Two primary issues are emphasized in this commentary. One is my alternative proposal that foresight should not be considered an all-or-none phenomenon, as suggested, but rather as a continuum. Second, I argue that the requirements from the memory that mediates foresight seem contradictory: It has to encode specific experiences, while at the same time it should be flexible enough to be applied to new situations. Keeping both issues in mind as this field advances will help fine-tune the welcome contribution of Suddendorf & Corballis (S&C).

The claim that foresight is unique to humans is common, and is based on reasonable intuition, but there is no evidence to support it unequivocally. In fact, as S&C review, evidence suggests that nonhuman animals might actually possess some limited foresight. For example, scrub jays show planning ability with their food-caching behavior (Raby et al. Reference Raby, Alexis, Dickinson and Clayton2007), fish can learn to anticipate feeding time and place (Reebs & Lague Reference Reebs and Lague2000), rats have shown successful anticipatory behavior in response to complex feeding schedule (Babb & Crystal Reference Babb and Crystal2006), and apes save tools for future use (Mulcahy & Call Reference Mulcahy and Call2006). Although these demonstrations do not necessarily reflect foresight as complex as ours, they nevertheless reflect future-related thoughts and actions. Do these mechanisms differ on a qualitative level from human foresight? The evidence does not support a definite positive or negative answer to this question. Nevertheless, given how much more sophisticated and intelligent we are compared with other animals, intuition might dictate that animals are not able to think about the future, as suggested by S&C as well as by others. But foresight is not necessarily an all-or-none phenomenon. Therefore, rather than surrendering to intuition, I propose an alternative whereby foresight is a continuum in that it is present in animals to the extent that it is needed for survival in their own environments. Indeed, if a costly faculty such as foresight developed in humans for survival, it makes sense that nonhumans could benefit from it as well. Until definitive evidence to support the claim that nonhumans lack foresight is potentially reported, this alternative of a continuum is as plausible and thus should be assigned a comparable likelihood, which gives less weight to intuition, during treatments such as this presented by S&C. In other words, instead of providing complex explanations for behavior that is reminiscent of mental time travel in animals (e.g., “a combination of predispositions and specific learning algorithms”; target article, sect. 3.2, para. 2), we can follow the logic of Occam's razor and simply call it foresight of reduced magnitude.

That said, it is easy to see that animals would require foresight that is considerably less complex and rich in scenarios than that required by humans. Consider the range of activities in a life of a nonhuman animal, which is the “pool” of elements that can be combined to create a foresight. This pool includes sleeping, eating, grooming, mating, and usually not much more. In other words, no one but humans is expected to be able to imagine scenarios as complex as sipping a piña colada on the beach while checking e-mail on a laptop connected to the Internet via satellite. As reviewed in the target article, animals nevertheless show meaningful “precursors” of foresight ability. According to the alternative I proposed here, however, instead of a precursor, these manifestations may represent actual foresights, albeit of lower complexity. What might be seen as “limited ability for mental time travel” could indicate that animals have simply evolved as much of this ability as their environment and way of life require.

This issue of an all-or-none versus a continuum is analogous to another important issue. In the context of future-related processes, there is foresight that involves elaborate future-related thought, simulation and planning, and there are the simpler associations (e.g., fire=hot), which can be completely automatic and accomplished outside of awareness (Bar, in press). Are these two processes qualitatively different, relying on independent mechanisms, or do they reflect different extremes on a continuous spectrum of complexity? Although it is not clear if nonhuman animals possess foresight, it is clear that they can use simple predictions that are based on associations (i.e., they learn from experience and apply it in the future). Therefore, they might simply lack the presumably other mechanism required for elaborated foresight (Gilbert Reference Gilbert2006). On the other hand, one could argue that while simple associations are merely atomic elements, when accumulated, they could serve the basis for phenomena as complex as foresight simulation (Bar et al. Reference Bar, Aminoff, Mason and Fenske2007). Let us consider a reasonably complex example: You need to get your friend a birthday gift. This does not sound like a simple associative operation. However, when we examine the constituent operations required to achieve this goal, it becomes apparent that experience-based associations may suffice: Your knowledge of this friend's taste as well as needs could be linked in an associative cluster (e.g., a context frame; Bar Reference Bar2004); the items that correspond to this associative cluster (e.g., your friend likes but does not have an iPod, a vegetarian cookbook, a Toyota FJ Cruiser, 1980s memorabilia) are each associated in your memory with a certain price range and with knowledge of how to get each of these items, and they can be further linked with knowledge about your budget, time constraint, and so on. Therefore, a task that initially seemed quite remote from being based on associations, could be achieved by relying on associations and their appropriate integration. If foresight is indeed derived from associations of varying degree of complexity (which we might not always be aware of), this would support the proposal that foresight could be seen as a continual phenomenon of different levels of complexity and thus possibly present in animals, though to a lower degree than in humans.

The issue of associations is also related to the second part of this commentary. A critical characteristic required from the memory system that subserves foresight is flexibility. This is derived primarily from the notion that in creating foresights we frequently have to rely on past experience while modifying their details to adhere to novel situations. At the same time, it is widely endorsed, and presented logically by S&C, that episodic memory is the “database” from which foresights are drawn. Episodic memory encodes specific experiences (Tulving 1984; Squire & Zola Reference Squire and Zola1998) using associative representations (Eichenbaum & Fortin Reference Eichenbaum and Fortin2005) and, as S&C review, contains the particularities of specific episodes (although episodic memory is far from being a perfect depiction of reality; see Schacter & Addis's commentary in this issue). That episodic memory is crucial for foresight is evident from reports of loss of foresight by patients with a compromised episodic memory, typically in medial-temporal amnesia (Klein et al. Reference Klein, Loftus and Kihlstrom2002b; Hassabis et al. Reference Hassabis, Kumaran, Vann and Maguire2007). What begs an explanation is how would a system that is so specific in its representation of relations can at the same time be flexible enough to be applied in novel situations. This paradox emphasizes that more work is needed before we understand the operations done on memory to derive foresight (Bar, in press). Addressing the computational and neural operations used for reconstructing and translating an existing memory into generalized future “memory” (Ingvar Reference Ingvar1985) would result in knowledge with implications leading to important clinical, theoretical, and technological advances.