State-dependent memory: Remembering Heather Graham's phone number

A few days ago I gave a talk at the Cambridge Memory Film Festival, introducing some of the scientific themes raised in the Hollywood comedy, The Hangover, starring Bradley Cooper and Heather Graham.  Here's a brief summary.

“Why is there a tiger in the bathroom?”

The basic idea of the film is that Doug and a group of his best friends are in Las Vegas drunkenly celebrating the fact that he is soon to be married.  The next morning, Doug’s friends wake up in their hotel suite with no memory of the previous night, and soon realise that Doug is missing.  Furthermore, there is a baby in the wardrobe, a tiger in the bathroom, and a chicken is wandering around the suite.  Hilarious consequences ensue.

The memory-impairing effects of alcohol are a staple topic for light-hearted treatment in popular culture, probably because most people can relate to the notion of waking up after a night on the tiles, not entirely sure about their recollection of everything that transpired.  In psychology, this phenomenon is termed state-dependent memory.

One aspect of state-dependent memory, and the idea that the plot of The Hangover is based on, is that if you experience an event in one physiological state (e.g., drunk), your memory for the event is likely to be impaired if you later try to remember it in another physiological state (e.g., sober).

A further aspect of the phenomenon can be illustrated by imagining another scenario.  Imagine that last night you were out drinking with your friends, and you happened to strike up a conversation with the Hollywood actress, Heather Graham (or, if you prefer, the Hollywood actor, Bradley Cooper). (This sort of scenario occurs quite often in Cambridge, incidentally.  Feel free to apply to work with us here).

You vaguely remember Heather (or Brad) leaning over to you at one point in the evening and whispering her (or his) phone number in your ear.  The next morning, you wake up feeling terrible, but remember the conversation and decide that you must immediately call her (or him).  However, of course, you cannot now for the life of you remember the number.  The key question is this: Will you be more likely to remember the number, and save your chances of a date with Heather (or Brad), if you drink a whole lot more alcohol?

The evidence

Believe it or not, despite the fairly limited likelihood of the average psychologist ever getting Heather Graham’s phone number, let alone needing to know how best to remember it, there are hundreds of studies out there that have investigated the issue.  One of the first was by Donald Goodwin and colleagues, published in Science in 1969, who asked male volunteers to perform memory tasks that involved learning and remembering words while either sober or under the effects of alcohol.

Data from Goodwin et al. (1969)

As displayed in the figure, Goodwin et al. found, as would be expected, good retention of the words if the volunteers had been sober at learning and sober at recall.  Perhaps unsurprisingly, if volunteers were sober at learning but drunk at recall, their memory was relatively impaired.  And, as illustrated in The Hangover, volunteers who were drunk at learning and sober at recall were also amnesic.

The really interesting finding was that the group of volunteers who were intoxicated at learning (and we’re talking a mean of 111 mg of alcohol per 100 ml of breath, or roughly 3 times the UK drink driving limit), and were similarly inebriated during the retention test, nevertheless recalled a comparable amount to those who had been sober on both occasions.  As Goodwin et al. concluded, the results indicated “that learning which the subject acquires while he is intoxicated may be more available to him while he is intoxicated than when he is sober.”

This striking result encouraged a whole assortment of follow-up studies, seeking to determine whether the effect of alcohol on memory could be generalised to other physiological states.  Just to take a few examples, Kelemen and Creeley showed that drinking coffee at learning and recall led to just as good memory as drinking a placebo drink on each occasion, both of which were significantly better than if there was a change of drink between phases.  Kenealy demonstrated similar results by playing music to volunteers to induce a happy or sad mood at learning and recall.  In a famous study, Godden and  Baddeley showed that the same effect could be elicited in deep sea divers who learned and subsequently tried to recall information either on land or 20 ft under water.  Finally, Grant et al. found that noise while studying might not subsequently impair memory if testing occurred in a noisy environment, but that if, for example, an exam was to take place in a quiet exam hall, revising with music or other noise in the background might not be the most sensible policy.  You can find other tips from psychology research for effective exam studying in my previous post here.

What’s going on?

To explain these phenomena, Endel Tulving proposed the Encoding Specificity Principle, according to which memory performance depends on the similarity between the information comprising a memory trace and the information available at recall.

The encoding specificity principle

When we are encoding an event into memory, the memory trace is made up of details about the event (who was there, what they said, etc), but also of the context in which the event occurred.  Context in this sense is a broad term, encompassing elements such as where and when the event happened, who else was there, and also thoughts and feelings we had while experiencing the event.  These internally-generated thoughts and feelings are likely to be influenced by many factors.  For example, was it dark or light, warm or cold, noisy or quiet, were we happy or sad, drunk or sober, etc.  All these elements are bound together to form the memory trace relating to that event.

What Tulving realised was that this isn’t the whole story, however.  According to the encoding specificity principle, the context we are in when we try to retrieve a memory can also have a substantial impact on our likelihood of successfully accessing the correct memory trace.  Specifically, the chances of retrieval success are directly determined by the overlap between the encoding and retrieval contexts.  Thus, bizarre as it may seem, if we are drunk at encoding, our subsequent memory will be more successful if we are also drunk at retrieval.

Data from Park & Rugg (2008)

Brain imaging evidence from Park and Rugg supports the notion that memory performance depends, at least in part, on the overlap between processing operations at encoding and retrieval.  They had participants learn everyday objects (e.g., apple) that were presented either as words or as pictures of the objects.  At test, participants were asked to distinguish between studied and non-studied items.  Each studied item was presented in a form that was either congruent (e.g., word at study and test) or incongruent (e.g., word at study, picture at test).  Park and Rugg found that the highest memory success occurred in the congruent conditions, and was associated with overlap between the brain areas activated during learning and retrieval.

Memory as a reconstructive process

What this all means is that it’s not just what’s happening during the encoding of a memory that determines what we remember of an event.  Our memories can be influenced greatly by factors at the time of retrieval.  In this way, as William James noted over a century ago, memory “retrieval” is something of a misnomer.  Remembering an event is not like picking a DVD off the shelf and re-playing it, but involves a reconstructive process.  We store assorted sensory elements of an event, but to experience the subjective “re-living” of that event, we must construct a narrative structure at the time of retrieval that incorporates all the stored elements in a plausible, satisfying way.

The fact that this narrative construction occurs at retrieval, and is thus subject to influence from our biases and expectations at the time of retrieval, is supported by a great deal of experimental evidence.  Among the most famous is a series of studies by Elizabeth Loftus, who investigated the effect of leading questions on eyewitness testimony.  She had participants watch films of various crime scenes, such as a car accident, and then asked them to recall details of the event.  She found that the form of the question could have a considerable influence on the way the event was remembered.  For example, if participants were asked “How fast were the cars going when they smashed into each other?”, they recalled the speed as significantly faster than if the word “contacted” was used instead of “smashed”.

For more on the nature of reconstructive memory, see this excellent blog post by Mo Costandi.

The implication of this, though, is that if you and some friends do decide to try to remember what happened last night by getting drunk again the next morning, there is a good chance you will all remember it differently.