An experiment was conducted to investigate ways of increasing the likelihood of an offender confessing. Ninety participants were asked to commit a mock-crime that involved them stealing a wallet. Later the mock-offenders were presented with evidence from a witness who was said to have seen the offence. Participants were randomly assigned to one of three conditions in which they were presented with a witness statement that either contained detailed correct information, not-detailed correct information, or incorrect information. Further, half in each condition were told the witness had identified them, while the other half were told the witness had not identified them. Participants were asked about their likelihood of confessing to the crime, and were asked what had influenced their decision. The results indicate that incorrect information made offenders less likely to confess, while being identified or not, had no impact. The results are discussed in relation to police interviewing techniques.