This thesis investigates the application of crime linkage to sexual assaults with an adult sample where the offender is a stranger to the victim. Chapter 1 gives an overview of crime linkage and sexual offending. Additionally, it examines the theories that crime linkage is based on, with focus on situations that may impact on the effectiveness and usefulness of crime linkage. Chapter 2 compares serial and one-off sexual offenders; finding that there is almost little difference between the behaviours during an offence. Chapter 3 tests the two basic assumptions of crime linkage; behavioural consistency and distinctiveness. Using a larger sample than most current research, and then extending the dataset to include one-off offenders, excellent levels of predictive accuracy were achieved, thus providing support for the assumptions underpinning crime linkage. Chapter 4 test if the stage during the sexual offences has an impact on the predictive accuracy of the behaviours. However, the complete sample of behaviours still performed best. Chapter 5 investigates the performance of five different similarity coefficients with crime linkage analysis. None of the coefficients significantly outperformed the current one. Chapter 6 concludes the thesis, summarises the findings, and discusses the implications on both academic and practical applications of crime linkage.