Purpose. The release on licence of prisoners who have committed serious violent and/or sexual offences requires rigorous risk assessment and risk management. This study evaluates the ADViSOR project, designed to examine the contribution of prison behaviour monitoring to community supervision of a sample of the highest risk offenders released in England and Wales under Multi-Agency Public Protection Arrangements (MAPPA).Method. The offence-related behaviour of a total group (n = 25) of MAPPA prisoners in one prison, due for release in the following year to two adjacent probation trust areas, was monitored. Their behaviours in the community were followed up for 1 year. A comparison group (n = 36) was formed of the total number of MAPPA prisoners released from prisons nationally to the same two probation trusts.Results. The frequencies of ADViSOR negative behaviours in prison and the community were strongly correlated, r s (25) = .55, p = .004, as were positive behaviours, r s (25) = .56, p = .004. No statistically significant correlations were found either under usual MAPPA processes in the ADViSOR prison or comparison group prisons. The frequency of ADViSOR negative behaviours statistically significantly predicted, with 92% accuracy, the offenders who would reoffend or be recalled to prison (n = 8). Statistically significant similarities in types of behaviour were also identified. Conclusion.Results are discussed in terms of the contribution of behavioural monitoring to risk prediction with high-risk offenders, consistency of cross-situational behaviours, and implications for policy and practice.Prediction of reoffending is a major topic of psychological and criminological research, and research evidence has greatly enhanced the accuracy of offence prediction methodologies. Risk prediction has progressed from an early reliance on clinical judgement