Date of Award


Level of Access Assigned by Author

Campus-Only Dissertation

Degree Name

Doctor of Philosophy (PhD)




Geoffrey Thorpe

Second Committee Member

Douglas W. Nangle

Third Committee Member

Jeff Hecker


Although firesetting behavior has long attracted the interest and attention of psychologists, it has relatively rarely been subjected to empirical investigation. Most recent research has been disjointed focusing on either firesetting as a type of conduct problem (CP) or as a special type of behavior or syndrome requiring unique assessment and intervention. The present study sought to bridge this gap by examining violence and firesetting-specific risk factors drawn from the literature in samples of juvenile offenders who have engaged in firesetting and/or other conduct problems/criminal behaviors. To this end 138 cases were selected from the Maine State Forensic Service evaluations of juveniles who had been charged with arson and other crimes. Data on risk factors was coded from the evaluations using a Coding Manual that included firesetting risk-specific items and items from the Structured Assessment of Violence Risk in Youth (SAVRY; a comprehensive assessment of violence risk comprised of items drawn from the literature; Borum et al., 2006).

Results demonstrated that these items could be reliably coded from the forensic evaluations according to the procedures of the Coding Manual with specific training in its use (Kappa statistics of over .4 obtained for all violence and firesetting-specific risk items). Exploratory factor analysis to address the issue of unidimensionality for modem test analyses revealed a three factor solution best fit the violence risk items: Disobedience/Overt CP, Family Dysfunction and Self-Harm/Suicidality, and Nonviolent/Covert CP. Firesetters (juveniles charged with arson or having a history of firesetting) and general offenders (juveniles charged with other offences and no firesetting history) did not differ on any of these subscales. Modem test theory analyses indicated that all of the code-able violence risk items were significantly related to the implied latent trait of violence risk and suggested that some items might be better able to discriminate juvenile defendants at varying levels than others. Notably although a history of firesetting behavior was related to the implied latent trait of violence risk, its discriminating ability (within a sample of severe CP) was low and did not appear to meet adequate levels. Limitations and future directions are discussed.