On pattern completion, cues and future-oriented cognition

Abstract

Barzykowski and Moulin's view on involuntary autobiographical memory focuses on automatic activation of representations and inhibitory control mechanisms. We discuss how and when a known neural mechanism – pattern completion – may result in involuntary autobiographical memories, the types of cues that may elicit this phenomenon and consider interactions with future-oriented cognition.

Barzykowski and Moulin (B&M) synthesize considerable evidence to support their primary thesis that involuntary autobiographical memories (IAMs) occur in the context of constant and automatic activation of autobiographical fragments in the absence of inhibitory control mechanisms; when the activation of a given memory representation passes a threshold, it is involuntarily retrieved. While we agree with their proposed model, here we discuss how another critical mechanism in direct retrieval – pattern completion – may result in IAMs, the types of cues that may elicit this phenomenon and how this process interacts with future-related cognition.

Pattern completion, the process of recalling an entire memory representation when cued with a subset of its elements, was originally proposed by Marr (1971) as a fundamental computation of the hippocampus along with pattern separation. He argued that whether incoming information elicits pattern completion depends on the degree to which it matches a previously stored representation, suggesting that as few as one-third of the elements may be sufficient to elicit recall (Becker, 2016). We suggest that B&M's proposal is broadly consistent with this and later computational models of pattern completion and separation (McClelland, McNaughton, & O'Reilly, 1995; Norman & O'Reilly, 2003) in two ways. First, the hippocampus is argued to behave in a thresholded manner depending on the overlap of the cue with the stored memory: If the threshold is met, the cue is pattern completed and retrieval ensues (Elfman, Aly, & Yonelinas, 2014; see Eichenbaum, Yonelinas, & Ranganath, 2007, for a review of neuroimaging evidence of thresholded hippocampal signals). Second, once this threshold is met, retrieval occurs automatically via an autoassociation network mediated by the dense recurrent connections within CA3, resulting in a pattern of excitation in the output layer which is then back-projected to the complete set of neocortical regions representing the memory (Rolls, 2016; for a recent review of empirical evidence, see Becker, 2016).

Therefore, we argue that IAMs occur when content received by the hippocampus overlaps sufficiently with an existing memory as to trigger the automatic “completion” of the whole representation in neocortical regions. While conceptually similar to the proposed threshold model, in that overlap between the cue and the content must pass a threshold, pattern completion reflects a known neural mechanism by which automatic retrieval can occur. As a fundamental retrieval mechanism, pattern completion should act at all levels of the memory hierarchy (Conway & Pleydell-Pearce, 2000), including not just specific events but also general events and lifetime periods shown to engage the hippocampus (e.g., Ford, Addis, & Giovanello, 2011).

Although B&M suggest that cues are central to their account of spontaneous memory phenomena, little consideration is given to why some cues, whether present in the external environment or generated internally by the individual, are more likely to evoke IAMs than other cues. It is likely that, in everyday situations, a specific combination of cues is required to provide sufficient overlap with the stored representation in order for pattern completion to occur (Marr, 1971), explaining why IAMs do not occur all the time. Moreover, it may be that the interaction between cues and internal states is important. Along these lines, Klinger and colleagues have suggested that current concerns, or the state of an organism between commitment to a goal and later accomplishment or abandonment of that goal (Klinger, 1978), make goal-relevant cues especially salient and likely to evoke spontaneous thoughts and memories (e.g., Klinger & Cox, 2004).

Consider an individual who takes the same train ride into work each day. While looking out the window of the train, the individual is exposed to many cues (e.g., a billboard featuring an airplane; a junkyard full of old automobiles) that could, but typically do not, produce spontaneous retrieval of past experiences (Berntsen, 2009). However, if the individual were to set a goal (e.g., a planned trip) related to one of those cues (e.g., a billboard featuring an airplane), then that cue may be more likely to evoke not only thoughts about the goal, but conceivably some relevant experience from the past (e.g., a previous trip). Of course, in this scenario, the memory that comes to be evoked by the now-relevant cue may be more likely to surpass some threshold of activation due to its relation to the goal, but it is important not to overlook the fact that the cue itself is also more salient (Klinger, 2013). Understanding how goals, cues, and representations independently and/or jointly determine the occurrence of spontaneous memory phenomena, such as IAMs, are issues warranting further attention in the literature (e.g., Cole & Berntsen, 2016).

Consideration of goals alongside cues and representations may help shed light on other related phenomena. For instance, one important function of memory is to retain simulations related to predictions, intentions, and plans for the future (Ingvar, 1985; Szpunar, Addis, McLelland, & Schacter, 2013). Recent studies have shown that simulations of the future are especially likely to come to mind in a spontaneous manner if they have been thought about on some previous occasion (e.g., Jeunehomme & D'Argembeau, 2016). This is likely the case because simulations result in memory representations that can be directly elicited by goal-relevant cues (for similar considerations in the prospective memory literature, see McDaniel & Einstein, 2000; for more general considerations of the role of pattern completion mechanisms in future-oriented cognition, see Falandays, Nguyen, & Spivey, 2021). Although more speculative, it is possible that in some situations, goal-relevant cues only weakly evoke memories of the future, along the lines of the continuum proposed by B&M, giving rise to a spontaneous familiarity-based feeling that one should be doing something in the present moment (a “déjà vu for the future”) as opposed to the more classic feeling that one has already experienced the present moment.

In sum, for a more complete account of IAMs and related phenomena, it will be important to consider more precisely the neural mechanism(s) governing the interaction between cues and representations, and why specific cues and representations are especially relevant and/or accessible at any moment in time.