Higher Levels of Intelligence and Executive Functioning Protect Maltreated Children Against Adult Arrests: A Prospective Study

Abstract

Research shows that maltreated children are at elevated risk for arrest as adults and that higher verbal intelligence, reading ability, and executive functioning (abstract reasoning and cognitive flexibility) may be protective against criminal behavior. The current study examines this hypothesis using data from court substantiated cases of child abuse and neglect and demographically matched controls followed prospectively into middle adulthood (N = 1,196). At age 29, verbal intelligence was assessed with the Quick Test and reading ability with the WRAT-R. At age 41, abstract reasoning was assessed with the Matrix Reasoning test and cognitive flexibility with the Trail Making Test-B. Arrest records were gathered from law enforcement agencies through mean age 51. Data were analyzed with binomial logistic regressions. The results indicated that maltreated children were at increased risk for arrest for nonviolent and violent crime. Higher verbal intelligence, reading ability, nonverbal reasoning, and cognitive flexibility were protective against arrest for violent crime. The protective effects of neuropsychological functions were more pronounced for violent than nonviolent crime, for the control than maltreated children, and differed by gender and race. These results suggest that interventions targeting improved cognitive and neuropsychological functions may serve an important role in reducing risk for crime.

Over three million children were referred for suspected child abuse and neglect in fiscal year 2015 and over 700,000 of these children were found to be victims of substantiated child maltreatment in the United States (U.S. Department of Health & Human Services, 2017). Numerous studies have demonstrated that maltreated children are at increased risk for antisocial behavior, including violent and non-violent offending and arrest (Jonson-Reid et al., 2010; Lansford et al., 2007; Malinosky-Rummell & Hansen, 1993; Smith & Thornberry, 1995; Widom, 1989; Maxfield & Widom, 1996). However, it is also clear that not all victims of childhood maltreatment grow up to become perpetrators of crime and violence (McGloin & Widom, 2001). Jaffee, Caspi, Moffitt, Polo-Thomas, & Taylor (2007) have suggested that better neuropsychological functioning may be protective against antisocial behavior for maltreated children. The goal of the current investigation is to examine whether neuropsychological functioning mitigates the relationship between childhood maltreatment and adult non-violent and violent crime. Neuropsychological functions encompass a range of cognitive tasks and the current manuscript focuses on verbal intelligence, reading ability, and executive functions of cognitive flexibility (assessed by set shifting) and abstract nonverbal reasoning.

Childhood Maltreatment and Adult Antisocial and Criminal Behavior

The strongest evidence for the risk child maltreatment poses for antisocial behavior is based on longitudinal and prospective investigations. These studies show that in childhood and adolescence, maltreated children are at elevated risk for oppositional defiant and conduct disorders and engage in delinquency and antisocial acts at higher rates than their peers (Cohen, Brown, & Smailes, 2001; Famularo, Kinscherff, & Fenton, 1992; Jaffee, Caspi, Moffitt, & Taylor, 2004; Jonson-Reid et al., 2010; Lansford et al., 2002, 2007; Manly, Kim, Rogosch, & Cicchetti, 2001; Moylan et al., 2010; Stouthamer-Loeber, Loeber, Homish, & Wei, 2001). These externalizing difficulties persist into adulthood. When followed up into adulthood, previously maltreated children are more likely to meet criteria for Antisocial Personality Disorder than non-maltreated children (Johnson, Cohen, Brown, Smailes, & Bernstein, 1999; Luntz & Widom, 1994). Several studies report a relationship between child maltreatment and criminal behavior (Mersky & Reynolds, 2007; Ryan & Testa, 2005; Thornberry, Henry, Ireland, & Smith, 2010; Widom, 1989), including higher rates of arrests for non-violent and violent crime (Maxfield & Widom, 1996; Widom, 1989a). Overall, reviews of longitudinal (Jaffee, 2017) and longitudinal and cross-sectional literature (Maas, Herrenkohl, & Sousa, 2008) have concluded that the relationship between childhood maltreatment and subsequent antisocial and violent behavior is robust.

However, criminal behavior and arrest for crime are not inevitable outcomes of childhood abuse and neglect. Maltreated children have significantly greater odds, ranging from 1.6 to 2.1, of engaging in criminal behavior as adults relative to controls (Maxfield & Widom, 1996; Thornberry et al., 2010) and 52 % of maltreated individuals do not have any arrest records and 82% do not have records for violent criminal arrest (Maxfield & Widom, 1996). Although there is an increased risk, most maltreated children do not go on to perpetrate criminal acts, leading researchers to consider factors that protect them from engaging in criminal behavior.

Neuropsychological Functioning and Antisocial and Criminal Behavior

The current conceptualization of neuropsychological functions is guided by a modified Cattell-Horn-Carroll (CHC), theory of cognitive abilities (Floyd, Bergeron, Hamilton, & Parra, 2010; McGrew, 2005). The CHC theory includes, empirically-derived, three strata of intelligence that build on each other, including the general cognitive ability (third stratum), broad cognitive abilities (second stratum) and basic skills (first stratum). Here we measure a sampling of broad cognitive skills including, nonverbal reasoning, cognitive flexibility, reading and verbal abilities, which allows us to assess the protective effects of a range of key abilities.

Poor neuropsychological functioning is theorized to play a causal role in the development of antisocial behavior (Moffitt, 1993). Individuals with lower levels of cognitive functioning underperform academically, feel isolated in schools and develop low self-esteem, and gravitate towards antisocial behaviors and deviant peers. They are more likely than other students to fail or drop-out of school and engage in a life of criminal activity (Isen, 2010; Lynam, Moffitt, & Stouthamer-Loeber, 1993; Moffitt, 1993). The protective pathways may differ for the different neuropsychological functions measured in the current manuscript. For example, cognitive flexibility may enable individuals to generate prosocial solutions to conflict whereas reading skills may help individuals excel academically.

In empirical studies, criminal offenders tend to score lower on measures of overall intellectual functioning (Isen, 2010; Lynam et al., 1993). Specifically, poor language abilities, including reading, have been shown to predispose individuals to later antisocial behavior (Donnellan, Ge, & Wenk, 2000; Isen, 2010; O’Cummings, Bardack, & Gonsoulin, 2010; Rucklidge, McLean, & Bateup, 2013). Additionally, meta-analytic studies have reported higher levels of executive dysfunction, including on measures of cognitive flexibility and abstract reasoning, among individuals engaging in antisocial and criminal behavior (Morgan & Lilienfeld, 2000; Ogilvie, Stewart, Chan, & Shum, 2011). Thus, research points to an association between poorer cognitive functioning and antisocial and criminal behavior in adulthood. However, higher levels of neuropsychological functioning have not been extensively examined as protective factors against antisocial and criminal behavior among victims of child abuse and neglect.

The Protective Role of General and Verbal Skills

A number of studies have reported protective effects of intelligence when focusing on childhood adversity broadly defined, not exclusively childhood maltreatment. For example, in a birth cohort of Danish men, boys who were presumed to be at high risk for criminal behavior because they had fathers with extensive criminal histories, but did not grow up to be criminally involved, had high overall IQ and Verbal IQ scores (Kandel et al., 1988). In a German sample of high risk adolescents with extensive adverse childhood experiences, intelligence predicted resilience to behavior problems in cross-sectional and longitudinal analyses over a two year period (Losel & Bliesener, 1994). In a large high-risk sample of Swedish 18 year-old males with histories of childhood adversity and prior behavioral problems, higher intelligence was associated with reduced likelihood of criminal arrest records over the next 10 years (Stattin, Romelsjo, & Stenbacka, 1997). Similarly, in a longitudinal study of US children with high levels of adverse childhood experiences, higher IQ was protective against self- and parent-reported conduct problems in late adolescent/emerging adulthood (Masten et al., 1999). However, few studies have examined the protective role of neuropsychological functions on the criminal behavior of maltreated individuals (Paschall & Fishbein, 2002). In one study, intelligence was protective against negative outcomes for maltreated children assessed in high school (Herrenkohl, Herronkohl, & Egolf, 1994). A second longitudinal study examined the protective effects of intelligence on antisocial behavior of maltreated children following them into middle childhood and found that boys with above average intelligence were resilient (Jaffee et al., 2007).

The Protective Role of Executive Functioning

Researchers have considered the protective role of executive functions in the associations between childhood adversity and negative outcomes from an emotion regulation perspective (Masten, 2001). Executive functions of abstract reasoning and cognitive flexibility contribute to effective regulation of emotional and behavioral responses in stressful situations (Blair & Raver, 2012) and may be associated with reduced aggression and criminal activity (Lösel & Farrington, 2012; Moffitt, 1993). It is also the case the relationship can be bi-directional, where better emotional regulation is associated with better problem solving skills. However, there are no empirical studies known to the authors specifically testing the role of these executive functions in protecting abused and neglected children from engaging in adult criminal behavior. [Notably, studies of individuals with pronounced psychopathic traits have sometimes found them to score higher on measures of executive functions (Sellbom & Verona, 2007)].

Existing studies of more general exposure to adversity and stressful life experiences across childhood and adulthood may provide insight into the protective role of executive functions for maltreated children. In a sample of kindergarten children who were homeless, higher levels of executive functioning skills were associated with fewer aggressive and defiant behaviors (Masten et al., 2012). Similarly, in a different sample of children living in shelters, effortful control (the ability to inhibit impulse in favor of carrying out a different behavior, which requires executive functioning skills of inhibition, set shifting, and focusing of attention) was associated with fewer externalizing skills over and above other important factors (Obradović, 2010). In a study of college students and low-income participants from the community, individuals with more developed executive functioning skills, including cognitive flexibility, reported less hostile and aggressive behavior in response to stress than individuals who exhibited lower executive functioning abilities (Sprague, Verona, Kalkhoff, & Kilmer, 2011).

In sum, executive functioning seems a promising avenue of investigation as a buffer against the negative effects of childhood maltreatment on risk for antisocial and criminal outcomes. Recognizing the paucity of research in this area, Paschall & Fishbein (2002) highlighted the need to study the moderating role of neuropsychological functions in the relationship between childhood abuse and aggression.

Gender and Racial Differences

In general, rates of criminal activity for women tend to be lower than for men, although women have higher rates of minor property crime and men have higher rates of serious person or property crime (Schwartz & Steffensmeier, 2008). Some research has shown that many of the factors that contribute to criminal behavior (i.e., the role of poverty, substance abuse, etc.) tend to be similar for women and men (Islam, Subrata, & Nurjahan, 2008; Darrell Steffensmeier & Allan, 1996), gendered theories of criminality suggest that traditional male and female roles and cultural norms may contribute to different pathways of criminality across genders. Women tend to be more concerned about the preservation of relationships than men and this characteristic is thought to act as a barrier to crime. In addition, it is suggested that the societal subjugation of women may result in fewer opportunities for crime and individual differences among females may not be as relevant for criminal involvement as those of men (Schwartz & Steffensmeier, 2008). The role of neuropsychological function and the type of neuropsychological functions that may protect men and women differently against criminal behavior has not been previously studied, although it is possible that differences may be observed based on the role of cognition in processing of cultural and relationship factors that may contribute to gender differences in criminal activity.

Race differences in criminal arrests have been reported and there is extensive research showing that minorities are disproportionally represented in the criminal justice system (Chilton & Triplett, 2007; Engen, Steen, & Bridges, 2002). However, the protective role of neuropsychological skills in regard to these race differences have not been previously examined. It is possible that a different set of cognitive abilities protects minority and White individuals from arrest, however, no directional hypothesis can be generated based on the limited past literature.

Current Study

The current prospective study of court-substantiated maltreated children and controls followed into middle adulthood builds on existing research and has several advantages. The design includes an unambiguous operationalization of child abuse and neglect and a comparison group of children matched as closely as possible on age, sex, race and approximate social class background. The assessment of risk for criminal behavior, assessed by arrest records, extends into middle adulthood. We have a diverse sample of males and females and African Americans and Whites and examine race and gender differences because the consequences of childhood maltreatment have been shown to differ for individuals from different backgrounds (Widom et al., 2013). We use official arrest history data from multiple levels of law enforcement to provide a comprehensive assessment of criminal histories. Based on evidence that violent antisocial behavior is more likely to be associated with neurocognitive deficits than nonviolent behavior (Barker et al., 2011; Moffitt et al., 2008), we examine whether the protective effects of neuropsychological functioning are different for violent and nonviolent crimes.

In this paper, we focus on three research questions: 1) Does childhood maltreatment predict adult arrest for non-violent and violent crime? 2) Do neuropsychological functions of verbal intelligence, reading ability, abstract nonverbal reasoning and cognitive flexibility buffer maltreated individuals from arrest for non-violent and violent crime in adulthood? 3) Do the protective effects of neuropsychological functions differ by race and gender?

Methods

Design and Participants

The data used here were collected as part of a large prospective cohort design study (Leventhal, 1982; Schulsinger, Mednick, & Knop, 1981) in which abused and/or neglected children were matched with non-abused and non-neglected children and followed into adulthood. Since it is not possible to assign subjects randomly to groups, the assumption of equivalency for the groups is an approximation. For complete details of the study design, see Widom (1989a).

The original sample of maltreated children (N = 908) was composed of all substantiated cases of childhood physical, sexual abuse and neglect processed from 1967 to 1971 in the county juvenile (family) or adult criminal courts of a Midwestern metropolitan area. To capture maltreatment that occurred before adolescence, cases of abuse and neglect were restricted to children ages 0-11 at the time of the incident.

A comparison group of children without documented histories of childhood abuse and/or neglect (N = 667) was matched with the abuse/neglect group on the basis of age, sex, race/ethnicity, and approximate family social class during the time that the abuse and neglect cases were processed. This matching was important because it is theoretically plausible that any relationship between child abuse and neglect and subsequent outcomes is confounded with or explained by social class differences (Adler et al., 1994; Bradley & Corwyn, 2002; Conroy, Sandel, & Zuckerman, 2010; MacMillan et al., 2001; Widom, 1989a). The matching procedure used here is based on a broad definition of social class that includes neighborhoods in which children were reared and schools they attended (Watt, 1972). Shadish, Cook, and Campbell (2002) recommended using neighborhood and hospital controls to match on variables that are related to outcomes, when random sampling is not possible. Children who were under school age at the time of the abuse and/or neglect were matched with children of the same sex, race, date of birth (± 1 week), and hospital of birth through the use of county birth record information. For children of school age, school records were used to find matches with children of the same sex, race, date of birth (± 6 months), class in elementary school during the years 1967 to 1971, and home address. The investigators attempted to find matches for the entire sample; however, matches were found for 74% of the abused and neglected children. The controls represent a group of children with similar demographic characteristics and from childhood neighborhoods and hospitals of birth but who do not have documented histories of maltreatment.

The initial phase of the study compared the abused and/or neglected children to the matched comparison group on juvenile and adult criminal arrest records (Widom, 1989a). Subsequent phases of the study involved interviewing the abused and/or neglected and comparison groups during 1989-1995 (N = 1,196; Mage = 29), 2000-2002 (N = 896, Mage = 40) and 2003-2005 (N = 808; Mage = 41). Law enforcement records were also searched in 1994 (Mage = 33) and 2013-2014 (Mage = 51).

Although there was attrition associated with death, refusals, and our inability to locate people over the various waves of the study, the demographic composition of the sample across the four interviews has remained about the same (see Table 1). The current study uses data from the first and third interviews. The sample is heavily represented by individuals at the lower end of the socioeconomic spectrum: 60% completed high school, 54.9% held unskilled or semiskilled jobs, and only 13.7% held semi-professional or professional jobs (Hollingshead, 1975). Extensive attrition analyses using several methods (e.g. Heckman 2-step method, Structural Equation Models) to assess for selection bias indicate that maltreatment status did not influence participation in later interviews (see Widom, Czaja, & DuMont, 2015, for a detailed description).

Table 1.

Participant characteristics across waves of the study

	Records	Interview 1	Interview 2	Interview 3
		1989 – 1995	2000-2002	2003 – 2005
	(N = 1575)	(N = 1196)	(N = 896)	(N = 808)
	CHARACTERISTICS
Sex N (% male)	776 (49)	614 (51)	439 (49)	382 (47)
White N (%)	1043 (66)	752 (63)	557 (62)	488 (60)
Black N (%)	513 (33)	417 (35)	315 (35)	299 (37)
Other N (%)	19 (1)	26 (2)	23 (3)	21 (3)
Hispanic N (%)	4 (0.3)	45 (4)	36 (4)	32 (4)
Ethnicity N (% White, non-Hispanic)	1043 (66)	763 (64)	568 (63)	499 (62)
Abuse/neglect N (%)	908 (58)	676 (57)	501 (56)	457 (57)
Physical abuse N (%)	160 (10)	110 (9)	81 (9)	78 (10)
Neglect N (%)	697 (44)	543 (45)	403 (45)	369 (46)
Sexual abuse N (%)	153 (10)	96 (8)	72 (8)	61 (8)
Mean age at petition (SD)	6.4 (3.3)	6.3 (3.3)	6.2 (3.3)	6.3 (3.3)
Mean age at interview (SD)		29.2 (3.8)	39.5 (3.5)	41.2 (3.3)

Note: SD = standard deviation. Abuse/neglect= includes physical and sexual abuse and neglect; Mean age at petition= mean age at which maltreatment was identified in records.

Procedures

Participants were interviewed in person in their home or other quiet location of their choosing. Interviewers were blind to the purpose of the study and to the inclusion of an abuse/neglect group. Participants were also blind to the purpose of the study and were told that they had been selected to participate as part of a large group of individuals who grew up in that area during the late 1960s and early 1970s. Institutional Review Board (IRB) approval was obtained for each wave of the study, including the IRB of the City University of New York. Participants provided written or verbal (for those with limited reading ability) informed consent.

Measures and Variables

Childhood Maltreatment.

Childhood maltreatment was assessed through a review of official records processed during the years 1967 to 1971 when children were ages 0-11. Neglect cases reflected a judgment that the parents’ deficiencies in childcare were beyond those found acceptable by community and professional standards at the time. These cases represented extreme failure to provide adequate food, clothing, shelter, and medical attention to children. Physical abuse cases included injuries such as bruises, welts, burns, abrasions, lacerations, wounds, cuts, bone and skull fractures, and other evidence of physical injury. Sexual abuse charges included fondling or touching, felony sexual assault, sodomy, incest, and rape. Eleven percent of the sample experienced more than one type of maltreatment.

Verbal Intelligence.

Verbal skills were assessed during the 1989-1995 interviews when participants were mean age 29. The Quick Test (Ammons & Ammons, 1962), an easily administered measure of current level of verbal intelligence where the subject can point to a picture on a card, was used. The normed Quick Test has an average normed score of 100 (SD = 10) and in the current sample M = 87 (Control: M = 90; Maltreated: M = 84) (Perez & Widom, 1994) . The normative sample was stratified across socioeconomic (SES) backgrounds and has been successfully used with ethnically diverse samples (Vance, Hankins, & Brown, 1988). Quick Test scores correlate highly with WAIS full scale and verbal (r = .79 - .86) IQs (Dizzone & Davis, 1973) and have shown good reliability and validity (Joesting & Joesting, 1972; Zagar et al., 2013).

Reading Ability.

The Wide Range Achievement Test-Revised (WRAT-R Level II) reading test was used to assess reading ability at age 29. The WRAT-R reading test is a reasonably accurate predictor of reading levels and classroom academic achievement as reported by teachers (Webster, Hewett, & Crumbacker, 1989) and has adequate concurrent, content and construct validity (Jastak & Wilkinson, 1984). The test was normed on a diverse sample of over 15,000 individuals, including minorities. On the normed sample, the average score on the WRAT is 100 (SD = 10) and the standard scores of the majority of the current sample ranged from 75-79, indicating Borderline Average reading ability (Controls: M = Low Average; Maltreated: M = Borderline) (Perez & Widom, 1994). Comparisons of the WRAT-R with other achievement and ability tests have yielded correlations in the high .60’s through .80’ s. According to Jastak and Wilkinson (1984), across more than 20 different studies involving a total of 1,000 subjects, the WRAT-R reading test correlated with the Peabody Individual Achievement Test (PIAT) grade level scores in reading recognition at an average of .87. Correlations with the California Achievement Test total reading score showed an average correlation of .81 with the WRAT-R reading. When administered to male adolescent juvenile delinquents, Prewett, Lillis, and Bardos (1991) found that the K-TEA Brief Form (K-TEA-BF) Reading subtest (Kaufman & Kaufman, 1985) yielded standard scores similar to those yielded by the WRAT-R-level II reading subtest.

Cognitive Flexibility.

Trail Making Test B was used to assess cognitive flexibility at age 41. This assessment (Reitan, 1958) requires participants to alternate sequencing of letters and numbers and has been used extensively as an assessment of executive functioning, particularly switching of cognitive sets (Kortte, Horner, & Windham, 2002; Mitrushina, Boone, & D’Elia, 1999; Reitan, 1958). This assessment is widely-used nationally and internationally and has been used with low-SES and minority populations (Lucas et al., 2005; Tombaugh, 2004). The current sample completed the measure in an average of 88 seconds (Controls: M = 80; Maltreated: M = 96), comparing to the 75 second normed average. Trail Making B scores were natural log transformed to correct for high positive skew. In the current study, higher scores generally indicate better performance.

Nonverbal Reasoning.

The Matrix Reasoning test, also administered when participants were 41, is a subtest of the WAIS-III (Wechsler, 1997) to assess non-verbal reasoning and problem solving, requires participants to identify a pattern and complete it with a missing part selected from presented options. The WAIS-III was normed on a nationally-representative sample and included individuals across different races and SES levels. The normed test has an average score of 10 (SD = 1) and in the current sample, the average scaled score was 7.4 (Controls: M = 8; Maltreated: M = 7). Scores on Matrix Reasoning are highly correlated with other assessments of non-verbal reasoning, including the Halstead Category Test subtests (Titus, Retzlaff, & Dean, 2002) and arithmetic subtest of the Luria-Nebraska Neuropsychological Battery (Devaraju-Backhaus, Espe-Pfeifer, Mahrou, & Golden, 2001), and executive functioning assessments (Hill et al., 2010; Zook, Welsh, & Ewing, 2006).

Arrest.

Records from three levels of law enforcement (local, state, and federal) agencies were searched for arrests during 1987–1988 (Widom, 1989a), in 1994 (Maxfield & Widom, 1996), and in 2013-2014 (Widom, Fisher, Nagin, & Piquero, 2017). Official criminal history information was used because it is a reliable assessment of serious offending (Geerken, 1994) and does not suffer some of the limitations of self-report information and based on considerable prior research into the association of maltreatment and arrest (Mersky & Reynolds, 2007; Ryan & Testa, 2005; Widom, 1989) Dichotomous (yes/no) variables were created for any adult (18 years or older) non-traffic arrest, arrest for a nonviolent crime, and arrest for a violent crime, which included arrests for the following crimes and attempts: assault, battery, robbery, manslaughter, murder, rape, and burglary with injury. In order to more accurately differentiate between individuals who perpetrated violent crime and those who did not, anyone with a history of arrest for violence was included in the violent arrest category, irrespective of history of arrest for nonviolent crime. Therefore, the violent and nonviolent arrest groups are mutually exclusive. History of arrest for any (non-traffic, non-status offense) crime while a juvenile (younger than 18 years) was controlled in multivariate analyses.

Data Analytic Plan

Bivariate associations between childhood maltreatment and any arrest, arrest for a nonviolent crime, and any arrest for a violent crime, intelligence, reading ability, nonverbal reasoning and cognitive flexibility were first assessed with Chi-Square and t-tests with SPSS v.24. Subsequently, Hierarchical Binary Logistic regression analyses in SPSS were used to examine whether childhood maltreatment predicted arrest for any crime, arrest for nonviolent crime, and arrest for violent crime in adulthood, and whether neuropsychological functions of verbal intelligence, reading ability, nonverbal reasoning and cognitive flexibility moderated these relationships. Separate regressions were run for each neuropsychological construct and for any arrest, arrest for non-violent crime, and arrest for a violent crime. Step 1 of each regression included participant age, race, sex, and the main effects for childhood maltreatment and one neuropsychological function. Step 2 included the term representing the interaction between child maltreatment and the neuropsychological function. When significant interactions were found, additional regressions were run for the maltreatment and control groups separately to determine the relationship of the neuropsychological function with arrest for each group. Race and gender differences in the protective effects of neuropsychological functions on arrest for a nonviolent crime and violent crime were assessed with separate logistic regressions for Nonwhites and Whites and Males and Females.

To address potential problems with attrition and because the third interview (2003-2005, N = 808) involved a smaller sample than the initial interview (1989-1995, N = 1,196), the regression analyses were rerun in Mplus v. 8 (Muthen & Muthen, n.d.) using the Full Information Maximum Likelihood (FIML) estimator (Enders & Bandalos, 2001). FIML takes into consideration all available data points and allows for estimation of parameters for the sample of 1,196 (see Table 1 for sample characteristics) regardless of missing values.

Results

Table 2 shows the characteristics of the maltreated and control groups on the tests of neuropsychological functioning and risk of arrests and tests for significant differences. Maltreated individuals performed significantly worse on neuropsychological functioning measures and were more likely to be arrested for any crime, nonviolent crime, and violent crime in adulthood. Supplemental Table A shows bivariate correlations among neuropsychological functioning and arrest variables. Notably, the correlations among the cognitive measures ranged from .33 to .69, suggesting considerable overlap among the measures, but evidence for independent constructs as well.

Table 2.

Descriptive statistics for arrest and neuropsychological functions for individuals with histories of child maltreatment and controls

	Total Sample (N = 1196)	Control (N = 520)	Maltreated (N = 676)
ARREST	N (%)			χ2 (df)
Any arrest	609 (51)	226 (44)	383 (57)	20.48 (1) ***
Any Nonviolent arrest	333 (28)	129 (25)	204 (30)	10.01 (1)**
Any Violent arrest	276 (23)	97 (19)	179 (27)	10.14 (1) **
NEUROPSYCHOLOGICAL FUNCTION		M (SD)		t (df)
Verbal Intelligence (age 29)	36.99 (6.00)	38.75 (5.52)	35.64 (6.00)	9.16 (1183) ***
Reading Ability (age 29)	49.36 (16.47)	53.91 (16.26)	45.77 (15.75)	8.68 (1176) ***
Nonverbal Reasoning a (age 41)	14.19 (6.00)	15.45 (5.67)	13.22 (6.07)	5.31 (757) ***
Cognitive Flexibility a. b (age 41)	88.37 (47.59)	79.14 (41.23)	95.66 (50.94)	4.97 (768) ***

Note: Neuropsychological functions are in raw scores

^aNonverbal reasoning and cognitive flexibility analyses are based on a reduced sample size of 792, (controls = 342 and maltreated = 450).

^bCognitive flexibility is based on speed of accurate task completion, and higher scores represent worse performance in the current table. χ² (df) = Chi Square (degrees of freedom); t (df) = t statistic (degrees of freedom); M (SD) = mean (standard deviation)

^***p < .001

^**p < .01

Arrest for Any Crime in Adulthood

Table 3 shows that child maltreatment predicted increased risk for arrest and that higher verbal intelligence and reading ability reduced the likelihood of an arrest in adulthood. There were no significant interactions between group (maltreatment versus control) and verbal intelligence (OR = 1.01, p > .05), reading ability (OR = 1.01, p > .05), nonverbal reasoning (OR = 1.05, p > .05) or cognitive flexibility (OR = 1.21, p > .05) in risk of arrest for any crime in adulthood.

Table 3.

Results of logistic regression analyses examining the roles of child maltreatment and four neuropsychological functions in predicting any crime in adulthood

	Neuropsychological Function
	Verbal Intelligence	Reading Ability	Nonverbal Reasoning	Cognitive Flexibility
	Odds Ratio (95% Confidence Interval)
Age	0.99 (.96, 1.02)	0.99 (.95, 1.02)	1.03 (.98, 1.07)	1.02 (.98, 1.07)
Nonwhite	2.01 (1.54, 2.63) ***	1.92 (1.47, 2.56) ***	2.43 (2.78, 3.33) ***	2.33 (1.64, 3.23) ***
Female	0.29 (.22, .37) ***	0.28 (.22, .37) ***	0.26 (.19, .36) ***	0.25 (.18, .35) ***
Juvenile Arrest	4.18 (2.95, 5.93) ***	4.00 (2.81, 5.69) ***	4.13 (2.64, 6.45) ***	4.49 (2.82, 7.15) ***
Maltreatment	1.57 (1.20, 2.05) **	1.58 (1.21, 2.07) **	1.45 (1.05, 2.00) *	1.51 (1.08, 2.09) *
Neuropsychological Function	0.97 (.94, .99) ***	0.98 (.97, .99) ***	0.99 (.96, 1.01)	0.71 (.50, 1.02)
χ2 (df)	269 (6) ***	279 (6) ***	183 (6) ***	185 (6) ***
Nagelkerke’s R2	0.27	0.28	0.28	0.29

Note: There were no significant interactions between maltreatment and neuropsychological functions. χ² (df) = Chi Squared (degrees of freedom).

^***p < .001

^**p < .01

^*p < .05

Arrest for Nonviolent Crime in Adulthood

Table 4 shows that child maltreatment predicted risk of arrest for a nonviolent crime in adulthood and only reading ability reduced the likelihood of arrest for a nonviolent crime. There were no significant interactions between group (maltreatment versus control) and verbal intelligence (OR = .99, p > .05), reading ability (OR = 1.01, p > .05), nonverbal reasoning (OR = 1.02, p > .05) or cognitive flexibility (OR = 1.40, p > .05).

Table 4.

Results of logistic regression analyses examining the roles of child maltreatment and four neuropsychological functions in predicting arrest for any nonviolent crime in adulthood

	Neuropsychological Function
	Verbal Intelligence	Reading Ability	Nonverbal Reasoning	Cognitive Flexibility
	Odds Ratio (95% Confidence Interval)
Age	0.98 (.96, 1.01)	0.99 (.95, 1.02)	1.03 (.98, 1.08)	1.02 (.97, 1.08)
Nonwhite	1.61 (1.19, 2.17) **	1.54 (1.12, 2.08) **	1.85 (1.28, 2.63) ***	1.79 (1.23, 2.63) **
Female	0.43 (.32, .57) ***	0.41 (.30, .55) ***	0.38 (.26, .54) ***	0.36 (.25, .52) ***
Juvenile Arrest	2.94 (1.99, 4.35) ***	2.78 (1.87, 4.14) ***	2.59 (1.56, 4.30) ***	2.73 (1.61,4.61) ***
Maltreatment	1.55 (1.15, 2.09) **	1.53 (1.13, 2.06) **	1.39 (.97, 1.99)	1.45 (1.01, 2.09) *
Neuropsychological Function	0.98 (.96, 1.01)	0.99 (.98, .99) **	1.00 (.97, 1.03)	0.84 (.56, 1.25)
χ2 (df)	96 (6) ***	100 (6) ***	62 (6) ***	63 (6) ***
Nagelkerke’s R2	0.14	0.14	0.13	0.14

Note: There were no significant interactions between maltreatment and neuropsychological functions. χ² (df) = Chi Squared (degrees of freedom).

^***p < .001

^**p < .01

^*p < .05

Arrest for Violent Crime in Adulthood

Childhood maltreatment predicted increased risk of arrest for a violent crime in adulthood (see Table 5). These new results show that verbal intelligence and cognitive flexibility were protective and reduced the risk of arrest for a violent crime in adulthood. However, there were two significant interactions indicating that the protective role of reading ability and nonverbal reasoning differed for the control and maltreated groups. Analysis of these interactions revealed that reading ability and nonverbal reasoning only significantly protected control participants, (OR = .96, 95% CI (.94, .97), p < .001) and (OR = .88, 95% CI (.82, .94), p < .001) respectively, and not maltreated participants, (OR = .99, 95% CI (.97, 1.00), p > .05) and (OR = 1.02, 95% CI (.96, 1.07), p > .05) respectively, against arrest for a violent crime.

Table 5.

Results of regression analyses predicting any arrest for a violent crime

	Neuropsychological Function
	Verbal Intelligence	Reading Ability	Nonverbal Reasoning	Cognitive Flexibility
	Odds Ratio (95% Confidence Interval)
Step 1
Age	1.00 (.96, 1.05)	0.99 (.95, 1.05)	1.03 (.96, 1.11)	1.03 (.96, 1.10)
Nonwhite	3.03 (2.08, 4.35) ***	3.03 (2.04, 4.35) ***	4.76 (3.03,7.69) ***	4.76 (2.77, 7.69) ***
Female	0.15 (.10, .22) ***	0.15 (.10, .22) ***	0.13 (.08, .21) ***	0.12 (.07, .20) ***
Juvenile Arrest	6.64 (4.37, 10.09) ***	6.46 (4.22, 9.87) ***	8.76 (5.06, 15.18) ***	10.25 (5.71, 18.28)***
Maltreatment	1.54 (1.06, 2.24) *	1.59 (1.09, 2.33) *	1.42 (.88, 2.28)	1.59 (.97, 2.59)
Neuropsychological Function	0.94 (.91, .97) ***	0.97 (.96, .99) ***	0.96 (.92, .99) *	0.50 (.29 .83) **
Step 2
Age	1.00 (.96, 1.05)	1.00 (.95, 1.05)	1.02 (.95, 1.10)	1.03 (.96, 1.10)
Nonwhite	3.03 (2.08, 4.35) ***	3.03 (2.08, 4.54) ***	5.26 (3.23, 9.09) ***	4.76 (2.78, 7.69) ***
Female	0.15 (.10, .22) ***	0.15 (.10, .22) ***	0.13 (.08, .22) ***	0.12 (.07, .20) ***
Juvenile Arrest	6.69 (4.40, 10.17) ***	6.68 (4.35, 10.27) ***	9.77 (5.54, 17.22) ***	10.24 (5.73, 18.29)***
Maltreatment	0.24 (.02, 2.53)	0.39 (.12,1.28)	0.19 (.05, .68) *	1.75 (.02, 187)
Neuropsychological Function	0.91 (.87, .96) ***	0.96 (.94, .98) ***	0.88 (.82, .94) ***	0.49 (.21 ,1.11)
Maltreatment x Neuropsychological Function	1.05 (.99, 1.12)	1.03 (1.01, 1.06) *	1.16 (1.06, 1.26) **	0.98 (.34, 2.81)
χ2 (df)	332 (7) ***	344 (7) ***	255 (7) ***	246 (7) ***
Nagelkerke’s R2	0.45	0.47	0.51	0.51

Note: χ² (df) = Chi Squared (degrees of freedom)

^***p < .001

^**p < .01

^*p < .05

Gender and Race Differences

In analyses of gender differences (Table 6 and Supplemental Table B, C), the results show that the protective effects of verbal intelligence, reading ability, and nonverbal reasoning are significant for males only and significant maltreatment by neuropsychological functioning interactions show that it is the control males who are more strongly protected than maltreated males. Cognitive flexibility, however, protects both maltreated and control females against risk for violent crime arrest. Reading skills protect both maltreated and control males, but not females, against arrest for nonviolent crime. Maltreatment predicted increased risk of arrest for nonviolent crime for females and for violent crime for both genders.

Table 6.

Gender differences in the association of neuropsychological skills with nonviolent and violent arrest

	ODDS RATIO (95% CONFIDENCE INTERVAL)
	MALE	FEMALE
	Arrest for Nonviolent Crime
Group (maltreatment vs. control)	1.49 (.99, 2.23)	1.85 (1.22, 2.80)**
Verbal Intelligence	0.97 (.93, 1.00)	1.00 (.96, 1.03)
Reading Ability	0.99 (.97, .99) *	0.99 (.97, 1.00)
Nonverbal Reasoning	1.01 (.96, 1.06)	0.99 (.95, 1.03)
Cognitive Flexibility	0.93 (.51, 1.67)	0.79 (.46, 1.40)
	Arrest for Violent Crime
Group (maltreatment vs. control)	1.63 (1.05, 2.52)*	2.22 (1.14, 4.30)*
Verbal Intelligence	0.93 (.89, .96)***	0.97 (.92, 1.03)
Reading Ability	0.96 (.95, .98)***	1.00 (.98, 1.02)
Nonverbal Reasoning	0.95 (.90, 1.00) *	0.98 (.92, 1.05)
Cognitive Flexibility	0.57 (.29, 1.12)	0.39 (.17, .80) *
Verbal Int. x Group	1.11 (1.02, 1.21) *	1.00 (.90, 1.10)
Reading Ability x Group	1.03 (1.01, 1.07)*	1.01 (.96, 1.05)
Nonverbal Reasoning x Group	1.21 (1.08, 1.35)**	1.06 (.92, 1.24)
Cognitive Flexibility x Group	1.20 (.31, 4.76)	1.04 (.19, 5.88)

Note: Group = comparison of individuals with child maltreatment histories compared to those without. There were no significant interactions predicting arrest for nonviolent crime. Regressions were run separately by gender and controlled for race, age, and arrest as a juvenile.

^***p < .001

^**p < .01

^*p < .05

In analyses of race/ethnicity differences (Table 7 and Supplemental Table B, C), the results show that among Whites, controls are more strongly protected by higher levels of reading and nonverbal reasoning skills than individuals with histories of child maltreatment against arrest for violent crime. In contrast, for Nonwhites, the protective effects of verbal intelligence and reading ability against arrest for a violent crime did not differ by group (maltreated versus control). There were no race differences in the buffering effect of reading skills on risk of arrest for a nonviolent crime. Maltreatment significantly predicted increased risk for nonviolent crime for Whites and Nonwhites and for violent crime for Whites.

Table 7.

Race differences in the associations of neuropsychological functions with nonviolent and violent arrest

	ODDS RATIOS (95% CONFIDENCE INTERVAL)
	NON-WHITE	WHITE
	ARREST FOR NONVIOLENT CRIME
Group (maltreatment vs. control)	1.78 (1.09, 2.89)*	1.64 (1.14, 2.35)*
Verbal Intelligence	0.99 (.95, 1.03)	0.97 (.94, 1.01)
Reading Ability	0.98 (.87, .99)*	0.99 (.98, .99)*
Nonverbal Reasoning	0.97 (.92, 1.03)	1.07 (.98, 1.05)
Cognitive Flexibility	0.57 (.30, 1.11)	1.07 (.65, 1.78)
	ARREST FOR VIOLENT CRIME
Group (maltreatment vs. control)	1.73 (.99, 3.03)	1.79 (1.11, 2.89)*
Verbal Intelligence	0.94 (.89, .99) *	0.94 (.90, .97)**
Reading Ability	0.98 (.96, .99)*	0.97 (.96, .99)***
Nonverbal Reasoning	0.95 (.89, 1.01)	0.97 (.92, 1.02)
Cognitive Flexibility	0.46 (.21, 1.02)	0.54 (.27, 1.11)
Verbal Int. x Group	1.02 (.92, 1.12)	1.08 (.99, 1.18)
Reading Ability x Group	1.01 (.97, 1.05)	1.05 (1.01, 1.08)**
Nonverbal Reasoning x Group	1.13 (.99, 1.29)	1.19 (1.06, 1.36)**
Cognitive Flexibility x Group	0.35 (.07, 1.85)	3.13 (.68, 14.29)

Note: Group = comparison of individuals with child maltreatment histories compared to those without. There were no significant interactions predicting arrest for nonviolent crime. Regressions were run separately by Whites and Non-Whites and controlled for gender, age, and arrest as a juvenile.

^***p < .001

^**p < .01

^*p < .05

Additional Analyses with FIML

The results of FIML analyses were consistent with results presented above, except that with FIML, cognitive flexibility was protective of both Non-Whites (β = −.12, p < .001) and Whites (β = −.14, p < .05) against arrest for violent crime. (FIML results are available upon request.)

Discussion

Using a prospective longitudinal design, the current study replicated previous findings that childhood maltreatment leads to an increased risk for any arrest and arrest for a nonviolent (for females, minorities and Whites) and violent crime (for both genders and Whites). The current research also examined potential protective factors for maltreated and high-risk children and found that the protective effects of neuropsychological functions are more pronounced for violent than nonviolent crime. These new results add to the body of literature on the protective role of cognitive functioning in at risk children (Jaffee et al., 2004; Masten et al., 1999; Obradović, 2010; Sprague et al., 2011).

Higher levels of verbal intelligence and reading ability protected both maltreated children and controls against arrest for any crime in adulthood; however, when we examined these effects by type of arrest, only reading ability was protective against arrest for nonviolent crime. All four types of neuropsychological functions protect individuals against arrest for violent crime and we found that violent criminal activity was more strongly linked to neuropsychological functioning than nonviolent criminal behavior. These new findings provide some support for the theoretical (Moffitt et al., 2008) and empirical (Barker et al., 2011) work on the development of delinquency and the theorizing that better neuropsychological functioning contributes to emotional and behavioral inhibition and the development of prosocial interpersonal skills and, in turn, these contribute to the reduced likelihood of violence (Barker et al., 2007; Moffitt, 1993). Another possibility is that youth who have higher verbal intelligence, better reading ability and higher executive functions are more likely to stay in school longer and graduate with classmates, characteristics that have previously been found to mediate risk for being arrested as an adult (Allwood & Widom, 2013). Yet a third possibility is that the characteristics of higher verbal intelligence and reading ability are more likely to lead to improved opportunities for employment which would also be associated with decreased risk of arrest (Aaltonen, Macdonald, Martikainen, & Kivivuori, 2013).

Surprisingly, our results showed that the protective effects of reading and nonverbal reasoning were significant only for the controls. That is, the buffering effects of reading ability and nonverbal reasoning skills for violent crime were only found for the control and not individuals with histories of child maltreatment. These findings suggest that individuals without maltreatment histories may be more able to take advantage of their cognitive skills and to engage in fewer violent criminal activities. On the other hand, it is also possible that elevated rates of mental illness (e.g., post-traumatic stress disorder or substance abuse) among the previously maltreated children (Barker et al., 2007; Widom, 1999; Widom, Marmorstein, & White, 2006) interfere with their abilities to utilize their cognitive skills to resolve conflict nonviolently. Particularly because this is a community sample largely composed of individuals from lower socioeconomic status (SES) backgrounds, it is notable that the current findings show that cognitive abilities are protective against arrest even for individuals who are traditionally at higher risk for crime associated with lower SES (Males & Brown, 2014; Osborne, McCord, & Higgins, 2016).

We also found that these relationships differed by gender and race. The effects of verbal intelligence, reading ability, and nonverbal reasoning were protective against violent arrest for males, but not females. Conversely, higher levels of cognitive flexibility showed a large protective effect against being arrested for maltreated and control females. Specifically, one standard deviation increase in cognitive flexibility reduced the likelihood of arrest for a violent crime by 61% for these females. It is possible that these gender differences are associated with distinct biological and/or socialization demands that males and females face during development (Taylor et al., 2000; Zimmermann & Iwanski, 2014). For example, it has been suggested that females are more likely to affiliate with others in times of stress, while males engage in fight or flight behaviors (Taylor et al., 2002). Furthermore, gendered theories of criminal activity suggest that females are more concerned about preserving relationships than males which may protect women against engagement in criminal activity (Schwartz & Steffensmeier, 2008). It is possible that cognitive flexibility contributes to relationship maintenance and is protective against crime for women. Currently, knowledge about the development of delinquency in females is limited and lags behind that of males (Barker et al., 2007; Krupa & Childs, 2014). However, these findings suggest that future research ought to consider potential differences in the consequences of child abuse and neglect for males and females.

Past research has shown that some consequences of childhood maltreatment differ for individuals of different racial backgrounds (Widom et al., 2013). In this study, we examined whether the protective effects of neuropsychological functions on arrests are similar or different for Nonwhite and White maltreated children. In the current sample, the Nonwhite group is comprised mostly of Blacks, with a small group of Hispanics. We found that verbal intelligence protected both Nonwhites and Whites against risk of arrest for a violent crime. However, among Whites, the controls were more strongly protected by their reading ability and nonverbal reasoning skills from arrest for violent crime than those with histories of child maltreatment. These differences between maltreated and control groups were not found among the Nonwhite subsample, where the protective effects of reading ability did not differentiate the groups. While Whites and Nonwhites in our sample showed the buffering effect of better neuropsychological functioning on adult arrest for a violent crime, these results suggest that the impact is complex.

When considered in light of the theoretical framework on the relationship between adversity and criminal behavior of individuals from different backgrounds, these findings do not support the “racial invariance” hypothesis that predicts equal outcomes of maltreatment across individuals of different races (Sampson & Bean, 2006). Nonwhite maltreated individuals have higher rates of arrest in the current study, suggesting that they may be experiencing “double jeopardy” related to being members of minority groups and experiencing child maltreatment (McLeod & Owens, 2004). On the other hand, these findings provide some support for the resilience hypothesis among the Nonwhite maltreated children, in that they are able to utilize their reading ability and verbal skills to reduce the likelihood of arrest (Sampson, 2008; Steffensmeier, Ulmer, Feldmeyer, & Harris, 2010). Research has shown that minorities are treated more harshly by the criminal justice system and it is possible that their verbal skills are helpful in reducing the likelihood of arrest (Chilton & Triplett, 2007; Engen et al., 2002).

Our findings regarding gender and race differences in these relationships should be considered preliminary because of the smaller sample sizes when the data are examined by race and gender and should be replicated. However, it is worth noting that the differences in observed odds ratios (a measure of effect size) were largely consistent with the significance of the findings, suggesting that the results are not simply due to power but represent differences in the magnitude of associations.

The current findings suggest that interventions that promote cognitive development and increased cognitive flexibility for maltreated and non-maltreated children may have positive consequences for decreasing risk for violent criminal behavior. It is notable that cognitive flexibility skills were associated with the largest effect size for reducing violent crime for both White and Nonwhite participants and reduced the risk of arrest for a violent crime for both the control and maltreated groups by 50%. Therefore, efforts should be undertaken to make efficacy-based interventions associated with improved executive functions more accessible to at-risk children (Diamond & Lee, 2011). Cognitive skills develop within the context of families, schools, and neighborhoods and interventions that are trauma-informed and promote better educational opportunities for high risk families may be key in promoting neuropsychological skills and decreasing the likelihood of adult arrest (Ko et al., 2008).

Despite the multiple strengths of this study, some caveats are worth noting. These results are based on cases of childhood abuse and neglect drawn from official court records from the 1960’s and 70’s and most likely represent the most extreme cases processed in the system at the time (Groeneveld & Giovannoni, 1977). This means that these findings are not generalizable to unreported or unsubstantiated cases of child abuse and neglect (Widom, 1989a). Because these cases represent children and families predominantly from the lower end of the socioeconomic spectrum, these findings can also not be generalized to abuse and neglect that occurs in middle or upper- class children, for whom consequences may be different than children in the current study (Widom, 2000). There are a number of other factors that may be contributing to these findings. For example, socioeconomic status may have influenced the development of cognitive functions, particularly since the majority of cases in the study were neglect. Because poverty is closely associated with neglect and we did not specifically examine the role of poverty, it is possible that poverty plays a role as well (see Nikulina, Widom, & Czaja, 2011 for an examination of the role of poverty in estimating consequences of childhood neglect). The court records that were used to operationalize maltreatment do not permit an analysis of characteristics of the abuse or neglect experience (e.g. chronicity or severity). Further, arrest records do not capture self-reported criminal behavior. Finally, it would have been ideal to measure neuropsychological functioning in childhood and adolescence and follow participants prospectively. Although the design of this study did not allow for this type of assessment, neuropsychological functions remain fairly stable into the 20s and early 40s when they were measured in the current study (Broderick, & Blewitt, 2014; Deary et al., 2009).

In conclusion, the results of the current study suggest that higher levels of cognitive functioning are protective against the likelihood of arrest for violent crime for maltreated and control children, although these buffering effects may differ by gender and race. At-risk children may benefit from interventions designed to improve cognitive functioning, which may have long-lasting consequence in the prevention of violent crime in adulthood.