail at three years of age, they show a transi-tional state of performance at four years of age,and they pass at five years of age (Atance &Meltzoff, 2005, 2006; Busby & Suddendorf, 2005;Lemon & Moore, 2007; Moore, Barresi, &Thompson, 1998; Russell, Alexis, & Clayton,2009; Suddendorf & Busby, 2005; but seeSuddendorf, Nielsen, & von Gehlen, 2011, for acase in which children do pass at four years ofage). Recent studies in developmental cognitionhave also adopted the what-where-and-when para-digm to track the emergence of the retrospectivecomponent, episodic-like memory, in young chil-dren. For example, Hayne and Imuta (2011) devel-oped a hide-and-seek paradigm akin to the scrub-jaycaching paradigm to test the ability of three- andfour-year-old children to remember specific pastcaching events, and Bauer et al. (2012) investigatedthe age at which children could bind the “where”component of a personally experienced event withthe other details of the event. In line with the



results of other tests of episodic cognition in childrenof three to five years of age, these what-where-and-when memories also show the typical developmentaltrajectory outlined above, although rudimentary epi-sodic-like memory skills do appear even in three-year-old children (Hayne & Imuta, 2011). Workby Scarf, Gross, & Columbo (2013) suggests thatthree-year-olds can remember these episodic-likememories over short but not long delays, soperhaps it is the ability to retain episodic memories,as opposed to the ability to encode the details of theevent, that develops across this time window(Mullcally & Maguire, 2014).

My colleagues and I (Russell, Cheke, Clayton, &Meltzoff, 2011) also tested three-, four- and five-year-old children on a what-where-and-when taskinspired by the Clayton and Dickinson (1998)scrub-jay paradigm. In our current what-where-and-when task the young children were requiredto bind what (the type of food), where (the locationof the boxes in which the food had been placed) andwhen (how long ago), just as the jays were requiredto do. By contrast, however, the children were givena prospective task rather than a retrospective one. Sorather than being given a choice between the foodsin the two boxes after either a short or long delay,which would have been analogous to the scrub-jaywhat-where-and-when memory procedure, in factthe children were asked at the time of “caching”(i.e., baiting) to anticipate which box they wouldvisit to get something to eat when they came backlater, either after a short or a long delay.

The issue is whether or not children show thesame developmental trajectory on our what-where-and-when task as they do on the othertests of episodic foresight. The assumption is thata similar developmental trajectory would suggestthat our what-where-and-when task taps intosomething similar to what the other tests of episo-dic cognition measure, a finding that would providesupport for the hypothesis that episodic-like cogni-tion really is like episodic cognition. By contrast, ifthe developmental trajectory for the what-where-and-when tasks were faster, such that even childrenof three years of age passed the task, then thisfinding would suggest that the scrub-jay-inspiredparadigm taps into a simpler process.

In our what-where-and-when task, the childrenwere given two types of food to choose from, just asthe jays were. Here the preferred food is a chocolateteddy bear and the less preferred food is a rich teafinger biscuit, analogous to the wax worms andpeanuts in the Clayton and Dickinson (1998)study, respectively. In one location (a hot box),but not in the other (a cold box), the chocolateteddy bear melts after the long interval but notafter the short interval, and the children are toldthat they will not be allowed to eat the melted cho-colate because it will make too much mess. Themelting chocolate scenario parallels the degra-dation of the wax worms in the Clayton andDickinson study. By contrast the biscuits do notchange over time, to parallel the peanuts in thescrub-jay study. So if children can rememberwhat-where-and-when, then they should choosethe hot box that contains the chocolate after theshort delay because the chocolate teddy bear willstill be intact, but they should switch to choosingthe cold box containing the finger biscuit afterthe long delay because by then the chocolate willhave melted and therefore they will not be able toeat it.

As with the jays, the children were first given theopportunity to learn that food in the hot box melts,whereas that in the cold box does not. First theywere shown both boxes and requested to placetheir hand in both boxes so that they could feelthe difference in temperature between the twoboxes. The children were then told that itemsplaced in the hot box melt after the long delaybut not after the short delay and shown photos ofa melted snow man and a melted bar of chocolatethat had been placed in the hot box after a longdelay. They were also told “food keeps cool andfresh in the cold box”.

Following this familiarization phase, the chil-dren were given direct experience of the effects ofthe delay on the foods in the boxes. First theywere shown two saucers, one containing a milkchocolate teddy bear and the other containing arich tea finger biscuit. The children were thenasked to place one saucer in each box. Each childreceived experience with both delays, and theorder was counterbalanced. For the short delay


CLAYTON

condition they were told “Let’s see what happens ifwe leave the things for a short time” before beingescorted into an adjoining room for 3–5 minutes.On their return they found that all the itemsremained intact. For the long delay condition theinstruction was “let’s see what happens when weleave the things for a long time”, and then wentto play for 30–45 minutes. On their return the chil-dren discovered that although the biscuit was stillintact, the chocolate had melted and had made aterrible mess, and therefore they could not eat it.Having experienced the effects of both delays onthe foods “cached” in the boxes, the children wereallowed to choose either a chocolate teddy bear ora finger biscuit to eat now. Only those childrenthat chose the chocolate were invited to take partin the test trials, since a strong preference for theunmelted chocolate was an essential pre-requisiteof this design, in much the same way as thestrong preference for wax worms in the scrub-jaystudies of Clayton and Dickinson (1998).

The test trial was administered the followingday, and the children were divided into twogroups, with half the children in each group receiv-ing either the long or the short delay conditions.Unlike the jays, the children in each conditionwere only tested on one of the two delays,because children often just repeat the response tothe first delay they are tested on. In the Future-Self Group, the children saw the chocolate teddybear being placed in the hot box and the fingerbiscuit in the cold box. They were then asked toanticipate which box they would visit to get some-thing to eat when they came back later, whichwould either be after a short (3–5 min) or a long(30–45 min) delay. In the Self-Caching Groupthe children were only shown the hot box (theywere told that the cold box was not workingtoday) and had to choose which food they would“cache” in it, to be retrieved either after a short ora long delay. In this latter case, the children werenot asked to imagine what they would do in thefuture, but only to reason about the future palatabil-ity of the food which could be done using semanticknowledge-based future thinking as opposed toprojecting one’s self into the future (i.e., episodicfuture thinking).

In terms of the proportion of correct responses,all the children were highly accurate with the shortdelay, having no problem selecting the box contain-ing the chocolate teddy bear at a time when thechocolate would not have melted. What the chil-dren found difficult was the long delay condition.Indeed we found that the three-year-olds failedthe long delay condition, performing no betterthan chance in both the Self-Caching andFuture-Self groups, whereas the four- and five-year-olds had a greater proportion of correctresponses. In the Self-Caching Group, both thefour- and five-year-olds passed and there was nodifference between the two age classes in the pro-portion of correct responses. In the Future-SelfGroup their performance did not reach statisticalsignificance, although there was a trend (p = .074for both age classes). Perhaps this difference reflectsthe fact that it is easier to solve the task using thesemantic knowledge-based system as opposed toepisodic future thinking, as discussed in the originalpaper (Russell et al., 2011).

Why did the children in the Future-Self Grouphave such difficulty with the long delay condition?One possibility is that they found it hard to inhibittheir current desire for chocolate (Atance &Meltzoff, 2006), which was in conflict with theirknowledge of what they should choose for thefuture (biscuit). A second possibility is thatalthough the children knew that the chocolatewould be melted after the long delay there was anoverwhelming temptation to choose the chocolatesimply because it was their favourite food.

In order to explore these possibilities, we testedfour-year-old children on the equivalent of theripening food scrub-jay paradigm, in which thepreferred food is inedible after the short delay anddelicious after the long delay, whereas the less pre-ferred food is always edible (Russell et al., 2011).Again the less preferred food was a finger biscuitplaced in the fridge to keep it fresh but now thefavourite food was cake. After the short delay thecake was still in the form of a cake mix (inedible)that turned into delicious cake in the oven afterthe long delay. Familiarization, training andtesting were conducted in the same way as for theprevious experiment. If the children’s performance



in the previous experiment was affected by theexecutive challenge of having to inhibit their prefer-ence for the food type they currently prefer, thenthey should perform poorly on this second exper-iment, choosing for their current desire, thebiscuit. If, however, their choices were masked bythe overwhelming preference for the favouritefood then they should choose the cake after thelong delay as the currently less preferred cake mixwill have become a baked cake.

We found that the four-year-olds in theFuture-Self group performed poorly on the longdelay, as they did in the previous experiment.This suggests that they struggled with the chal-lenge of inhibiting reference to the currently pre-ferred/edible food, namely the chocolate teddybear in experiment 1 and the finger biscuit inexperiment 2. Taken together, the results ofthese two experiments suggest that the poor per-formance of the four- and five-year-old childrenwas affected by the executive challenge of havingto inhibit their reference to the item they currentlypreferred, irrespective of whether or not that wastheir favourite food.

Interestingly, the four-year-olds in the Self-Caching Group showed a different pattern ofresults in experiment 2. They performed poorlyon the short delay and well on the long delay ques-tion. Perhaps this was because it taps into a naturalprepotent response to put a cake into the oven,whereas putting a biscuit into the fridge is notsomething most children will have had experienceof. In other words, it seems likely that the four-year-olds “gained a kind of illegitimate success onthe long-delay condition of Self-Caching…basing their choice on semantic association ratherthan the logic of the task” (Russell et al., 2011,p. 367).

2.3. Other tests of episodic-like cognition

So far we have focused our attention on the what-where-and-when component of mental timetravel, and the extent to which the performanceof three-, four-, and five-year-old children on thewhat-where-and-when tasks shows the same sortof developmental trajectory as that found in

episodic future-thinking tasks. Although there isgood evidence that a number of animals canremember the what-where-and-when of pastevents, the results for the children on the what-where-and-when prospective task are mixed. Onthe one hand the performance of the Self-Caching Group suggests an improvement overthe same developmental period as other tests ofepisodic cognition—what we failed to detect inthe episodic-like memory task was a developmen-tal transition at age four, or clear success on thetasks for the five-year-old children in the Future-Self Group. As discussed earlier, one possiblereason for this is that the children’s performancewas compromised by the executive challenge ofinhibiting reference to the currently referreditem. This is particularly intriguing in the lightof the recent studies we have conducted on theability of corvids to dissociate current and futuremotivational states.

2.4. Dissociation of future and presentmotivational states in corvids

It was Suddendorf and Corballis (1997) who firstargued that the ability to dissociate current fromfuture motivational states is a critical feature of epi-sodic future planning. Mental time travel provides aprofound challenge to the motivational systembecause the subject has to suppress thoughtsabout their own current motivational state inorder to allow them to imagine future needs, andto dissociate them from current desires. Indeed,this is probably why the young children have somuch difficulty passing these tests of episodic cog-nition discussed above. The hypothesis wasinspired by a comparative and developmental per-spective, and indeed it has led to a number oftests of whether animals can dissociate currentfrom future motivational needs. For example,Naqshbandi and Roberts (2006) provided squirrelmonkeys with the opportunity to choose betweeneating one versus four dates. Eating dates makesmonkeys thirsty, and the monkeys were thengiven some water after a long or a short delay. Infact the monkeys received water after a shorterdelay if they had chosen the one date rather than


CLAYTON

the four dates. The monkeys gradually reversedtheir natural preference to choose four dates overone date, suggesting that they could anticipatetheir future thirst. The problem is that one cangive a simple alternative associative explanation interms of reinforcement of the anticipatory actbecause the monkeys had received repeated trialsin which they had the opportunity to learn aboutthe consequences of their choices.

More convincing evidence for a dissociationbetween current and future motivational statescomes from a study by Correia et al. (2007) onthe food-caching choices of western scrub-jays.Here we capitalized on the specific satiety effectwe first discussed in the context of food-sharingto test whether the jays would choose to cache thefood they want now or whether they wouldchoose the food they think they will want whenthey come to recover their caches at a later date.At the start of the experiment the birds cachedthe food they desired at the time, but the birdsrapidly switched to caching the food they wouldwant at the time of recovery, despite continuingto eat the food they desired at the time. Thisresult suggests that the jays can plan futureactions based on what they anticipate they willdesire in the future as opposed to what they neednow. Subsequent studies (Cheke & Clayton,2012) show that Eurasian jays can also dissociatefuture from current motivational needs, and thatthey can differentiate between two distinct futuredesires and plan for each one appropriately,despite experiencing a conflicting currentmotivation.

It should be noted, however, that the bird doesnot need to imagine a future situation to solvesuch tasks. Clayton, Russell, & Dickinson (2009)argued that the act of recovering a particular foodmight trigger the memory of the time the birdcached that food. If the bird is hungry for that par-ticular food then recovering that food type will berewarding, thereby directly reinforcing the act ofcaching that particular food type through thememory of doing so. Such memory-mediatedreinforcement does not require the jay to imaginewhat its future motivational state will be. Indeed,it may be for this very reason that the jays are

better at dissociating current and future motiva-tional states than young children.

2.5. Tulving’s Spoon Test

Tulving has argued that it is possible to testwhether or not animals are capable of such episodicfuture thinking. He devised the “Spoon Test”,based on an Estonian children’s story tale, whichhe argues is a “future-based test of autonoetic con-sciousness that does not rely on and need not beexpressed through language” (Tulving, 2005,p. 43). In the original story, a young girl dreamsabout going to a birthday party. All of her friendsare eating her favourite desert, a delicious chocolatemousse. All she can do is watch, however, becausethe rules of the game are that no one is allowed toeat it without their own spoon, and she has notbrought one with her. As soon as she is backhome, she goes to the kitchen to find a spoonwhich she caches under her pillow for safekeeping,knowing she will have one to hand in the future—be it for real birthday parties or even dreams offuture birthday parties.

To pass the Spoon Test, the subject must actanalogously to the little girl, using past experienceto take action now for a future event, caching thespoon and then carrying it to a new party.Tulving (2005) pointed out that it is importantthat the spoon that has been obtained in anotherplace and at another time, in a distinctly separateevent and not just a continuation of the currentepisode. At issue then is whether there is any evi-dence that animals and young children can passthis Spoon Test.

Mulcahy and Call (2006) were the first to devisea Spoon Test for animals. They trained bonobosand orang-utans to use a tool to obtain a foodreward that would otherwise have been out ofreach, and then gave the apes the opportunity tochoose one of the tools from the experimentalroom, which they could carry into their sleepingquarters to be used the following morning.Although most of them sometimes chose thecorrect tool, the individual pattern of success foreach subject was not consistent across subsequenttrials, as one would expect if the apes had a



proper understanding of the task. Furthermore, asimpler explanation in terms of reinforcement ofthe anticipatory act cannot be ruled out becausethe apes received a number of training trials(Raby et al., 2007a; Shettleworth, 2007;Suddendorf, 2006). A more convincing case offuture planning was provided by Osvath andOsvath (2008), who showed that chimpanzeesand orang-utans are capable of overriding immedi-ate drives in favour of future needs when selecting atool for use in the future.

Food-caching scrub-jays also pass the SpoonTest, spontaneously planning for tomorrow’s break-fast without reference to their current motivationalstate (Raby et al., 2007a). The birds were given theopportunity to learn across six days that theyreceived no food for breakfast in one end compart-ment (the hungry room), and that they receivedfood for breakfast in the other end compartment(the breakfast room). For the rest of the day thebirds had free access to both compartments andalso to a central arena. During this time the birdswere maintained on powdered food, which theycould eat but not cache. This was to ensure thatthe birds were neither rewarded nor punished forcaching, and therefore one could not invoke areinforcement of the anticipatory act to explain theresults. Having been confined to each end compart-ment at breakfast time for an equal number of times,the birds were unexpectedly provided with theopportunity to cache food in both compartmentsone evening, at a time when there was plenty offood for them to eat and therefore no reason forthem to be hungry, by placing a bowl of wholefood items in the central arena and two cachingtrays, one in each end compartment. If the jays canplan for the future then they should cache food inthe hungry room, given that they did not knowwhich compartment they would find themselves inat breakfast the following morning and thereforewhether or not breakfast would be provided.

That is precisely what the birds did, suggestingthat the jays could anticipate their future need forbreakfast tomorrow, at a time when they wouldbe hungry, and even though they were not hungryat the time when they had the opportunity tocache. Furthermore, as mentioned above, we can

rule out an explanation in terms of mediatedreinforcement of the anticipatory act because thebirds had not been given the opportunity to cacheduring training. Indeed, Shettleworth has arguedthat

two requirements for genuine future planning are that the behav-

iour involved should be a novel action or combination of actions

… and that it should be appropriate to a motivational state other

than the one the animal is in at that moment…Raby et al.

describe the first observations that unambiguously fulfil both

requirements. (Shettleworth, 2007, p. 825)

Taken together these results suggest that the scrub-jays and chimpanzees can pass the Spoon Test.Furthermore, in an analogous task with young chil-dren, in which they were given the opportunity toplan for tomorrow’s playtime as opposed to break-fast, we found that the children did not pass thetask until they were four years of age (Atance,Louw, & Clayton, in press). At issue, however, iswhether or not these tasks require the use of episo-dic future thinking in the first place. In the absenceof language there is no way of knowing whether thejays’ ability to plan for future breakfasts is based ona projection of the self in time, which is what wouldbe required to satisfy an episodic future-thinkingaccount. The same reasoning applies to the apes’ability to choose tools and the children’s ability toplan whether or not to hide a toy to play with inthe future. It is entirely possible that the jays, apesand children rely on a semantic knowledge-basedsense of the future, in which they take prospectiveaction but without any personal mental timetravel into the future (Raby et al., 2007a; Raby,Alexis, Dickinson, & Clayton, 2007b). To solvethe Spoon Test, all the subject has to do is todecide what has to be done to ensure that theimplement will be at hand, be it a spoon, anothertool, or a food cache, without the need to imagineone’s self in possible future episodes or scenarios.What these studies do demonstrate, however, isthe capacity of non-linguistic animals and youngchildren to plan for a future motivational statethat stretches over a timescale of at least tomorrow.The results on corvids therefore challenge theassumption that the ability to anticipate and actfor future needs evolved only in the ape lineage(Raby et al., 2007a).


CLAYTON

3. PHYSICAL PROBLEM SOLVING

The third, and final, source of evidence for the cog-nitive capacities of corvids comes from studies ofphysical problem solving, and in particular theirability to manufacture tools and use them to reachfood that would otherwise be out of beak reach.Although a number of species of corvids occasion-ally use tools in the wild (Lefebvre, Nicolakakis, &Boire, 2002), the New Caledonian crow is by farthe most prolific tool-user (Hunt, 2000). Thisspecies makes two types of tools, which are usedfor obtaining insect larvae from holes in livingand dead wood, from leaf litter, and from thebase of plants (Hunt, 1996; Hunt & Gray,2004a). They make stick tools, which often havea hook sculpted at one end, and stepped-cuttools, which are fashioned from Pandanus andfern leaves, resulting in a tool that has a thickbase and tapers through one or more steps into anarrow tip (Hunt & Gray, 2004b).

A number of researchers have discussed whetheror not the tool-using skills of corvids are a reflectionof physical cognition, as they are thought to be forprimates (e.g., Emery & Clayton, 2009b; McGrew,2013). Support for this claim comes from thefinding that there is a strong positive correlationbetween the frequency of tool-use and brain sizein birds (Lefebvre et al., 2002) and in primates(Reader & Laland, 2002). Furthermore, youngchimpanzees require years of observing expert tool-makers to learn how to use their tools efficiently: ittakes five-and-a-half years to learn how to fish fortermites (Lonsdorf, 2006) and three to five yearsto learn how to crack nuts using hammers andanvils (Biro et al., 2003). Although NewCaledonian crows spontaneously insert sticks intocrevices irrespective of whether they have had theopportunity to observe a tool-user (Kenward,Weir, Rutz, & Kacelnik, 2005), subsequentstudies have shown that social learning is essentialfor the transmission of tool-manufacturing tech-niques (Holzhaider, Hunt, & Gray, 2010;Kenward, Weir, Rutz, & Kacelnik, 2006).

3.1. Innovative tool-use

The final source of evidence to support the argumentthat tool-users are intelligent comes from studies ofinnovative tool-use. For example, Weir et al. (2002)observed one captive New Caledonian crow calledBetty spontaneously modifying a piece of wire tomake a hook-shaped tool, which she then used topull up a small bucket containing meat. Even morestriking is the finding that hand-raised rooks,which do not use tools in the wild, will spon-taneously craft these hooked shape tools in the lab-oratory and use them to obtain food (Bird &Emery, 2009a). This is far from trivial, as recentstudies investigating the hook-making abilities ofyoung children suggest that it is not until childrenare about eight years of age that they can solvesuch tasks reliably (Beck et al., 2011; Chappell,Cutting, Apperley, & Beck, 2013; Cutting,Apperley, & Beck, 2011; Tecwyn, Thorpe, &Chappell, 2014). For example, in the study byBeck et al. (2011), the five- to seven-year-old chil-dren performed between 30 and 50% correct,whereas the eight-year-olds were 80% correct.Subsequent studies on four- and six-year-old chil-dren suggest that one of the difficulties lies in theill-structured nature of these tool innovation pro-blems, in which the various components of a sol-ution need to be retrieved and coordinated(Cutting, Apperley, Chappell, & Beck, in press).The four-year-old children could not coordinatethe actions required to solve the task, even whenexplicit instructions were provided. Although thesix-year-olds could do so, they struggled to bringto mind the components of the solution.

Innovative tool-use is not restricted to captivity,however. Wild rooks have been observed at theMembury service station on the UK’s M4 motor-way pulling up the bin liners of rubbish bins inorder to obtain food.2 Interestingly it takes about20 pulls to reach the food. Similar behaviour hasbeen observed in the Tower of London ravens.3

A classic case of innovative tool-use in apes is thefloating peanut task of Mendes, Hanus, and Call

2First shown on BBC1 in Britain’s Smartest Animals, July 2003, and repeated on BBC1 in Super Smart Animals, Feb 2012.3BBC1, The One Show: Bird Brains, May 2013.



(2007). In the original experiment, orang-utans wereprovided with a peanut that was out of reach at thebottom of a transparent vertical tube, and the chal-lenge was to develop a method for raising thepeanut within their reach. All five orang-utanssolved the task on the very first trial, by collectingwater and spitting it into the tube. It took aboutthree mouthfuls of water to raise the water to alevel at which the peanut could be reached.Mendes et al. (2007, p. 2637) argued that both thespontaneous development and the timing of theaction “make this behaviour a likely candidate forinsightful problem solving”. A sceptic would arguehowever that in the absence of knowing about theprecise reinforcement history of the animals, it isnot clear that this was the first instance of this behav-iour, thereby questioning whether this example is atrue case of innovation. The very same argumentcould be used to question the behaviour of thewild rooks and Tower of London ravens.

Subsequent studies have shown that whilechimps also solve the task either by spitting wateror urinating into the tube to raise the floatingpeanut to a level at which it can be reached byhand, young children had surprising difficulty onthe floating peanut task, even though they weregiven a watering can containing water which theycould use to raise the water level (Hanus,Mendes, Tennie, & Call, 2011). Of the twenty-four four-year-olds tested, only two were successful,whereas ten of the twenty-four six-year-olds passedthe task, and fourteen of the twenty-four eight-year-olds passed.

3.2. The Aesop’s fable task

Although the trap-tube task (Visalberghi & Trinca,1989) has been traditionally used as the benchmarktest to investigate what animals understand aboutthe tools that they use (e.g., Limongelli, Boysen,& Visalberghi, 1995; Martin-Ordas, Call, &Colmenares, 2008; Povinelli, 2000; Seed, Call,Emery, & Clayton, 2009; Seed, Tebbich, Emery,& Clayton, 2006; Taylor, Hunt, Medina, &Gray, 2009; Teschke & Tebbich, 2011;Visalberghi & Limongelli, 1994), considerable pro-gress has been made in recent years using a variant

of Mendes’ floating peanut task, inspired by theAesop’s fable “The Crow and The Pitcher”. InAesop’s 2000-year-old tale, a crow drops stonesinto a half-full pitcher of water to raise the levelwithin beak reach so that it can quench its thirst.

Bird & Emery (2009b) first developed thisAesop’s Fable paradigm to test physical cognitionin rooks. Rather than making the birds thirsty,the birds were tempted with a worm that was float-ing on the top of a vertical transparent tube half-filled with water. Next to the tube was a pile ofstones. Rather than having to collect water andpour, spit or urinate into the tube, the birds hadto use the stones to raise the water level. Bird andEmery found that rooks that had experience ofdropping stones, but not in the context of water,would spontaneously put the stones into the tubeto raise the water level within beak reach andthereby obtain the worm. Furthermore when thewater level was varied, the birds matched thenumber of stones required to increase the waterlevel to within beak reach. The birds were alsoselective in their choice of stones, using those thatwould most efficiently raise the water level.

A strength of both the floating peanut test andthe Aesop’s Fable task lies in their ability toexamine how animals respond to novel problemsthat are not related to the animal’s typical tool-use behaviour. This means that the ability to under-stand the affordances of the task can easily beseparated from exposure to technology and finemotor capacity and thus makes it appropriate forinvestigating physical cognition in animals that donot regularly use tools in the wild, as well as inyoung children who may not yet have developedsuch fine motor skills. Furthermore, because theparadigms are simple, it is easy to manipulate anumber of variables systematically in order to inves-tigate what subjects do and do not understandabout the physics of the task. For example, wecan test whether or not the subject will only usestones when the tube contains a liquid, asopposed to a solid substrate such as wood chipsor sand, and whether the subjects will be selectivein the type of tool that they use to raise the waterlevel, preferentially choosing objects that sink asopposed to those that float. These issues were


CLAYTON

explored by testing the performance of othercorvids that do not use tools in the wild, namelyEurasian jays (Cheke, Bird, & Clayton, 2011),and subsequently comparing their performance tothat of the habitual tool-using New Caledoniancrows (Jelbert, Taylor, Cheke, Clayton, & Gray,in press; Taylor et al., 2011), as well as young chil-dren of four to ten years of age (Cheke et al., 2012).

3.3. Testing corvids’ understanding of theAesop’s Fable task

The basic paradigm was as follows. After learningto drop stones into an air-filled tube to receive aworm reward from the bottom of the tube, the sub-jects were presented with a choice between a tubehalf-filled with woodchips or air and a tube half-filled with water, both containing a worm thatwas resting on the top of the substrate, either natu-rally in the case of the liquid and solid substrate, orthanks to blue tack in the case of the air-filled tube(Experiment 1). The reward was a worm that wasout of beak reach. Having been hand-raised, wecan be sure that they have had no experience ofwater in this form or discovering the consequencesof dropping stones into water.

Of the four Eurasian jays that were tested, twoof them learned to drop significantly more stonesinto the water than into the sawdust tube over thecourse of Experiment 1 (the other two birds wereuninterested in the task). These two birds quicklylearned across five trials to drop significantly moresinking objects into the water than floating onesin Experiment 2, suggesting that they were selectivein the type of tool that they used. The jays’ behav-iour could be explained in terms of instrumentallearning, however—dropping a sinkable objectinto the water tube leads to the increased prob-ability of obtaining a worm and therefore learningwhich tube (water versus wood chips) and whichobject (sinking versus floatable) was more likely tobe rewarded. For example, dropping a stone intothe water tube would lead to the worm beingwithin beak reach on about 20% of trials, whereasdropping a stone into the sawdust tube wouldnever result in the bird receiving a worm reward.On any given stone drop, placing a sinking object

into water would also lead to the worm movingslightly closer, which may also be rewarding,while placing a sinking object into sawdust leadsto no movement of the reward. The same argumentapplies to the choice of whether or not to drop asinking or floating object into the water tube.

To investigate whether the jays’ performancecould be explained exclusively by instrumentallearning we conducted a further series of tests.We found that the birds failed to learn when thecolour of the substrate arbitrarily indicated whichtube would be rewarded (Experiment 3) yet theirbehaviour was clearly goal directed, for the birdsselectively placed the stones in the tube that wasbaited with a worm as opposed to a tube in whichthe worm was not floating on the surface of thewater (Experiment 4). The birds were able tolearn a mechanized version of the task, in whichstone-dropping resulted in the approach of aworm yet they were unable to learn if the rewardprobabilities remained the same, but the rewarddid not move (Experiment 5). Perhaps the birdslearned the relationship between stone drops andmovement rather than the causal mechanism ofthe Aesop’s Fable task.

In the final control test of non-causal cues(Experiment 6), we presented the birds with anapparatus that contained three tubes, two widetubes on either side that were large enough forstones to be inserted and a central narrow tubethat contained the worm but was too small forstones to be inserted. The apparatus consisted ofa “U-tube” on one side and a single tube on theother, but the bases were hidden beneath anopaque box so that the birds could not see whichtubes were connected to one another. As the U-tube contained a single body of water, a stoneinserted into the wide arm would raise the level ofwater in both arms, while a stone inserted intothe single, wide tube would only raise the waterlevel in that particular tube. This task was designedto provide the same movement cues as the originaltask (i.e., stone insertion into one tube caused theapproach of food, stone insertion into the othertube did not) but with counter-intuitive mechanismcues. We found that the jays were unable to learnthis “U-tube” task even when they received twice



as many trials as on the original task. So, whileinstrumental learning involving movement cueswas both necessary and sufficient for the jays tolearn the task, the presence of cues suggesting a“possible” or “impossible” causal mechanism wereable to enhance or retard learning respectively(Cheke, Bird, & Clayton, 2011).

Subsequent studies on the New Caledoniancrows (Jelbert et al., in press; Taylor et al., 2011)suggested a comparable pattern of results. Thecrows preferentially dropped stones into a water-filled tube instead of a sand-filled tube; they werehighly selective in which objects they droppedinto the water tube, choosing sinking and solidones rather than floating and hollow ones; andadding more objects into a tube with a high waterlevel rather than a low one (Taylor et al., 2011).Like the jays, the New Caledonian crows failedthe more challenging task that required them toattend to the counter-intuitive causal cues in a U-shaped apparatus. These results suggest that thejays and New Caledonian crows possess a sophisti-cated, but incomplete, understanding of the causalproperties of displacement.4

3.4. Testing children’s understanding of theAesop’s Fable task

We also tested children between four and ten yearsof age on the Aesop’s Fable task (Cheke et al.,2012). The children were trained to drop stonesinto tubes in a similar manner to the Eurasianjays, and presented with three of the same tasks,namely a tube containing water and one containingsawdust, a water tube and a choice of sinking andfloatable objects, and the three tube apparatus con-sisting of a U-tube and a single tube. In all of thesetasks the children were given five 2-minute trials inwhich to attempt to retrieve a floating token thatcould be exchanged for a sticker.

Overall the performance of the children showeda clear developmental trajectory. Although childrenfrom four years of age were able to learn over thecourse of five trials to drop marbles into the tubecontaining water rather than the one containing

substrate in order to raise the level of a token, thistask was only solved reliably at the outset by chil-dren over eight years of age. Children from fiveyears or age were able to learn over the course offive trials to drop marbles, rather than cork balls,into the water tube to raise the level in order toreach the token but, in common with the firsttask, it was only children of eight years of age andabove that spontaneously solved this task. Asimilar developmental trajectory was also foundfor the U-tube task in which dropping a stoneinto one tube of water apparently caused a rise inthe level of the token floating in an adjacent tubeof water.

3.5. Comparing corvids and children on theAesop’s Fable task

How do these results compare with the performanceof the corvids? Taken together, the findings suggestthat rooks (Bird & Emery, 2009a, 2009b), Eurasianjays (Cheke, Bird, & Clayton, 2011) and NewCaledonian crows (Taylor et al., 2011) were allable to learn to drop stones into water rather thansubstrate within about five trials, a performancethat equates roughly with four- to seven-year-oldchildren (Cheke et al. 2012). Furthermore,Eurasian jays (Cheke, Bird, & Clayton, 2011) andNew Caledonian crows (Taylor et al., 2011) werealso able to learn to drop sinking items into waterrather than floating items within five trials, andagain this is a performance that equates roughlywith five- to seven-year-old children (Cheke et al.,2012). These comparisons are useful only to theextent to which they allow us to investigate thepossible presence of some common mechanism bywhich the children and corvids might solve suchtasks. Consequently, an interesting and importantdifference between the performance of the corvidsand the children emerges in the U-tube task. Onthis task the Eurasian jays (Cheke, Bird, &Clayton, 2011) and the New Caledonian crows(Jelbert et al., in press) performed substantiallyworse than the children, and much worse thanthey did on the other two tasks. Unlike the

4Note that the rooks in Bird & Emery’s (2009b) experiments were not presented with the U-tube task.


CLAYTON

corvids, the children’s performance on the U-tubetask was not affected by counter-intuitive mechan-ism cues.

It seems likely that the corvids possess a rudi-mentary concept of the causal mechanism under-lying the relationship between their stonedropping and the movement of the food reward,and the causal relationship in the U-tube task vio-lated the assumptions of what was possible accord-ing to this mechanism. By contrast it is interestingto note that the children’s performance was equiv-alent on this task to the other tasks, even in thoseindividuals that did not infer the presence of theU-tube. The children who were successful on theU-tube task were those that were able to noticeand describe the causal relationships betweenputting a stone in a particular place and theapproach of the food. These children could besaid to be learning using the rule “Do the actionthat causes the movement of the reward” whereasthe corvids’ performance was more in line withthe rule “Do the action that causes the movementof the reward, where the choice of action is affectedby, but not reliant on, some concept ofmechanism”.

Furthermore, the fact that children were notimpaired on the U-tube task relative to the othertasks may indicate that the they did not interpretthe event as “impossible” because they did notunderstand that insertion of an item into onebody of water cannot raise the level of anotherbody of water, and this is what allows them tosolve the U-tube task in a way that the corvidscannot. It seems quite likely that the childrenignored the “impossible” causal cues.Interestingly, children as old as eleven years ofage prioritize contingency and contiguity as evi-dence of causality above information about mech-anism, and may ignore information aboutmechanism altogether if this conflicts with appar-ent contingency information (Koslowski,Okagaki, Lorenz, & Umbach, 1989;Schlottmann, 1999, 2001). So unlike the corvids,the children’s willingness to believe their actionsto be causal was not affected by the presence ofcues indicating that the reward was not in thesame body of water as the stones.

It may be that children with no comprehensionof the mechanism did not explicitly attribute causa-tion, but simply allowed co-variation informationto guide their actions. That a bias to prioritize co-variation above mechanism information exists inchildren is extremely intriguing and warrantsfurther investigation. It may be that it is a usefuldevelopmental stage that exists to allow childrento learn about causation unfettered by ideas ofwhat is and is not “possible”. Equally intriguing isthe idea that such a bias could conceivably comeabout as a product of extensive technological encul-turation: children have considerable experience ofdevices with hidden mechanisms that make appar-ently impossible events happen (e.g., phones, com-puters, and light switches, to name but a few) andfor which children are not encouraged to investigatemechanism (indeed, many of us do not understandanything about the mechanism even as adults!) butsimply learn to use relying on co-variation betweenaction and outcome. One possibility is that for chil-dren growing up in heavily mechanized and electri-cally powered societies, an understanding of causalmechanisms has become somewhat separated fromthe development of causal judgements. To disen-tangle these two points, it would be interesting toinvestigate the performance of children and adultsfrom both mechanized and non-mechanized cul-tures to discover in which populations this biasexists.

3.6. Wrapping up and moving on

What can we conclude about “ways of thinking”from the comparisons of the performance ofcorvids and children on these cognitive tests? Toaddress this question I shall first consider whatthe research on developmental cognition tells usabout how a young child’s understanding of socialcognition, mental time travel and physicalproblem solving develops. The findings reviewedhere suggest that young children do not passthese tests until relatively late in development, gen-erally between four and six years of age in the caseof most of the experiments on social cognition andmental time travel, but it is not until they reacheight years of age that they spontaneously and



reliably solve the tasks that tap into physicalproblem solving.5

At first sight this pattern of results seems sur-prising: intuitively one might have thought thatthe tests on desire state attribution, what-where-and-when memory and future planning wouldrequire a more complex form of cognitive processthan the physical problem-solving tasks, namelythe ability to engage in perspective taking, toproject one’s self into other times (mental timetravel) and other minds (mental attribution). Intests of episodic and social cognition the typicaldevelopmental trajectory is between three and fiveyears of age, such that three-year-olds fail, four-year-olds show a transitional state of performance,and it is not until they reach five years of age thatchildren pass. The work reviewed here howeversuggests that this developmental trajectoryextends to six years of age in those cases wherethe child’s current self-desires are in conflict withwhat needs to be chosen for one’s future self (e.g.,Russell et al., 2011) or for another self (e.g., Legget al., 2014). Young children struggle to inferdesires whenever they have to represent the subjec-tive nature of these desire states as opposed todirectly experiencing them. It is particularly chal-lenging when this desire state is in direct conflictwith their current desire because this creates theadditional challenge of having to inhibit referenceto their present self.

Perhaps the fact that children pass these per-spective-taking tasks earlier than the water tasks(e.g., Cheke et al., 2012 for the Aesop’s Fable para-digm and Hanus et al., 2011 for the floating peanutexperiment) and the hook-tool tasks (e.g., Becket al., 2011; Cutting et al., 2011, in press) is areflection of the effects of extensive technologicalenculturation on cognitive development ingeneral, and physical problem solving in particular.It seems plausible that for children growing up inheavily mechanized and electrically poweredsocieties, an understanding of causal mechanismshas become somewhat separated from the develop-ment of causal judgements. Indeed, in the previous

section on physical problem solving I argued thatchildren show a bias to prioritize co-variation (con-tingency and contiguity) above mechanistic infor-mation, unfettered by ideas of what is and is not“possible”. It would be fascinating to conductcross-cultural studies to investigate these issuesfurther. In Western mechanistic societies childrengain considerable experience of devices withhidden mechanisms that make apparently imposs-ible events happen, such as turning on a smartphone, computer, or light. Often, children are notencouraged to investigate the mechanism butsimply learn to rely on co-variation betweenaction and outcome, thereby learning how topress the on-button but not how the on-buttonworks. There is a reliance on being the consumersof existing technology as opposed to the creatorsof new devices, which may be problematic for thedevelopment of some aspects of physical problemsolving. As Gerver Tully argued in his TED talkentitled “Five dangerous things you should letyour kids do”, learning to manipulate as well as toarticulate provides the foundation for the develop-ment of what he termed “real life intelligence”, anidea eloquently expressed by Bronowski’s famousphrase, “the hand is the cutting edge of the mind”(Bronowski, 1974, p. 116).

As alluded to in the previous section, it would befascinating to carry out ethnographic studies ofwhen these abilities develop in children from bothmechanized and non-mechanized cultures to dis-cover in which populations this bias exists and theextent to which it is a product of technical encul-turation or a universal feature of the human mind.These comparisons would allow us to investigatewhether, and to what extent, a child’s understand-ing of the physical world is developmentally delayedcompared to that of perspective taking (mental timetravel and mental attribution) and whether or notthis applies specifically to children that have beenraised in mechanized cultures. Such comparativestudies would also us to gain a better understandingof the mechanisms controlling the various behav-ioural decisions that children make in these and

5Of course, in the case of our experiments on experience projection it is not established when these abilities develop, as we were only

able to test two-year-olds.


CLAYTON

related physical problem-solving tasks, includinghow they bring to mind and coordinate thevarious actions required of the more complex phys-ical tasks involved in the hook manufacture tests(Cutting et al., in press).

This research also has a number of implicationsfor the corvids. The work on comparative cognitionsuggests that these birds have a rather sophisticatedunderstanding of social cognition, mental timetravel and physical problem solving. The fact thatonly a select few groups of animals possess thesecomplex cognitive abilities suggests that these abil-ities have not evolved by common descent, other-wise they should be prevalent in most, if not all,birds and mammals. Consequently it is arguedthat the cognitive similarities between corvids andchildren must be a case of convergent evolutionresulting in similar psychological functions,despite striking differences in their brain architec-ture and overall morphology (e.g., Emery &Clayton, 2004; van Horik et al., 2012).

What can this tell us about the evolution of boththe human and the corvid mind? By investigatingtwo distantly related kinds of mind, ones thathave very different neural architectures (Jarviset al., 2005) yet similar patterns of large-scalenetwork organization (Shanahan et al., 2013), wecan address the question “what exactly is it thathas evolved convergently?” How can these twokinds of mind, the nucleated bird brain thatcorvids possess and the laminar brain of children(and other mammals for that matter), produce thesame (or similar) psychological functions?Furthermore, by integrating the fields of compara-tive cognition and developmental cognition we canask questions about “what it is that evolves anddevelops”, focusing on both the similarities andthe differences in the performance of corvids andchildren on various experimental interventions inthe hope of establishing what mechanisms arebeing used, when, and by whom.

One issue is whether these corvids really aresolving the problems using human-like reasoningabilities or whether their decision-making processes

are in fact fundamentally different from those usedby children. The comparisons on the Aesop’s Fabletask suggest there are some important differences.The children are more ready than the birds “tobelieve in magic”, namely to accept contingencyas evidence for causality because of their inter-actions with mysterious hidden mechanisms. Forthe corvids, however, seeing is believing; theircausal reasoning abilities are constrained to therealm of the observable.6 This is likely to explaintheir failure on the “U-tube task” despite theirimpressive performance on the other physicalproblem-solving tasks and their spontaneousability to manufacture novel tools, even in thosespecies that have not been under strong evolution-ary selection pressure to do so in the wild.

Furthermore the comparisons of performanceon the future planning and desire state attributiontasks suggest that birds are more able to disengagefrom the current motivation state to make rationalchoices about a future desire for themselves or forthat of another individual in the present. Onepossible reason for this is that the birds employ adifferent mechanism to the children, perhapsusing rule-based knowledge (semantic associations)to think about other minds and other times ratherthan relying on the episodic cognition system(e.g., Cheke, Thom, & Clayton, 2011), which isby definition subjective and therefore more suscep-tible to errors that arise from overvaluing thecurrent self (Gilbert, 2006).

A second issue relates to the relationshipbetween intelligence and encephalization. Thecorvid brain is small in absolute size, about thesize of a walnut. Yet when expressed in relationto body size, the corvid brain is as large as thatfound in the great apes, and much larger thanthat of birds of similar body size such as pigeons(Emery & Clayton, 2004). As with apes, the enlar-gement occurs largely in the prefrontal cortex,which in the avian brain is now called the nidopal-lium caudale, as mentioned in the introduction(Jarvis et al., 2005). Recent findings suggest thatpatterns of neuronal connectivity may be key, and

6See Povinelli (2000) for a more detailed elaboration of the difference between the human and animal mind in reasoning about

observable and unobservable causes.



that the avian forebrain shows a similar large-scalenetwork organization to that of mammals, despitethe differences in the neural architecture ofnucleated avian brains and laminar mammalianones (Shanahan et al., 2013). Here, then, thenotion is that there is convergence in cognition,with similarities in terms of psychological functionand network organization (see also Güntürkün,2012), despite divergence in neuronal architectureand morphology, and differences in the decisionprocesses these two kinds of mind make whensolving these cognitive tasks. One could make ananalogy here with Apple and PC computers—atone level of analysis they do the same things in asimilar way, but viewed from another perspectivetheir operating systems are indeed different.

In order to further explore these hypothesesabout what exactly has converged in the cognitiveabilities of corvids and children, future researchwill aim to identify the cognitive milestones facedby both corvids and children, by developing tasksthat allow us to unravel the mechanisms thatoperate in these two kinds of minds with apparentlysimilar psychological functions. The quest is tosearch for both similarities and differences, whichis a key principle in convergent evolution.Understanding such patterns lies at the very heartof humanity, facilitating new ways of seeing andtriggering new ways of thinking.

Original manuscript received 20 March 2014

Accepted revision received 23 May 2014

First published online 10 September 2014

REFERENCES

Atance, C. M., Bernstein, D. M., & Meltzoff, A. N.(2010). Thinking about false belief: It’s not justwhat children say, but how long it takes them to sayit. Cognition, 116, 297–301.

Atance, C. M., Louw, A., & Clayton, N. S. (in press).Thinking ahead about where something is needed:New insights about episodic foresight in preschoolers.Journal of Experimental Child Psychology.

Atance, C. M., &Meltzoff, A. N. (2005). My future self:Young children’s ability to anticipate and explainfuture states. Cognitive Development, 20, 341–361.

Atance, C. M., & Meltzoff, A. N. (2006). Preschoolers’current desires warp their choices for the future.Psychological Science, 17, 583–587.

Babb, S. J., & Crystal, J. D. (2006a). Discrimination ofwhat, when, and where is not based on time of day.Learning and Behaviour, 34, 124–130.

Babb, S. J., & Crystal, J. D. (2006b). Episodic-likememory in the rat. Current Biology, 16, 1317–1321.

Bauer, P. J., Doydum, A. O., Pathman, T., Larkina,M., Güler, O. E., & Burch, M. (2012). It’s allabout location, location, location: Children’smemory for the where of personally experiencedevents. Journal of Experimental Child Psychology,113, 510–522.

Beck, S. R., Appeley, I. A., Chappell, J., Guthrie, C., &Cutting, N. (2011). Making tools isn’t child’s play.Cognition, 119, 301–306.

Bednekoff, P. A., & Balda, R. P. (1996a). Observationalspatial memory in Clark’s nutcrackers and Mexicanjays. Animal Behaviour, 52, 833–839.

Bednekoff, P. A., & Balda, R. P. (1996b). Social cachingand observational spatial memory in Pinyon jays.Behaviour, 133, 807–826.

Bird, C. D., & Emery, N. J. (2009a). Insightful problemsolving and creative tool modification by captive non-tool-using rooks. Proceedings of the Natural Academy ofSciences, 106, 10370–10375.

Bird, C. D., & Emery, N. J. (2009b). Rooks use stones toraise the water level to reach a floating worm. CurrentBiology, 19, 1410–1414.

Biro, D., Inoue-Nakamura, N., Tonooka, R.,Yamakoshi, G., Sousa, C., & Matsuzawa, T.(2003). Cultural innovation and transmission of tooluse in wild chimpanzees: Evidence from field exper-iments. Animal Cognition, 6, 213–223.

Bronowski, J. (1974). The ascent of man. London: LittleBrown.

Bugnyar, T., & Heinrich, B. (2005). Ravens, Corvuscorax, differentiate between knowledgeable and ignor-ant competitors. Proceedings of the Royal Society B:

Biological Sciences, 272, 1641–1646.Bugnyar, T., & Kotrschal, K. (2002a). Observational

learning and the raiding of food caches in ravens,Corvus corax: Is it “tactical” deception?. Animal

Behaviour, 64, 185–195.Bugnyar, T., & Kotrschal, K. (2002b). Scrounging

tactics in free-ranging ravens. Ethology, 108, 993–1009.

Bugnyar, T., & Kotrschal, K. (2004). Leading a conspe-cific away from food in ravens Corvus corax?. AnimalCognition, 7, 69–76.


CLAYTON

Bugnyar, T., Stöwe, M., & Heinrich, B. (2007). Theontogeny of caching in ravens, Corvus corax. AnimalBehaviour, 74, 757–767.

Busby, J., & Suddendorf, T. (2005). Recalling yesterdayand predicting tomorrow. Cognitive Development, 20,362–372.

Cassidy, K. W., Cosetti, M., Jones, R., Kelton, E., Rafal,V. M., Richman, L., & Stanhaus, H. (2005).Preschool children’s understanding of conflictingdesires. Journal of Cognition and Development, 6,427–454.

Chappell, J., Cutting, N., Apperley, I. A., & Beck, S. R.(2013). The development of tool manufacture inhumans: What helps young children make innovativetools?. Philosophical Transactions of the Royal Society B:Biological Sciences, 368, 20120409.

Cheke, L. C., Bird, C. B., & Clayton, N. S. (2011a).Tool-use and instrumental learning in the Eurasianjay (Garrulus glandarius). Animal Cognition, 14,441–455.

Cheke, L., Thom, J. M., & Clayton, N. S. (2011b).Prospective decision making in animals: A potentialrole for inter-temporal choice in the study of prospec-tive cognition. In M. Bar (Ed.), Predictions in the

Brain. Chapter 24 (pp. 325–354). Oxford: OxfordUniversity Press.

Cheke, L. C., & Clayton, N. S. (2012). Eurasian jays(Garrulus glandarius) overcome their current desiresto anticipate two distinct future needs and plan forthem appropriately. Biology Letters, 8, 171–175.

Cheke, L. C., Loissel, E., & Clayton, N. S. (2012). Howdo children solve the Aesop’s fable task?. Public

Library of Science One, 7, e40574.Clary, D., & Kelly, D. M. (2011). Cache protection

strategies of a non-social food-caching corvid,Clark’s nutcracker (Nucifraga columbiana). Animal

Cognition, 14, 735–744.Clayton, N. S., Bussey, T. J., & Dickinson, A. (2003).

Opinion: Can animals recall the past and plan forthe future?. Nature Reviews Neuroscience, 4, 685–691.

Clayton, N. S., Dally, J. M., & Emery, N. J. (2007).Social cognition by food-caching corvids Thewestern scrub-jay as a natural psychologist.Philosophical Transactions of the Royal Society B:

Biological Sciences, 362, 507–522.Clayton, N. S., & Dickinson, A. (1998). Episodic-like

memory during cache recovery by scrub jays. Nature,395, 272–274.

Clayton, N. S., Russell, J., & Dickinson, A. (2009). Areanimals stuck in time or are they chronesthetic crea-tures?. Topics in Cognitive Science, 1, 59–71.

Clayton, N. S., Yu, K. S., & Dickinson, A. (2001). Scrubjays (Aphelocoma coerulescens) form integratedmemories of the multiple features of caching episodes.Journal of Experimental Psychology: Animal Behavior

Processes, 27, 17–29.Clayton, N. S., Yu, K. S., & Dickinson, A. (2003).

Interacting cache memories: Evidence for flexiblememory use by Western scrub-jays (Aphelocoma cali-

fornica). Journal of Experimental Psychology: Animal

Behavior Processes, 29, 14–22.Corballis, M. C. (2013). Mental time travel: A case for

evolutionary continuity. Trends in Cognitive Sciences,17, 5–6.

Correia, S. P. C., Dickinson, A., & Clayton, N. S. (2007).Western Scrub-Jays (Aphelocoma californica) anticipatefuture needs independently of their current motiva-tional state. Current Biology, 17, 856–861.

Cutting, N., Apperly, I. A., & Beck, S. R. (2011). Whydo children lack the flexibility to innovate tools?.Journal of Experimental Child Psychology, 109,497–511.

Cutting, N., Apperly, I. A., Chappell, J., & Beck, S. R.(in press). The puzzling difficulty of toolinnovation: Why can’t children piece their knowl-edge together?. Journal of Experimental Child

Psychology.Dally, J. M., Emery, N. J., & Clayton, N. S. (2004).

Cache protection strategies by western scrub-jays(Aphelocoma californica): Hiding food in the shade.Proceedings of the Royal Society B: Biological Sciences,271, S387–S390.

Dally, J. M., Emery, N. J., & Clayton, N. S. (2005a).Cache protection strategies by western scrub-jays,Aphelocoma californica: Implications for social cogni-tion. Animal Behaviour, 70, 1251–1263.

Dally, J. M., Emery, N. J., & Clayton, N. S. (2005b).The social suppression of caching in western scrub-jays (Aphelocoma californica). Behaviour, 142,961–977.

Dally, J. M., Emery, N. J., & Clayton, N. S. (2006).Food-caching western scrub-jays keep track of whowas watching when. Science, 312, 1662–1665.

Dere, E., Huston, J. P., & De Souza Silva, M. A. (2005).Episodic-like memory in mice: Simultaneous assess-ment of object, place and temporal order memory.Brain Research, 16, 10–19.

Doherty, M. J., & Anderson, J. R. (1999). A new look atgaze: Preschool children’s understanding of eye-direction. Cognitive Development, 14, 549–571.

Dunham, P., Dunham, F., & O’Keefe, C. (2000). Two-year-olds’ sensitivity to a parent’s knowledge state:



Mind reading or contextual cues?. British Journal of

Developmental Psychology, 18, 519–532.Emery, N. J., & Clayton, N. S. (2001). Effects of experi-

ence and social context on prospective caching strat-egies by scrub jays. Nature, 414, 443–446.

Emery, N. J., & Clayton, N. S. (2004). The mentality ofcrows. Convergent evolution of intelligence in corvids

and apes. Science, 306, 1903–1907.Emery, N. J., & Clayton, N. S. (2008). How to build a

scrub-jay that reads minds. In S. Itakura & K.Fujita (Eds.), Origins of the social mind: Evolutionaryand developmental views. Chapter 4, (pp. 65–98).Springer: Tokyo.

Emery, N. S., & Clayton, N. S. (2009a). Comparativesocial cognition. Annual Review of Psychology, 60,87–113.

Emery, N. J., & Clayton, N. S. (2009b). Tool use andphysical cognition in birds and mammals. CurrentOpinion in Neurobiology, 19, 27–33.

Feeney, M. C., Roberts, W. A., & Sherry, D. F. (2009).Memory for what, where, and when in the black-capped chickadee (Poecile atricapillus). Animal

Cognition, 12, 767–777.Gilbert, D. (2006). Stumbling on happiness. Knopf:

New York.Griffiths, D. P., Dickinson, A., & Clayton, N. S. (1999).

Episodic memory: What can animals rememberabout their past?. Trends in Cognitive Sciences, 3,74–80.

Grodzinski, U., & Clayton, N. S. (2010). Problems facedby food-caching corvids and the evolution of cognitivesolutions. Philosophical Transactions of the Royal SocietyB: Biological Sciences, 365, 977–987.

Güntürkün, O. (2005). The avian ‘prefrontal cortex’ andcognition.CurrentOpinion inNeurobiology,15, 686–693.

Güntürkün, O. (2012). The convergent evolution ofneural substrates for cognition. Psychological

Research, 76, 212–219.Hanus, D., Mendes, N., Tennie, C., & Call, J. (2011).

Comparing the performance of apes and human chil-dren in the floating peanut task. Public Library of

Science, 6, e19555.Hare, B., & Tomasello, M. (2005). Human-like social

skills in dogs?. Trends in Cognitive Sciences, 9,439–444.

Hayne, H., & Imuta, K. (2011). Episodic memory in 3-and 4-year old children. Developmental Psychobiology,53, 317–322.

Heinrich, B., & Pepper, J. W. (1998). Influence of com-petitors on caching behavior in the common ravens,Corvus corax. Animal Behaviour, 56, 1083–1090.

Heyes, C. M. (1998). Theory of mind in nonhuman pri-mates. Behavioral Brain Sciences, 21, 101–114.

Holekamp, K. E., Sakai, S. T., & Lundrigan, B. L.(2007). The spotted hyaena (Crocuta crocuta) as amodel system for study of the evolution of intelli-gence. Journal of Mammalogy, 88, 545–554.

Holzhaider, J. C., Hunt, G. R., & Gray, R. D. (2010).Social learning in New Caledonian crows. Learning& Behavior, 38, 206–219.

Hughes, C., & Russell, J. (1993). Autistic children’s dif-ficulty with mental disengagement from an object: Itsimplication for theories of autism. Developmental

Psychology, 29, 498–510.Hunt, G. R. (1996). Manufacture and use of hook-tools

by New Caledonian crows. Nature, 379, 249–251.Hunt, G. R. (2000). Tool use by the New Caledonian

crow Corvus moneduloides to obtain Cerambycidaefrom dead wood. Emu, 100, 109–114.

Hunt, G. R., & Gray, R. D. (2004a). The crafting ofhook tools by wild New Caledonian crows. BiologyLetters, 271, 88–90.

Hunt, G. R., & Gray, R. D. (2004b). Direct observationsof pandanus-tool manufacture and use by a NewCaledonian crow (Corvus moneduloides). Animal

Cognition, 7, 114–120.James, W. (1890). The principles of psychology. New York:

Holy.Jarvis,E.D.,Güntürkün,O.,Bruce,L.,Csillag,A.,Karten,

H., Kuenzel,W.,…Butler, A. B. (2005). Avian brainsand a new understanding of vertebrate brain evolution.Nature Reviews Neuroscience, 6, 151–159.

Jelbert, S. A., Taylor, A. H., Cheke, L. C., Clayton, N.S., & Gray, R. D. (in press). Using the Aesop’s Fableparadigm to investigate causal understanding of waterdisplacement by New Caledonian crows. Public

Library of Science ONE.Jozet-Alves, C., Bertin, M., & Clayton, N. S. (2013).

Evidence of episodic-like memory in cuttlefish.Current Biology, 23, R1033–R1035.

Kenward, B., Weir, A. A. S., Rutz, C., & Kacelnik, A.(2005). Behavioural ecology: Tool manufacture bynaive juvenile crows. Nature, 433, 121.

Kenward, B., Weir, A. A. S., Rutz, C., & Kacelnik, A.(2006). Development of tool use in NewCaledonian crows: Inherited action patterns andsocial influences. Animal Behaviour, 72, 1329–1343.

de Kort, S. R., Dickinson, A., & Clayton, N. S. (2005).Retrospective cognition by food-caching Westernscrub-jays. Learning and Motivation, 36, 159–176.

Koslowski, B., Okagaki, L., Lorenz, C., & Umbach, D.(1989). When covariation is not enough – The role of


CLAYTON

causal mechanism, sampling method, and sample-sizein causal reasoning.ChildDevelopment, 60, 1316–1327.

Kulahci, I. G., & Bowman, R. (2011). Recachingdecisions of florida scrub-jays are sensitive to ecologi-cal conditions. Ethology, 117, 700–707.

Lefebvre, L., Nicolakakis, N., & Boire, D. (2002). Toolsand brains in birds. Behaviour, 139, 939–973.

Legg, E. W., & Clayton, N. S. (in press). Eurasian jays(Garrulus glandarius) conceal caches from onlookers.Animal Cognition.

Legg, E. W., Ostojic,́ L., Hughes, C., & Clayton, N. S.(2014). How behavioural cues influence children’sdesire state attribution. Manuscript in preparation.

Lemon, K., & Moore, C. (2007). The development ofprudence in the face of varying future rewards.Developmental Science, 10, 502–511.

Limongelli, L., Boysen, S. T., Visalberghi, E. (1995).Comprehension of cause-effect relations in a tool-using task by chimpanzees (Pan troglodytes). Journalof Comparative Psychology, 109, 18–26.

Lonsdorf, E. V. (2006). What is the role of mothers inthe acquisition of termite-fishing behaviors in wildchimpanzees (Pan troglodytes schweinfurthii)?.Animal Cognition, 9, 36–46.

Lurz,R. (2009). If chimpanzees aremindreaders, could be-havioral science tell? Toward a solution to the logicalproblem. Philosophical Psychology, 22, 305–328.

Marino, L. (2002). Convergence of complex cognitiveabilities in cetaceans and primates. Brain, Behaviorand Evolution, 59, 21–32.

Martin-Ordas, G., Call, J., & Colmenares, F. (2008).Tubes, tables and traps: great apes solve two functionallyequivalent trap tasks but show no evidence of transferacross tasks. Animal Cognition, 11, 423–430.

Martin-Ordas, G., Haun, D., Colmenares, F., & Call, J.(2010). Keeping track of time: Evidence for episodic-like memory in great apes. Animal Cognition, 13,331–340.

McGrew, W. C. (2013). Is primate tool use special?Chimpanzee and new Caledonian crow compared.Philosophical Transactions of the Royal Society B:

Biological Sciences, 368, 20120422.Mendes, N., Hanus, D., & Call, J. (2007). Raising the

level: Orangutans use water as a tool. Biology Letters,3, 453–455.

Miklosi, A., Topal, J., & Csanyi, V. (2004). Comparativesocial cognition: What can dogs teach us?. AnimalBehaviour, 67, 995–1004.

Moore, C., Barresi, J., & Thompson, C. (1998). Thecognitive basis of future-oriented prosocial behaviour.Social Development, 7, 198–218.

Moore, C., Jarrold, C., Russell, J., Lumb, A., Sapp, F., &MacCallum, F. (1995). Conflicting desire and thechild’s theory of mind. Cognitive Development, 10,467–482.

Mulcahy, N. J., & Call, J. (2006). Apes save tools forfuture use. Science, 312, 1038–1040.

Mullally, S. L., & Maguire, E. A. (2014). Learning toremember: The early ontogeny of episodic memory.Developmental Cognitive Neuroscience, 9, 12–29.

Naqshbandi, M., & Roberts, W. A. (2006). Anticipationof future events in squirrel monkeys (Saimiri sciureus)and rats (Rattus norvegicus): Tests of the Bischof-Kohler hypothesis. Journal of Comparative

Psychology, 120, 345–357. (2006).O’Neill, D. K. (1996). Two-year-old children’s sensi-

tivity to a parent’s knowledge state when makingrequests. Child Development, 67, 659–677.

Ostojic,́ L., Legg, E. W., Shaw, R. C., Cheke, L. G.,Mendl, M., & Clayton, N. S. (2014). Can maleEurasian jays disengage from their own currentdesire to feed the female what she wants?. BiologyLetters, 10, 20140042.

Ostojic,́ L., Shaw, R. C., Cheke, L. G., & Clayton, N. S.(2013). Evidence suggesting that desire-state attribu-tion may govern food sharing in Eurasian jays.Proceedings of the Natural Academy of Sciences, 1101,4123–4128.

Osvath, M., & Osvath, H. (2008). Chimpanzee (Pantroglodytes) and orang-utan (Pongo abelii) forethought:Self-control and pre-experience in the face of futuretool use. Animal Cognition, 11, 661–674.

Penn, D. C., & Povinelli, D. J. (2007). On the lack ofevidence that non-human animals possess anythingremotely resembling a ‘theory of mind’. PhilosophicalTransactions of the Royal Society B: Biological Sciences,362, 731–744.

Perner, J., Lang, B., & Kloo, D. (2002). Theory of mindand self-control: More then a common problem ofinhibition. Child Development, 73, 752–767.

Perner, J., & Ruffman, T. (2005). Psychology: Infants’insight into the mind: How deep?. Science, 308,214–216.

Povinelli, D. J. (2000). Folk physics for apes: The chimpan-zee’s theory of how the world works. Oxford: OxfordUniversity Press.

Raby, C. R., Alexis, D. M., Dickinson, A., & Clayton,N. S. (2007a). Planning for the future by WesternScrub-Jays. Nature, 445, 919–921.

Raby, C. R., Alexis, D. M., Dickinson, A., & Clayton,N. S. (2007b). Empirical evaluation of mental timetravel. Behavioral Brain Sciences, 30, 330–331.



Reader, S. M., & Laland, K. N. (2002). Social intelli-gence, innovation and enhanced brain size in pri-mates. Proceedings of the National Academy of

Sciences, 99, 4436–4441.Repacholi, B. M., & Gopnik, A. (1997). Early reasoning

about desires: Evidence form 14- and 18-month olds.Developmental Psychology, 33, 12–21.

Russell, J. Alexis, D. M., & Clayton, N. S. (2009).Episodic future thinking in 3- to 5- year-old-chil-dren: The ability to think of what will be neededfrom a different point of view. Cognition, 114,56–71.

Russell, J. R., Cheke, L. C., Clayton, N. S., & Meltzoff,A. M. (2011). What can What-Where-WhenWWW) binding tasks tell us about young children’sepisodic future foresight?. Theory and two exper-iments. Cognitive Development, 26, 356–370.

Scarf, D., Gross, J., & Colombo, M. (2013). To have andto hold: episodic memory in 3- and 4-year-old chil-dren. Developmental Psychobiology, 55, 125–132.

Schacter, D. L., Addis, D. R., & Buckner, R. L. (2007).Remembering the past to imagine the future: Theprospective brain. Nature Reviews Neuroscience, 8,657–661.

Schacter, D. L., Addis, D. R., & Buckner, R. L. (2008).Episodic simulation of future events: Concepts, dataand applications. Annuals of the New York Academy

of Sciences, 1124, 39–60.Schacter, D. L., Addis, D. R., Hassabis, D., Martin, V.

C., Spreng, N., & Szpunar, K. K. (2012). The futureof memory: Remembering, imagining, and the brain.Neuron, 76, 677–694.

Scheid, C., & Bugnyar, T. (2008). Short-term observa-tional spatial memory in Jackdaws (Corvus monedula)and Ravens (Corvus corax). Animal Cognition, 11,691–698.

Schlottmann, A. (1999). Seeing it happen and knowinghow it works: How children understand the relationbetween perceptual causality and underlying mechan-ism. Developmental Psychology, 35, 303–317.

Schlottmann, A. (2001). Perception versus knowledge ofcause and effect in children: When seeing is believing.Current Directions in Psychological Science, 10,111–115.

Seed, A. M., Call, J. Emery, N. J., & Clayton, N. S.(2009). Chimpanzees solve the trap problem whenthe confound of tool-use is removed. Journal of

Experimental Psychology: Animal Behavior Processes,35, 23–34.

Seed, A. M., Clayton, N. S., & Emery, N. J. (2008).Co-operative problem solving in rooks (Corvus

frugilegus). Proceedings of the Royal Society B:

Biological Sciences, 275, 1421–1429.Seed, A. M., Emery, N. J., & Clayton, N. S. (2009).

Intelligence in corvids and apes: A case of convergentevolution?. Ethology, 115, 401–420.

Seed, A. M. Tebbich, S., Emery, N. J., & Clayton, N. S.(2006). Investigating physical cognition in rooks,Corvus frugilegus. Current Biology, 16, 697–701.

Shanahan, M., Bingman, V. P., Shimizu, T., Wild, M.,& Gunturkun, O. (2013). Large-scale networkorganization in the avian forebrain: A connectivitymatrix and theoretical analysis, Frontiers inComputational Neuroscience, 7, 1–17.

Shaw, R. C., & Clayton, N. S. (2012). Eurasian jays(Garrulus glandarius) flexibly switch caching and pil-fering tactics in response to social context. AnimalBehaviour, 84, 1191–1200.

Shaw, R. C., & Clayton, N. S. (2013). Careful cachersand prying pilferers: Eurasian jays (Garrulus glandar-ius) limit auditory information available to competi-tors. Proceedings of The Royal Society of London Series

B, 280, 1–7.Shaw, R. C., & Clayton, N. S. (in press). Pilfering

Eurasian jays use visual and acoustic inofrmation tolocate caches. Animal Cognition.

Shettleworth, S. J. (2007). Planning for breakfast.Nature, 445, 825–826.

Stulp, G., Emery, N. J., Verhulst, S., & Clayton, N. S.(2009). Western scrub-jays conceal auditory infor-mation when competitors can hear but cannot see.Proceedings of the Royal Society of London Series B

Biological Letters, 5, 583–585.Suddendorf, T. (2006). Behavior: Enhanced: Foresight

and evolution of the human mind. Science, 312,1006–1007.

Suddendorf, T., & Busby, J. (2005). Making decisionswith the future in mind. Dveelopmental and com-parative identification of mental time travel.Learning and Motivation, 36, 110–125.

Suddendorf, T., & Corballis, M. C. (1997). Mental timetravel and the evolution of the human mind. Genetic,Social, and General Psychology Monographs, 123,133–167.

Suddendorf, T., Nielsen, M., & von Gehlen, R. (2011).Children’s capacity to remember a novel problem andsecure its future solution. Developmental Science, 12,1–8.

Taylor, A. H., Elliffem, D., Hunt, G., Emery, N. J.,Clayton, N. S., & Gray, R. D. (2011). NewCaledonian crows learn the functional properties ofnovel tool types.Public Library of ScienceOne, 6, e26887.


CLAYTON

Taylor, A. H., Hunt, G., Medina, F., & Gray, R. D.(2009). Do new caledonian crows solve physical pro-blems through causal reasoning?. Proceedings of the

Royal Society of London Series B, 276, 247–254.Tecwyn, E. C., Thorpe, S. K. S., & Chappell, J. (2014).

Development of planning in 4- to 10- year-old chil-dren: Reducing inhibitory demands does notimprove performance. Journal of Experimental ChildPsychology, 125, 85–101.

Teschke, I., & Tebbich, S. (2011). Physical cognition andtool-use: Performance of Darwin’s finches in the two-trap tube task. Animal Cognition, 14, 555–563.

Teufel,C.R., Clayton,N. S.,&Russell, J. R. (2013). Two-year-old children’s understanding of visual perceptionand knowledge formation in others. Journal of

Cognition and Development, 14, 203–228.Thom, J. M., & Clayton, N. S. (2013). Re-caching by

Western scrub-jays (Aphelocoma californica) cannotbe attributed to stress. Public Library of Science One,8, e52936.

Tulving, E. (2002). Episodic memory: From mind tobrain. Annual Review of Psychology, 53, 1–25.

Tulving, E. (2005). Episodic memory and Autonoesis:Uniquely Human?. In H. Terrace & J. Metcalfe(Eds.), The missing link in cognition. Origins in self-

reflective consciousness (pp. 3–56). Oxford: OxfordUniversity Press.

Van der Vaart, E., Verbrugge, R., & Hemelrijk, C. K.(2012). Corvid re-caching without ‘theory of mind”:A model. Public Library of Science One, 7, e32904.

Vander Wall, S. B., & Jenkins, S. H. (2003). Reciprocalpilferage and the evolution of food-hoarding behav-ior. Behavioral Ecology, 14, 656–667.

Van Horik, J., Clayton, N. S., & Emery, N. J. (2012).Convergent evolution of cognition in corvids, apesand other animals. In T. Schackelford & J. Vonk(Eds.), The Oxford handbook of comparative evolution-ary psychology, Chapter 5 (pp. 80–101). OxfordUniversity Press.

Visalberghi, E., & Limongelli, L. (1994). Lack of com-prehension of cause-effect relations in tool-usingcapuchin monkeys (Cebus apella). Journal of

Comparative Psychology, 108, 15–22.Visalberghi, E., & Trinca, L. (1989). Tool use in capu-

chin monkeys: Distinguishing between performingand understanding. Primates, 30, 511–521.

Watanabe, S., & Clayton, N. S. (2007). Observationalvisuospatial encoding of the cache locations ofothers by western scrub-jays (Aphelocoma californica).Journal of Ethology, 25, 271–279.

Weir, A. A. S., Chappell, J., & Kacelnik, A. (2002).Shaping of hooks in new caledonian crows. Science,297, 981.

Wheeler, M. A. (2000). Episodic memory and autonoe-tic awareness. In E. Tulving & F. I. M. Craik, (Eds.),The Oxford handbook of memory (pp. 597–625).Oxford: Oxford University Press.

Zinkivskay, A., Nazir, F., & and Smulders, T. V. (2009).What-where-when memory in magpies (Pica pica).Animal Cognition, 12, 119–125.



ways of thinking: from crows to children and back again · 2017-01-09 · of the mammalian...

Documents