Mahr & Csibra (M&C) argue that episodic memory should be defined by its metarepresentational format, and that this format endows it with an additional communicative function over and above that provided by non-metarepresentational “event memory.” Although we are sympathetic to the idea that metarepresentations may be centrally involved in episodic memory (cf. Redshaw Reference Redshaw2014; Suddendorf Reference Suddendorf, Corballis and Lea1999), and that the construction of nested mental scenarios and our urge to exchange them may be central to what distinguishes humans from other animals (Suddendorf Reference Suddendorf2013), we have concerns with M&C's characterization of adaptive function.

Although M&C state that they are not making claims about the actual evolutionary history of episodic memory, we do not see how a serious discussion of adaptive function can simply sidestep this central issue. In doing so, the authors fail to recognize the important distinction in evolutionary biology between (1) the forces that initially brought about a trait in the past and (2) the current utility or present-day fitness consequences of a trait. Whereas the former concerns adaptive function in the sense envisioned by Tinbergen (Reference Tinbergen1963), the latter may also be considered adaptive function but is hotly debated (e.g., see Bateson & Laland Reference Bateson and Laland2013; Nesse Reference Nesse2013). Furthermore, the authors' interpretation does not quite fit either of these evolutionary conceptions of function. Instead, they refer to the forces that have allowed a capacity to be “retained in the selection process” (sect. 2, para. 1). Retention is a separate issue from original function or current utility, and traits can remain within a gene pool as long as they are not detrimental to fitness (e.g., even if they no longer add adaptive value).

Tinbergen's (Reference Tinbergen1963) original sense of adaptive function is probably the most widespread and perhaps the most challenging question to answer. Without a time machine, it is difficult to determine why any trait, let alone a cognitive trait that leaves no direct fossil record, originally evolved. Episodic memory may have evolved because it provided our ancestors with a direct survival and/or reproductive benefit, but it may just as well have evolved as a by-product of another capacity with benefits in other domains. For instance, although we agree that the capacity to form metarepresentations may have fundamentally transformed the nature of memory (e.g., Cheng et al. Reference Cheng, Werning and Suddendorf2016; Redshaw Reference Redshaw2014), we note that it may have also fundamentally transformed foresight (e.g., Redshaw & Suddendorf Reference Redshaw and Suddendorf2016; Suddendorf & Corballis Reference Suddendorf and Corballis2007), communication (e.g., Suddendorf Reference Suddendorf2013; Wilson Reference Wilson and Sperber2000), and the understanding of minds (e.g., Perner Reference Perner1991; Stone & Gerrans Reference Stone and Gerrans2006). Therefore, it is possible that episodic memory is a by-product of a metarepresentational capacity that initially evolved in another domain, even if it may have current utility.

Furthermore, when considering current utility (or trait retention), there is no reason to presuppose that episodic memory must have only one significant effect on fitness. Dextrous hands and opposable thumbs, for instance, may have originally evolved to allow primates to thrive in forest environments, but later on they allowed hominins to create tools for use in the savannah (and now they allow us to write and to engage in many other activities that may result in fitness benefits). Likewise, even if episodic memory has a communicative function, as M&C argue (and we agree that it has), it may also have many other current uses. For example, it enables people to learn from an event more than once and to draw lessons based on the discovery of connections between two or more events (Klein et al. Reference Klein, Cosmides, Gangi, Jackson, Tooby and Costable2009). And despite M&C's protestations, episodic memory may also have a future-oriented function.

One of us was involved in putting forward the case that episodic memory is but one instantiation of a general, mental time travel system that allows us to mentally construct scenarios situated in the past or future (Suddendorf & Corballis Reference Suddendorf and Corballis1997; Reference Suddendorf and Corballis2007, cf. Addis et al. Reference Addis, Wong and Schacter2007). Supporting this view, episodic memory and episodic foresight share close links in the brain (e.g., Schacter et al. Reference Schacter, Addis and Buckner2007), in development (e.g., Busby & Suddendorf Reference Busby and Suddendorf2005), in phenomenology (e.g., D'Argembeau & Van der Linden Reference D'Argembeau and Van der Linden2004), and in their dysfunctions (e.g., Kwan et al. Reference Kwan, Carson, Addis and Rosenbaum2010; Race et al. Reference Race, Keane and Verfaellie2011). And although episodic memory probably has (current) functions of its own, the principal benefits of this general system may lie in its open-endedness, which allows us to imagine various future alternatives and subsequently shape the future to our own design.

M&C argue that this constructive perspective leaves retention of past episodes as an accident and cannot explain why episodic memory is frequently reliable, but this an unfair characterization of the position. Episodic memory may well be an adaptive design feature of the capacity for foresight (Suddendorf & Busby Reference Suddendorf and Busby2003), but this does not mean it is an accident and its common veracity is a mystery. Roughly accurate mental representations of past episodes are required as an arsenal of experiences from which to extrapolate. Furthermore, to understand the present and predict how the future may unfold, it is often necessary to construct narratives that link events across time. Narratives are of course also critical to human communication – and language capacities may have co-evolved with mental time travel to allow the sharing of episodic information (Suddendorf et al. Reference Suddendorf, Addis and Corballis2009). This process does not require that we represent previous experiences with perfect accuracy, but it does require some level of veracity. So we tend to store the gist of past events, and even though reconstructions of the same event often vary over time (e.g., Schmolck et al. Reference Schmolck, Buffalo and Squire2000), we can draw on these roughly accurate representations to construct future alternatives and evaluate them. By comparing several options in terms of their likelihood and desirability, humans are in the position to actively pursue one path over another and to prepare for multiple contingencies (Redshaw & Suddendorf Reference Redshaw and Suddendorf2016). Thus, in contrast to M&C's proposal, episodic memory may have many current functions, and a principal one might well be to enable the creation of roughly accurate representations of potential future events that can drive prudent action.