Rethinking the definition of episodic memory

The definition of episodic memory, as proposed by Tulving, includes a requirement of conscious recall. As we are unable to assess this aspect of memory in non-human animals, many researchers have referred to demonstrations of what would otherwise be considered episodic memory as “episodic-like memory.” Here the definition of episodic memory is re-considered based on objective criteria. While the primary focus of this re-evaluation is based on work with non-human animals, considerations are also drawn from converging evidence from cognitive psychology, neuropsychology, and cognitive neuroscience. Implications of this rethinking are discussed, as well as considerations of familiarity, indirect measures of memory, and generally what should be viewed as necessary for episodic memory. This perspective is intended to begin an iterative process within the field to redefine the meaning of episodic memory and to ultimately establish a consensus view.


Introduction 1
Most of us can think back to the previous week (e.g., 2 a recent dinner with friends) and events of importance as considerations of familiarity, indirect measures of 15 memory, and generally what should be viewed as nec-16 essary for episodic memory. Episodic memory is stud-17 ied in a variety of subfields within psychology, includ-18 ing cognitive psychology, neuropsychology, and cog-19 nitive neuroscience, in addition to comparative psy-20 chology. In this article, I draw from converging evi-21 dence from all of these subfields to suggest a path for-22 ward that may be consistently and broadly used. By 23 adopting a new definition that can be externally ver-24 ified, objectively evaluated, and less reliant on self-25 report, episodic memory can be more consistently as-26 sessed across a variety of subfields. This perspective is 27 intended to begin an iterative process within the field 28 to redefine the meaning of episodic memory and to ul-29 timately establish a consensus view. tively evaluated, rather than rely on introspective re-89 ports. Thus, the approach proposed here is to disentan-90 gle consciousness from the definition of episodic mem-91 ory and adopt methods that are more objective (e.g., 92 source-memory test vs. remember/know procedure), 93 as well as turn to neuropsychological dissociation find-94 ings, to remove this dependency. This is not necessarily 95 to suggest that episodic memory and consciousness are 96 not related, but rather to allow for separations in their 97 definitions that can be individually evaluated indepen-98 dently and rely less on anthropocentric thinking. Ad-99 mittedly, Tulving (2005) makes a similar argument, but 100 in the opposite direction-suggesting that it is anthro-101 pocentric to think that animals' memory works simi-102 larly to our own. While I agree with the premise of 103 this perspective, my goal is to move further towards 104 objective tests of memory, in contrast to Tulving's use 105 of introspective judgments. 106 When first proposed, Tulving (1972, p.384) did not 107 have a strong view of episodic and semantic mem-108 ory as being distinctly different: "I will refer to both 109 kinds of memory as as two stores, or as two systems, 110 but I do this primarily for the convenience of com-111 munication, rather than as an expression of any pro-112 found belief about structural or functional separation 113 of the two." Later, Tulving (1983) strengthened this 114 distinction by suggesting that these two memory sys-115 tems differed in their biological basis. In subsequent 116 decades, Tulving shifted his view further-Tulving 117 (2002, p.5) stated: "Episodic memory is a recently 118 evolved, late-developing, and early-deteriorating past-119 oriented memory system, more vulnerable than other 120 memory systems to neuronal dysfunction, and prob-121 ably unique to humans." Importantly, Tulving (2002) 122 does acknowledge that his perspective on episodic 123 memory as a concept has changed drastically over the 124 thirty years since his initial 1972 work, so I consider the 125 2002 perspective as the primary definition of episodic 126 memory. (For a more exhaustive overview of how the 127 concept of episodic memory has developed over the 128 decades, see Tulving, 2002; for a brief anecdotal de-129 scription of the very beginnings of the term episodic 130 memory, see Tulving, 2016.)

131
To reiterate, my goal here is not to suggest that there 132 is no distinction or continuum between episodic and 133 semantic memory, but rather to re-evaluate the defini-134 tion of episodic memory and our current understand-135 ing of memory in non-human animals, and to con-136 sider how we can test for episodic memory more ob-137 jectively. Tulving (2002) himself comments that some 138 other well-known memory researchers do not agree 139 with his strong distinction between episodic and se-140 mantic memory, providing us with insight that even 141 ter article on memory, or consult book after 160 book, without encountering the term "con- . Indeed, without turning to self-report, it 213 is difficult to measure consciousness in humans, as the 214 definition of consciousness itself is nebulous at best.

215
While dissociating episodic memory from con-216 sciousness, it is useful to be reminded that there are 217 terms that may benefit from increased use in the lit-218 erature, particularly in more precisely characterising 219 the link between episodic memory and conscious-220 ness. The act of conscious recollection has been termed 221 . I do not dispute that ecphory is a cognitive 226 process that humans can engage in; but here I suggest 227 that conscious recall, ecphory, should not be considered 228 as a requirement for demonstrating episodic memory 229 ability, and consider that it is unclear whether non-230 human animals can engage in ecphory. In relation to 231 the Tulving (2002) definition of episodic memory, an al-232 ternative is to consider that studies provide evidence of 233 episodic memory either with or without ecphory.

234
Defining boundaries of episodic memory 235 Apart from removing the requirement of conscious rec-236 ollection, a more nuanced discussion of what should be 237 considered as episodic memory is also warranted. An 238 independent, but largely convergent approach may be 239 to make this discrimination based on neurobiological, 240 functional specialization and consider episodic mem-241 ory to be hippocampal dependent. This approach has 242 the additional benefit of being atheoretical, directly 243 testable, and ensures that cognitive theory is linked 244  (Tulving, 1983(Tulving, , 1984. This view is also con- ter the addition of the requirement of autonoetic expe-298 riences by Tulving (1985Tulving ( , 2002. Thus, for reasons of 299 both parsimony and objective measurement, here I pro-300 pose a consistent view also be taken for human mem-301 ory research-and adopted as the formal definition of 302 episodic memory. I consider the hippocampus as be-303 ing central to episodic memory based on it's consistent 304 involvement in recollection, memory for contextual in-305 formation, and mental time travel. Admittedly, the 306 hippocampus is not a unitary neurobiological structure 307 (e.g., see   based on being distinct in salience/congruity within a 402 local, episodic context (e.g., akin to an oddball or von-403 Restorff paradigm); secondary distinctiveness is in-404 stead in distinct relation to general knowledge or expe-405 riences, thus corresponding to semantic memory. One 406 approach that has been demonstrated to be successful 407 with both humans and non-humans in decreasing the 408 influence of familiarity in recognition procedures is to 409 ensure that all items have been pre-familiarised such 410 that the lures are not entirely 'new' to the experimental 411 task and to match lures as closely as possible to the old 412 items (e.g.  Given this boundary criterion of familiarity being in-415 sufficient for qualifying as episodic memory, what re-416 mains and follows from dual-process theory is recol-417 lection. More broadly, this can be considered as mem-418 ory for items along with contextual information, such 419 as location, time, value, or any of a myriad of other 420 attributes that can be considered evidence of remem-421 bered source information. Though few studies have 422 done so, source memory can be tested in non-human 423 animals (e.g., Crystal et al., 2013).  More broadly, Tulving (2002) expressed that there 512 was no evidence that non-human animals can engage 513 in mental time travel. While this may have seemed 514 more clear-cut at the time, research in the past two 515 decades has provided evidence of mental time travel in 516 non-human animals. For instance, Clayton and Dick-517 inson (1998) conducted a well-designed study where 518 scrub jays cached perishable food (wax-moth larvae) 519 and non-perishable food (peanuts). When allowed to 520 retrieve the stored food after a short delay, they pre-521 ferred the fresh larvae. When recovering food after 522 a long interval, the jays quickly learned to avoid the 523 now-decayed larvae and to instead recover the cached 524 peanuts. This study provides relatively clear evidence 525 that the jays not only remembered where they had 526 cached food, but also which food was cached at each 527 location and when it has been cached-allowing the jay 528 to infer the fresh or decayed state in the case of the per-529 ishable larvae.

530
Providing a more direct example of mental time 531 travel, we can also turn to the behaviour labeled 'vi-532 carious trial-and-error learning.' Here, researchers 533 had observed that when navigating complex T-534 mazes, rats would sometimes stop at the choice-535 point/intersections and look at either potential path-536 way, seemingly to deliberately consider the potential 537 outcomes of the decision (Muenzinger, 1938;Tolman, 538 1939Tolman, 538 , 1948. More recent work has explored this be-539 haviour further, more explicitly linking it to prospective 540 hippocampal firing rate activity (sometimes referred 541 to as "preplay") and mental time travel (Redish, 2016;542 Dragoi & Tonegawa, 2013; Ólafsdóttir et al., 2015; de 543 la Prida, 2020; Kay et al., 2020). While it is difficult 544 or even impossible to gain true insight into the intro-545 spective process that this behaviour corresponds to, it 546 is plausible that it corresponds to a conscious recollec-547 tion experience. Regardless, it must be conceded that 548 this procedure has provided more objective evidence 549 of episodic memory than the subjective criterion that is 550 currently used with humans based on self-report in a 551 remember/know procedure.

552
Despite there being some evidence of mental time 553 travel in non-human animals, this ability may not be 554 necessary to demonstrate memory for contextual infor-555 mation (i.e., what, when, where, which) in these tasks 556 in some or all other species. That is, even when mental 557 time travel is possible, it may not be the same 'strategy' 558