Popular perceptions about the recidivism of homicide offenders are contradictory, varying from one extreme - that such offenders rarely commit further violent offences - to the opposite, where it is thought that they remain at a high risk of serious reoffending. The present study draws on the records of 1088 persons arrested in Western Australia over the period 1984-2005 for domestic murders and other types of homicides (robbery and sexual murder), including attempted murder, conspiracy to murder, manslaughter (unintentional homicide) and driving causing death. Our database provides up to 22 years follow-up time (for those arrested in 1984) and accounts critically for the first and any subsequent arrests, if they occur. Of the 1088 persons, only 3 were subsequently arrested and charged with a homicide offence event in the follow-up period. Among those arrested for a murder and subsequently released, we estimate a probability of 0.66 (accounting for censoring) of being rearrested for another offence of any type. The corresponding probabilities for those originally arrested for manslaughter or for driving causing death were equal, at 0.43. A dynamic analysis of the longitudinal data by survival analysis techniques is used to reliably estimate these probabilities. Having a prior record increased the risk of re-arrest; for example male non-Aboriginals arrested for murder with at least one prior arrest have an estimated probability of 0.72 of being rearrested for another offence of any type. Their estimated probability of being rearrested for another serious offence was 0.33. These findings should be of interest to courts and correctional agencies in assessing risk at various stages of the administration of criminal justice.