This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Table 1 COGNITIVE ABILITIES AND CRIME Cognitive Ability Scores by Offender Classification for Caucasians Test Adolescent limited Life course persistent Cohen's M SD M SD F P d CAT Reading total grade 8.73 2.58 8.28 2.62 5.359 .021* .17 Arithmetic total grade 8.23 2.21 7.85 2.05 5.434 .020* .18 Language total grade 8.15 2.51 7.50 2.49 12.118 .001** .26 Total grade 8.37 2.29 7.91 2.20 7.597 .006** .20 Sample size 597 256 CTTM Language 9.41 1.63 9.06 1.62 7.75 .006** .21 Nonlanguage 9.89 1.41 9.58 1.40 8.86 .003** .22 122 Sample size 565 242 GATB General 99.02 18.70 95.29 16.45 7.535 .006** .21 Verbal 92.90 16.72 89.86 14.74 5.650 .010* .19 Numeric 96.07 20.01 92.20 18.49 6.203 .013* .20 Spatial 108.53 20.76 107.54 18.53 0.382 .537 .05 Perceptual 105.27 18.15 104.93 17.89 0.055 .814 .02 Sample size 525 228 Raven 45.31 7.66 45.19 7.11 .043 .836 .02 Sample size Note. CAT 524 231 California Achievement Test; CTMM = California Test of Mental Maturity; GATB = General Aptitude Test Battery. Degrees of freedom for CAT, CTMM, GATB, and Raven are (1, 851), (1, 805), (1,751), and (1, 753), respectively. *p This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 400 Table 2 DONNELLAN, GE, AND WENK Cognitive Ability Scores by Offender Classification for Hispanics Adolescent Life course limited persistent Cohen's Test M SD M SD F P d CAT Reading total grade 6.71 2.55 6.20 2.40 3.440 .064 .21 Arithmetic total grade 6.78 2.00 6.37 1.94 3.623 .058 21 Language total grade 6.79 2.81 6.10 2.70 5.073 .025* 25 Total grade 6.78 Sample size 133 13 2.32 6.28 2.17 4.135 .043" .22 210 CTTM Language 8.11 1.62 7.91 Nonlanguage 9.05 1.38 8.88 1.50 1.136 .287 .13 1.39 1.038 .309 .12 12 Sample size 122 161 191 GATB General 86.39 15.04 80.97 14.45 9.639 .002** Verbal 81.60 11.59 77.96 11.54 7.041 .008** .31 Numeric 85.16 19.09 78.80 18.03 10.033 .002** .34 Spatial 99.27 18.28 96.18 19.73 1.830 .177 .16 Perceptual 98.66 17.97 92.80 18.93 7.053 .008** .31 33333 .36 Sample size 116 181 Raven 42.30 8.57 41.57 8.14 0.549 .459 .09 Sample size 116 181 Note. CAT California Achievement Test; CTMM = California Test of Mental Maturity; GATB = General Aptitude Test Battery. Degrees of freedom for CAT, CTMM, GATB, and Raven are (1, 341), (1, 311). (1,295), and (1,295), respectively. *p < .05. **p < .01. Hispanic adolescents, then it is reasonable to suggest that individ- ual differences in cognitive ability may not have the same influ- ence on criminal careers. Rather, differences in individual charac- teristics should interact with both the distal and the proximal environments in which they are embedded. This person-context interaction has long been proposed (Caspi & Bem, 1990) but less frequently studied. Hence, person-context interaction should be an important part of research and theory on crime and delinquency. This is an especially important perspective to bear in mind when discussing the sensitive issue of cognitive abilities and crime because it takes into account the complexities of human development. As noted earlier, there were large numbers of individuals who did not fit into either the adolescent-limited or the life-course persistent offender category. Although the number classified as "other" in our study might be considered too high, it is comparable to the number not classified (58%) in one of Moffitt et al.'s studies comparing childhood-onset and adolescent-onset antisocial indi- viduals (Moffitt, Caspi, Dickson, Silva, & Stanton, 1996). This fact suggests that a two-category system may not capture all the variations in criminal careers. Indeed, recent work by D'Unger, Land, McCall, and Nagin (1998) and Nagin, Farrington, and Moffitt (1995) suggests that a simple dual taxonomy may not adequately reflect the diversity of patterns in criminal careers. Using data from three studies of crime and delinquency conducted in London, Philadelphia, and Racine, Washington, they were able to identify at least four consistent categories of criminal careers: a nonoffending pattern, a low-rate chronic offending pattern, a high- rate chronic offending pattern, and an adolescent-limited pattern. Future work will need to investigate the role that cognitive abilities may play in these more complicated categories. We also noted interesting and important ethnic differences in the classification system used here. Hispanics and African Americans had higher incidences of life-course-persistent offenders within their ethnic groups (66.3 and 61.7%, respectively) than did Cau- casians (30.2%). This observation, in conjunction with the results reported here, indicate that ethnicity itself may play a larger role in life-course-persistent offending than suggested by Moffitt's (1993a) proposal (we thank an anonymous reviewer for this ob- servation). Future work will need to address the undoubtedly complex role that ethnicity plays in the development and mainte- nance of criminal behavior over the life course. It is possible that taxonomies that are developed to apply to some ethnic groups will not apply to all ethnic groups. A few caveats should be noted when considering the findings reported here. First, this study did not control for the role of SES. SES may play a role in the relationships under investigation. Unfortunately, this information was not available, and we have no foolproof method to reconstruct this variable. However, the lack of consideration of SES probably would not have significantly changed the impact of offender classification on cognitive abilities in the analyses reported here. A more likely scenario is that the ethnic differences on test scores would have been mitigated by the inclusion of SES. In addition, it is not clear that our sample This article is intended solely for the p I is not to be disseminated broadly. COGNITIVE ABILITIES AND CRIME Table 3 Cognitive Ability Scores by Offender Classification for African Americans Adolescent limited Life course persistent Test M SD M SD F Cohen's d CAT Reading total grade 6.63 2.49 6.30 2.37 1.970 .161 .14 Arithmetic total grade 6.59 1.69 6.39 1.82 1.292 .256 .11 Language total grade 6.19 2.50 6.05 2.60 0.315 575 .05 Total grade 6.51 2.06 6.30 2.07 1.015 .314 .10 *=30 Sample size 151 767 291 CTTM Language 7.99 1.41 8.04 1.36 0.127 .721 -.04 Nonlanguage 8.80 1.29 8.93 1.32 1.081 .299 -.10 Sample size 147 279 GATB General 80.39 14.47 80.17 13.16 0.025 .875 .02 Verbal 79.62 11.04 77.86 10.92 2.314 .129 Numeric 77.81 17.95 77.24 16.98 0.095 .758 Spatial 92.35 20.24 92.20 17.58 0.006 .941 485 .16 .03 .01 Perceptual 89.52 20.27 90.63 18.51 0.302 .538 -.06 Sample size 135 266 Raven 40.55 8.37 40.48 9.50 0.005 946 10 .01 Sample size 127 244 Note. CAT California Achievement Test; CTMM = California Test of Mental Maturity; GATB = General Aptitude Test Battery. Degrees of freedom for CAT, CTMM, GATB, and Raven are (1,440), (1, 424), (1,399), and (1, 369), respectively. contained appreciable variability in SES, given that it was com- posed of delinquent juveniles. A second limitation of this study is that the arrest record follow-up included arrests only from the state of California. It is possible that this method overlooked some out-of-state arrests. On the other hand, it is not unreasonable to assume that a majority of the individuals in our sample stayed in California for long periods of time. In fact, because of financial and logistical constraints, few studies of crime prediction use arrest records outside of a single state. Accordingly, this study is comparable to most other studies of crime and delinquency. A third caveat concerns the operationalization of Moffitt's (1993a) taxonomy in this investigation. Although we believe that we have captured the essence of Moffitt's taxonomy in our clas- sification system, alternative operationalizations of life-course- persistent and adolescent-limited offenders might have different results. Finally, the archival nature of this data set does place limitations on the generalizations that can be drawn from our findings. Future research should attempt alternative operational definitions of Moffitt's taxonomy and use longitudinal samples that are expressly designed to empirically test the developmental taxonomy. In sum, the results reported here are consistent with Moffitt's (1993a) proposal for Caucasians and Hispanics but not for African Americans. As such, it is important to emphasize the role of context and process in theory and research on crime and delin- quency. Individuals become (and remain) criminals as a result of complex developmental processes. One single factor, such as cognitive ability, will not determine an individual's fate; however, it may increase or decrease an individual's chances and length of criminal involvement in certain contexts. In that regard, cognitive ability is a potentially important contributor to criminal behavior. However, we wish to emphasize that cognitive ability is by no means the only contributor to criminal behavior, nor is it even the most important contributor. In closing, we suggest that viewing criminal career development from a probabilistic standpoint that pays close attention to process and context will help to advance this field of study. Accordingly, future research should move to the empirical study of the processes of how, when, and where cognitive abilities are related to crime and delinquency. Future research should also pay more attention to the role that ethnicity plays in the development of criminal careers. The search for types of criminal offenders will probably not end up "carving nature at its joints," but it may serve important heuristic purposes in guiding research and theory. In that spirit, exploring the mechanisms and contents of typologies may ultimately im- prove scientific understanding of crime and human development. 401 100