Careful what you ask for: When deja vu is not deja vu
When I teach memory to my undergraduates, I point out that one of the leading theories of memory retrieval and the very nature of memory itself hinges on what it feels like to access memories. Some information feels like it is being remembered, whereas other information comes to mind as a fact. Likewise, when making judgements of whether we’ve encountered something before — recognition — we can sometimes retrieve specifics to justify our endorsement of prior occurrence, or other times we can be certain we’ve encountered the information before without retrieving such context – our recognition is driven by a feeling of familiarity. I’ve made a career out of researching such feelings – and when they go wrong, such as with deja vu – and I have a lot invested in it. In short, I think that memory is guided by subjective states which operate at retrieval. As I am teaching, having outlined this theory, I like to pause at this point and ask how we might measure such subjective feelings as experimental psychologists.
I tell them that the simple answer is that we can just ask the participant in our experiment to tell us what they think or feel*. In the 1980s, according to some of my elder and better colleagues, this came as something as a revelation for the experimental study of memory, and was even viewed as something of a return to introspectionism. I didn’t know all that when I started a PhD looking at subjective evaluations – metacognition – in Alzheimer’s disease. Alongside advances in patient-centred care, it seemed a reasonable thing to do to ask patients to gauge their own memory capacity, and the measures yielded sensible data and meaningful results, if you got around the fact that the accuracy of such metacognitive evaluations was plagued by floor effects for the memory task. Likewise, the classification of recognition memory responses according to subjective states yields replicable and robust effects. Levels of processing manipulations effect the experience of ‘remembering’ in a predictable fashion, for instance, as I tell my students.
Anyway, my career developed to include the study of deja vu and delusional states of familiarity. Deja vu, it should be noted, is a subjective experience par excellence. Because of the earlier work on subjective experiences of memory and the mainstream use of subjective reports, it has become acceptable to just ask people about the states they are experiencing. Again in the 1980s, we started ‘inducing’ Tip-of-the-Tongue (ToT) states by merely administering a set of tough general knowledge questions and asking people if they had the answer on the tip of their tongue (it might be nice in this era of embodied cognition to see if there actually IS something on the tip of the tongue). Some experimenters do neat things like look at the susceptibility of having another ToT after having just had one, or the reaction time to respond to a question the trial after reporting a ToT, but such things are rare. As you might have noticed, ToTs vary with age (but how they differ seems to differ in the lab and the real world) and other ‘objective’ factors like that – all that helps us triangulate on what such states are, and more importantly, it helps us know whether we actually are measuring something real but intangible . Nonetheless, you can never really know if someone is experiencing a ToT or not. Likewise for deja vu – as we (the handful of deja vu researchers who will assemble shortly for the first time at ICOM 6) are found of saying – it has no behavioural corollary.
The holy grail of deja vu research is the experimental induction of the experience. A few sensible, apparently testable theories of deja vu formation exist, and so we should put them to the test. If we know what causes deja vu, we can provoke it in experiments. Clearly the most compelling and elegant of these experiments have been run by Anne Cleary, and they represent the very best we have in understanding deja vu-like illusions of familiarity. We are probably all searching for the paradigm which does for deja vu what the DRM did for false memory, but I don’t need to go into any detail here about the experiments carried out on deja vu because what I am addressing here is something more fundamental than that: can we trust someone’s judgement when they report having had a deja vu experience?
All of which is a necessarily long preamble to describe what seems to me to be the culmination of my research work so far, five well-powered experiments (or four experiments and 2B) run with somebody else’s graduate student, Radka Jersakova at the University of Leeds (hard-working first author) and my own ex-student Akira O’Connor at the University of St Andrews (hard-working senior author). These data have just been published in PLOS ONE and made available on line. It seems to me they raise some very alarming issues about subjective reports in cognitive psychology.
A long time ago, when I still lived in the UK, we started out with the aim of measuring how we asked questions about deja vu influenced the report of deja vu. It seemed to us that in our own experiments and in those published elsewhere, it was pretty easy to generate deja vu. Or at least, it was suspiciously too easy to have students report they were experiencing deja vu. Or at the very least, it was relatively easy to have students report that they were having what they thought to be deja vu experiences. These tortuous sentences underlie the whole problem of subjectivity (we shall come to this later): is what I say is deja vu the same as what you say deja vu is? More importantly, how could we know we were really giving people deja vu? I suggest that if we were really capable of reproducing the intensity and strangeness of the deja vu experience in the laboratory, then we should probably have a few spontaneous reports of deja vu in our cubicles – and, probably, a long queue of eager undergraduates willing to do our experiments. We don’t have either.
In our PLOS ONE article we stress the fact that even in control trials (where deja vu is not supposed to be generated) reports of experience can be as high as 23%. We thought that the use of repetitive questions on each trial about deja vu were responsible for this high baseline report of deja vu and we were concerned that it would lead to the exaggeration of the real rate of deja vu, which is, admittedly, always significantly (in the statistical sense of the word) higher in the experimental condition. We were worried by the possibility of ‘false positive’ deja vu experiences – and thus a misunderstanding of what deja vu is.
In our first experiment we set out to look at the effect of repeated asking about deja vu on the reported rates of deja vu formation. We had four between subject conditions. In one, participants were asked repeatedly about deja vu formation, in another they were asked about the ToT experience, in a further condition they were asked about the colour of the font the words were written in, and finally, in a control condition, they weren’t questioned about any subjective experience. Now, here’s the critical design issue. We did all this embedded in a continuous recognition task. To our knowledge, no one ever reported or hypothesised that a continuous recognition task should give rise to any particular feeling, and certainly not the ToT or the deja vu experience. To be clear: we purposefully chose this task because it should NOT generate feelings of deja vu or ToT. It’s a basic task. People see a stream of words, some of which repeat. Their task is just to report if they have already seen the word before or not. We’ve been running such tasks for a long time in experimental and clinical settings. We’re not given to believe that they provoke deja vu. Of course, there’s no reason why someone shouldn’t have deja vu carrying out such a task, but given that its experienced by most young people about 3-6 times a year, it would be a very rare occurence during our memory task that lasts less than an hour. To reiterate, we chose such a task just to see if we could provoke the report of deja vu when we weren’t expecting to generate it, and we were, perhaps naively thinking that reports of deja vu would be zero or close to it, in the conditions when we weren’t asking about deja vu.
Critically, at the end of the experimental part, with this recognition task and all these intermittent questions about subjective experience, we asked all participants – regardless of condition – the same questions about deja vu, ToT, and the colour of the words. We were surprised to find that even in our control condition, deja vu was reported by 41% of our sample. And, this level of deja vu reporting did not change according to the experimental condition. That is, a lot of people reported having deja vu. A lot of people also reported having had ToT during this task. And remember, we didn’t design the task expecting to find deja vu – and much less ToT. One reviewer suggested that we actually had given people deja vu, even though we had not intended to; perhaps continuous recognition does give people deja vu, but no-one had ever thought to ask.
I must admit, after we analysed the results of this first experiment – which were pretty unequivocal, I felt panicked and thought the whole study of the subjective experience of memory might collapse about my ears like an imaginary house of cards. We found what we had predicted, but more so. I was torn between either thinking I had unwittingly conjured up a very simple way of inducing deja vu or tip-of-the tongue; or that people just obligingly experienced whatever strange sensation I asked them about. In the end, given my knowledge of memory and the hypotheses we had made – I erred towards the second explanation. We kept going to test a few other hypotheses about why we might have such a high rate of deja vu reporting. In essence, we kept trying to reduce this high rate of reporting deja vu – and kept failing.
Perhaps people just didn’t understand what we meant by ‘deja vu’? After all, the phrase is now repeated in the media as just meaning tedious repetition, which was pretty apt for our experiment. However, across our experiments, even when presented with a clear contemporary definition**, between 32% and 58% of participants nonetheless reported experiencing deja vu. Changing the definition of deja vu or asking participants to bring to mind a real-life instance of déjà vu or TOT before completing the recognition task had no impact on the report of deja vu – it stayed high. Only in one experiment did we have any impact on the rate of reporting deja vu: changing the method of requesting subjective reports from the commonly used retrospective questioning (e.g. “Have you experienced deja vu?”) to free report prospective instructions (e.g. “Indicate whenever you experience deja vu.”) reduced the total number of reported subjective experiences. That’s the take-home message for researchers working on similar topics.
Where does this leave me and subjective experience? I still believe that my deja vu experiences are real and that we all, from time-to-time, have the experience. Have I fallen out of love with subjective report? It seems to me that when designing experiments we need to be really careful what we ask for***. The more nebulous and infrequent the experience, and the less concrete the definition, the more carefully we have to tread around issues of subjective report. In our experiment, the salience, intensity, and emotionality were all rated significantly higher for the real world experience of deja vu than for the experimentally generated deja vu (if indeed it was actually deja vu) – but people nonetheless thought that what they were experiencing was deja vu, or that it could be described as like that. That’s the great thing about subjective report: perhaps they’re all right and we’re wrong. They did have deja vu, but I think they didn’t. All of which just means we need some objective measure of deja vu – a reaction time difference, a neural signature, a cost to processing in a secondary task, anything. Until then it’s open season on deja vu research: it is, after all, pretty easy to get someone to say they have had a deja vu experience.
Finally, it is perhaps now that we realise why it is so important to study deja vu. It is such an intensely subjective and strange experience that to master the measurement and induction of the experience will be very challenging for current paradigms and will help us generate better tools and procedures for cognitive psychology more generally. But I would say that. I’m a deja vu researcher.
* Please note: other more objective paradigms are available. I won’t go into those here, but needless to say, as with any domain, over-use of any one paradigm is likely to be misleading. They are very useful and I’ve used them myself.
** “Déjà vu is a feeling of familiarity with a situation (e.g. seeing a word) combined with an awareness that this familiarity is inappropriate (i.e. you know you have not experienced the situation before)”. “The tip-of-the-tongue sensation is the failure to bring a word to mind, combined with partial recall of some of its characteristics (e.g. the starting letter) and the feeling that you will bring it to mind soon”.
*** The original title of the article was indeed ‘Careful what you ask for: etc.,’ But an early reviewer took exception with this turn of phrase and we dropped the pithy aphorism-colon-dry science structure so favoured by all academics.