2.1. Toward a definition of mental time travel

Episodic memory, in contrast to semantic memory, provides access to the personally experienced event, rather than just the knowledge extracted from the event. The phenomenological experience of remembering, or what Tulving (Reference Tulving1985) calls autonoetic or self-knowing consciousness, such as recollecting where and when one learned that Wellington is the capital of New Zealand, is different from merely knowing that fact. This distinction is supported by experimental manipulations and imaging studies, and by dissociations in impairments following brain injury (Gardiner et al. Reference Gardiner, Ramponi and Richardson-Klavehn2002; Henson et al. Reference Henson, Rugg, Shallice, Josephs and Dolan1999; Klein et al. Reference Klein, Loftus and Kihlstrom2002b; Tulving Reference Tulving, Terrace and Metcalfe2005). Thus, amnesic patients, such as K.C., may know facts (e.g., the difference between stalagmites and stalactites) and procedures (e.g., how to play chess), without being able to recall a single personally experienced event leading to knowing them (Tulving Reference Tulving, Terrace and Metcalfe2005). Episodic memory is not about regularities, but about reconstructing particularities of specific events that have happened to the individual.

In effect, then, episodic memory implies a mental reconstruction of some earlier event, including at least some of the particularities of that event, such as the principal characters involved, the actions that took place, the setting, and the emotional reactions. Metaphorically speaking, it might be regarded as the result of a mental journey into the past. This idea is readily extended to the future. Based on previous experiences, we can imagine specific events in the future, including the sorts of particularities that have characterized events in the past. Mental time travel into the future might include the planning of some specific event, such as a dinner party, or it might involve the mental anticipation of some event that we know to be scheduled for some future date, such as a job interview. Again, though, there is a distinction between merely knowing that some event will occur, such as that the sun will set, and mentally creating an event, such as a sunset actually experienced, with gradual fading of light, and the blue flash on the horizon as the last image of the sun disappears.

The mental reconstruction of past events and construction of future ones may have been responsible for the concept of time itself, and the understanding of a continuity between past and future. Having a concept of time allows us to understand that past and future are on the same dimension, and what was the future eventually becomes the past (that is, unless the universe comes to an end). Mental time travel allows us to imagine events at different points along this continuum, even at points prior to birth or after death. This means that mental time travel is a generative process, incorporating known elements but arranged in particular ways to create the experience of events that are actually occurring. Even episodic memory may not be a faithful recreation of a past event. False memories have been widely documented (e.g., Loftus & Ketcham Reference Loftus and Ketcham1994) and are readily created in the laboratory (e.g., Roediger & McDermott Reference Roediger and McDermott1995). This means that mental time travel cannot be defined in terms of the veracity of the content. We know what mental time travel is because we can introspectively observe ourselves doing it and because people spend so much time talking about their recollections and anticipations.

A major challenge, though, is to establish a definition, or set of criteria, that might identify mental time travel in nonhuman animals that cannot express their experiences in words. This problem is akin to the decades-long search for behavioral criteria for theory of mind (see sect. 4.3) in nonhuman animals (Heyes Reference Heyes1998; Povinelli & Vonk Reference Povinelli and Vonk2003; Premack & Woodruff Reference Premack and Woodruff1978; Suddendorf & Whiten Reference Suddendorf, Whiten, Sterelny and Fitness2003; Tomasello et al. Reference Tomasello, Carpenter, Call, Behne and Moll2005; Whiten & Byrne Reference Whiten and Byrne1988). Tulving (Reference Tulving, Tulving and Donaldson1972) originally defined episodic memory in terms of the kind of information it appears to store: namely, what happened where and when. This is the so-called www criterion; that is, demonstrated understanding of what, where, and when might be considered sufficient evidence that an imagined event is an instance of mental time travel. For example, if an animal that caches food can be considered to demonstrate understanding of what was cached, where it was cached, and when it was cached, then it has episodic memory for the caching event itself. Similarly, if its behavior indicates specific knowledge of a future event, such as what food it will retrieve, where it will find it, and when it will retrieve it, then it may be said to have travelled mentally into the future to a retrieval event itself. However, although useful, the www criterion is neither necessary nor sufficient for mental time travel. Personal experience, as well as research on false memories, show that one can mentally revisit an event without accurate information as to what, where, and when, and conversely one can know what, where, and when something happened (e.g., one's birth) without remembering the event itself (Suddendorf & Busby Reference Suddendorf and Busby2003b). Thus, Tulving changed his definition to emphasize the phenomenological experience (autonoetic consciousness) of episodic retrieval (Tulving Reference Tulving1985; Reference Tulving, Terrace and Metcalfe2005), and www memory in animals has been more cautiously termed “episodic-like memory” (Clayton & Dickinson Reference Clayton and Dickinson1998).

Phenomenological experience, though, is a private mental capacity, and one may rightfully wonder if one could ever know of another organism's mental world. It is conceivable that neurophysiological criteria may eventually suffice, but there is as yet no agreement as to the conditions defining phenomenological awareness, even in humans. Nevertheless, for mental time travel to have evolved, there must have been something natural selection could work on, that is, some effect on survival or reproduction (unless it was merely an accidental side effect of some other adaptation). The answer we propose is that mental time travel provides increased behavioral flexibility to act in the present to increase future survival chances (Suddendorf & Busby Reference Suddendorf and Busby2003b; Reference Suddendorf and Busby2005). For example, one might prepare for a forthcoming job interview based on past experiences of interviews, and on imagining the questions one may be asked and on one's possible replies to them.

If this argument is correct, then the crux of mental time travel lies in its role in enhancing biological fitness in the future, so that mental time travel into the past is subsidiary to our ability to imagine future scenarios. Episodic memory may also inform semantic prospection by, for example, providing boundary conditions for the scope of generalizations. Indeed, memories for particular contradicting episodes are primed when people retrieve semantic generalizations (Klein et al. Reference Klein, Cosmides, Tooby and Chance2002a). But we argue that the primary role of mental time travel into the past is to provide raw material from which to construct and imagine possible futures. In establishing whether a given behavior is indicative of mental time travel, then, we should consider application of that behavior to the future rather than to the past (Suddendorf & Busby Reference Suddendorf and Busby2005). We therefore advance the following description of conditions that, if met, may entitle us to invoke the construct of mental time travel to explain it.

Mental time travel is evident in voluntary behavior that solves a problem that the organism will encounter at a future point in time, where “future” entails that the problem is not already manifest (as when acting to satisfy a current hunger, for example). To ascertain that a given behavior was driven by mental time travel, it is necessary to rule out chance, innate predispositions, procedural and semantic prospection, or any combination of these. We postulate that the crucial selective advantage mental time travel provides is flexibility in novel situations and the versatility to develop and adopt strategic long-term plans to suit individual selected goals. Thus, paradigms that use transfer tests and cross different domains would be strongest in making cases for mental time travel. Such tests provide evidence for mental time travel into the past only by virtue of the evidence linking past and future mental time travel, to which we turn next.

2.2. Evidence for continuity of past and future mental time travel

Since past is fact and future is fiction, common sense might suggest that different cognitive mechanisms underlie recollection of past events and construction of future ones. There is a fundamental causal asymmetry, and one simply cannot know the future as one knows the past. However, various lines of evidence suggest that mental time travel into the past shares cognitive resources with mental construction of potential future episodes (Suddendorf & Corballis Reference Suddendorf and Corballis1997). Normal adults report a decrease in phenomenological richness of both past and future episodes with increased distance from the present (D'Argembeau & Van der Linden Reference D'Argembeau and Van der Linden2004). The temporal distribution of past events people envisage follows the same power function as the temporal distribution of anticipated future events (Spreng & Levine Reference Spreng and Levine2006). Amnesic patients who are unable to answer simple questions about yesterday's events have been found to be equally unable to say what might happen tomorrow (Klein et al. Reference Klein, Loftus and Kihlstrom2002b; Tulving Reference Tulving1985), and it is not until around age 4 that children are able to accurately answer both such questions (Busby & Suddendorf Reference Busby and Suddendorf2005). Patients with depression who have trouble retrieving specific memories from their past also have trouble imagining specific future episodes (Williams et al. Reference Williams, Ellis, Tyers, Healy, Rose and MacLeod1996). Finally, brain imaging has shown that both remembering the past and imagining the future are associated with frontal and temporal lobe activity, although there are specific areas in the frontal pole and medial temporal lobes that are more involved with the future than with the past (Okuda et al. Reference Okuda, Fujii, Ohtake, Tsukiura, Tanji, Suzuki, Kawashima, Fukuda, Itoh and Yamadori2003).

The constructions of future and past episodes both depend in part on semantic memory, since one must construct events that are consistent with one's general knowledge of the world. Nevertheless, it is the episodic component that provides the particularities that can fit one's future plans precisely to the occasion – the seating arrangement, the clothes one might wear, the menu, the topics of conversation. In allowing us to foresee unique events, mental time travel offers the ultimate step in adaptation to the future. Like human language, it is open-ended and generative, so there is no end to the number of potential future scenarios one might envisage.

Although episodic memory preserves something of the particularities of individual events, it is often unreliable and subject to distortion, as we have seen. In cases of amnesia, moreover, it is episodic memories that are the most vulnerable (Wheeler et al. Reference Wheeler, Stuss and Tulving1997). The fact that episodic memory is fragmentary and fragile suggests that its adaptiveness may derive less from its role as an accurate record of personal history than from providing a “vocabulary” from which to construct planned future events (and perhaps to embellish events of the past). It may be part of a more general toolbox that allowed us to escape from the present and develop foresight (Suddendorf & Corballis Reference Suddendorf and Corballis1997), and perhaps create a sense of personal identity (Schacter Reference Schacter1996). Indeed, our ability to revisit the past may be only a design feature of our ability to conceive of the future (Suddendorf & Busby Reference Suddendorf and Busby2003b).

2.3. Neurophysiological evidence

We have suggested that neurophysiological evidence may eventually serve as a proxy for the phenomenology of mental time travel. There is increasing evidence concerning the neurophysiology of mental time travel in humans, but little from nonhuman animals. Some of this evidence suggests, in fact, that mental time travel may be uniquely human, since the brain areas involved have apparently undergone changes that are not evident in other primates. For example, the key to mental time travel may lie in part in the expansion of the brain, and the prefrontal cortex in particular, in human evolution. Brain size varies between species partly as a function of body size, and an appropriate comparative measure is the encephalization quotient (EQ), devised by Jerison (Reference Jerison1973), which is based on the regression of brain weight on body weight. The human EQ is about three times that of the chimpanzee. Further, it has been claimed that the increase is larger in the prefrontal cortex, known to be critically involved in episodic memory (see review by Wheeler et al. Reference Wheeler, Stuss and Tulving1997) than in other brain areas (Deacon Reference Deacon1997). Although this disproportionate increase has been disputed (e.g., Semendeferi et al. Reference Semendeferi, Damasio and Frank1997; Uylings Reference Uylings1990), Deacon (Reference Deacon1997) argues that other studies have failed to measure prefrontal cortex independently of motor and premotor areas. A more recent study confirms the disproportionate enlargement of the prefrontal cortex in humans, but indicates that it is restricted to white matter, and does not apply to grey matter (Schoenemann et al. Reference Schoenemann, Sheenan and Glotzer2005).

Regardless of the question of size, there is also evidence that the prefrontal cortex has been substantially reorganized in the course of human evolution. For example, Area 13, which appears to be a subdivision of Area 11, is only about half that expected on the basis of brain size, and Semendeferi et al. (Reference Semendeferi, Armstrong, Schleicher, Zilles and van Hoesen1998) suggest that this diminution occurs because other regions of Area 11, along with regions of Area 47, have been enlarged to accommodate a large number of specialized subdivisions. The frontal pole is also considerably enlarged relative to that in apes, especially in the right hemisphere (Semendeferi et al. Reference Semendeferi, Armstrong, Schleicher, Zilles and van Hoesen2001). Summarizing this evidence, Flinn et al. (Reference Flinn, Geary and Ward2005) suggest that these changes to Area 11 and the right prefrontal cortex appear to be involved in “self awareness, social problem solving, the ability to recall personal experiences, and the ability to project oneself into the future” (pp. 30–31).

The involvement of the prefrontal cortex may depend on neural loops that also encompass the basal ganglia and perhaps other regions as well. In one functional magnetic resonance imaging (fMRI) study, for example, brain activity was recorded while human subjects learned actions in a Markov decision task that would bring either immediate or future rewards (Tanaka et al. Reference Tanaka, Doya, Okada, Ueda, Okamoto and Yamawaki2004). In both cases significant activity was seen in the lateral orbitofrontal cortex and striatum, but for actions concerning future rewards, there was additional activation in the dorsolateral prefrontal cortex, inferior parietal cortex, dorsal raphe nucleus, and cerebellum. Regression analysis also suggested graded maps of time scale within the insula and striatum, with ventroanterior regions involved in predicting immediate rewards and dorsoposterior regions in predicting future rewards. In another study involving appetitive conditioning in humans, fMRI recordings showed activity in the orbitofrontal cortex and striatum to be dependent on the time between a present conditioned stimulus and an anticipated reward (O'Doherty et al. Reference O'Doherty, Deichmann, Critchley and Dolan2002; see also O'Doherty Reference O'Doherty2004).

3.1. Mental time travel into the past

Despite our earlier reservations, the www criterion for episodic memory has featured prominently in recent comparative research. Though one might expect to find precursors of mental time travel in our closest living relatives, the most vigorously argued case has come, not from apes, but from birds that cache food in various locations and later retrieve it. Scrub jays can select food locations not only according to the type of food, but also according to how long it has been stored. For example, they will recover recently cached worms in preference to nuts, since fresh worms are more palatable; but if the worms have been cached for too long, they will retrieve nuts, because the worms will have decayed and become unpalatable (Clayton et al. Reference Clayton, Bussey and Dickinson2003). Scrub jays seem to know what has been cached, where it was cached, and when it was cached, thus apparently meeting the www criterion. As noted earlier, however, investigators have used the term “episodic-like memory” in acknowledgment that the www criterion need not imply true episodic memory, in part because there need not be any implications about autonoetic consciousness (Clayton et al. Reference Clayton, Bussey and Dickinson2003; Emery & Clayton Reference Emery and Clayton2004).

When humans talk about past events, they typically exchange information about even more “w”s, such as “who did what to whom, and when, and where, and why, and what happened next” (Suddendorf & Corballis Reference Suddendorf and Corballis1997, p. 159). Recent evidence indicates that jays may indeed also store information about who observes them cache (Dally et al. Reference Dally, Emery and Clayton2006b). The birds were more likely to move the food to new locations if a more dominant bird observed the caching than if a less dominant bird did so. The outcome of this re-caching behavior was presumably a reduction in the likelihood of the food being pilfered in future.

This innovative research program on scrub jays has inspired similar work on other species, with varying success (see Table 1 for a summary). Rhesus monkeys, for example, seemed unable to learn that a preferred food was available only after a short interval, but not after a long one (Hampton et al. Reference Hampton, Hampstead and Murray2005), whereas rats were able to learn that a particular arm of a maze contained chocolate pellets after a long delay but not after a short delay (Babb & Crystal Reference Babb and Crystal2005).

Table 1. Recent studies suggesting knowledge of different “w” information in various species that form part of “episodic-like” memory

The cases listed in Table 1 explicitly link their results to one of the “w”s of episodic-like memory. Numerous earlier studies might be included when re-categorized in line with this interpretation (see, e.g., Schwartz & Evans Reference Schwartz and Evans2001). No doubt this table will be expanded and altered further as more comparative research is conducted with the explicit aim of uncovering evidence for episodic-like memory. We believe that this approach may provide one of the first areas in which research systematically maps the types of mental content in different species. Although we look forward to a more comprehensive table, we remain unconvinced that any of these cases shows true mental time travel (Hampton & Schwartz Reference Hampton and Schwartz2004; Roberts Reference Roberts2002; Suddendorf & Busby Reference Suddendorf and Busby2003b).

Suddendorf and Busby (Reference Suddendorf and Busby2003b) put forward various reasons why the data from jays and other species are not yet conclusive. In the present article, we have noted earlier the distinction between the type of information stored and the experience of mentally revisiting an episode that led Tulving to change his definition of episodic memory. The information listed in Table 1 may be known rather than remembered. There is a difference between remembering having put something somewhere, some time ago, and simply knowing right now what is where, and how fresh it is. In fact, episodic-like memory may result from feed-forward mechanisms that need not be about the past at all. Episode A may cause cognitive change B, which may in turn affect later behavior C, without B carrying any information about A itself (Dretske Reference Dretske1982). For example, performance that depends on time since an event may rely on the strengths of memory traces, which fade in time, and may therefore provide a direct cue without implying mental travel through time. Trace strength or some other time-dependent process may act as a clock to determine a “use-by” date for consumption.

One argument against this specific idea that www memory may depend on trace strength is that rats, at least, still show the effects of trace strength in the accuracy of recognition after hippocampal lesions, yet lose the ability to discriminate temporal order (Eichenbaum et al. Reference Eichenbaum, Fortin, Ergorul, Wright and Agster2005), a finding that is also consistent with the hippocampus being a primary site for episodic memory in humans. Even so, the hippocampus may be involved in access to trace strength as a discriminative cue, so failure to make temporal discriminations after hippocampal damage need not rule out the trace-strength hypothesis. Whether or not this specific hypothesis is correct, it should be clear that there are viable alternatives to the notion that evidence for www, or even wwww, memories is evidence for mental time travel. Indeed, even the original proponents of this approach concede that these memories may exist without the jays mentally reconstructing the past (Dally et al. Reference Dally, Emery and Clayton2006b).

A few researchers have tried to tackle the question with different sets of innovative methodologies (Zentall Reference Zentall2006). For example Zentall et al. (Reference Zentall, Clement, Bhatt and Allen2001) taught pigeons to press one or the other of two new keys depending on whether they had previously responded or not responded on another earlier key – which suggests an ability to refer to a previous behavioral episode. Similarly, dolphins appear to remember their own previous behavior, as shown by an ability to respond to commands equivalent to sentences like “do something not recently done,” or “repeat the most recent response” (Mercado et al. Reference Mercado, Murray, Uyeyama, Pack and Herman1998). Rats show evidence for temporal order. When exposed to sequences of four different odors in distinct locations, rats later showed above-chance performance on tests in which they were to choose which of two odors or locations had occurred earlier in the sequence (Eichenbaum et al. Reference Eichenbaum, Fortin, Ergorul, Wright and Agster2005). However, all of these approaches run into problems of interpretation similar to those that afflict the “www” approach. As we suggested earlier here and elsewhere (Suddendorf & Busby Reference Suddendorf and Busby2005), the solution may be to focus on future-directed behavior instead of on episodic memory. We argue that it is action with the future in mind that provides increased selective advantage and is hence visible to evolution. If we are right, it might also be visible to clever researchers (if nonhuman animals have it).

3.2. Mental time travel into the future

Animals meet significant recurrent problems such as seasonal food shortages in different ways (e.g., through migration, torpor, increased fat storage). Caching food is a solution that has evolved among some mammals and birds that provide for their young (Smith & Reichman Reference Smith and Reichman1984). This strategy rests on some memory capacities and suggests orientation to future consumption, but it need not imply that the animal actually envisages that future, or even explicitly plans for it. For example, one study showed that rats continued to cache food in locations in a maze where food was later repeatedly degraded or pilfered, even though they avoided those locations in retrieving the food (McKenzie et al. Reference McKenzie, Bird and Roberts2005). This suggests that they had little sense of what would happen in the future. Similarly, young scrub jays vigorously cache all kinds of objects well before they develop the behavior of retrieving them. When caching and retrieval develop in such predictable fashion, it appears to be based on instinctive or implicitly learned predispositions that may be modulated by semantic memory but need not involve mental time travel (Suddendorf & Busby Reference Suddendorf and Busby2003b).

A possible test of “episodic-like” prospection might run as follows. The birds would first be taught that if they select a cache location under one condition (say, a green light) they will be allowed to retrieve from that location at any time, but if they select under another condition (say, a red light) they must wait at least a day before being allowed to retrieve. Assuming that they could learn this, they are then given the choice of worms or nuts, and the question is whether they would select fresh worms when allowed immediate access, but would anticipate decayed worms when access is delayed, and so choose nuts. If such anticipation were possible, it might indicate an understanding that how long food has been cached projects into the future as well as into the past. However, the same difficulties as with “episodic-like” memory apply. The birds might seem to understand the state of the world tomorrow (rotten worms) only through a combination of predispositions and specific learning algorithms that have evolved in the caching context (Dally et al. Reference Dally, Emery and Clayton2006b; Suddendorf & Busby Reference Suddendorf and Busby2003b).

Whatever the outcome of such an experiment, adult jays have been shown to modulate caching behavior in ways that seem to take the possibility of future pilfering into account. Earlier, we alluded to evidence that they re-cache food if observed by a more dominant bird (Dally et al. Reference Dally, Emery and Clayton2006b). Another study has shown that, when observed caching food, jays re-cache food only if they themselves had previously stolen the food of others. Birds with experience of theft seem to anticipate a future in which their own food might be stolen, and act accordingly (Emery & Clayton Reference Emery and Clayton2001). It takes a thief to know a thief.

Prevention of pilfering is a common adaptive problem for species that store food, and various strategies have evolved to deal with it (for a review, see Dally et al. Reference Dally, Clayton and Emery2006a). Both scatter-hoarders (e.g., ravens) and larder-hoarders (e.g., acorn woodpeckers) may carefully disguise their caches or actively defend their stores (e.g., Bugnyar & Heinrich Reference Bugnyar and Heinrich2006). Strategies of pilferage avoidance are increasingly being documented. For example, Eastern grey squirrels space caches further apart when conspecifics are present and preferentially dig oriented away from them (Leaver et al., in press). Theft is not the only future problem that hoarders may address. Gray squirrels bite out the seeds of white oak (but not red oak) acorns before storage, which prevents germination (Steele et al. Reference Steele, Turner, Smallwood, Wolff and Radillo2001). Treating food before storage (e.g., pine squirrels dry mushrooms in the sun) or keeping it alive but unable to escape (e.g. moles biting worms) are strategies that reduce spoilage and hence increase future survival chances (e.g., Smith & Reichman Reference Smith and Reichman1984).

Are these future-oriented behaviors fundamentally different from non-behavioral species-specific adaptations to the same problem, such as seasonally storing food as blubber or body fat? Perhaps not. Advantageous behavior can lead to evolutionary change (Bateson Reference Bateson2004) and can become instinctive as descendants of animals displaying the behavior face selection pressure for reliable, less resource-intensive shortcuts. New Caledonian crows, for example, manufacture and use tools (Hunt Reference Hunt1996). Although there is evidence for social transmission (Hunt & Gray Reference Hunt and Gray2003), the birds appear to have inherited a predisposition for tool use. They show such behavior without any experience or opportunity for observation (Kenward et al. Reference Kenward, Weir, Rutz and Kacelnik2005). Adaptations for certain types of memory and learning have also been reported in regent honeyeaters. In line with ecological factors (i.e., replenishment rates), these birds easily learn to avoid recently rewarding food locations and to return after long periods, but do poorly when the contingency is the other way around (Burke & Fulham Reference Burke and Fulham2003). Though this learning is appropriately future-oriented to the contingencies of its habitat, such constraints distinguish it from the flexible future-oriented behavior characteristic of human mental time travel.

3.3. The Bischof-Köhler hypothesis

We argued earlier that the selective advantage of mental time travel is the increased flexibility in acting in the present to secure future needs. First, therefore, one has to be capable of conceiving having different future needs, such as imagining being thirsty when currently quenched. One hypothesis, proposed by Bischof-Köhler (Reference Bischof-Köhler, Eckensberger and Lantermann1985) and Bischof (Reference Bischof1985), is that it is in fact an inability of nonhuman animals to flexibly entertain future need or drive states that limits their capacity for mental time travel. That is, nonhuman animals are unable to differentiate future states from present ones (Bischof-Köhler Reference Bischof-Köhler, Eckensberger and Lantermann1985; Bischof Reference Bischof1985; Suddendorf & Corballis Reference Suddendorf and Corballis1997). Superficially, at least, this hypothesis cannot be correct, since many species act to secure future needs, as illustrated by the caching of food. As noted, though, such behaviors may typically be largely instinctive, and experience may only allow changes within specific domain parameters. The Bischof-Köhler hypothesis may still apply to more individual, flexible situations involving non-instinctive behaviors. One study suggests that children anticipate future needs by around age 4 (Suddendorf & Busby Reference Suddendorf and Busby2005), and the challenge is now to document its development in more detail and to test nonhuman species.

Nonverbal tests of the Bischof-Köhler hypothesis have been proposed (Suddendorf Reference Suddendorf1994; Suddendorf & Busby Reference Suddendorf and Busby2005; Tulving Reference Tulving, Terrace and Metcalfe2005), but have yet to be carried out. They involve controlling a current need state, such as ensuring that the subject is no longer thirsty, and then providing an opportunity to secure a future need, such as obtaining a drink for a future situation that will induce thirst. There are various caveats to consider in order to rule out alternative explanations based on instinct or associative learning (Suddendorf & Busby Reference Suddendorf and Busby2005). For example, instinctual explanations can be ruled out if the experimental scenario does not involve species-typical behavior and if similar problems can be solved in a variety of domains. If the same animal were to pass variations of such tests across different needs (e.g., hunger, thirst, temperature) and contexts (locations, time-intervals), the Bischof-Köhler hypothesis would become increasingly untenable.

Mulcahy and Call (Reference Mulcahy and Call2006) recently came closest to implementing such a test. They trained bonobos and orangutans to obtain grapes from an apparatus using a tool. Access to the apparatus was then blocked and the animals were presented with a selection of two suitable and six unsuitable tools which they could take into a waiting room from where the apparatus was still visible. An hour later, they were allowed back into the testing room and given access to the apparatus. In 7 out of 16 trials, on average, the apes carried a suitable tool into the waiting room and returned with it to obtain grapes an hour later. There were strong individual differences in performance, with one orangutan achieving 15 out of 16 correct. This orangutan and the best performing bonobo were then tested again, but with an overnight delay between tool selection and return. They still returned with a suitable tool in more cases than expected by chance. A third experiment showed that the apes could pass the task even when they could not see the apparatus during tool selection. The final control study investigated whether the animals merely associated the tool with the reward. Subjects again received a grape reward if they returned with the right tool, but were not actually given an opportunity to use the tool. Performance in this condition was poorer, suggesting that they did plan ahead in the other studies (Mulcahy & Call Reference Mulcahy and Call2006).

Nevertheless, there are some concerns about this conclusion (Suddendorf Reference Suddendorf2006). The same tools were appropriate over trials, so apes could have just learned to always return with these same tools. This highlights the importance of the final control condition designed to rule out explanations based on associations. However, this control condition was not given to the successful animals of Experiments 1 to 3, but to a new group of four animals. Two of these never brought the tool back and hence could never have experienced the reward that may have facilitated performance in the previous studies. Thus, their data do not inform us about the power of the reward. The other two performed identically to two animals in Experiment 1. Thus, contrary to what the authors claimed, association cannot be ruled out (Suddendorf Reference Suddendorf2006).

But even if subsequent studies confirm that great apes can select and keep a different tool for a specific future use, it does not show anticipation of future needs as proposed by the Bischof-Köhler hypothesis. The “future need” potentially anticipated in these studies refers to the “need” for a tool to satisfy a current desire for the treat, not anticipation of a different internal drive or need state (e.g., such as a future desire that is different from present). The studies did not control or manipulate the drive or need state of the subjects, and it is not unreasonable to assume that they all had a desire for grape rewards throughout testing (Suddendorf Reference Suddendorf2006). Animals that are not capable of conceiving of future drive and need states would have little reason to concern themselves with a remote future, as all they would care about is satisfaction of current needs. More research is required to determine the extent of animal foresight, but at present the limit proposed by the Bischof-Köhler hypothesis has not been falsified.

We are not making these killjoy arguments because of any preconceived notion about the way the world should be. We would be delighted if it could be established that other species have mental time travel. It would be a serious challenge to many humans' anthropocentric worldview and would have profound moral implications (Suddendorf & Busby Reference Suddendorf and Busby2003a). Those species that can be shown to ponder what has happened, and to speculate about what might happen in the future, would require fundamentally different welfare considerations. For example, they would suffer not just from current pain, but also from revisiting past and anticipating future pain (Lea Reference Lea2001). At present, there is no need to abandon the null hypothesis, but neither are we in a position to firmly conclude that other animals do not possess this capacity. We admit that a substantial body of comparative data is required before it can be concluded that any trait is uniquely human (Hauser et al. Reference Hauser, Chomsky and Fitch2002).

Given the selective importance of future-oriented behavior, species are likely to have evolved a wealth of different mechanisms. The growing research interest in the question of animal mental time travel may help identify and describe these mechanisms. Our closest surviving relatives, the great apes, may be expected to have precursors of the human ability (e.g., foresight limited to serving current needs). Mulcahy et al. (Reference Mulcahy, Call and Dunbar2005), for example, have shown that gorillas and orangutans can select an appropriate tool in one location to solve a problem at another location that was not simultaneously visible. There is a range of evidence that great apes can imagine alternative states of the world (see sect. 4.1) and may hence offer a link between animal and human minds (for a review, see Suddendorf & Whiten Reference Suddendorf and Whiten2001). It is possible, then, that antecedents of mental time travel had evolved in the common great ape ancestor. It has been argued that living in fission-fusion societies, as modern chimpanzees do, may have selected for an increased capacity to monitor across time and space the comings and goings of other group members (Barrett et al. Reference Barrett, Henzi and Dunbar2003). Alternatively, ecological pressures on large arboreal apes, like modern orangutans, feeding on spatially and temporally variable food sources (e.g., fruit) might have led to selection for increased tracking of these patterns and planning of travel routes in our common ancestor. But it remains to be established what exactly the nature and extent of modern apes' foresight (and that of other animals) is. We maintain that the data so far continue to suggest that mental time travel is unique to humans. Next we make the case that future research should focus not only on purported instances of mental time travel, but also on subsidiary mechanisms that are typical of human mental time travel and that may be necessary to sustain it.

4.1. The stage

To entertain a future event one needs imagination – some kind of representational space in our mind for the imaginary performance. In cognitive psychology the concept of working memory is usually conceived of as such a space (or workbench) where information is temporally combined and manipulated (Baddeley Reference Baddeley1992). Human working memory appears to comprise, in addition to a central executive, two distinct subsystems: a phonological loop and a visuospatial sketchpad. Although non-lingual species may not have anything like the inner voice and inner ear (i.e., the phonological loop), they may very well have an image-based system (i.e., the visuospatial sketchpad).

To represent the future, the stage has to allow combination not only of recently presented material, but also of information stored in long-term memory. It must allow offline processing. Dreaming is an example of offline processing, and mammals appear to have dreams. Offline replay during sleep may be involved in consolidation of declarative memories in the hippocampus (Kali & Dayan Reference Kali and Dayan2004; Wilson & McNaughton Reference Wilson and McNaughton1994). In fact, dreaming or day-dreaming may be regarded as a form of “scenario building” that social mammals may have evolved to allow practice/play without suffering consequences (Alexander Reference Alexander, Mellars and Stringer1989).

To make such offline processing, or mental play, directly relevant to situations, however, one also has to be able to collate it with perceptions of the present real world. So-called secondary representations are thought to allow just that (Perner Reference Perner1991; Suddendorf Reference Suddendorf, Corballis and ea1999). In addition to a primary representation of current reality, a secondary representation of a goal constellation, for example, allows an individual to test potential moves by mental rather than physical trial and error. This capacity is also evident in pretend play (where one entertains an imaginary world while maintaining awareness of the real), in Piagetian invisible displacement tests (where one needs to reason about an invisible past trajectory of a target), and in various other basic cognitive tasks. During the second year of life, human children begin to display evidence for secondary representation in each of these contexts (Perner Reference Perner1991; Suddendorf & Whiten Reference Suddendorf and Whiten2001). A review of the literature found that only our closest relatives, the great apes, have so far provided some evidence for secondary representation in all of these areas (Suddendorf & Whiten Reference Suddendorf and Whiten2001; Reference Suddendorf, Whiten, Sterelny and Fitness2003; Whiten & Suddendorf, in press) and may therefore have the mental stage to imagine past or future events. Though convergent evolution may have produced similar skills in other species (e.g., corvids; see Emery & Clayton Reference Emery and Clayton2004), our own capacity for secondary representation appears to have evolved in the common great ape ancestor some 14 million years ago (Suddendorf & Whiten Reference Suddendorf and Whiten2001).

4.2. The playwright

To generate content, imaginary events need a script or narrative. This requires access to a declarative database – that is, stored information that is not stimulus bound, but can be made available top-down. As noted earlier, some nonhuman species appear to have declarative memory that could serve this function. To generate new content, however, one further needs to be able to combine and recombine existing elements. Like language, imaginary narratives may operate according to the principle of “discrete infinity” (Hauser et al. Reference Hauser, Chomsky and Fitch2002), involving the recursive application of rules to create an unlimited set of potential future scenarios (Suddendorf & Corballis Reference Suddendorf and Corballis1997).

Recursion is a computational procedure that calls itself, or calls an equivalent kind of procedure – as exemplified in sentences with embedded clauses, such as “The malt that the rat ate lay in the house that Jack built.” Recursion lies at the heart of grammar, and enables us to create, from a finite set of elements, a potential infinity of sentences that convey a potential infinity of meanings. Recursion may also be said to underlie other aspects of human thought, including perhaps music, manufacture, navigation, social relationships, and numbering systems (Corballis Reference Corballis1991; Hauser et al. Reference Hauser, Chomsky and Fitch2002). Theory of mind, which is the understanding that others have beliefs, desires, or intentions, for example, can involve several levels of recursion, as in such propositions as I think that you think I think you're stupid, which involves double recursion. Counting, as distinct from subitization or number estimation, involves principles that can be used recursively, enabling us to enumerate indefinitely. We humans, at least, use the concept of time in recursive ways, as in our understanding of such sentences like “Next Thursday he will have arrived in Mexico.” The future perfect is but one of some 30 different tenses in English, and reflects the close relation between language and mental time travel. Children begin to show a recursive, generative capacity in the early preschool years.

There is as yet little, if any, evidence for recursion in any nonhuman species (Corballis Reference Corballis, Sterelny and Fitness2003). Many birds sing songs of considerable complexity, and individual birds may have repertoires of many songs, involving different sequences of notes, but individual songs tend to be fixed and are repeated with remarkable consistency (e.g., Kroodsma & Momose Reference Kroodsma and Momose1991). There has been at least one claim that birds can generate songs using grammatical principles: Hailman and Ficken (Reference Hailman and Ficken1986) reported that the songs of chickadees follow rules governing the selection and sequencing of notes, but those rules appear to conform to a simple finite-state grammar rather than a context-free grammar like that underlying human language (Corballis Reference Corballis1991, p. 140). There is no evidence of recursion.

A recent claim that starlings can learn to recognize recursive patterns of sounds (Gentner et al. Reference Gentner, Fenn, Margoliash and Nusbaum2006) is, in our opinion, without foundation. The experimenters generated recursive sequences of up to eight sounds by embedding pairs of sounds, each consisting of a rattle and a warble, into another pair in recursive fashion. For example, if A _i denotes a rattle and B _i a warble, two levels of embedding would create the sequence A ₃(A ₂(A ₁B ₁) B ₂)B ₃. Exact sequences were seldom repeated, since the actual sounds were randomly selected from a population of eight rattles and eight warbles. The birds were able to discriminate sequences based on this recursive embedding from sequences not obeying the rule. This need not mean, though, that they understood recursion, since each sequence could be regarded as simply comprising a sequence of rattles followed by a sequence of warbles (e.g., A ₃A ₂A ₁ followed by B ₁B ₂B ₃). That is, the birds may simply have established the number of sequential rattles and the number of sequential warbles, and responded positively if the two were equal. Number perception for auditory sequences in birds may actually surpass that in humans (Thompson Reference Thompson1969), so a strategy of “estimate-and-match” may well have been within the capacity of the starlings. An earlier study showed that tamarins were unable to perform an analogous discrimination task (Fitch & Hauser Reference Fitch and Hauser2004), but even if they had been able to make the discrimination, it need not have implied that they recognized recursion itself (see Corballis [in press] for more detailed discussion).

4.3. The actors

Events typically involve characters, and mental time travel might well have evolved in order to predict (and aid or thwart) the behaviors of others (discussed further on in this section). To represent self and others realistically requires declarative knowledge of individuals and some folk psychology (i.e., theory of mind) to predict how they act (e.g., knowing that they typically act on the basis of their beliefs to fulfill their desires). With such knowledge, one can engage in more sophisticated social competition and cooperation, and one of the most effective ways to do so is to act like an actor: “to see ourselves as others see us, so as to cause them to see us as we would like them to rather than as they would like to” (Alexander Reference Alexander, Mellars and Stringer1989, p. 491). Self-awareness is difficult to define and measure. Alexander (Reference Alexander, Mellars and Stringer1989) argues that one important aspect of self-awareness springs directly from an ability to consider one's self in various alternative versions of future events. This is the notion of “free will,” implying that one can deliberately choose to pursue one of several entertained paths.

It is established that great apes, like 2-year-old children and unlike many other animals tested, can at least recognize themselves in mirrors (e.g., Bard et al. Reference Bard, Todd, Bernier, Love and Leavens2006; Gallup Reference Gallup1970; Patterson & Cohen Reference Patterson, Cohen, Parker, Mitchell and Boccia1994; Suarez & Gallup Reference Suarez and Gallup1981). They have an expectation of what they look like from the outside, and this expectation is rapidly updatable (Nielsen et al. Reference Nielsen, Suddendorf and Slaughter2006). Together with evidence for some understanding of what others can see (Call et al. Reference Call, Hare and Tomasello1998; Hare et al. Reference Hare, Call, Agnetta and Tomasello2000) or did see (Hare et al. Reference Hare, Call and Tomasello2001), this strongly suggests that they can “see themselves as others see them” at least in the literal sense. But it is not so clear whether they care about manipulating how others see them (even in this literal sense, only humans regularly adorn themselves with jewellery and clothes).

Nevertheless, there is a sizable body of evidence to suggest that primates do engage in various forms of tactical deception in social interactions (Whiten & Byrne Reference Whiten and Byrne1988). Primate social knowledge is impressive. Many primates recognize other individuals, as well as their social rank and relations (Seyfarth et al. Reference Seyfarth, Cheney and Bergman2005). Great apes and 2-year-old children also appear to have some limited folk psychology, but the precise nature of their capacities remains to be established (e.g., Suddendorf & Whiten Reference Suddendorf, Whiten, Sterelny and Fitness2003; Tomasello et al. Reference Tomasello, Call and Hare2003). Currently, the available data suggest that they do not have a representational theory of mind which includes the capacity to metarepresent that others may hold representations contradicting their own (Call Reference Call2001; Whiten & Suddendorf Reference Whiten and Suddendorf2001). It is only between ages 3 and 4 that humans pass tests (e.g., false-belief tasks) that demonstrate such understanding (e.g., Perner Reference Perner1991; Wellman et al. Reference Wellman, Cross and Watson2001). We have argued that it may require this level of folk psychology to be able to identify with one's future self, understand that this future self may have mental states that differ from one's current states, and care about them (Suddendorf & Corballis Reference Suddendorf and Corballis1997). As Hazlitt (Reference Hazlitt1805) observed: “The imagination … must carry me out of myself into the feelings of others by one and the same process by which I am thrown forward as it were into my future being, and interested in it” (p. 1).

4.4. The set

The mental play also needs a physical context that operates according to real-world principles. This requires some “folk physics.” Although one prominent researcher recently concluded that not even apes can mentally represent unobservable causal factors such as gravity and force (Povinelli Reference Povinelli2000), many other researchers are not convinced that their folk physics is quite so limited (Allen Reference Allen2002; Hauser Reference Hauser2001; Whiten Reference Whiten2001). Chimpanzees, for example, have passed double invisible displacement (object-permanence) tasks that require reasoning about the movement of objects that are no longer perceptible (Collier-Baker & Suddendorf Reference Collier-Baker and Suddendorf2006).

Especially important to mental time travel is some appreciation of the time dimension itself. Some fundamental timing processes (e.g., circadian rhythms) that link behavior to regular environmental changes are clearly widespread in the animal kingdom and are fairly well understood (e.g., Albrecht & Eichele Reference Albrecht and Eichele2003). Other, more cognitive processes are less well understood. There appear to be various different timing mechanisms in the human brain that track key information about temporal distance, order, and location within time patterns and humans appear to draw on different combinations of those (Friedman Reference Friedman2005).

Proposals for time measurement mechanisms range from discrete accumulating neural pacemakers (e.g., in scalar expectancy theory) to models involving continuous decay strength of memory traces. As noted earlier, fading of memory may itself provide temporal information that could account for the when component of www memory in scrub jays. In humans, cognitively controlled, rather than automatic, tracking of the passage of time typically activates the right hemispheric dorsolateral prefrontal cortex, an area also implicated in working memory (Lewis & Miall Reference Lewis and Miall2006). Lewis and Miall argue that both working memory and judging temporal distance may depend on the same dopamine-modulated neuronal system. Children begin to accurately discriminate distances of past events from around age 4 onwards (Friedman & Kemp Reference Friedman and Kemp1998).

Mental time travel may itself make such timing based on memory-decay functions less reliable. Since mentally constructing and reconstructing episodes reactivates memory traces, it may strengthen these memory traces and make them appear to be “fresher” in memory than warranted by age alone. Other timing mechanisms may therefore have evolved for the construction of a more accurate chronology. Indeed, chronology appears not to be a basic property of human memory, but rather, depends on active repeated construction (Friedman Reference Friedman1993). Some information about the order in which events occurred (e.g., before and after) may, however, be stored in memory. As we saw earlier, even rats show evidence of tracking temporal order (Eichenbaum et al. Reference Eichenbaum, Fortin, Ergorul, Wright and Agster2005).

How internal clocks, order codes, or other processes give rise to adult human concepts of the time dimension remains unclear. Human cultures have developed different semantic representations of time patterns (e.g., hours, days, weeks, months, etc.) that enable people to locate the time of events. Actively reconstructing locations of events in such explicit time patterns is probably the most common human approach to timing (Friedman Reference Friedman1993). Location-based processes are evident in scale effects where judgment of time of day of a past event may be more accurate than judgment of the month in which the event occurred – distance-based mechanisms cannot explain such phenomena. Friedman (Reference Friedman2005) recently reviewed the evidence and concluded that humans use both verbal and image-based processes to reason about the temporal location of events and that children use image codes to differentiate daily events by around age 5, and verbal lists, such as days of the week, by age 7. Children's relatively late acquisition of cultural time concepts suggests that they are not easy to acquire.

Once established, such time patterns offer the opportunity to place future events at particular points in time. Of course, neither order codes nor distance information in memory can explain the development of a temporally differentiated sense of the future. Children begin to show a rough temporal differentiation of future events by around age 5 (for a recent review, see Friedman Reference Friedman2005). Nonhuman animals seem severely limited in learning temporal information, and performance appears to be based on non-declarative mechanisms, rather than any such explicit concepts of time (Roberts Reference Roberts2002). Nonetheless, the work on scrub jays reviewed earlier suggests temporal-tracking mechanisms more sophisticated than previously thought to exist in any nonhuman species.

4.5. The director

Mental time travel is of course not to be mistaken for clairvoyance. The future contains many possibilities, and hence mental time travel involves entertaining different versions of scenarios and evaluating their likelihood and desirability. Like a director trying out alternative ways of presenting a scene, effective mental time travel requires rehearsals and evaluations. This entails some level of dissociation and metacognition (i.e., thinking about thinking), which first emerges in children around age 4 (Perner Reference Perner1991; Suddendorf Reference Suddendorf, Corballis and ea1999; Suddendorf & Corballis Reference Suddendorf and Corballis1997). There is at best limited evidence for very basic metacognitive capacities in nonhuman animals (Hampton Reference Hampton2001; Shettleworth & Sutton Reference Shettleworth, Sutton, Hurley and Nudds2006).

In studies involving extensive training on a simple task, such as determining which of two stimuli is the larger, some monkeys given the option of not taking a trial eventually learn to avoid trials they are likely to fail (Smith et al. Reference Smith, Shields and Washburn2003; Son & Kornell Reference Son, Kornell, Terrace and Metcalfe2005). Such uncertainty monitoring arguably implies metacognition, in that the monkeys may know that they do not know, and may even imply self-reflective consciousness. However, Son and Kornell concede that their results need not mean that monkeys are aware of their judgments. It remains unclear to us how any such performance is “meta” over and above other undisputable and common confidence judgments that animals may hold about their own ability. For example, judging how far it can jump is surely essential for any monkey traversing the canopy. The nature of what is supposed to be (meta-) represented may play a role in comparative cognition. Great apes have shown some signs of competence in regard to basic mental states. Call and colleagues (Call Reference Call2004; Call & Carpenter Reference Call and Carpenter2001), for example, recently provided some evidence that great apes know something about what they have or have not seen, as reflected in whether they do or do not seek further information before acting. But Call (Reference Call, Terrace and Metcalfe2005) notes that it is not known whether any nonhuman species attribute more complex states such as knowledge or beliefs to themselves.

A key role of a director is to orchestrate rehearsals. Humans rehearse significant future events and practice their own performance not only in their minds, but also in action. In fact, this trait would seem essential to any account of the enormous diversity of human expertise. Humans choose to learn and practice things they want to get better at. A function of play in many species may be that it prepares the young for the future (e.g., fighting skills). However, unlike species-typical play, humans choose what to practice and often do so in anticipation of specific future events (e.g., rehearsal for a theater performance). We know of no evidence that nonhuman animals deliberately practice for specific anticipated events (Suddendorf & Busby Reference Suddendorf and Busby2003b).

4.6. The executive producer

Enacting a planned event requires voluntary control, including executive functions such as the ability to inhibit other stimulus-driven responses in favor of one that suits the anticipated events best. Impaired mental time travel capacity following frontal-lobe damage has been associated with impairment of such self-regulation (Levine Reference Levine2004). Even inhibiting a simple response in order to increase one's total future reward is a difficult task for young children, as research on delay of gratification amply illustrates (Mischel & Mischel Reference Mischel and Mischel1983; Moore et al. Reference Moore, Barresi and Thompson1998). Adult humans, on the other hand, regularly forego instant rewards because of a multitude of anticipated long-term gains (e.g., weight loss, longevity, afterlife).

To override current drives in favor of acting to secure some anticipated future outcome, it would help to overestimate the positive or negative effects of future reward or punishment. In fact, humans show a systematic bias towards overestimating the intensity of future affective responses to future events (e.g., the dreaded interview with the bank manager often turns out to be more benign than expected). The nature of this biased “affective forecasting” is currently a topic of investigation in social psychology (e.g., Gilbert Reference Gilbert2006; Wilson & Gilbert Reference Wilson and Gilbert2005). Accurately anticipating one's own future mental states is no easy feat and is often influenced by one's current states – which is why one buys more treats when shopping while hungry (Gilbert et al. Reference Gilbert, Gill and Wilson2002). Even children show this effect (Atance & Meltzoff Reference Atance and Meltzoff2006).

Plans further require prospective memory – remembering to perform a future action at the appropriate time. Purported evidence for prospective memory in nonhuman animals has been disputed (Thorpe et al. Reference Thorpe, Jacova and Wilkie2004). Executive control may be required not only to implement strategic action plans, but also to manage the motivational and goal system itself (Conway et al. Reference Conway, Singer and Tagini2004; Suddendorf & Busby Reference Suddendorf and Busby2005). For mental time travel to become the flexible and effective future-oriented strategy that it is, cognition had to truly take the driver's seat of behavior (Suddendorf Reference Suddendorf, Corballis and ea1999). Modern humans often juggle a multitude of different goals and must decide when to do what to achieve which aims. The development of these sophisticated, resource-intensive, and error-prone mechanisms continues well into adulthood, and we know of no evidence for a comparable capacity in nonhuman animals.

4.7. The broadcaster

Finally, temporally displaced events are not only entertained in private but are often communicated. A play is not just a metaphor for mental time travel, it is often also a public expression of it. More generally, humans use language to exchange and complement their mental travels into the past and their ideas about future events, as well as to cooperatively coordinate plans and strategies (Suddendorf & Corballis Reference Suddendorf and Corballis1997). There is as yet no evidence that nonhuman animals communicate such mental travels. Such communication does not require language, as pantomime and dance can demonstrate. Chimpanzees appear to have some capacity to imitate (Whiten Reference Whiten1998) and to recognize when someone else is copying them (Nielsen et al. Reference Nielsen, Collier-Baker, Davis and Suddendorf2005). Animals with such abilities might in principle also be able to re-enact past episodes, using mime, if they wanted to communicate those to others.

Although language is not necessary for mental time travel, it provides the clearest evidence of it. Indeed, it may not be too far-fetched to suppose that mental time travel was a prerequisite to the evolution of language itself, since, as we saw earlier, language is exquisitely equipped to express events that are distant in both space and time from the present. If we are correct in supposing that mental time travel is uniquely human, then this may explain why language itself is unique to our species, at least in the strong sense defined by Hauser et al. (Reference Hauser, Chomsky and Fitch2002). The adaptive utility of the capacity to transcend the present in an open-ended, flexible manner may have driven these same properties in language (for a discussion, see Corballis & Suddendorf, in press).

4.8. Components of successful mental time travel

In summary, the theater metaphor implies that mental time travel requires a constellation of skills and is not simply an isolated capacity. Adult humans may fail to act now for a particular future event because of a deficiency in any one of these components (e.g., one may fail to predict one's own future mental states, or one may fail to override more immediate impulses). Young children may also fail to do so because one or the other of these components is not yet fully developed. Further study of how these requisite capacities emerge in human development may hold clues to the relationship between them.

Some of the foundations, such as imagination, self-recognition, and semantic memory, begin to emerge early in childhood; whereas other capacities, such as recursive thought and representational theory of mind, develop first in children between ages 3 and 4. It is around this latter time that unequivocal evidence for episodic memory emerges, childhood amnesia ceases, and implicated prefrontal regions mature (e.g., Levine Reference Levine2004). Between ages 3 and 5, children start to reason about future and past states (Atance & Meltzoff Reference Atance and Meltzoff2005; Gopnik & Slaughter Reference Gopnik and Slaughter1991; Suddendorf & Busby Reference Suddendorf and Busby2005), make plans (Atance & O'Neill Reference Atance and O'Neill2005; Hudson et al. Reference Hudson, Shapiro and Sosa1995), delay gratification (Mischel & Mischel Reference Mischel and Mischel1983; Thompson et al. Reference Thompson, Barresi and Moore1997), accurately report events of yesterday and tomorrow (Busby & Suddendorf Reference Busby and Suddendorf2005; Harner Reference Harner1975), and temporally differentiate events in the past (Friedman & Kemp Reference Friedman and Kemp1998) and in the future (Friedman Reference Friedman2000).

Nonhuman animals may also be limited because of deficiencies in any of these components. Although there is evidence in some species (especially from our closest relatives) for basic components, such as imagination and self-recognition, support for most others is very slim. Should a component be present only weakly, if at all, in other species, mental time travel would be severely limited, at best. For example, without a representational theory of mind, foresight might be restricted to serving satisfaction of current needs, as the Bischof-Köhler hypothesis proposes. Our suggestion, then, is that when studying the possibility of mental time travel in other species, the research agenda should include work on these components to human mental time travel. Determining to what extent specific components are necessary and sufficient for what aspects of foresight to emerge will be an important task.