Eyewitnesses often wrong: Swedish study

Eyewitnesses cannot always be trusted, a psychologist at the Lund University has revealed in a recent study.

• Swedish breakthrough could slow Parkinson's (19 Jan 11)
• Swedish universities prep for tuition fee fallout (1 Dec 10)
• Police baffled by man with 'no identity' (21 Nov 10)

Farhan Sarwar, a witness psychologist at Lund University, disclosed the findings in his dissertation scheduled for defence on Friday.

The more often a witness narrated and discussed a story, the greater the risk of error became, according to Sarwar.

Eyewitnesses are weak at reproducing details such as the attire of criminals or what weapons were used, but they can be used to grasp the main points of a course of events, said Sarwar.

Sarwar's thesis investigated the effects of eyewitness retellings and discussions with non-witnesses on eyewitness memory and meta-memory judgments.

The first study examined the effect of eyewitness discussions with non-witnesses on eyewitness memory and meta-memory realism for the overall information about an event.

The results suggest that discussions of an experienced event may reduce some of the beneficial memory and meta-memory effects from mere retellings, but may not have substantial negative effects compared to a control condition.

"Analysis of the type of questions asked suggests listeners ask more about the peripheral details as compared with the central details," Sarwar said in a statement.

One year later, a follow-up study of participants in the retell condition showed no evidence of memory and meta-memory benefits present in the original final test after about 24 days.

However, participants in the retell condition recalled more accurate items than participants in the control condition.

A second study evaluated the effect of eyewitness discussions with non-witnesses for different forensically central, forensically peripheral and non-forensic information. These are the types of information that police may ask about at the start of a crime investigation.

The results from the two experiments showed that participants had better memory and meta-memory realism for forensically central and non-forensic information than for forensically peripheral information.

Moreover, participants in the four conditions were equally capable of distinguishing between correct and incorrect items.

In addition, in the first experiment, participants in conditions involving retelling and discussing the event reported more items, as well as the number of correct forensically central items, compared to the control condition.

The third study investigated whether retellings and discussions caused more reminiscences and hypermnesia, or an elevated level of memory recall, than mere retelling accounts.

The results showed that discussions cause more reminiscences and hypermnesia over the five sessions as compared to mere retellings. They also revealed that the more times something was repeated over the sessions, the higher the probability was for it being retrieved at the final recall.

"Interestingly, the retelling or discussion of information in an earlier or later session did not predict if it would be reported in the testing session," said Sarwar.

The results showed that forensically peripheral information, but not forensically central information was affected by the reiteration effect, or the effect that confidence tends to increase when a person asserts the same statement many times.

This may be due to peripheral information being less integrated than central information.