BACKGROUND Persistent offending has been a topic of interest in the criminal justice literature for some time. Despite its popularity, the concept remains without a consistent or agreed upon definition. For example, persistent offending has been variably defined as offending before and during adulthood (e.g., Bergman & Andershed, 2009), frequent offending (e.g., McGloin & Stickle, 2011), or an early age of first offence (e.g. Hay & Forrest, 2009). In turn, inconsistent methods have been used to operationalise and identify persistent offenders, such as offending before and after the age of 21 (Farrington, Ttofi, & Coid, 2009), the five percent most frequent offenders (Piquero & Lawton, 2002), and multiple offences committed before the age of 14 (Hagell & Newburn, 1994). The definitional inconsistency surrounding the concept of persistent offending poses a significant threat to the generalisability of research, accuracy of theory, and efficacy of policy and interventions. Although previous authors have highlighted their concerns regarding the methodological inconsistencies pertaining to research on persistent offending (Hagell & Newburn, 1994; Piquero, 2009), these concerns seem to have fallen on deaf ears. AIMS This dissertation argues that the use of inconsistent definitions and operationalisations of persistent offending are contributing to the inconsistent findings and competing explanations on the phenomenon. Therefore, more conceptual discussions and empirical observations drawing attention to the ramifications of this issue, as well as methods for rectifying the problem, are needed. Through a series of published and unpublished papers, this dissertation attempts to meet this need by: (1) Highlighting the prevalence of inconsistent definitions, operationalisations, and measures of persistent offending in the literature, and the consequent need for consistency; (2) Empirically demonstrating the flaws associated with these inconsistencies, and; (3) Proposing how to best define, operationalise, and measure persistent offending. The underlying position of this dissertation is that conceptually, persistent offending is best defined and measured by the duration of the criminal career. The arguments and empirical findings in this dissertation support this premise. DATA AND ANALYSES This dissertation uses data from the Cambridge Study in Delinquent Development (CSDD). The CSDD is a longitudinal, population-based study that, to date, has observed the development of offending behaviours in 411 South London males from the age of eight to 56. Boys were interviewed in school at ages eight, ten, and 14 years. Conviction data was recorded annually from the age of 10 to 56. All offences leading to a conviction, excluding minor offences such as traffic infractions and public intoxication, were included in the analyses. The vast majority of men (91 percent) were at risk of conviction at 56 years of age. Chi-square, multiple, and logistic regression analyses were used to examine the association between childhood risk factors and conviction frequency, criminal career duration, and different offending pathways. Descriptive statistics and odds ratios were conducted in order to examine the overlap in the number of offenders identified by different operationalisations of persistent offending. Finally, Pearson’s and partial correlation were used to examine the relationship between criminal career duration, conviction frequency, and age of first conviction. RESULTS Five key findings are reported in this dissertation. First, reviews of the published literature indicated that many of competing empirical findings can be attributed to the use of different definitions and operationalisations of persistent offending. Second, competing measures of persistent offending (i.e., criminal career duration and conviction frequency) are associated with different types and numbers of childhood risk factors. Indeed, not only did offenders with the longest criminal careers have fewer childhood risk factors than offenders with the most convictions, but the childhood risk factors associated with these offenders did not differ to those experienced by one-time offenders. Third, depending on the key measure used, different operationalisations of persistent offending generally identify vastly different offenders as persistent. Fourth, when controlling for offence frequency, onset age is not associated with criminal career duration. Finally, persistent offenders identified by the duration of the criminal career tend to have the longest criminal careers, a more normative age of onset, and vary in their conviction rates. CONCLUSION The collective results of this dissertation support the idea that persistent offending should logically and consistently be defined and measured by the duration of the criminal career. More so, it is proposed that persistent offending should be defined as a criminal career that exceeds the average duration for a criminal career in a population or offender based sample. Nonetheless, a fundamental limitation of proposing a specific definition of persistent offending is that, due to the concepts ambiguity, there are no clear right answers. It may therefore be some time before there is accord on how to define and identify this phenomenon. Nevertheless, it is hoped that, if nothing else, the arguments and findings in this dissertation will spur more scholarly discussions, and help pave the way towards establishing a consistently used definition of persistent offending.