Involuntary memories are not déjà vu

Abstract

The proposed framework can benefit from integrating predictive processing into the explanation of déjà vu which corresponds to interrupted prediction. Déjà vu is also accompanied by familiarity. However, considerable ambiguity is inherent in familiarity, which necessitates elaboration of this construct. Research findings on involuntary autobiographical memories and déjà vu show discrepancies, and clustering these constructs can be counterproductive for research.

We believe that the authors have proposed a comprehensive approach to integrating the process of memory retrieval to account for both involuntary autobiographical memories (IAMs) and déjà vu experiences. In the current commentary, we focus on four points that may contribute to the discussion. These are: (a) The need to integrate predictive processing into the explanatory framework for déjà vu, (b) the need to further specify the concepts of weak activation and familiarity, (c) the findings conflicting with the idea of common origins for déjà vu and IAM, and (d) the possible detriments of clustering constructs.

In recent years, the concept of prediction emerged with high impact to the extent that some researchers consider the study of prediction a new paradigm (Hutchinson & Barrett, 2019). Predictive processing emerges in areas of study from perception and imagery (Grush, 2004) to visual recognition (Bar et al., 2006) and decision-making (Doya, 2008). Cleary and Claxton (2018) suggested that déjà vu, too, could be a function of predictive processing. Even in novel circumstances, top-down processes seek contextual cues to past contexts to activate relevant past experiences. We use these past experiences or schema representations to predict current experiences. The déjà vu experience may happen when this prediction is erroneous. The mismatch between the prediction and the actuality would dissipate the fleeting feeling of familiarity, and the individual reports a déjà vu. Cleary et al. (2021) emphasized the familiarity component and showed that not all déjà vu experiences are accompanied by a feeling of prediction, but the feeling of familiarity typically characterizes them. Predictive processing studies rely on self-report making them difficult to observe because prediction remains implicit in daily life, whereas déjà vu is defined by awareness.

Researchers employing event-related potentials to study familiarity more directly obtained varying results. For example, familiar faces show N250, whereas participants' own faces do not (Sommer et al., 2021; but also Wiese et al., 2022). Caharel and Rossion (2021) showed that when long-term familiarity exists, the response is much sooner. Leynes and Upadhyay (2022) used familiar words and showed that responses changed depending on the context. Therefore, one must use caution when considering familiarity as a singular construct.

The concepts of familiarity and weak activation form the main tenets of the authors' hypotheses. These constructs require further elaboration because several factors may be influential, for example, insufficient cues, the inadequacy of cues, threshold for activation, or activation level. In addition, weak familiarity may arise because memory cells or ensembles have weak connectivity, cell clusters are partially damaged, or the representation is temporarily inaccessible (Josselyn & Tonegawa, 2020). If familiarity is a function of contextual cues, there must be a sufficient level of cues to trigger familiarity, but the configuration of these cues may not match the activated representation. We have also seen that schema activation and remembering specific memories cannot be distinguished easily in recalling past events (Ece & Gülgöz, 2021). In déjà vu, it is likely that the cues activate a schema rather than a specific event, and thus, an actual memory may not be available. In IAMs, on the other hand, what is remembered is a previously experienced specific event. It is also possible that déjà vu results from the hippocampus being unable to reinstate the original memory representation or to reconstruct it (Barry & Maguire, 2019; Frankland, Josselyn, & Köhler, 2019). With IAMs, on the other hand, memory is reinstated without any deliberation.

Another point is the discrepancies between IAM and déjà vu findings. Berntsen (2010) proposed that IAM and voluntary recall share the same encoding and storage principles with four claims. First, the IAM recall experience is universal, and they are observed in most individuals (Ball & Little, 2006; Rubin & Berntsen, 2009), whereas Brown (2003) reported the rate of people experiencing déjà vu at least once in their lifetime to be 67%. Another discrepancy is the decrease in the frequency of déjà vu experience with age. No such decrease is observed in IAM experience (Berntsen et al., 2015). One would expect that the likelihood of having a déjà vu would increase with experience and the possibility of an associative match, according to configural similarity and familiarity theories. Berntsen (2010) also proposed that IAM recall is frequent, with around 20 memories per day on average (Rasmussen & Berntsen, 2011), but déjà vu experience is very rare, a few times a year (Brown, 2003). If familiarity increases déjà vu experience, then déjà vu experience should be more frequent because our environment is full of cues overlapping with our experience (Cleary et al., 2012; Cleary & Claxton, 2018). Berntsen (2010) argued that involuntary memories belong to the same memory system as voluntary ones. However, IAMs differ in retrieval processes (Barzykowski & Staugaard, 2016). Déjà vu phenomenon is difficult to categorize within the memory system because it does not include any key processes observed in memory, such as encoding, maintenance, or retrieval. In déjà vu, what is lacking is not “access to content” (see Fig. 1 in the main article), but the content itself. Conversely, there is an actual recall of “content” in IAM. Therefore, déjà vu is merely a spontaneous and false belief of familiarity.

Although trying to explicate memory mechanisms within a single framework and to have a general model of retrieval is valuable, clustering qualitatively distinct processes as on a continuum may blind researchers to nuances in the underlying processes. There seem to be similar problems with other constructs. For example, false memory is an overarching term that subsumes memory implantation, recovered memories, imagination inflation, misinformation effect, and the Deese, Roediger and McDermott (DRM) task although there is evidence that they may be products of different mechanisms. Patihis, Frenda, and Loftus (2018) studied the correlations between different false memory paradigms and observed no relationship. In order to avoid such pitfalls, we should focus on particular mechanisms rather than constructing general frameworks, at least for now.