Barzykowski and Moulin (B&M) suggest that involuntary autobiographical memories (IAMs) and the experience of déjà vu can be described and explained as natural phenomena resulting from memory processing, particularly retrieval. Both are based on the same basic retrieval processes, they argue, and can be considered as “involuntary” or spontaneous cognitive events. The target paper discusses similarities and differences between the two phenomena, and outlines how these unusual subjective experiences may be explained by relating them to known mechanisms of memory retrieval. This leads the authors to conclude that IAMs and déjà vu lie on a continuum. Here, I expand on why inhibitory (dys)function may play a significant role in such a continuum.

In memory systems, distinct patterns of neuronal activity (Atallah & Scanziani, Reference Atallah and Scanziani2009; Colgin, Reference Colgin2016) are modulated intricately by inhibitory neurons in the hippocampus and cortex. More generally, inhibitory neurons are important for the structural formation and proper functioning of neuronal assemblies (Holtmaat & Caroni, Reference Holtmaat and Caroni2016). This is due to their fast-spiking behaviour which allows them to provide powerful feedforward and feedback inhibition to excitatory neurons (Gan, Ming Weng, Pernía-Andrade, Csicsvari, & Jonas, Reference Gan, Ming Weng, Pernía-Andrade, Csicsvari and Jonas2017) and constrain the size or identity of such assemblies (Colgin, Reference Colgin2016; Gan et al., Reference Gan, Ming Weng, Pernía-Andrade, Csicsvari and Jonas2017; Isaacson & Scanziani, Reference Isaacson and Scanziani2011). This limits the realm of excitatory plasticity for neurons in such assemblies, subsequently providing increased efficiency for memory storage and greater robustness against over-excitation (Mongillo, Rumpel, & Loewenstein, Reference Mongillo, Rumpel and Loewenstein2018), which may also help reduce the number of spurious attractors. From a computational perspective, it therefore seems natural that inhibition will be implicated in IAMs and déjà vu.

From a psychological perspective, Conway and Pleydell-Pearce (Reference Conway and Pleydell-Pearce2000) describe a model wherein a cognitive inhibitory control mechanism may actively supress the activation of memories while they are not relevant. Such a mechanism could prevent such memories from becoming fully activated in our conscious awareness, while leaving room for the experience of familiarity in the absence of the underlying memory content. Retrieval of IAMs could occur via a similar mechanism. While B&M do not categorically reject this as an explanation, they point to recent studies (Barzykowski, Radel, Niedźwieńska, & Kvavilashvili, Reference Barzykowski, Radel, Niedźwieńska and Kvavilashvili2019; Barzykowski, Staugaard, & Mazzoni, Reference Barzykowski, Staugaard and Mazzoni2021) which suggest a retrieval threshold or activation account. For example, IAMs may be particularly effective in capturing memory-related attention due, for example, to certain phenomenological properties such as emotional valence or particularity. However, it seems more likely that both accounts are true and work in concert.

In computational associative memory models where memories are spatially correlated (Burns, Haga, & Fukai, Reference Burns, Haga and Fukai2022; Haga & Fukai, Reference Haga and Fukai2019), memory capacity and the range of retrieval across related memories is dramatically altered by subtle modulations of inhibition – without any change to the underlying excitation or memory structures. While there is still a critical threshold or level of activation which must be obtained for memories to become activated, the dynamics of these mechanisms are changed by inhibitory modulation. In the context of humans, where inhibition protects against memory interference (Koolschijn et al., Reference Koolschijn, Emir, Pantelides, Nili, Behrens and Barron2019), this may explain why clinical studies have reported that delusions and hallucinations associated with schizophrenia are due to impaired inhibitory regulation (Vogels & Abbott, Reference Vogels and Abbott2007; Yizhar & Fenno, Reference Yizhar and Fenno2011).

Learning of context-dependent associations has also been shown in rats to be correlated with changes in inhibition (Kuchibhotla et al., Reference Kuchibhotla, Gill, Lindsay, Papadoyannis, Field, Sten and Froemke2017) and inhibition may play an active role in helping to forget competing memories of a retrieved memory (Wimber, Alink, Charest, Kriegeskorte, & Anderson, Reference Wimber, Alink, Charest, Kriegeskorte and Anderson2015), perhaps pruning the relevant assemblies to improve or maintain the strength of the retrieved memory. It seems likely that IAMs and déjà vu may be implicated or are affected by this “pruning” role, in combination with other inhibitory mechanisms, for example, disinhibition (Letzkus, Wolff, & Lüthi, Reference Letzkus, Wolff and Lüthi2015), and that these phenomena will subsequently occur in different frequencies for persons with inhibitory neuron disorders. Inhibitory function is also known to change when exposed to different drugs, and indeed we see that recreational uses of alcohol and other substances often increase mind wandering (Sayette, Reference Sayette, Reichle and Schooler2009), a phenomenon the authors argue is related to IAMs and déjà vu.

In summary, the role of inhibitory neurons in memory systems is well-established. These neurons are important for the structural formation and proper functioning of neuronal assemblies, which allows for efficient memory storage and increased robustness against over-excitation. Inhibition also plays a role in protecting against memory interference and competition. Dysfunction in inhibitory regulation has been linked to delusions and hallucinations, and subtle modulation of inhibition shows dramatic changes in memory retrieval dynamics in computational models. These findings suggest that IAMs and déjà vu experiences may be related to the function of inhibitory neurons or their cognitive control. These phenomena may occur at different frequencies for individuals with disorders affecting inhibitory function.

Unifying the threshold/activation and inhibitory control accounts of IAMs and déjà vu thus appears a sensible goal; it is encouraging the authors state that “our position is relatively inclusive” in this regard and do not categorically reject the inhibitory account. The challenge will be to design suitable experiments with relevant human populations and in relevant conditions to connect these high-level subjective phenomena with the rapidly growing biological and computational literature non-trivially implicating inhibition in memory processes.