Déjà vu experiences have also attracted some interest in the clinical neurological liter- ature, primarily in epilepsy (Gloor, 1990; Spatt, 2002). It is well known that déjà vu experiences are symptomatic of the aura of simple and complex partial seizures in a subset of TLE patients. In these ictal events, déjà vu is usually felt as a vivid and com- pelling mnemonic experience with a subjective sense that the feeling of familiarity is inappropriate. Given these characteristics, the scientific investigation of déjà vu in TLE can offer unique insight into the neural mechanisms that mediate familiarity assessment, and how they interact with other memory and cognitive control processes.
Here, we review research on déjà vu in TLE that speaks to these issues with a focus on phenomenology, localization, and behavioral correlates.
Phenomenology
Descriptions of déjà vu are often provided spontaneously by TLE patients when report- ing their subjective experiences during the aura of their seizures in a clinical interview.
Phenomenological descriptions provided by such patients suggest that ictal déjà vu is neither evoked by, nor bound to, specific stimuli in the immediate environment, a point that was previously emphasized by O’Connor and Moulin (2008). For example, one patient included in our recent investigation of déjà vu in TLE (Martin et al., 2012) described her experience as follows: “It’s highly visual. Things will suddenly become very familiar. There isn’t a progression from vaguely to highly familiar; it’s just highly familiar.
It’s initially object‐specific, but when I focus my attention on something else it too
becomes familiar. I will even search the room for something that isn’t familiar but everything seems to be so.” This account illustrates that déjà vu is initially experienced for whatever object or scene that is at the focus of attention at the moment of seizure onset, but readily spreads to other aspects of the environment as the seizure progresses.
Given that the mechanism of elicitation appears to be different in spontaneously occurring and ictal déjà vu, one may wonder whether the subjective phenomenology in terms of mnemonic experience is indeed the same (see also Adachi et al., 2010).
Spontaneous reports are often consistent with the formal definition of the phenomenon without any prior provision of related criteria (e.g., Martin et al., 2012; O’Connor and Moulin, 2008). For example, one TLE patient whom we recently examined offered the following description of déjà vu at the onset of his seizures: “It’s a really strange feeling. I’ll feel like I’ve been somewhere before – like I’m looking at a snapshot of a scene that I’ve seen previously. The strange bit is that I also know that I have not been to that place before.” This account was similar in spirit to the descriptions provided by six other patients included in that experimental study (Martin et al., 2012). Generally, the experience was described as consisting of strong feelings of familiarity for the immediate visual environment that persist for only a few seconds and that are accom- panied by a conflicting sense that they are inappropriate. Similar experiences in other modalities were rarely reported in this sample (audition – two individuals, taste – one individual). To the extent that similar characteristics have also been reported in retro- spective studies of healthy individuals (Brown, 2004), we believe that the study of déjà vu in TLE patients can inform our understanding of the phenomenon more broadly;
it offers a unique opportunity to gain better insight into its neural mechanisms, and its relation to familiarity and recollection, as probed in cognitive experiments.
When considering research on déjà vu in TLE, it should be kept in mind that clinical neurology has historically made reference to this phenomenon sometimes with other, often less specific terms (e.g., dreamy state). Accordingly, in this literature déjà vu has not always been distinguished clearly from other experiential phenomena that can accom- pany seizures originating in the temporal lobes (see Illman et al., 2012a, for discussion).
In this context, it is worth noting that déjà vu experiences are typically described as being static in their phenomenology in that they only pertain to the environment currently perceived. There are other mnemonic experiences that are more dynamic in nature and that may be considered to be pathological states of recollection. Such experiences have been referred to as déjà vécu in research on TLE and other neurological conditions, and relate to the progression of sequences of events (Moulin et al., 2005). Whether the presence of déjà vécu is related to other structural brain abnormalities than déjà vu is an issue that, to our knowledge, has not been addressed in any systematic manner as of yet.
Despite this challenge, however, several research findings on relatively well‐defined déjà vu experiences in TLE patients point to a surprisingly consistent picture concerning the neuroanatomical basis of this specific phenomenon in the MTL.
Neural correlates
Research addressing the neural correlates of déjà vu in TLE has generally been con- ducted with one of two experimental approaches. The first, and historically most widely employed approach in clinical settings, has been to evoke déjà vu through electrical intracranial stimulation of cortical tissue pre‐surgically or intraoperatively.
With the second approach, researchers have compared TLE patients with and without
déjà vu with respect to distinct seizure characteristics that manifest in ictal scalp or intracranial electrical recordings, and/or neuroimaging measures of lasting differ- ences in brain morphology or focal brain metabolism. Modern epilepsy research that has attempted to localize mnemonic experiential phenomena accompanying seizures in TLE based on stimulation or ictal EEG recordings has primarily, although not exclusively, linked them to the MTL (Bancaud et al., 1994; Gloor et al., 1982; Vignal et al., 2007; Weinand et al., 1994). Several studies conducted with intracranial EEG recordings, for example, have demonstrated that seizure activity that is associated with mnemonic experiential phenomena, including déjà vu, is more consistently localized to medial than lateral temporal lobe structures (Gil‐Nagel and Risinger, 1997; Vignal et al., 2007; Weinand et al., 1994). Gloor et al. (1982) reported similar findings using stereotactically implanted depth electrodes to stimulate the MTL and lateral temporal lobes in 35 TLE patients, four of whom experienced déjà vu with their seizures. In the TLE patients whose seizures were accompanied by déjà vu, virtually all evoked déjà vu experiences were obtained through stimulation of the MTL, including the amygdala, but rarely from stimulation of the lateral temporal lobes.
Evidence from other studies conducted in TLE patients suggests that, within the MTL, déjà vu may be specifically associated with activity in extra‐hippocampal struc- tures in the anterior parahippocampal region (i.e., PRC and entorhinal cortex, ERC;
Bartolomei et al., 2004; Guedj et al., 2010). Such findings are of particular relevance to the dual‐process model of MTL organization that has been advocated by Aggleton and Brown (1999) and others. Bartolomei et al. (2004) employed intracranial cortical stimulation to examine the role of specific MTL structures in the genesis of déjà vu experiences. Through the use of surgically implanted depth electrodes, they targeted the amygdala, anterior HC, and PRC and ERC (collectively referred to as rhinal cortex). Stimulation applied to rhinal cortex manifested in phenomenological impres- sions of déjà vu in 11% of all stimulations. Comparison of effects for ERC and PRC revealed that stimulation of ERC was more frequently associated with déjà vu and PRC with the recovery of specific visual memories regarding distinct, frequently encoun- tered objects. By contrast, only 2.2% of stimulations of the amygdala and 2.1% in the anterior HC yielded déjà vu. More recently, data that link déjà vu to processes in rhinal cortex have also been reported by Guedj et al. (2010) based on an examination of interictal brain metabolism. The authors conducted an 18‐FDG‐PET investigation that directly compared lasting, interictal glucose metabolic abnormalities in unilateral TLE patients, with and without déjà vu, and healthy control participants. The results revealed more pronounced hypometabolism in ipsilateral rhinal cortex of those patients whose seizures were typically accompanied by déjà vu as compared to those who did not experience déjà vu with their seizures. No differences were found between the patient groups with respect to metabolic activity in either the HC or amygdala. The only other structure that exhibited such differences was a region in the ipsilateral lateral superior temporal gyrus. Compared to healthy control participants, both patient groups did show hypometabolism in ipsilateral HC. This finding suggests that lasting functional changes in the hippocampus can be detected in the context of TLE with déjà vu, but may not be uniquely associated with the generation of these experiences.
That MTL structures other than rhinal cortices also play a role once the seizure has been triggered is suggested by additional evidence from studies that have focused on the synchronization of intracranial electroencephalography (EEG) signals in the MTL in relation to déjà vu (Bartolomei et al., 2012; see also Barbeau et al., 2005).
Taken together, the evidence reviewed regarding the localization of déjà vu in TLE primarily links the phenomenon to rhinal cortex, as opposed to other MTL structures such as the HC or amygdala. Considered in relation to the cognitive neuroscience liter- ature of recognition memory reviewed previously, these findings point to overlap concerning the neural mechanism of familiarity. The dual‐process model of recognition memory represents a theoretical framework within which déjà vu experiences in TLE can be readily interpreted to the extent that an erroneous familiarity signal that arises from seizure activity in PRC is at the core of the déjà vu experience. We note that this account, in and of itself, does not offer any explanation as to why the familiarity experi- ence in déjà vu is experienced as subjectively inappropriate. We will return to this point after reviewing research on the behavioral correlates of the presence of déjà vu in TLE.
Behavioral correlates
The presence of interictal hypometabolism in rhinal cortex that was found to be uniquely associated with the presence of déjà vu in TLE patients (Guedj et al., 2010) raises the interesting question as to whether unique interictal behavioral impairments might also be present in this group. Based on the dual‐process model of MTL orga- nization, one might predict that any such lasting impairments would be reflected in selective familiarity impairments on recognition memory tasks that probe familiarity and recollection separately. Although mild memory impairments in pre‐surgical patients with TLE are well documented in general, this specific issue has received sur- prisingly little empirical investigation in the neuropsychological literature so far.
In our own work, we initially became interested in examining déjà vu in TLE within the dual‐process model framework following research we undertook on a single case, patient N.B., whom we tested post‐surgically. Prior to surgical intervention, this individual suffered from intractable TLE caused by a circumscribed ganglioglioma in the left amygdala. Interestingly, the aura of her complex partial seizures was frequently accompanied by phenomenological feelings of déjà vu. To control these seizures, N.B. underwent a rare surgical resection that targeted anterior regions of lateral and medial temporal cortex in the left hemisphere. Critically, her resection included large portions of PRC and ERC, but spared the HC; the extent of the resection was quantified in our laboratory using magnetic resonance (MR) volumetry, which confirmed intact structural integrity of the HC. In our post‐surgical examination of N.B.’s recognition memory performance, we discovered that her ability to discriminate between previously studied and novel stimuli was selectively impaired when based on familiarity assessment.
In four different experiments, conducted with three different methodological approaches to probe familiarity and recollection for verbal materials (R/K paradigm, receiver operating characteristics of confidence‐based recognition decisions, and a response deadline procedure), we showed that N.B. exhibited impairments in familiarity assessment with preserved recollection (Bowles et al., 2007). Evidence from an fMRI study also suggested that, despite impoverished inputs from PRC and ERC, N.B.’s left HC showed clear signs of functional integrity, as reflected in novelty responses that were comparable to those we observed in control participants (Bowles et al., 2011). Although this study provided critical new support for the dual‐process model of MTL organization, the suggested link between déjà vu and selective familiarity impairments must be considered indirect, given that N.B. no longer experienced seizures at the time we revealed her memory impairments.
In research that has focused on behavioral correlates of déjà vu and other mnemonic phenomena in association with TLE pre‐surgically, one study failed to identify any significant differences between patients whose seizures were or were not accompanied by the phenomenon (Vederman et al., 2010). In this investigation, however, memory performance was evaluated using standard neuropsychological tests with verbal mate- rials only. Thus the authors could not examine familiarity assessment, specifically.
Moreover, memory was tested with verbal stimuli presented in the auditory modality.
Such an arrangement may not be optimal to reveal behavioral correlates of a phenomenon that is predominantly visual in nature. Interictal behavioral deficits that are unique to TLE patients with déjà vu may only be revealed through the use of memory tasks that probe recognition memory for visual stimuli in particular.
In recent research conducted in our own laboratory, we employed experimental tasks of visual recognition memory to examine whether the presence of ictal déjà vu in TLE is indeed associated with selective impairments when judgments are based on familiarity assessment (Martin et al., 2012). Toward this end, we compared two groups of pre‐surgical patients with intractable unilateral TLE; patients in the first group (n = 7) consistently experienced déjà vu as part of their seizures (TLE+) while individuals in the second group (n = 6) had never experienced any mnemonic phenomenon during their seizures (TLE–). Semi‐structured interviews confirmed that ictal déjà vu in TLE+ patients could indeed be characterized as feelings of familiarity for the immediate visual environment that were experienced as subjectively inappro- priate. We conducted two experiments that involved the presentation of novel complex visual scenes in a study phase and a requirement to make recognition judgments in a subsequent test phase. The scenes employed were sampled from a set of discrete cat- egories (i.e., three types of rooms) so as to prevent memory discrimination based on gist and so as to encourage reliance on perceptual characteristics of objects that dif- fered across scene exemplars. Using an R/K procedure (Tulving, 1985), we first sought to determine whether impairments in unilateral TLE with déjà vu would be selective for familiarity assessment and would spare recollection when these processes were probed with meta‐memory judgments in the test phase (“How do you know that this scene was presented previously?”). In a second experiment, we used a variant of the process dissociation procedure developed by Jacoby and colleagues (i.e., an exclusion task), to probe the interplay between familiarity and recollection when they were placed in opposition by way of experimental manipulation (see Yonelinas and Jacoby, 2012, for review). Specifically, this experiment allowed us to determine whether TLE patients with déjà vu could still engage recollection processes success- fully when it was necessary to counteract feelings of familiarity. To get at this issue, participants were first presented with a set of scenes in a study phase. In the subsequent test phase they were presented with a mixed list of previously studied and novel items.
Critically, all of the novel lures were also repeated at various delays during the test phase, with participants being asked to endorse only items as old that had indeed been encountered during the study phase rather than as repeated lures during the test phase (Jennings and Jacoby, 1997). Inasmuch as repetition of lures at test will gen- erate a sense of familiarity for these items, accurate exclusion of these items from
“old” responses is thought to rely on recollection of contextual detail.
Results from the R/K task for categorized visual scenes revealed evidence for impairments that were limited to familiarity‐based assessment in the TLE+ patient group. These impairments were reflected in reduced accuracy of discrimination and
reflected increased false‐alarm rates as well as increased “miss” responses for old items.
We also observed a strikingly selective pattern of impairments in TLE+ patients with the exclusion task. Despite displaying recognition impairments that affected the ability to discriminate between previously studied and novel stimuli, TLE+ patients showed evidence of a spared ability to counter feelings of familiarity for repeated lures, presumably based on intact recollection. The selective deficits observed in these patients contrasted with the broader pattern of recognition memory impairments in the TLE– patient group. Specifically, these individuals showed deficits that affected both recollection and familiarity in the R/K paradigm; it also affected the ability to counter familiarity in the exclusion task by way of recollection. Interestingly this behavioral pattern across patient groups paralleled noticeable differences in ipsile- sional MTL volumes. Volumetric measures of T1‐weighted anatomical MR images revealed that ipsilateral MTL structures were overall less affected in TLE+ than in TLE– patients, with hints of more focal volume reductions in the rhinal cortices of patients who displayed déjà vu in their seizure profile (i.e., TLE+).
To our knowledge, the findings reported in Martin et al. (2012) represent the first dem- onstration of an empirical link between déjà vu and behavioral indices of familiarity assessment on experimental tasks of recognition memory. This link is suggestive of a shared neural mechanism that is common to feelings of familiarity in ictal déjà vu and feelings of familiarity outside of the ictal context (see Spatt, 2002, for related discussion). We will con- sider the nature of this proposed shared mechanism as part of our concluding section.