Differential Relationships Between Personal and Community Stressors and Children’s Neurocognitive Functioning

Abstract

Early adversity can alter development of neurocognition, including executive cognitive and emotional regulatory functions. This is the first study to explore differential relationships between personal (physical and emotional abuse and neglect, school and parental stressors) and community (neighborhood problems and witnessing neighborhood violence) stressors and neurocognition. Predominantly Latino children (n = 553) aged 10 to 12 years completed tasks measuring intelligence, impulsivity, problem solving, cognitive flexibility, decision making, and emotion attributions. Adjusting for age and parent education, bivariate regression analyses found exposure to personal stressors to be associated with relative deficits in at least one neurocognitive function. Community stressors were related to relative deficits in emotion attributions and problem solving. In multivariate analyses, neglect was related to misattributions of emotion and IQ deficits, and physical abuse was related to problem solving. Community stressors were not correlated with neurocognition when viewed relative to personal stressors. Stressor types were differentially associated with performance on specific neurocognitive tasks.

Evidence is accumulating for the association among various types of psychosocial stressors and altered states of neurobiological functioning (Carrey, Butter, Persinger, & Bialik, 1995; Cicchetti & Blender, 2006; De Bellis, 2005; Heim & Nemeroff, 2002; van der Kolk, 1997). Findings suggest that stressful or traumatic experiences may directly affect psychophysiological, neuroendocrine, cognitive, and genetic activities, which can subsequently increase vulnerability to psychopathology, including substance abuse, affective disorders, and aggression (e.g., Anisman & Zacharko, 1986; Fumagalli, Molteni, Racagni, & Riva, 2007; Meaney et al., 1996; Sousa, Cerqueira, & Almeida, 2007). Child maltreatment and other interpersonal types of severe stress may play an especially distinct and significant role in the increased incidence of developmental delays and behavioral disorders because of the severity of social (e.g., impaired relationships, poor school performance) and psychological (e.g., depression, anxiety) dysfunction that results from such maltreatment (Maxfield & Widom, 1996). On the other hand, not all children who experience adversity exhibit these problems, suggesting there is individual variation in their associations with various types of psychosocial stressors. Alterations in neurobiological processes have been implicated as mediators of these effects (Dinan, 1996; Giaconia et al., 2000; Heffernan et al., 2000; Hibbard, Uysal, Kepler, Bogdany, & Silver, 1998; Kendler et al., 2000; Morales & Guerra, 2006; Wills, Sandy, & Yaeger, 2000), which may help to explain these differential developmental pathways (Caspi et al., 2002).

Both neurodevelopment and enduring brain plasticity are dependent on social inputs, providing a plausible basis for understanding putative underlying mechanisms in the relationship among psychosocial stressors, cognitive function, and behavior. Specific associations between a history of adversity and integrity of prefrontally modulated executive cognitive functions (ECFs) may be particularly relevant given the reliance of these neural systems on environmental influences in early childhood and throughout the life span (Enslinger, 1999). In addition, prefrontal functions are intimately interconnected with activity in structures of the limbic system (e.g., amygdala, hippocampus, hypothalamus) to form the basis for integrating motivation, goal-directed behavior, sensitivity to consequences, perception of social cues, and inhibition. Thus, development of prefrontal-limbic circuitry underlying cognitive abilities and their emotional corollaries may be especially affected by adversity given their exquisite sensitivity to psychosocial influences (Bremner, 1999; Bremner & Vermetten, 2001; Critchley et al., 2000; Koenen et al., 2001; Mizoguchi et al., 2000; Skosnik, Chatterton, Swisher, & Park, 2000; Steckler & Holsboer, 1999).

Although several recent studies cited above have focused on stress-related alterations in neurobiological systems, there has been relatively little attention devoted to the possible functional manifestations of stress, particularly involving executive cognitive and emotional regulatory processes. Furthermore, many existing studies include individuals who have been diagnosed with post-traumatic stress disorder (PTSD). Evidence suggests those with PTSD compose a distinctive subgroup of those experiencing trauma, particularly in terms of physiological discriminators. Unfortunately, the relative lack of studies including individuals who have experienced adversity without PTSD compels us to review the PTSD literature for guidance in formulating hypotheses for this initial investigation. Also considered were studies that focus on social and emotional deprivation, which may have more import to measures of neglect in the present study.

Koenen et al. (2001) compared adult participants with PTSD resulting from abuse to controls without an abuse history using several tasks, five of which are sensitive to frontal lobe dysfunction; specifically, measures of working memory, visual spatial attention, and concept learning. Relative to controls, PTSD participants were impaired on frontally mediated tasks, suggesting a specific pattern of ECF deficits, whereas other task performances not specific to the prefrontal cortex were not significantly different between groups. Beers and De Bellis (2002) found children with PTSD performed more poorly on measures of attention and abstract reasoning (a dimension of ECF) than did those without PTSD. Carrey et al. (1995), on the other hand, assessed various physiological responses to emotional and cognitive stimuli in children who were abused but did not necessarily have a diagnosis of PTSD, relative to children who were not abused. Abused children exhibited lower skin conductance responses during all presentations of emotional and cognitive stimuli and also exhibited lower verbal and full-scale IQ scores, suggesting delays in cognitive development and reduced physiological responsiveness to environmental input. Additional studies of individuals who were not selected on the basis of a PTSD diagnosis have specifically focused on the relationship between domestic violence and cognitive development in children, finding strong and consistent support for the negative impact of psychosocial stress on cognitive and school performance (e.g., Huth-Bocks, Levendosky, & Semel, 2001; Kerouac, Taggart, Lescop, & Fortin, 1986; Koenen, Moffitt, Caspi, Taylor, & Purcell, 2003; Wolfe, Zak, Wilson, & Jaffe, 1986).

The present study further examines a relatively unexplored area: that various types of psychosocial stressors (e.g., personal versus community) are differentially related to executive cognitive and emotional regulatory development; for example, existing studies either combine stressor types or focus on one type to the exclusion of others. An exception includes studies on child neglect versus abuse that suggest that neglect is associated with impairment in neurobiological and cognitive systems, perhaps more so than abuse (R. E. Allen & Oliver, 1982; De Bellis, 2005; Eckenrode, Laird, & Doris, 1993). Erickson and Egeland (1996) further reported that children who had been neglected were socially withdrawn, inattentive, and cognitively underachieving, although many were also physically abused, making it difficult to partition the effects. Other studies have compared the effects of sexual versus physical abuse, but they have not included cognitive assessments; instead, they have tended to show differences in endocrine outputs (King, Mandansky, King, Fletcher, & Brewer, 2001). Another investigation disaggregated stressors into types and chronicity (Morales & Guerra, 2006), unlike previous studies, but focused on behavioral maladjustments and not cognition.

In sum, although there is an extensive body of literature on the effects of exposure to stress on neurobiological functioning, primarily in adults and in both children and adults with PTSD, there are few studies of stress that focus on cognitive function, even fewer that include children and those without PTSD, and none comparing relationships between cognition and stressor types. Once we better understand the relationship between stressor types and cognitive outcomes, more effective interventions can be developed for children who have experienced various stressors and present with differing deficits and needs. The malleability of cognitive processes, as well as the potential to ameliorate or strengthen parenting, family, and neighborhood conditions that influence them (see Fishbein, 2000), may translate to targeted interventions that protect against maladaptive outcomes (McGloin & Widom, 2001).

The Present Investigation

The conceptual model on which this study was based postulates that exposure to severe stressors such as child abuse, neighborhood violence, or poor parenting compromises cognitive and emotional development that provides the groundwork for enduring stress response styles (see Davidson, 1994; De Bellis et al., 1999; De Haan, Luciana, Malone, Matheny, & Richards, 1994; Fumagalli et al., 2007; Sinha, 2001). Executive cognitive and emotional regulatory mechanisms modulated by the cortico-limbic circuitry, henceforth referred to as neurocognitive functions, are the focus of this investigation given their developmental vulnerability. Deficits in neurocognition may impair the ability to generate viable coping options and evaluate their consequences, whereas deficits in emotional processes (e.g., perception of social cues) may interfere with the processing of cognitive stimuli, leading to ineffective or inefficient coping strategies such as avoidance, denial, or emotional ventilation. Although there is an increasing number of studies on various neurobiological effects of child maltreatment and adversity, neurocognitive and emotional regulatory functions potentially affected by particular types of stressors have yet to be elucidated. Furthermore, the majority of existing studies on stress have focused on Caucasians. There is a need to examine neurocognitive correlates of adversity and maltreatment in other cultures and ethnic groups.

The present study is a first step toward disaggregating associations between neurocognitive functioning and community and interpersonal stressors in children in an effort to eventually understand how stress exposure may influence cognitive function and, potentially, risk for psychopathology. Accordingly, the primary hypothesis under investigation was that children reporting a greater level of stress exposures would present with poorer neurocognitive functioning than children with lower levels of stressor exposures. Also, expectations were that specific types of stressors would be differentially associated with relative neurocognitive deficits, specifically that personal stressors would be more significantly related to lower levels of neurocognitive functioning than community stressors. This hypothesis is based on research suggesting that chronic stress, which more aptly characterizes interpersonal stressful experiences, particularly when originating in the home, may have a greater impact on developmental outcomes than community stressors, which tend to be more sporadic events with a lesser impact on stress responses (Printz, Shermis, & Webb, 1999).

Method

The present study on childhood stressors was carried out as one component of a longitudinal study to identify the cognitive and psychosocial precursors and consequences of drug abuse in a group of youth aged 10 to 12. Baseline assessments of neurocognitive functioning were conducted prior to the onset of any drug or alcohol use and are the focus of the present analyses. Additional baseline measures included the youth’s demographics, psychological functioning, behavioral traits, lifetime stressors, and level of neurocognitive and emotional functioning.

Sampling and Recruitment Procedures

Children aged 10 to 12 were randomly sampled in five public schools to enhance the representativeness of the sample with respect to the larger community in Cicero, Illinois, which is predominantly Latino, an understudied population in general. This age group was selected because of the unlikelihood of drug or alcohol use that would influence our measures, enabling us to establish a true baseline (i.e., prior to any substance use). Thus, those who reported any previous drug use were excluded (32 children total—12 [2.2%] reported inhalant use, 23 [4.2%] reported alcohol use, and 4 [0.7%] reported marijuana use; percentages are not cumulative). At the same time, this age group includes children who are on the threshold of manifesting behavioral problems of interest to the larger study because of increasing social demands and autonomy characterizing entry into middle school as well as the ability to more aptly measure emerging executive and emotional regulatory skills. The school district provided information including name, date of birth, school attended, grade level, parents’ names, home address, telephone number, and learning disability status of all of the fourth, fifth, and sixth grade students who were enrolled in the target schools in the spring of 2003. This file contained data on 2,454 students. Of these students, 40 who fell outside of the specified age range were excluded. Also excluded were 284 students who had an identified learning disability and were enrolled in special education classes. The remaining file, containing information on 2,142 students, was then divided into batches for monthly random sampling over a 12-month period.

Letters were mailed and telephone calls placed to the homes of sampled dyads (parent or guardian and child). Stringent follow-up activities were conducted when addresses were incorrect or phone numbers disconnected to maintain a representative sample, such as accessing computerized tracking systems maintained by the schools, contacting the post office, and computer searching telephone books through CompuServe’s PhoneFile. Once contact was made with a primary caregiver, the project and its procedures were explained, and an appointment was scheduled to obtain consent. Of the 658 dyads who were fielded and locatable, 553 (84%) participated, 98 (15%) refused, and 7 (1%) were ineligible or incapable of participating. A rudimentary nonresponse bias analysis was conducted to determine whether the 553 youth who were successfully recruited differed from the 1,589 youth who did not participate, examining a variety of dimensions, including age, school attended, and grade level. The only measure on which our sample of youthful participants differed from the nonparticipants was age; that is, our participants were slightly younger. The sample was fairly evenly split across males and females—48% of the sample were male, 52% were female. Of the sample, 85% was Hispanic or Latino, 8% were white, approximately 2% were Black or some other race, and 5% did not report their racial/ethnic background. The majority of respondents were 11 years old (43%).¹ And the yearly combined income reported by the primary caregiver was approximately $33,000.

Measures

The youth interview instruments were administered in English and Spanish using laptop computers and computer-assisted personal interviewing technology (Baker, 1992; Bradburn et al., 1991; Weeks, 1992). Each interview item was forward and back translated. Instructions for neurocognitive testing were provided in Spanish for the 31.7% of children in this sample who spoke primarily Spanish. In almost all cases, interviews were conducted in their homes.

Personal stressors.

Youth were administered various instruments to assess self-reported exposure to specific types of stress. The Childhood Trauma Questionnaire (Bernstein & Fink, 1998) includes subscales for Physical Abuse (α = .60), Physical Neglect (α = .40), Emotional Abuse (α = .65), Emotional Neglect (α = .75), and Sexual Abuse (α = .65). There were few cases of self-reported sexual abuse, and one participant was clearly an outlier; thus, this variable was removed from the analyses. These individual subscales were used for the bivariate analyses, but for the multivariate analyses Physical and Emotional Abuse scales were summated, as were the Physical and Emotional Neglect scales to produce two subscales: Any Abuse (α = .73) and Any Neglect (α = .74). Five items to gauge experiences with school stressors (e.g., having trouble with a teacher, getting suspended, being bullied) were combined into a single indicator (α = .68), and seven items to assess stressful experiences with parents (e.g., fighting with parents, involvement of parents in supervising, helping with homework, etc., feeling that they cared) were combined into a single indicator (α = .54). The latter two sets of items were extracted from the National Survey on Drug Use and Health instrument.

Community stressors.

Community stressors were measured by two variables. An index of witnessing neighborhood violence was created by summating five survey questions from the Community Violence Survey (Thomson, Roberts, Curran, Ryan, & Wright, 2002); for example, “During the past 12 months, how often … did you see or hear or read about a fight in this neighborhood in which a weapon was used?” “… have you witnessed a violent argument in the neighborhood?” “… did you see or hear or read about a sexual assault or rape?” (α = .77). Seven items were summated to assess children’s perception of problems within their neighborhood; for example, “How serious do you think the following problems are for your neighborhood as a whole?” “… excessive noise?” “… heavy traffic or speeding cars?” “… lack of access of adequate food shopping?” “… lack of recreation areas?” “… trash and litter?” “… no sidewalks or poorly maintained sidewalks?” “… violence?” (α = .78). There was no redundancy among items on the different personal or community scales.

Cognitive tasks.

The integrity of prefrontally modulated cognitive and emotional regulatory functions was measured using a noninvasive, developmentally appropriate, and specially designed battery of tasks. These instruments were selected based on their conceptual relatedness to constructs of interest in the larger study; for example, conduct problems, early onset drug use, and other risky behaviors. Four of the six tasks were administered in computerized format using E-Prime programming, and the other two were administered manually. Tasks were presented in a game-like scenario, reducing fatigue, increasing motivation, and enhancing validity of the data. The six tasks included the Raven’s Coloured Progressive Matrices, a revised version of the Cambridge Decision-Making Task, the Tower of London test, the Stroop Color Word Test, the Logan Change Task, and the Ekman Facial Recognition Test.

The Raven’s Coloured Progressive Matrices (Raven, Raven, & Court, 1998) is an age-normed, nonverbal measure of general intelligence frequently used with children. The Raven’s IQ task is administered manually and consists of 36 incomplete colored matrices with four options from which respondents must select the one that correctly completes the matrix. The Raven’s task is considered to be a reasonable measure of general intellectual ability and is adaptable to administration in non-English-speaking individuals. Age-adjusted percentile scores were used in analyses.

The Cambridge Decision-Making Task in computerized form was developed to dissect particular cognitive components believed to affect sensitivity to consequences and risk taking (Rogers, Blackshaw, et al., 1999; Rogers, Owen, et al., 1999) and consistently activates the orbital portion of the prefrontal cortex in neuroimaging studies of normal controls (Fishbein et al., 2005; Rogers, Everitt, et al., 1999; Rogers, Owen, et al., 1999). In this version adapted for children, the participant is told that the computer has hidden, on a random basis, a ring inside one of six red or blue boxes arrayed at the top of the screen and that he or she has to decide which set of boxes contains it. The participant is also told that there is an equal probability that the ring is hidden inside any of the six boxes. In addition, this decision involves gambling a certain number of points associated with each choice (10 vs. 90, 20 vs. 80, 30 vs. 70, 40 vs. 60, and 50 vs. 50). Immediately following a selection, one of the red or blue boxes opens to reveal the location of the ring. If the participant chooses the correct color, the points associated with that choice are added to the total points score; if the participant chooses the wrong color, the same points are subtracted. At the start of each sequence, the participant is given 100 points and instructed to make choices that will increase the number of points won. The ratio of colored boxes (5:1, 4:2, and 3:3) and the balance among the associated rewards vary independently from trial to trial according to a fixed pseudorandom sequence. This sequence ensures that each balance of reward and each ratio of colored boxes co-occur an equal number of times, with the restriction that on all trials with an unequal ratio of red and blue boxes (i.e., 5:1 or 4:2), the larger reward is always associated with the least likely outcome, thus capturing the conflict inherent in risk-taking situations. The central performance score generated by this task is the percentage of safe decisions selected (i.e., least risky in terms of odds of winning vs. losing points).

The Tower of London (Culbertson & Zilmer, 2001) measures implicit and skill (procedural) memory and problem solving (Spreen & Strauss, 1998) and has been shown to activate the dorsolateral prefrontal cortex, cerebellum, premotor cortex, cingulate, precuneus, and globus pallidus (Schall et al., 2003). Administered and scored manually, participants move colored beads from their initial position on upright and aligned pegs affixed to a small board to form a new, predetermined arrangement in as few moves as possible. Unlike many tests used for similar purposes, this task has the advantage of measuring both implicit and declarative memory. Performance of the Tower of London is thought to require successful planning, execution, monitoring, and revision of a series of actions using working memory as well as the selection of counterintuitive moves, which relies on inhibitory processes. There are 10 problems of increasing difficulty requiring from two to seven moves to solve. The participant is given 2 min to solve each problem. The task includes percentile age norms for the central outcome variable and total problems solved in a minimum number of moves; higher scores indicate better performance.

The Stroop Color Word Test (Golden, 1978) taps cognitive processes such as flexibility and resistance to interference and recruits the anterior cingulate cortex (Critchley, Tang, Glaser, Butterworth, & Dolan, 2005). This test was computerized and involves identifying words or a series of Xs displayed in color on the screen. In the first stage, color words (“red,” “blue,” and “green”) are shown in the color that the word spells; in the second condition, the Xs are presented in the colors; and in the third condition, the interference phase, words are presented in colors, but the color the word spells differs from the color in which the word is displayed. Participants are instructed to press the 1 key for red, the 2 key for blue, and the 3 key for green–always identifying the color in which the words are printed. For each trial, the display remains until the participant hits the correct key. For each of the three conditions, participants are given 45 s to complete as many trials as possible. The primary outcome variable is the interference score that is calculated by summing the number of responses to the first two conditions and dividing the resultant number by the product of multiplying the first two conditions. Higher scores reflect less cognitive interference.

The Logan Stop-Change Task measures an aspect of inhibition that involves the ability to shift responses in light of new information (Logan & Burkell, 1986), a function that has been shown to activate the right hemispheric anterior cingulate cortex, supplementary motor area, and inferior prefrontal and parietal cortices, which modulate error monitoring, interference control, and task management (Rubia et al., 2000). The task consists of two parts: (a) the baseline (36 trials) to assess the participants’ choice reaction time and (b) the main task (96 trials) to assess the participants’ ability to inhibit a prepotent response and initiate a new response. Each of these parts included instructions and several practice trials to ensure that the participants understood.

In the baseline portion, participants must press the 1 key when a star is displayed on the screen or press the 3 key when a circle is displayed. The baseline reaction time is calculated as the average reaction time for all correct responses to the 36 baseline trials. Feedback is provided after each trial, and the message “too slow” is displayed in a red square if the participant makes no response within 1,800 ms after the imperative stimulus is presented. The main task was identical to the baseline task except that participants were instructed to press the 2 key when they heard a tone rather than the 1 or 3 key as above. Tones were presented in 25% of the trials at four different times relative to the individual participant’s expected reaction time based on his or her reaction times in the baseline trials. It was relatively easy to change the prepotent response as instructed when the tones were presented close to the imperative stimulus (i.e., the star or circle). In contrast, it was very difficult to change the prepotent response when the tones were presented close to the anticipated key press for that individual. Each tone delay (500, 350, 250, 100 ms) was presented before the expected key press based on the individual’s responses during baseline trials. For the 500 ms trials, it was relatively easy to inhibit the prepotent response (i.e., to press either the 1 or 3 key) and to change to the new response (i.e., to press the 2 key), whereas on the 100 ms trials it was very difficult. Tones presented at 350 ms and at 250 ms were of intermediate difficulty. The need to respond as quickly as possible was emphasized. Central measures generated by this task include percentage of correct responses for each tone delay (i.e., 100, 250, 350, 500 ms) and reaction times for all tone delay trials combined.

The Ekman Facial Recognition Test (Ekman & Friesen, 1971) measures the ability to accurately identify emotional expressions in other people’s faces. This ability has been directly related to the function of the brain’s amygdala (responsible for perception of negative emotions) and is shown to be impaired in children and adults with externalizing disorders, such as conduct disorder, violence, and drug abuse (Blair, Morris, Frith, Perrett, & Dolan, 1999; Calder et al., 1996; Phillips, Roberts, & Lessov, 1997). Participants are instructed to identify the emotion (happy, anger, disgust, surprise, sadness, and fear) that best describes the facial expression. Practice trials are administered, after which 60 pictures of the six different emotional expressions are presented. The total number of correct responses was used in analyses.

The Dysregulation Inventory (Mezzich, Tarter, Giancola, & Kirisci, 2001), used in children, college students, and adults, was also included. This instrument was designed to measure components of three types of dysregulation, including emotional, behavioral, and cognitive. In the present study, only the Impulsivity, Inattention, and Hyperactivity subscales were included and summated as an additional measure of cognitive function relating to overt symptoms of attention-deficit/hyperactivity disorder (ADHD). This measure is different from the other measures of cognitive functioning in that it is generated by self-reports of a variety of symptoms rather than task performances that are specific to distinctive cognitive functions. Thus, the ADHD measure provides an indication of the extent to which general symptoms of ADHD are related to stress exposures.

Statistical Techniques

Ordinary least squares regression models were used to evaluate differential associations among self-reported stressful experiences, including both personal and community stressors, and selected dimensions of neurocognitive functioning. Bivariate models were used to examine direct relationships between various types of stressors and neurocognitive functions. Multivariate regression analyses were used to assess the impact of each personal and community stressor, relative to all other personal and community stressors, on each neurocognitive measure. A final model included personal and community stressors simultaneously. The multivariate models also controlled for child’s age and parent level of education (i.e., noncompletion vs. completion of high school), which is a reasonable proxy measure for socioeconomic status (Grzywacz, Almeida, Neupert, & Ettner, 2004). Given the distribution of variables, standardized coefficients were reported. In all models, a Bonferroni correction was used to adjust the alpha levels for a multiple comparison error rate or .10 (a probability of .90 that all confidence intervals are simultaneously correct), generating an alpha of .0125 for the eight outcomes being tested. Given that Bonferroni corrections are quite conservative and that such stringent adjustment increases the risk of Type II errors (Feise, 2002; Perneger, 1998), we also discuss theoretically relevant findings that were significant using the more liberal, unadjusted alpha levels, but with the appropriate qualifications.

Results

Means and standard deviations for each stressor measure and neurocognitive function are shown in Table 1. However, variations among task formats and administration, population characteristics (e.g., cultural, clinical), and age ranges make comparisons of means from similar instruments difficult. Nevertheless, the average scores on novel neurocognitive tasks for this sample are similar to those achieved in other studies with adolescents (e.g., Fishbein et al., 2005), and our means appear to be within normal ranges relative to scores for those instruments with normative data. Correlations among stressor types were examined to determine the degree to which children tend to be exposed to multiple types of stress (not shown). All of the stressor variables were significantly related to one another, which was expected. Although these correlations are highly significant (p < .001), the extent to which these stressors co-occur ranges from 15% to 35%, suggesting that measures are not redundant, eliminating concerns about multicollinearity, and that many children have not experienced multiple stressor types. The exception to this is physical and emotional abuse and physical and emotional neglect, which were highly correlated with each other; therefore, single summated abuse and neglect measures were generated from each set for use in the multivariate analyses. Correlations among cognitive task scores were also computed. Several of the cognitive tasks were correlated, but none greater than r = .2.

Table 1.

Descriptive Statistics: Personal and Community Stressors and Neurocognitive Functioning, Mean, Minimum, and Maximum

	M	SD	Min	Max
Personal stressors
Emotional abuse	1.69	2.37	0.00	14.00
Physical abuse	0.43	1.20	0.00	8.00
Emotional neglect	3.58	3.72	0.00	18.00
Physical neglect	1.65	2.01	0.00	11.00
Parent stress	11.97	2.90	7.00	22.00
School stress	7.81	2.35	5.00	20.00
Community stressors
Neighborhood violence	2.37	2.86	0.00	15.00
Neighborhood problems	5.13	4.25	0.00	19.00
Neurocognitive functions
Ekman: Total correct	41.65	6.59	6.00	57.00
Raven: Estimated IQ	47.00	30.01	5.00	95.00
Cambridge Decision-Making Task: % risky decisions	0.67	0.12	0.24	1.00
Tower of London: % problems solved in minimum moves	0.43	0.26	0.02	0.99
Stroop: Interference score	18.48	3.01	1.00	27.00
Stop-Change: % correct	0.58	0.28	0.00	1.00
Stop-Change: Reaction time (ms)	794.80	57.32	534.00	1057.00
Dysregulation Inventory: # ADHD symptoms	21.27	11.57	0.00	66.00

Note: N = 520.

Table 2 illustrates the results of the bivariate linear regression models assessing the relationship between each individual stressor and each measure of neurocognitive function. Asterisks indicate significance using the Bonferroni correction, and findings that were significant using a less stringent standard are also mentioned as marginal or trends (Feise, 2002; Perneger, 1998). All stressors, personal and community, were significantly and positively associated with ADHD symptoms in the bivariate analyses. Emotional abuse was significantly negatively associated with reaction time on the SCT. There was a marginal negative association between physical abuse and problem solving, as measured by the Tower of London task. Emotional and physical neglect were negatively associated with the ability to accurately attribute emotions (Ekman task) and estimated IQ (Raven). Experiencing stress from one’s parents was negatively albeit marginally associated with estimated IQ, risky decision making, as measured by the Cambridge Decision Making Task, and impulsivity. Similarly, trends were found for negative relationships between stress from school and emotion attribution, risky decision making, and impulsivity. Witnessing neighborhood violence was also marginally negatively associated with problem solving.

Table 2.

Bivariate Relationships Between Personal and Community Stressors and Neurocognitive Functioning, Ordinary Least Squares Standardized Betas

	Ekman			Raven			Cambridge Decision-Making Task			Tower of London			Stroop			Stop-Change			Stop-Change			Behavioral Dysreg.
	Emotion Attribution			Estimated IQ			Risky Decisions			Problem Solving			Cognitive Flexibility			Impulsivity: % Correct			Impulsivity: Reaction Time			ADHD Symptoms
	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p
Personal stress
Emotional abuse	−.03	(0.12)	.48654	−.07	(0.55)	.1108	−.04	(0.002)	.3621	−.08	(0.005)	.0529	.04	(0.11)	.3907	−.04	(0.005)	.3310	−.13**	(1.06)	.0026	.40***	(0.20)	< .0001
Physical abuse	−.003	(0.24)	.9514	−.02	(1.10)	.6208	−.07	(0.005)	.1059	−.09	(0.01)	.0326	.08	(0.10)	.0611	−.05	(0.01)	.2371	−.01	(2.11)	.7805	.28***	(0.41)	< .0001
Emotional neglect	−.19***	(0.08)	< .0001	−.11*	(0.35)	.0112	−.06	(0.001)	.1390	−.06	(0.003)	.1383	−.04	(0.04)	.4215	−.09	(0.003)	.0406	−.07	(0.68)	.1250	.14**	(0.14)	.0016
Physical neglect	−.24***	(0.14)	< .0001	−.16***	(0.64)	.0002	−.01	(0.003)	.8537	−.03	(0.006)	.3774	−.06	(0.07)	.2035	−.06	(0.006)	.2036	−.07	(1.25)	.0993	.20***	(0.25)	< .0001
Parent stress	−.09	(0.10)	.0515	−.10	(0.45)	.0208	−.10	(0.002)	.0289	−.07	(0.004)	.1288	.002	(0.05)	.9617	−.08	(0.004)	.0650	−.10	(0.87)	.0184	.26***	(0.17)	< .0001
School stress	−.09	(0.12)	.0322	−.06	(0.56)	.1640	−.09	(0.002)	.0331	−.03	(0.005)	.4443	−.04	(0.06)	.3818	−.06	(0.005)	.1754	−.11	(1.07)	.0136	.36***	(0.20)	< .0001
Community stress
Neighborhood violence	−.02	(0.10)	.5955	−.004	(0.46)	.9216	−.01	(0.002)	.8357	−.10	(0.004)	.0296	−.02	(0.05)	.7274	−.06	(0.004)	.1513	−.04	(0.89)	.4212	.21***	(0.17)	< .0001
Neighborhood problems	−.08	(0.07)	.0618	−.03	(0.31)	.4560	−.002	(0.001)	.9730	−.05	(0.003)	.2257	−.04	(0.03)	.3207	−.04	(0.003)	.3514	−.04	(0.60)	.4206	.17***	(0.12)	< .0001

Note: N = 552.

Adjusted alpha:

^*p < .0125.

^**p < .00625.

^***p < .00125.

When personal stressors were analyzed in the same model to assess their relative relations with neurocognitive function, many relationships that were significant in the bivariate analyses were no longer significant. Model 1 in Table 3 displays the relationship between personal stressors and emotion attribution. Neglect was the only stressor significantly associated with emotion attribution, where a 1 standard deviation increase in the neglect score was associated with a decrease of 0.18 standard deviations in the emotion attribution score. Neglect was also marginally associated with estimated IQ. Abuse and parent and school stressors were positively associated with symptoms of ADHD, with abuse having the strongest relative effect (β = .28, p < .0001). There was also a trend for abuse to be associated with problem solving when all other stressors were controlled for (Model 4). When analyzed together, no personal stressors were significantly associated with risky decision making, cognitive flexibility, or impulsivity.

Table 3.

Multivariate Relationships Between Personal Stressors and Neurocognitive Functioning, Ordinary Least Squares Standardized Betas

	Model 1			Model 2			Model 3			Model 4			Model 5			Model 6			Model 7			Model 8
	Ekman			Raven			Cambridge Decision-Making Task			Tower of London			Stroop			Stop-Change			Stop-Change			Behavioral Dysreg.
	Emotion Attribution			Estimated IQ			Risky Decisions			Problem Solving			Cognitive Flexibility			Impulsivity: % Correct			Impulsivity: Reaction Time			ADHD Symptoms
	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p
Personal stress
Any abusea	.03	(0.09)	.5259	−.01	(0.44)	.7886	−.04	(0.002)	.4387	−.10	(0.004)	.0408	.06	(0.05)	.1991	−.04	(0.004)	.3838	−.05	(0.86)	.2781	.32***	(0.15)	< .0001
Any neglect	−.18**	(0.06)	.0002	−.11	(0.29)	.0325	.02	(0.001)	.5761	−.01	(0.003)	.8827	−.05	(0.03)	.3359	−.03	(0.003)	.6000	−.01	(0.57)	.8897	−.03	(0.10)	.4674
Parent stress	−.01	(0.10)	.8007	−.05	(0.49)	.2789	−.08	(0.002)	.0984	−.04	(0.004)	.4097	.01	(0.05)	.8277	−.06	(0.01)	.2214	−.07	(0.94)	.1593	.14***	(0.17)	.0009
School stress	−.05	(0.13)	.2639	.01	(0.60)	.8388	−.08	(0.003)	.1008	.01	(0.005)	.8960	−.05	(0.06)	.2832	−.04	(0.01)	.4502	−.07	(1.16)	.1264	.25***	(0.20)	< .0001
Parent education	.19***	(0.58)	< .0001	.11*	(2.78)	.0119	.06	(0.01)	.1517	.09	(0.02)	.0495	.02	(0.28)	.6087	.05	(0.03)	.2756	.08	(5.36)	.0651	−.04	(0.95)	.3190
R 2	.09			.04			.03			.03			.03			.05			.03			.22

Note: N = 552. Models control for child age and parent education.

^a.Physical and emotional abuse and physical and emotional neglect were both correlated at r > .5; therefore, a combined measure of each type was used for the multivariate analyses to reduce multicollinearity concerns.

Adjusted alpha:

^*p < .0125.

^**p < .00625.

^***p < .00125.

In the bivariate analyses (Table 2), community stressors were associated with emotion attribution, problem solving, and ADHD symptoms. However, when witnessing neighborhood violence and perceiving neighborhood problems were modeled together, as seen in Table 4, they were only significantly associated with ADHD symptoms. Not only were the effect sizes quite small, these indicators contributed very little to explaining the overall variance of the model for ADHD (R² = .06).

Table 4.

Multivariate Relationships Between Community Stressors and Neurocognitive Functioning, Ordinary Least Squares Standardized Betas

	Model 1			Model 2			Model 3			Model 4			Model 5			Model 6			Model 7			Model 8
	Ekman			Raven			Cambridge Decision-Making Task			Tower of London			Stroop			Stop-Change			Stop-Change			Behavioral Dysreg.
	Emotion Attribution			Estimated IQ			Risky Decisions			Problem Solving			Cognitive Flexibility			Impulsivity: % Correct			Impulsivity: Reaction Time			ADHD Symptoms
	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p
Community stress
Neighborhood violence	.02	(0.11)	.9595	.02	(0.48)	.6666	−.01	(0.002)	.7709	−.08	(0.004)	.0699	.01	(0.05)	.8882	−.06	(0.004)	.1841	−.03	(0.94)	.5893	.17***	(0.18)	.0001
Neighborhood problems	−.07	(0.07)	.1149	−.04	(0.33)	.3990	.01	(0.001)	.8142	−.02	(0.003)	.6001	−.03	(0.03)	.4737	−.01	(0.003)	.8253	−.03	(0.63)	.5082	.13**	(0.12)	.0057
Parent education	.20***	(0.60)	< .0001	.13	(2.74)	.0022	.07	(0.01)	.0918	.08	(0.02)	.0538	.03	(0.28)	.5487	.06	(0.03)	.1756	.09	(5.31)	.0464	−.03	(1.03)	.4975
R 2	.05			.03			.01			.02			.02			.04			.02			.06

Note: N = 552. Models control for child age and parent education.

Adjusted alpha:

^*p < .0125.

^**p < .00625.

^***p < .00125.

To test for differential relationships between neurocognitive functioning and personal and community stressors, each cognitive measure was regressed on the complete set of stressors, and adjustments were made for child age and parental education (see Table 5). After controlling for the effects of all other personal and community stressors, only neglect was significantly and negatively associated with the ability to attribute emotion, as seen in Model 1. The same was true for Model 2, estimated IQ, although only a trend was found. In addition, abuse and parent and school stress were positively associated with ADHD symptoms, with abuse having the strongest effect (β = .26, p < .0001). No stressors were significantly associated with risky decision making, cognitive flexibility, or impulsivity.

Table 5.

Multivariate Relationships Between Personal and Community Stressors and Neurocognitive Functioning, Ordinary Least Squares Standardized Betas

	Model 1			Model 2			Model 3			Model 4			Model 5			Model 6			Model 7			Model 8
	Ekman			Raven			Cambridge Decision-Making Task			Tower of London			Stroop			Stop-Change			Stop-Change			Behavioral Dysreg.
	Emotion Attribution			Estimated IQ			Risky Decisions			Problem Solving			Cognitive Flexibility			Impulsivity: % Correct			Impulsivity: Reaction Time			ADHD Symptoms
	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p	β	SE	Unadj. p
Personal stress
Any abusea	.03	(0.10)	.4741	−.03	(0.47)	.5605	−.05	(0.002)	.3473	−.08	(0.004)	.1215	.08	(0.05)	.1301	−.03	(0.004)	.5680	−.06	(0.91)	.2470	.31***	(0.16)	< .0001
Any neglect	−.17***	(0.06)	.0005	−.10	(0.30)	.0364	.03	(0.001)	.6252	−.001	(0.003)	.9867	−.04	(0.03)	.4132	−.02	(0.003)	.6661	−.01	(0.57)	.8994	−.03	(0.10)	.4675
Parent stress	−.02	(0.10)	.6827	−.05	(0.49)	.2814	−.08	(0.002)	.1012	−.04	(0.004)	.4358	.01	(0.05)	.8996	−.06	(0.005)	.1955	−.07	(0.95)	.1601	.13**	(0.17)	.0018
School stress	−.05	(0.13)	.2604	−.02	(0.60)	.6795	−.08	(0.003)	.0923	.01	(0.005)	.8675	−.05	(0.06)	.3316	−.03	(0.005)	.5093	−.07	(1.17)	.1256	.24***	(0.20)	< .0001
Community stress
Neighborhood violence	.01	(0.11)	.8876	.06	(0.52)	.2417	.02	(0.002)	.6521	−.06	(0.005)	.2311	−.02	(0.05)	.6591	−.02	(0.005)	.6156	.04	(1.00)	.4734	.01	(0.17)	.7776
Neighborhood problems	−.03	(0.07)	.5425	−.02	(0.33)	.7338	.02	(0.001)	.7264	−.01	(0.003)	.8859	−.03	(0.03)	.4892	−.02	(0.003)	.7118	−.03	(0.64)	.5164	.08	(0.11)	.0593
Parent education	.19***	(0.59)	< .0001	.11*	(2.79)	.0105	.06	(0.01)	.1491	.09	(0.02)	.0429	.02	(0.28)	.6425	.05	(0.03)	.2938	.08	(5.39)	.0747	−.03	(0.94)	.4019
R 2	.09			.05			.03			.03			.03			.05			.03			.23

Note: N = 552. Models control for child age and parent education.

^a.Physical and emotional abuse and physical and emotional neglect were both correlated at r > .5; therefore, a combined measure of each type was used for the multivariate analyses to reduce multicollinearity concerns.

Adjusted alpha:

^*p < .0125.

^**p < .00625.

^***p < .00125.

Discussion

The literature suggests that various types of stressors are associated with similar cognitive outcomes, even in the absence of PTSD, from decrements in general intelligence to higher order neurocognitive impairments (Navalta, Polcari, Webster, Boghossian, & Teicher, 2006; Saigh, Yasik, Oberfield, Halamandaris, & Bremner, 2006). However, only a few studies have focused on neurocognitive functioning in children or adults relative to levels of exposure to maltreatment or adversity. The present investigation was guided by numerous reports that maltreatment and adversity are associated with disruptions to the activities of various brain systems and alterations in stress responses, which, in turn, have been related to memory deficits, impulsivity, and other neurocognitive impairments (Jay et al., 2004; Mizoguchi, Ishige, Takeda, Aburada, & Tabira, 2004; Pine et al., 1996; Pine et al., 1997; Rosenblum & Andrews, 1994). We anticipated that given the young age of our participants and, thus, the particular vulnerability of their developing brains, the cognitive and emotional correlates of exposure to stressors may already be measurable.

Overall, our results are supportive of the proposition that exposure to psychosocial stressors is associated with relatively lower levels of neurocognitive functioning in this sample of children. In bivariate analyses, experiencing psychosocial stressors, such as emotional and physical neglect and stress with parents and in school, was variably associated with poorer general intelligence, emotion attribution, and executive decision making. Reaction time during the impulsivity task was also faster in children reporting greater emotional abuse and, to a lesser extent, parent and school stress. Witnessing neighborhood violence and problems in the neighborhood were not significantly associated with performance on any of the cognitive tasks. Behavioral (“active”) symptoms of ADHD were strongly associated with all types of stressors, which is interesting in light of numerous reports that ADHD is prevalent in populations of children and adults with a history of maltreatment or PTSD (e.g., Endo, Sugiyama, & Someya, 2006; Weber & Reynolds, 2004). The strength of relationships between stressors and ADHD relative to our other neurocognitive outcomes suggests that the narrower and more specific task performance measures may be differentially and more weakly related to stressor exposures, whereas a more general measure of behavior (ADHD symptoms) may encompass more of the phenomenon under study. In either event, several authors have speculated about the temporal sequence; that is, stress may play an etiologic or exacerbating role in ADHD, however the presence of externalizing behaviors in ADHD may also increase the risk for maltreatment (Briscoe-Smith & Hinshaw, 2006).

Despite methodological differences among various studies of neurocognition, the present findings are consistent with expectations of a variety of relative cognitive deficits in participants with a history of exposure particularly to personal stressors. In response to both human and preclinical studies showing neurobiological abnormalities resulting from chronic or severe stress (Bremner et al., 1997; Stein, Koverola, Hanna, Torchia, & McClarty, 1997), most investigations employing cognitive testing have included measures of learning (in some cases with academic achievement as a proxy), memory, verbal skills, and comprehension and have uncovered corresponding functional deficits (D. M. Allen & Tarnowski, 1989; Eckenrode et al., 1993; Kurtz, Gaudin, Wodarski, & Howing, 1993; Teicher et al., 1997). Thus, for most stressors measured in the present study, support for their relationship to a variety of prefrontally modulated cognitive functions was found, although given the cross-sectional nature of the data used in this study causality cannot be assumed. It is also important to note that the measure of stress from relations with parents had relatively lower reliability than measures of abuse and neglect, suggesting that findings pertaining to the latter two indices may be more interpretable.

Our second aim was to determine whether exposure to personal stressors was more strongly associated with relatively lower levels of neurocognitive functioning than community stressors. Focused multivariate analyses found support for this hypothesis. The combined measure of abuse was associated with active symptoms of ADHD and, to a less extent, relative deficits in problem solving. Parent and school stress were also associated with symptoms of ADHD. Higher levels of reported neglect (emotional and physical) were marginally associated with lower IQ and, to a greater extent, with less accuracy in the perception of emotion in facial expressions; this latter finding held strong when stressors were considered both individually and in the presence of other stressors in the model. In ancillary analyses where responses were broken down by type of emotion expressed, misattributions were significantly related to neglect across all six emotions. Consistent with this finding, another study found that neglected children were less able to distinguish among emotional expressions (Pollak, Cicchetti, Hornung, & Reed, 2000). In contrast, studies conducted to date on child abuse fairly consistently have reported an enhanced ability to distinguish emotional expressions (Pollak & Sinha, 2002). It is possible that neglect specifically impairs the ability to correctly identify adults’ expressions and intentions because of an inconsistency between facial expressions and behaviors often exhibited by neglectful caregivers. Prior research has shown that the amygdala and its connections with the orbitofrontal cortex are largely responsible for this ability (Hamann & Mao, 2002; Lee, Farrow, Spence, & Woodruff, 2004; Stark et al., 2004). Given evidence from preclinical and human neuroimaging and other physiological studies that the pathophysiology of stress exposure involves this circuitry (Sinha, 2001; Taylor, Eisenberger, Saxbe, Lehman, & Lieberman, 2006; Teicher, Andersen, Polcari, Anderson, & Navalta, 2002; Teicher et al., 2003; Williams et al., 2006), impairments in emotional perception are expected. Thus, the increased number of errors on this task in the context of neglect suggests that these relative deficits may be related to either compromised amygdala function or prefrontal cortex modulation.

As theorized, community stressors played a lesser role in our models than did personal stressors; witnessing neighborhood violence and perceiving neighborhood problems were not associated with neurocognitive functioning when observed relative to personal stressors. Witnessing violence in particular has been associated with externalizing behaviors such as aggression (Bradshaw & Garbarino, 2004; Schwartz & Proctor, 2000) and perceived positive outcomes for use of aggression (Shahinfar, Kupersmidt, & Matza, 2001). These behaviors were tapped by our measure of ADHD symptoms, which was associated with both neighborhood problems and witnessing violence in the bivariate analysis and the separate multivariate model including only community stressors. Witnessing violence and poor behavioral outcomes have been further linked to social information processing abilities (Bradshaw & Garbarino, 2004; Schwartz & Proctor, 2000). Emotion perception, considered a dimension of social information processing, was found to be associated with neighborhood problems in the bivariate analysis. Again, however, relationships between community stressors and neurocognition diminished in the presence of personal stressors in the model.

Given these preliminary and correlational findings, future studies may shed further light on these questions by addressing whether chronic stress activation, such as with parental abuse, may be more likely to result in permanent perturbations in “wiring” and allostasis, thereby contributing to cognitive impairment more so than sporadic stress exposures that often characterize occasional exposure to community violence (for relevant discussions, see Bowman, Beck, & Luine, 2003; Sapolsky, 2003). One relevant aspect of the differences between sporadic versus chronic stress involves “controllability.” External stressors (e.g., community) may be perceived as more controllable, avoidable, or more distal than personal stressors (e.g., home and family), hence not as disabling from a cognitive perspective, particularly when not in the presence of other stressors in the home. In effect, for children sporadically experiencing these circumstances, a “stress reaction” may not occur that might otherwise be associated with neurocognitive decrements. Supportive of this possibility, DuMont, Widom, and Czaja (2007) reported that neighborhood factors did not exert a direct effect on resilience against adversity but instead moderated relationships between family stability and resilience in adolescence and between cognitive ability and resilience in young adulthood. Elucidating a more precise role for neighborhood influences may be a fruitful line of future investigations.

Implications

As an initial study, this study has several inherent limitations that temper definitive conclusions. The absence of formal psychiatric diagnoses precluded an estimate of the extent to which comorbidity may contribute to relationships among relative neurocognitive deficits and psychosocial stressors. Furthermore, the heterogeneity within the sample regarding other relevant traits (e.g., cultural factors, personality traits such as anxiety or low arousability, parenting techniques, peer influences, etc.) and the developmental phase during which the abuse occurred were not considered in the analyses. Similarly, functional manifestations of cognitive skills and emotion attribution (e.g., peer relationships and school grades) were not assessed and in future research may help to shed light on whether the lower scores for children with higher levels of stress translate into clinically significant effects in the children’s daily lives. Also important, the potential for heritable influences that may moderate relationships between parental maltreatment and child outcomes was not directly measured, although adjustments were made for parental education. One caveat relevant to this issue is that the absence of parental variables such as IQ and cognitive function that appear to be related to abuse and neglect (Nayak & Milner, 1998) may have resulted in underspecified models that could have an impact on the tested correlations. Another issue that is both a strength and a weakness of this research is the composition of the sample, which was largely Latino. This ethnic and cultural group has not been represented in existing studies on stress and is notably absent from research in general on neurocognitive outcomes. Thus, this study provides a small window into experiences within this population. However, whether cultural differences in Latino children can be expected to differentially affect children’s response to stress in any way cannot be addressed by a single study.

Although much additional research is needed, the extant literature suggests that vulnerability of the brain, particularly in the early years, plays a role in the ability of psychosocial stressors to alter the developmental progression of higher-order cognitive functions. Given the vulnerability of gradually emergent prefrontal cortex to psychosocial stress, the ability of this brain region to modulate executive and emotional functions may be compromised or delayed as a consequence of the neurodevelopmental effects of undue stress (Koenen et al., 2001). It is, therefore, essential that future research explore the possibility that higher-order cognitive functions may be affected by stress in more complex ways than have previously been considered. Such research should take advantage of recent advances in instrumentation, which have led to the development of more sensitive and specific measures of executive and emotional regulatory functions. Furthermore, given that neurocognitive deficits are likely related not only to the effects of psychosocial stress exposures but also to preexisting, possibly genetic, vulnerability factors (Caspi et al., 2002), studies are needed to determine the nature and extent to which genetic factors moderate the effects of early stress and the quality of the caregiving environment over time.

Elucidating the relationship among psychosocial stressors and ECF and emotional regulation using techniques that provide a window into brain systems that are associated with cognition, emotion, and stress responses may enhance our general understanding of the vulnerability of prefrontal cortical development and its limbic circuitry to psychosocial conditions early in life (Schore, 1997) and further reveal how various types of stressors may differentially affect these systems. Exposure to adversity in childhood is at least partially preventable in response to appropriate educational, preventive, and public health measures, particularly those that focus on reducing parenting and family stressors (Masten et al., 1999). Also, to enhance early mental development and adaptive success, children facing such serious personal challenges may benefit from individualized programs designed to strengthen compromised functions (Fishbein et al., 2005; Greenberg & Kusche, 2006). Thus, advancing this knowledge base may contribute to the development of interventions more appropriately targeted to specific deficits associated with exposure to various types of psychosocial stressors during childhood, thereby increasing resiliency and psychological well-being across the life span.

Biographies

Diana Fishbein, PhD, is a senior fellow and director of the Transdisciplinary Behavioral Science Program at RTI International in the Baltimore, Maryland, office. Her research area is cognitive and behavioral neuroscience, and the foci of her studies are the precursors and consequences of drug abuse and underlying neurobiological mechanisms in differential responses to interventions for high-risk behaviors.

Tara Warner is a research analyst at RTI International and a doctoral student in the Department of Sociology at Bowling Green State University. Her research interests include substance use treatment in the criminal justice system, sexual violence and victimization of college women, and risk and protective factors for substance among Hispanic immigrant youth.

Christopher Krebs, PhD, is a senior research social scientist at RTI International. His research interests include criminal and delinquent behavior, substance abuse epidemiology and treatment, intimate partner and sexual violence, justice systems, and program evaluation.

Nancy Trevarthen was a survey specialist at RTI International in the Chicago office. She has supported numerous studies in conducting field supervision of data collection efforts.

Barbara Flannery, PhD, is a psychology researcher at RTI International in the Baltimore, Maryland, office. Her background is in psychopharmacology of alcohol and drug addiction, and she is currently studying the effects of cognitive functioning on intervention responsivity.

Jane Hammond, PhD, is a developmental psychologist in the Transdisciplinary Behavioral Science Program at RTI International. Her research interests include neurodevelopmental outcomes related to substance abuse in children and adolescents including children prenatally exposed to licit and illicit substances.