Risk Factors for Institutional Misconduct Among Incarcerated Women: An Examination of the Criterion-Related Validity of the Personality Assessment Inventory

Abstract

Although widely researched in male offender samples, relatively little is known about the clinical utility of the Personality Assessment Inventory (PAI; Morey, 1991) among female prisoners. In this study, we examined the utility of various theoretically relevant PAI scales to predict 3 types of institutional misconduct (general, aggressive/defiant, and covert infractions) in a sample of 113 female inmates incarcerated for at least 1 year. The Antisocial Features (ANT) scale was the most consistent and effective predictor of misbehavior, with limited evidence to suggest that other PAI scales could demonstrate any incremental validity beyond this measure. More important, ANT continued to be associated with institutional misconduct even after controlling for criminal background variables such as prior convictions and a history of violence.

Notes

¹The background information, descriptive statistics, and analyses we report in this article considerably expand on the information reported in the PAI Interpretive Report for Correctional Settings manual for female offenders in several regards. For example, in this article, we address the theoretical and empirical link between trauma and antisocial behavior and report analyses in which we examine this relationship in regards to the PAI and institutional misconduct. We also provide more extensive incremental validity analyses for this particular sample of women by examining the predictive utility of the PAI beyond important criminal history variables known to be associated with institutional adjustment problems. As well, in this article, we examine an additional category of disciplinary infractions, termed “covert” misconduct (described later), which is theoretically and empirically linked with antisocial personality and psychopathy.

^a n = 112.

^bGiven in percentages.

*p < .05, two-tailed.

**p < .01, two-tailed.

²Length of incarceration did not affect the relationship between any of the PAI indicators and outcome categories we describe following.

³In addition to the significant total score correlations reported in Table 2, significant associations also were obtained for the ANT subscales in relation to the three classes of infractions and for some BOR and AGG subscales in relation to general infractions. ANT–A predicted all three categories of infractions (r _pb = .30, p < .01; r _pb = .22, p < .01; and r _pb = .18, p < .05 for general, aggressive/defiant, and covert, respectively). ANT–E was associated with general (r _pb = .24, p < .01) and aggressive/defiant (r _pb = .24, p < .05) infractions as was ANT–S (r _pb = .20 and .23, for general and aggressive/defiant, respectively). In addition, one of the AGG subscales(AGG–V)was significantly associated with general infractions (r _pb = .23, p < .05), as were two of the BOR subscales (BOR–N and BOR–S), r _pb = .21 and .22, respectively, p < .05.

⁴ Kraemer et al.'s (2003) use of the qualifier “typical” is intended to underscore the fact that effect sizes in applied research are relative.

⁵Although ARD–T was predictive of covert infractions at the bivariate level (Table 2), it was no longer associated with this criterion measure when entered into the model after the demographic/criminal history variables.

^aUnstandardized regression coefficients.

⁶It is noteworthy that unlike the institutional data reported for the same sample, among the ANT subscales, only ANT–E was significantly associated with recidivism in this sample (r _pb = .27), although a modest nonsignificant effect was reported for ANT–A (r _pb = .20). Such subscale differences across studies we believe bolster the argument that the ANT total score—rather than any of the individual subscales—is the most robust and generalizable indicator of socially deviant conduct on the PAI.