Self-report surveys are among the most widely used methods of collecting data on delinquent behaviour, despite persistent concerns over systematic bias. Further investigation into this area is necessary, because the increasing difficulty of obtaining detailed official criminal records is making researchers increasingly reliant on self-reports. Therefore, this analysis investigates the validity of self-reported conviction data in a community sample of young adults who are the offspring of the original male participants in the Cambridge Study in Delinquent Development. The analysis revealed a high level of concurrent validity, although those who had several convictions or convictions for serious offences were more likely to under-report them, whereas older offenders were less likely to over-report. Those living with their parents were less likely to under-report and more likely to over-report their offences. The predictive validity analyses demonstrated that self-reports significantly predicted subsequent convictions for property, violent, drug and driving offences. Users of self-report data need to take these results into consideration.