Those who have witnessed a crime would do best not to tell anyone about it. Contrary to what one might believe, a person’s memory of an event is not improved by retelling the story. Instead, the risk of an incorrect account increases the more the story is retold and discussed.
“The most accurate witness statements come from people who have seen a crime and then write down what happened before they recount it or discuss it with anyone”, says Farhan Sarwar.
However, it is quite unusual for witnesses to do this. On the contrary, many want to immediately discuss what they have seen. One example of how wrong they can be is the eyewitness descriptions of Anna Lindh’s murderer. Those who were there and saw the murderer were in agreement that he was wearing military clothing. When the pictures from the department store’s cameras were examined, it could be seen that he was wearing normal sports clothes.
Farhan Sarwar’s studies show that eyewitnesses are particularly bad at remembering details, such as what the perpetrator was wearing or what weapon was used. On the other hand, they are better at recalling the key events.
A witness who has told his story many times may become increasingly sure of the details of the crime. This could have devastating consequences for a criminal investigation, as the police place great importance on how confident the witness is, says Farhan Sarwar.
But if eyewitness accounts are so flawed, should they be used at all?
Yes, says Farhan Sarwar. Criminals are getting increasingly canny. They rarely leave any clues. Therefore, eyewitness accounts are still the most important thing the police have to go on. However, the police must be aware of how much importance they attribute to them.
Alongside the work on his thesis, Farhan Sarwar has developed a method to measure the likelihood of eyewitness accounts being correct, together with colleague Sverker Sikström. He has written a computer program that uses algorithms to give a reliable percentage figure for how likely it is that an eyewitness really has remembered correctly.
Testing on the method is not yet complete, but when it is ready the program will quickly be able to process a large number of recorded witness statements. He can already see a number of areas for use: courts, police questioning, security services, insurance companies, etc.