Challenges in the Assessment of Aggression in High-Risk Youth: Testing the Fit of the Form-Function Aggression Measure

Abstract

Recent efforts have focused on disentangling the forms (e.g., overt and relational) and functions (e.g., instrumental and reactive) of aggression. The Form-Function Aggression Measure (FFAM; Little, Jones, Henrich, & Hawley, 2003) shows promise in this regard; however, it is a new measure and its psychometric properties across different populations are unknown. The current study tested the underlying structure of the FFAM using confirmatory factor analysis in male and female high-risk adolescents (n= 381). Results indicated that none of the models tested demonstrated an acceptable fit in either males or females. However, a 6-factor model comprised of pure-overt, reactive-overt, instrumental-overt, pure-relational, reactive-relational, and instrumental-relational subtypes provided an improved fit relative to other models in both males and females. A multi-form, multi-function model equivalent to the model proposed by Little and colleagues (2003) also evidenced a relatively improved fit, highlighting the utility of disentangling form from function when examining aggression. Implications and challenges for assessing the forms and functions of aggression among high-risk adolescents are discussed.

Keywords:

• aggression
• adolescents
• self-report
• confirmatory factor analysis
• sex difference

Funding for this study was provided by the Canadian Institutes of Health Research (CIHR) New Emerging Team Grant (#54020) awarded to Dr. Marlene M. Moretti, a Social Sciences and Humanities Research Council of Canada (SSHRC) Postdoctoral Fellowship awarded to the first author, and a Social Sciences and Humanities Research Council of Canada (SSHRC) Doctoral Fellowship awarded to the second author. We would like to thank Marianne Brown, Mandi Burnette, Jessica Busch, Rosalind Catchpole, Preeti Chauhan, Kimberley DaSilva, Sara Driskell, Dharam Friedberg, Carole Jun, Jessica Klaver, Emily Marston, Jessica Meyer, Ingrid Obsuth, Jessica Owen, Melissa Funmilayo Ozim, Maya Peled, Amy Saltzman, and Kim Schloss for their assistance with data collection.

Notes

¹This sample also includes an additional group of males (n= 74, 31%) that were tested with a small subset of the larger protocol and were not part of the matching protocol.

²Males were not recruited in this sample as the focus was on female juvenile offenders.

³Although the ethnic composition of the Canadian and American offender samples varied, there were few notable differences between the samples. There were significant mean-level differences between Canadian versus American female offenders on three of the FFAM subscales: Canadian female juvenile offenders scored higher on the reactive-overt, instrumental-overt, and reactive-relational subscales. Similarly, although the Canadian sample comprised adolescents from both custodial and mental health settings, there were few differences between these groups. There were no significant mean-level differences between the custodial and mental health females on any of the FFAM subscales. However, males from the custodial setting scored higher on the reactive-overt and instrumental-overt subscales compared to males from the mental health setting.

⁴We did not test Model 1 because this model demonstrated the lowest relative fit indices. In addition, we did not test Model 3a because Models 2 and 3a were statistically equivalent when the models were split into overt and relational aggression. Finally, we did not test the composite model (i.e., Model 3b) because there were too few unique parameters to estimate model fit once split according to the form of aggression.