It is quite a different story when it comes to unfamiliar faces. In episode two of Eyewitness most of the contributors were unable to identify the unfamiliar faces of the perpetrators, even though there was hardly any delay between the crime and the VIPER parades (in a real investigation it is normal for the delay to be between 4 to 6 weeks). This tells us that their memory for the faces was not very good. But what if we removed the delay altogether? In other words, do you think you could tell whether two images that you could see at the same time were both of the same person, or of two completely different people? It may seem an odd question, but this is exactly the task involved in checking someone’s photo-identity card.

Dr Nicola Brace from The Open University’s Psychology Department explains a psychological experiment that investigated this very issue.


Copyright The Open University


Nicola Brace: The task facing an eye witness is to try to remember an unfamiliar face. However, even if we present people with a much simpler task that does not require memory, they still find it difficult to deal with unfamiliar faces. For example, checking the photo on a passport or an identity card does not require memory, but as we shall see it can still be a very difficult task. Richard Kemp, Graham Pike and myself carried out a study where we looked at whether putting a photograph on a credit card would reduce the fraudulent use of such cards. Photo credit cards are rare, so if you’ve not seen one, the photographs on them are about the size you see on driving licences.

In our study forty-four students acted as shoppers at a London supermarket, and they attempted to use four different photo credit cards, two of which were legitimate and two fraudulent. The students were aged between eighteen and forty-five and half were male and half were female. Six experienced shop assistants also volunteered to operate the tills. The supermarket was only open to our shoppers. So let’s describe the four credit cards that each shopper was supplied with.

Well, one had a photograph of the shopper similar to how they appeared when using the card, and we called this the unchanged appearance card. One had a photograph of the shopper with a small change made to their appearance, for example, a change to hairstyle, addition, removal of eyeglass or jewellery - the sort of changes that might naturally occur within a few weeks of a photograph being taken, and we called this the changed appearance card. One had a photograph of someone else, but who resembled the shopper in terms of sex, race, age and approximate hairstyles colouring and face shape, and this one we called the matched foil card. And the fourth had a photograph of someone else who was judged to be dissimilar in appearance, except they did match in terms of sex, race and age, and this we called the unmatched foil card.

The shoppers were told to select between ten and twelve items from the store, and attempt to purchase them using one of the four cards, and they were asked to repeat this procedure until they’d used all the cards at least once. On average half of the cards seen by each cashier were valid, ie they had a photograph of the shopper, and half were what we called fraudulent, and they had a photograph of somebody else. The cashiers were told beforehand that the study was looking at how quickly and accurately they could process credit cards that had a photograph, and that they should process the cards normally but also check the photograph to see it was of the shopper. If they wanted to reject a card because they were unhappy about either the signature or the photograph then they should call for a supervisor.

So what did we find? Well we found that the cashier’s decisions were largely correct when they were shown the unchanged cards. In fact, 93.3% of their decisions were correct when they were shown the card of the shopper with the photograph as they appeared on the night. They were also quite good when they were shown the changed cards. This is where they were showing a photo credit card with a photograph of themselves, but there was a slight difference. So there had been a change to their hairstyle, or the addition of eye glasses. In fact, 86.2% of the cards were correctly accepted by the cashiers.

Now what was surprising was when the cashiers were shown the matched foil cards. These are the cards, these are the photo credit cards that they should reject; however, only 36.4% of the cashier’s decisions were correct. So, in other words, they were mistakenly accepting a credit card that had a photograph of somebody else who did resemble the person. But only in terms of sort of hairstyle and face shape, it was not a photograph of the shopper.

Now even when the cashiers were shown the unmatched foil cards, these were the photo credit cards that had a photograph of somebody who didn’t resemble the person, other than in age, race and sex, only 65.9% of their decisions were correct. So overall we found that slightly more that 50% of the fraudulent cards were accepted by the cashiers. In many respects this result was quite shocking, as it showed that the cashiers were often unable to match a photograph to the face when both were right in front of them.

The key reason they found this task so difficult is that the faces were unfamiliar. This task involved no memory what so ever, as both photograph and face were visible at the same time. Think how much harder it would be to recognise an unfamiliar face from memory when you had previously seen it several weeks ago. It is therefore not surprising that eye witnesses often find it very difficult to select the perpetrator from an identity parade.