On the frequency and nature of the cues that elicit déjà vu and involuntary autobiographical memories

Abstract

Barzykowski and Moulin suggest that déjà vu and involuntary autobiographical memories recruit similar retrieval processes. Here, we invite the authors to clarify three issues: (1) What mechanism prevents déjà vu to happen more frequently? (2) What is the role of semantic cues in involuntary autobiographical retrieval? and (3) How déjà vu relates to non-believed memories?

Barzykowski and Moulin (B&M henceforth) offer an intriguing model to unify déjà vu and involuntary autobiographical memories (IAM). A critical claim of their model is that both psychological phenomena share, as a common cause, the automatic processing of ongoing cues in the environment. Specifically, they suggest that a perceptual overlap between an environmental cue and a particular mnemonic content is the starting point for both déjà vu and IAM. While their proposal is compelling, our comment raises three points that, we think, deserve further elaboration. The first and second points concern the nature of the cues leading to déjà vu and IAM experiences. The third point draws attention to the mechanism that differentiates déjà vu from non-believed memories.

The first concern can be put simply: There seems to be a tension between the fact that perceptual overlap is an extremely common phenomenon whereas déjà vu, as remarked by B&M, is extremely uncommon. How come such a rare psychological phenomenon is largely explained by such a common occurrence as the overlap between a memory representation and an environmental cue? According to B&M's model of déjà vu, our memory system involves a continuously active mechanism constantly scanning the environment for cues that could match memory representations. When this mechanism finds a cue that matches a stored representation, but the activation is not sufficient to elicit autobiographical recall, people experience only a sense of familiarity. This feeling is then metacognitively assessed in order to evaluate its source. If the system fails to find the source of the feeling of familiarity, it evaluates the cue as novel, and the output is thus experienced as déjà vu.

The problem is that the model is not sufficiently clear as to how exactly this underlying mechanism manages to discriminate between the very frequently experienced environmental cues that overlap with stored memory representations and that never get experienced as déjà vu, and the very infrequent instances in which they do. To be sure, we agree with B&M in that our memory system is constantly presented with overlaps between environmental cues and mnemonic representations (Fernández & Morris, 2018; Kafkas & Montaldi, 2018). What we find unclear is how and why exactly some of those overlaps manage to only elicit a sense of familiarity and a metacognitive evaluation of novelty whereas the vast majority elicit no feeling at all. One possible avenue to fill this gap in the model could be found in the schema literature, whereby novelty is proportional to the mismatch between the stimulus and its context (Van Kesteren, Ruiter, Fernández, & Henson, 2012). Another possible avenue is to appeal to prediction error, whereby novelty relates to the divergence between a prior likelihood and a posterior probability (De Brigard, 2012; Ergo, De Loof, & Verguts, 2020). Either way, we believe this bit of the model needs further development.

The second concern is about the relationship between semantic memory and IAM. When talking about IAM, the authors swiftly mention that semantic information can elicit IAM. Indeed, there is plenty of evidence suggesting that abstract and conceptual information are powerful triggers of IAM (Mace, 2005; Mace & Hidalgo, 2022). However, almost all of the examples B&M use to substantiate the model involve IAM that are triggered by perceptual cues. While we certainly agree that sensory cues are an important and frequent environmental trigger for IAM (Conway & Pleydell-Pearce, 2000), we also believe that B&M's model could be strengthened by clarifying the distinct contribution of the perceptual and the semantic properties of the environmental cues that can trigger IAM. Consider how semantic information can trigger IAMs in semantic-to-autobiographical memory priming. The work of Mace, McQueen, Hayslett, Staley, and Welch (2019) has shown that priming a concept (e.g., reading the word “frog”) increases the likelihood of having an IAM related to that concept. Moreover, perceptual cues contain both sensory and semantic information that can influence episodic memory (Davis et al., 2021; Hovhannisyan et al., 2021). Each of these could in turn trigger IAM in different ways. For example, when seeing a Darwin's frog, its sensory properties (such as its form or color) could trigger an IAM about a pet frog in my childhood, whereas the conceptual information associated with seeing such a frog could generate an IAM about the period of my life in which I studied the theory of evolution. By further exploring the properties of the cues that trigger IAM, B&M could generate a more thorough explanation about the memory processes that support IAM.

Our final point is a question about the relationship between déjà vu and another surprising memory phenomenon: “Non-believed memories.” In their paper, B&M state that déjà vu “arises out of a higher order interpretation of retrieval processes; it is not possible to have a déjà vu experience and not be aware of it. This critical feature distinguishes it from a false memory.” It is tempting to think that a similar metacognitive evaluation helps to explain “non-believed memories,” that is, recollective experiences of events one no longer believes that happened (Otgaar, Scoboria, & Mazzoni, 2014; Scoboria, Boucher, & Mazzoni, 2015). While unusual, non-believed memories have been shown to occur in around 20% of the population, and present similar phenomenological features as real autobiographical memories (Scoboria, Mazzoni, Kirsch, & Relyea, 2004). Moreover, like déjà vu, non-believed memories also involve the metacognitive awareness that what is experienced is not an actual memory. What it surprising, though, is that while in the case of déjà vu the metacognitive judgment that the event is not a memory is sufficient to remove its recollective experience, the same is not the case with non-believed memories: People know the events they thought were memories are not, but still experience them as such. What accounts for this difference? We think it would be interesting to try to expand the proposed model to see how it could accommodate non-believed memories.