Mental Time Travel in Animals?
October 30, 2010 3 Comments
By Thomas Suddendorf and Janie Busby
Are humans alone in their ability to reminisce about thepast and imagine the future? Recent evidence suggests that food-storing birds (scrub jays) have access to information about what they have stored where and when. This has raised the possibility of mental time travel (MTT) in animals and sparked similar research with other species. Here we caution that such data do not provide convincing evidence for MTT. Examination of characteristics of human MTT (e.g. non-verbal declaration, generativity, developmental prerequisites) points to other avenues as to how a case for animal MTT could be made. In light of the current lack of evidence, however,we maintain that MTT is a uniquely human characteristic.
Do animals reminisce about the good old days and ponder what the future might hold for them? Humans frequently engage in such mental time travel (MTT), reliving pastevents and entertaining possible future scenarios(Box 1). It has been argued that MTT is unique to humans, and that its emergence was a prime mover inhominid evolution. Recently, a series of innovative studies on food-storing scrub jays has raised doubt about this claim. In recovering stored food, these birds appear toact in ways that depend on what they stored where andwhen in the past, and on what they might expect to happen in the future. This has sparked interest in similar capacities in other species. Although we applaud these efforts, we argue here that current evidencedoes not yet warrant crediting other species with MTT. By examining other characteristics of MTT we point todifferent ways in which evidence could be obtained if thecompetence were to exist in animals.
Episodic memory and the scrub jay challenge
Travelling mentally into one’s past requires memory. Animals clearly have memory. But some memory researchers argue that there are distinct memory systems, raising the possibility that one of them is uniquely human.Tulving proposed that remembering an episode – what happened to me where and when – is processed in a different memory system from other types of information. Episodic memory is thought to be part of the larger explicit system (i.e. comprising the memories thatwe can declare) rather than the implicit system. Semantic memory – memory for facts about the world, not personal experiences – is the other explicit component. Episodic memory can be dissociated from semantic memory in that each can be selectively impaired andeach is associated with distinct cortical activity patterns. Tulving claims that only humans have episodic memory. Although the term is widely used, there has been some confusion about what episodic memory means, partly because Tulving modified his definition from the type of information stored (what, where and when) to anemphasis on what he terms ‘autonoetic’ (self-knowing) consciousness. He did this because it became clear that one can know (semantic memory) something about what happened where and when, without remembering that past episode.
Autonoetic consciousness is the subjective (recollective) experience associated with travelling back to a point in time and re-experiencing a pastevent. This requires concepts of self (the traveller)and subjective time (the dimension along which to travel). So whereas the earlier definition describes episodic memory in terms of the information encoded, the later emphasizes a specific type of recall; that is, MTT into the past.
We know that animals, from bees to monkeys, can learn from single events. But do they revisit the events that shaped their past? Innovative work by Clayton, Dickinson and their colleagues has produced perhaps the strongest case yet for episodic memory in animals. Their studies on the caching and retrieval behaviour of scrub jays (Fig. 1) have shown that these animals can form integrated memories of what was cached where and when. Scrub jays appropriately adjust recovery attempts of differentially perishable food caches depending on howlong ago they stored the food items. The authors concluded that the birds show all the behavioural criteria of episodic memory. But Clayton and colleagues cautiously refer to ‘episodic-like’ memory because the phenomenological experience defined by Tulving – autonoetic consciousness– could not readily be shown to exist. Is episodic-like memory anything like episodic memory? Clayton, Griffiths and Dickinson offer the following choice:
Option 1: Insist that autonoetic consciousness is crucial to episodic memory, which, they submit, probably makes it impossible to refute the claim of human uniqueness (because in the absence of language in animals there areno obvious behavioural manifestations).
Option 2: Characterize episodic memory in terms of the information encoded (i.e. Tulving’s earlier definition) and accept that scrub jays have it because they have beenshown to encode the what, where and when informationinto an integrated memory.
At first sight these seem to be the only options and, asscientists, it would seem clear which option we should adopt. Hypotheses need to be falsifiable. But let us be clear what is and what is not implied if we take Option 2 andsay that jays (and possibly other species) have episodic memory. It is implied that jays use what, where, when information; it is not implied that their memory retrieval is anything like remembering past episodes. In fact, the cognitive processes involved need not be about the past at all.
Box 1. Time in the human mind
More than half of adult conversation refers to past or future events. Who did what to whom and what happened next? What will happen, where and when?Many human actions are based on remotfuture goals. Although there is a growing literature on differences infuture time orientation, the basic capacity to entertain eventsremoved in time seems universal. Past events can be reconsidered, evaluated (e.g. regretted) and extrapolated to assess their consequences. Many aspects of human volition make sense only in thelight of MTT, including apparent biological paradoxes such assuicide (when the future outlook is particularly bleak), celibacy and hunger strikes. Humans construct personal and community histories and plans. Strategies (for individuals or even entire countries) can be coordinated, progress reviewed and adjustments made if necessary.Temporal concepts, time keepers and calendars have been developed to aid our orientation and plans. In short, it is safe to say that MTT is a significant human attribute, important to our dominance of the planet.
As Dretske noted: event A might cause cognitive change B that effects behaviour C at a later point in time, but this need not imply that B carries any information about A itself – the mediator B might be causal ratherthan informational. Thus, although jays perform actions C (recovery) that make sense only in the light of A (caching what, where and when) it need not imply that B represents the past event A. B might instead be a separate system(some implicit feed-forward algorithm) that causes certaincaching and recovery behaviour. In fact, even if one canshow that an animal has explicit access to the information,this need not imply MTT. There is a double dissociation here. One can know what happened where and when without being able to remember the event (e.g. yourbirth) and, conversely, one can travel back in time without access to accurate when and where information. I (TS) can vividly re-experience meeting a fascinating character once in the Philippines (or was that Indonesia?) sometime in the early nineties (or was that the late eighties?). Jays might know what food is hidden where and whether it is still good to eat, without knowing how or why they know it. If we take Option 2, then, there should be no implication(although it remains a possibility) that jays recollect the past episode and reason about the present state from there. Perhaps what–where–when or ‘www-memory’ would be a more neutral, descriptive term than episodic memory.
Although it is interesting that jays can encode, store and use such information, we suspect that many people are fascinated by this line of work largely because it raises the possibility of human-like episodic recall (Tulving’s later definition) in animals. According to Option 1, however, this might be impossible to show, in which case the human uniqueness claim would merely be a presumption. We argue, instead, that animals could provide evidence for MTT (even if we can not establish ‘the feel’, i.e. autonoetic consciousness). We might call this Option 1b .We take as aworking hypothesis that MTT is uniquely human because, although there could be evidence for MTT in animals, there has as yet been none forthcoming. So what could such evidence look like?
Avenues towards evidence for MTT in animals (if theyhad it)
Let us first see whether we can agree what phenomenonwe are in fact talking about. Although there are clearly problems studying MTT in others, we can observe it inourselves. Please recall the last conference presentationyou gave. Where and when was it? What were you talking about? Who was in the audience? Can you picture yourself back there? Now let’s travel forward to your next scheduled talk or lecture. Where and when will it be? What are you going to talk about? Who will be the audience? Can you picture yourself there? Now don’t get carried away…
Language and other ways of declaring MTT
Although the phenomenological experience of MTT isprivate and cannot be directly assessed (even in fellow humans), we do not, of course, always keep our travels asecret.We can declare it – that is what makes it part of the explicit system. Using language, we take mutual tripsdown memory lane and communicate our plans andvisions. Some certainly argue that language is the keydifference between humans and animals, and that episodicmemory can only emerge in an organism that possesses language. However, there have been efforts to teachlanguage to animals (chimpanzees, gorillas, orangutans, parrots, dolphins, seals), so evidence for MTT could comefrom there. But could MTT not also be declared without language? It is sometimes easier to express mental travels non-linguistically,for example, through pantomime, where onere-enacts events. Humans also practice their behavioural responses to foreseen, but not yet experienced,events. We can disinhibit the motor execution of mental simulations and overtly rehearse for the upcoming performance,or re-enact a fabulous goal-scoring move fromlast Sunday.
If animals engage in MTT, they might also express it to themselves or to others in such ways. Cleverexperimentation may be required to establish such evidence.Some ingenious research has already made progressat creating means for animals to declare their minds:showing that monkeys can report whether or not they remember. Non-verbal paradigms could be explored with children and compared with verbal tasks to validatethe measure. In summary, then, we argue that the claim that MTT is uniquely human is based on the current lack ofevidence from animals, not on the a priori impossibility of obtaining evidence. Memory for whatever, wherever and wheneverLet us briefly go back to that last talk you gave. What was the audience’s reaction? You might have some snapshots of what the scene looked like. But this imagery does not tend to unfold in an orderly fashion from start to finish. Instead,the chronology of our past ‘depends on a process of active, repeated reconstruction’.
Although we can describe (or re-enact) some aspects of the events, we fallshort on others (what did you wear that day…?). Extensive research on episodic retrieval clearly shows that we do not simply press the rewind button and replay all the actions and perceptions from beginning to end. We actively reconstructpast events from the gist or from visual snapshots,but draw heavily on our general semantic knowledge. Accuracy is not imperative. We are vulnerable to incorporating more recently acquired information and tend toreconstruct in ways that help justify our current attitudes. Our retrieval is often flawed, subject to interference,misinformation and updating. We argue that MTT isgenerative. That is, given a limited number of objects and actions, we can assemble and reassemble these elements into virtually unlimited constellations. MTT reflects ourability to imagine whatever, wherever and whenever, rather than our ability to faithfully record what, where and when information. Can you not go back to your lastlecture and insert the idea that your Mum was sitting inthe front row?
Why would humans, or any animal for that matter, haveevolved a system that is at times quite unreliable? Theanswer is that there is no selective advantage to reconstructingthe past per se, unless it matters for the presentor future. We propose that episodic reconstruction is justan adaptive design feature of the future planning system. Surprisingly, episodic future thinking has attracted farless research attention than episodic memory, but it has recently become in vogue. We know that patients like K.C. or D.B., who have lost their ability to mentally travel into the past, are equally impaired in their travels into thefuture. When you mentally travel forward to your upcoming talk, you can generate a variety of scenarios. Some of the elements are extrapolations from past events (is it going to be in the same lecture theatre?) and others draw on more general semantic knowledge. You can playthrough different possibilities just as you can play through‘what if ’ versions of past events (like inserting Mum inyour last lecture). Anticipating future scenarios clearlyhas selective advantages because you can prepare now (e.g. practice) for anticipated contingencies. One would think that such flexible foresight would become quiteobvious if animals had it (Box 2). So far there is little evidence for this for reviews).
Box 2. Limited foresightedness
One proposal, the Bischof-Ko¨ hler hypothesis, suggests that animals’ forethought is restricted because, unlike humans, they cannotanticipate future drives. Roberts recounts a revealing observation by D’Amato, which can be used to illustrate this hypothesis. His cebus monkeys were fed biscuits once a day andthe monkeys would hungrily eat to satiation. Then they indulged inan apparently irrational behaviour: they would throw the remaining food out of the cage, only to find themselves hungry again somehours later. Why not guard the food to satisfy future hunger? If one isnot hungry and can’t imagine being hungry again, then biscuits’utilitymay lie in their quality as projectiles. There is no point in acting now to secure a future need one can not conceive of. One reason forsuch a limit could be that travelling forward to the experiences and needs of a future selfmayrequire much the same processes as theory of mind. As Hazlitt noted 200 years ago:
‘The imagination by means of which alone I can anticipatefuture objects, or be interested in them, must carry me out ofmyself into the feelings of others by one and the same processby which I am thrown forward as it were into my future being,and interested in it.’
Neither future ‘self’ nor present (or future) ‘other’ can be experienceddirectly and are accessible only through an act of imagination(or mental simulation). A recent meta-analysis of theory ofmindresearch has shown that young children have as much difficultyimagining another’s false belief as they have recalling their own, suggesting a developmental synchrony between simulating one’sown past and another’s present. This has been supported by otherstudies reporting associations between theory of mind and behavioursthat appear to rely on MTT. So perhaps specific individual anticipations in animals are limited for the same reasons thattheir theory of mind seems limited. Humans clearly act towardssatisfaction of a multitude of future needs (from shopping for theweekend’s dinner to saving for retirement).
Prerequisite capacities for MTT
MTT draws on more general capacities. Open-endedgenerativity is achieved by use of recursive rules thatallow us to combine and recombine a finite set of elements. But we do not just jumble elements together. We can represent our representations as representations of the past, distinguish them from random hallucinations or dreams, and attempt to reconstruct the event flow as faithfully as possible. We can evaluate different scenarios as more- or less-likely future events. We think about thought. We might never know whether animals have flashbacks of previous perceptions and actions, like the snapshots we can conjure up spontaneously, but that is notthe question here. Could they think about these snapshots (if they had them) as past events and actively reconstruct,and reassemble them into future anticipations? Recursion in general, and meta-representation in particular, manifest in a host of other human capabilities (e.g. language,music, mathematics, theory of mind). Suddendorf and Corballis argued that MTT is uniquely humanlargely on the basis that these prerequisite abilities havenot yet been established in animals.
One also needs a sense of time if planning is to be effective, and a sense of self to project and relate to the present. Of the many species tested, only the great apes seem capable of recognizing themselves in mirrors, which is a controversial non-verbal test of self . Otherwise there is little to suggest animals have a self concept that they could project along a time dimension. Animals are by no means insensitive to matters temporal, however. Various timing abilities have been demonstrated,including tracking time of day, interval timing and recall oftemporal sequences. Yet Roberts’ review concluded that the capacity of animals to act on temporal information is very limited. What competence there is, he suggested, can be attributed to basic mechanisms, such as associations with states of the circadian cycle or other relatively primitive timing mechanisms, rather than a concept of time.
Conclusion, with a view to the future
The current evidence suggests that scrub jays have‘www-memory’ – they can encode, store and use informationabout what they cached where and when. This neednot imply that they travel mentally back to the originalcaching event or forward to the recovery. The hypothesis that MTT is uniquely human seems only worth upholding if it can – potentially – be refuted (hence Option 1b above). We identified avenues through which a case could be made.Given at least some indication of competence at prerequisites,our closest relatives, the great apes, might be themost likely candidates.
But it is certainly worth finding out more about scrub jays’ competence, and about their limits. For example, Clayton and colleagues recently provided some evidence for flexible memory us. However, can jays use their skill outside of the domain of caching and recovery? Can they, for instance, learn different decay functions (e.g. use differently coloured soil with preservatives or mould to change decay time) and use them as predictors of significant non-food related events? One could present jays with choice paradigms contrasting one caching tray now versus two trays later, with differentially perishable foods toinvestigate delay of gratification and contrast present with future needs. The paradigm could also be used to investigate MTT prerequisites. Emery and Clayton found that jays that had pilfered others’ caches would re-cache food in new sites if their own original caching had been observedby another jay. One could investigate jays’: (i) understandingof seeing (do they cache differentially in anarea not visible to a potential pilferer); (ii) self-recognition(does their own mirror image result in similar re-caching behaviour – if not, does a mirror image of another observerdo so?); and (iii) theory of mind (do they reduce re-cachingwhen the observer holds a false-belief, for example whenthe observer is not privy to a change in the hiding constellation)?
When evaluating new evidence for MTT, however, were mind scholars of Tinbergen’s four levels of explanation: function, causation, development and evolution. In terms of function, for example, MTT is clearly not unique in providing means for acting to enhance future survivaland reproduction. All animals that have memory and a capacity to learn possess a mechanism that is future oriented. Even the function implied by the Bischof-Ko¨hler hypothesis – to act before the adaptive problem or need is encountered – is clearly not unique to humans. Building a nest, preparing for hibernation and food caching make sense only in the light of the future needs. If we consider causation, however, we need to be alert that the same functional ends could be produced by different means. Hibernation, for example, is a species-specific universal that is displayed even if the animal has not yet experienced winter. The developmentally uniform emergence of this behavioural pattern suggests an innate, instinctual mechanism; a causal process that does not display the individual generativity evident in MTT. Scrubjays would benefit from having very accurate records of their caches, an objective that would perhaps not be served best by a generative reconstructive mechanism. Althoughfood storing evolved independently in several species, mechanisms homologous to those involved in human MTTare more likely to be found in our primate relatives. However, humans might have evolved MTT quite recently.The earliest potential evidence could be bifacial hand axes some 1.6 million years ago, which appear to have been made and kept for repeated future use. The only sure evidence of MTT, however, comes from writing.
In the earliest writings, the ancient Greeks describe the myth of Prometheus. Prometheus created humanity. He stole fire from heaven to give humans powers of thegods that distinguished them from other animals. He brought culture and technology. Prometheus literally means foresight.
[Source/credit: Mental Time Travel in Animals(download pdf)]