Executive Function Skills and School Success in Young Children Experiencing Homelessness

Abstract

The authors examined the role of executive function (EF) skills as a predictor of kindergarten or first-grade adjustment in 138 children living in shelters for homeless families. During the summer, children completed a battery of six EF tasks and three IQ measures. Teachers later rated children’s school adjustment in five domains of achievement and social conduct. Confirmatory factor analysis supported the construct validity of EF as distinct from the general factor in IQ tests. The differential predictive validity of EF scores for school adjustment was tested by hierarchical regression analysis in relation to IQ. Results supported the hypothesis that EF has unique predictive significance for homeless children. Findings also corroborate the feasibility and validity of EF assessments in community settings and contribute to growing evidence that EF skills are important for school success. Implications are discussed for addressing educational disparities for homeless and highly mobile children.

Keywords

achievement gap cognition poverty urban education at-risk students regression analyses factor analysis

Research spanning more than two decades indicates that homeless and highly mobile (HHM) children are at elevated risk for difficulties in school, including low achievement, conduct problems, and social problems (Buckner, 2008; Miller, 2011; National Research Council and Institute of Medicine [NRC], 2010; Samuels, Shinn, & Buckner, 2010). The recent economic and housing crisis in the United States, accompanied by increases in the number of students identified as HHM, has intensified concerns about their school success (Miller, 2011; Samuels et al., 2010). There is renewed interest in reducing risk and bolstering protective factors to promote the educational success of HHM students. The goal of the present study was to examine the potential importance of executive function skills as an indicator of academic success for homeless children entering kindergarten and first grade.

Childhood homelessness is a context of high cumulative risk, characterized by residential instability in conjunction with other risk factors for child development and educational disparities (Buckner, 2008; Masten, Miliotis, Graham-Bermann, Ramirez, & Neeman, 1993; Samuels et al., 2010). HHM children often experience the disruptions of midyear school and home changes, in conjunction with food insecurity, family and neighborhood violence, health problems, parental loss through separation or incarceration, and disconnections from family and friends. Children in currently homeless families experience more school disruptions and family stressors than children from comparable sociodemographic backgrounds who are not currently homeless (Masten et al., 1993). It is not surprising that HHM students generally have more academic problems than less mobile low-income peers. Recent studies using administrative data to compare achievement level and growth on standardized tests found that achievement for HHM students in reading and math is significantly and persistently lower than that of non-HHM students qualified for free or reduced-fee lunch (Cutuli et al., in press; Obradović et al., 2009).

Despite this evident risk, many currently or formerly homeless children manifest developmental competence, including academic success (Masten, 2011). For example, Obradović et al. (2009) and Cutuli et al. (in press) found that a substantial portion of students identified as HHM in a large, urban school district showed average or better reading and math achievement on a standardized national test. Understanding such resilience can inform efforts to reduce risk and promote educational success in HHM students who are not faring as well. Given the large numbers of HHM students in urban districts (e.g., a cumulative rate of 14% in Cutuli et al., in press), this knowledge also may be important for addressing persistent achievement gaps in urban schools. However, there are very few data on the possible origins of academic success in HHM students, particularly studies that focus on potentially malleable protective processes. This study focused on executive function (EF) skills as a key component of neurobehavioral function that shows considerable promise of both predictive validity for early school adjustment and plasticity in response to educational intervention.

EF refers to a broad set of cognitive control processes that enable individuals to manage and direct their attention, thinking, and actions to meet adaptive goals (Best & Miller, 2010; Blair & Raver, 2012; Diamond & Lee, 2011). These skills represent the “top-down,” voluntary aspect of self-regulation, including multiple and interrelated functions, such as working memory, inhibitory control, and cognitive flexibility (e.g., to deliberately shift one’s attention). EF skills develop and change with age and experience, in concert with the development and function of the prefrontal cortex and associated neural networks. Good school adjustment requires effective cognitive control in multiple ways. Children must be able to concentrate and ignore distractions, attend to the teacher, follow classroom rules, get along with other children, wait for rewards, and suppress impulses to play, aggress, or otherwise digress from expected behavior. Thus, it is not surprising that EF skills relate to school success and are viewed as a key element of school readiness (Blair, 2002; Blair & Razza, 2007; Diamond & Lee, 2011).

Two aspects of EF, often termed “cool” and “hot” EF (Brock, Rimm-Kaufman, Nathanson, & Grimm, 2009; Willoughby, Kupersmidt, Voegler-Lee, & Bryant, 2011; Zelazo & Müller, 2002), are recognized as distinct in that cool EF is involved in tasks that require little emotional control and relatively abstract problem solving, whereas hot EF is needed for tasks that require control of emotional reactions or excitement, delay of gratification, or resisting temptation. Cognitive control in a school context often involves both aspects of EF.

EF skills develop rapidly in the preschool years and continue to improve into early adulthood (Zelazo & Bauer, in press). Many influences likely play a role in EF development, including the interplay of genes, parenting, nutrition, play, and educational experiences as well as trauma, environmental toxins, and stress. Studies of individual differences indicate that disadvantaged children often perform worse on EF measures as well as measures of general intellectual functioning. Prefrontal neural networks associated with EF may be especially sensitive to the risks associated with poverty, adversity, trauma, toxins, and neglect, and this may be particularly true in early development (Blair & Raver, 2012; Shonkoff, 2011). At the same time, some children in high-risk contexts develop good EF skills.

EF skills may be a key protective influence for high-risk children, enabling them to succeed in a context of severe adversity or poverty (Blair & Raver, 2012; Sapienza & Masten, 2011). Buckner, Mezzacappa, and Beardslee (2003, 2009) found that more successful youth from low-income families (many of whom were formerly homeless) had better self-regulation skills according to interviewer ratings. Furthermore, Obradović (2010) found that children staying in emergency shelter who performed well on EF tasks also did well in kindergarten and first grade, with better achievement and fewer behavior problems rated by teachers.

In addition, there is growing evidence that EF skills are malleable and perhaps particularly so for higher-risk and lower-EF children. EF skills change in response to targeted intervention, including preschool education (e.g., Diamond & Lee, 2011). Moreover, changes in EF mediate positive behavior change in preventive interventions that explicitly target EF skills (e.g., Raver et al., 2011). The preschool years may well be a window of opportunity for preventive interventions that promote EF.

Although EF appears to be a key indicator of risk or resilience for school outcomes, it is not clear whether EF measures have distinct predictive validity with respect to related cognitive measures, particularly established IQ measures that include indices of verbal and nonverbal reasoning. Measures of EF show convergent validity via structural equation modeling methods (Willoughby, Blair, Wirth, & Greenberg, 2010), and there is observed covariance and internal consistency among multiple EF measures (Obradović, 2010; Zelazo & Bauer, in press). However, EF task performance often is correlated with scores on tests of vocabulary and novel problem solving, such as Block Design, Matrix Reasoning, and similar “fluid intelligence” tests (Blair, 2006). Performance on both EF and IQ tests may depend on basic attention, language, and memory functions that develop rapidly in the preschool years (Garon, Bryson, & Smith, 2008) and make it feasible to test young children on any kind of cognitive task that requires attention and following instructions.

Nonetheless, EF and IQ appear to become more differentiated as development proceeds. There is good evidence supporting discriminant validity of EF and traditional IQ tests in adults and adolescents, and three aspects of EF can be differentiated among adults and older children (Blair, 2006; Garon et al., 2008; Miyake et al., 2000). Less differentiation is observed in younger children, both among EF tasks and between EF and IQ measures (Zelazo & Bauer, in press), consistent with the suggestion that neurocognitive development in general involves greater functional specialization of both neural systems and cognitive functions (e.g., Johnson, 2011). Thus, with younger children, it is important to consider whether EF measures show not only convergence but also divergence from verbal and nonverbal measures of IQ.

The Present Study

The present study had two primary aims. The first aim was to examine the construct validity of EF assessed while children were staying in a homeless shelter in relation to the general factor in traditional IQ tests. The second aim was to test the differential predictive validity of EF assessed in shelter for subsequent school function in multiple domains of adjustment.

The present study is part of a programmatic and collaborative effort to understand and inform efforts to remedy the striking educational disparities observed for homeless and similarly low-income, highly mobile children in Minneapolis and similar urban school districts. The present study builds on preceding work to establish strong community collaborations and district partnerships, to pilot recruiting and assessment methods, and to test the reliability and validity of assessments conducted in a shelter context while maintaining a high level of respect and sensitivity in working with families undergoing challenging circumstances (e.g., Cutuli et al., in press; Herbers et al., in press; Masten et al., 1993; 2008; Obradović, 2010; Obradović et al., 2009). We examined EF skills as a harbinger of risk or resilience for early school problems or success among young homeless children, building on initial findings from our research program indicating that EF measures were feasible in a shelter context and showed promising validity regarding school function (Obradović, 2010).

We recruited a sample of currently homeless families with a child expected to begin kindergarten or first grade that fall. EF was assessed in shelter prior to the start of the school year, and teachers subsequently completed measures of classroom adjustment. The construct validity and predictive ability of EF was examined in relation to traditional IQ measures. We hypothesized that EF could be differentiated from general cognitive ability (IQ) and that better EF skills would forecast better school adjustment in five key domains of academic and social success or problems in the school context controlling for IQ: academic competence, peer acceptance, prosocial behavior, inattentive-impulsive behavior, and aggressive-defiant behavior.

EF skills were hypothesized to predict achievement, consistent with evidence linking EF to school achievement and readiness (Blair & Raver, 2012; Blair & Razza, 2007; Brock et al., 2009; Obradović, 2010; Raver et al., 2011; Willoughby et al., 2011). There also is considerable evidence linking EF skills to peer acceptance and rejection (Hughes & Ensor, 2011; Riggs, Jahromi, Razza, Dillworth-Bart, & Mueller, 2006; Willoughby et al., 2010). There is less evidence linking EF directly to prosocial behavior; however, the thoughtful and empathic behaviors that are central to ratings of prosocial behavior in school-age children have theoretical ties to theory-of-mind abilities in children (understanding the perspectives of other people), which have strong empirical, as well as conceptual, links to EF (e.g., Riggs et al., 2006). In contrast, EF task performance has been directly linked by theory and empirical findings to symptoms of inattention and impulsivity (Hughes & Ensor, 2011; Willoughby et al., 2010). Thus, we hypothesized that EF skills would be inversely predictive of these symptoms reported by teachers. Similarly, we expected that conduct problems characterized by aggressive and oppositional behaviors, often described as externalizing symptoms, would be predicted by EF. EF difficulties are associated with poorly controlled aggression in studies of preschoolers as well as older children (Hughes & Ensor, 2011; Riggs et al., 2006; Willoughby et al., 2010).

Given that EF develops rapidly during the preschool years, we also planned a priori to control for age in the key analyses linking EF skills to school adjustment. In this way, EF effects could be attributed to individual differences rather than uncontrolled age differences.

Method

Participants

Participants included 138 children entering kindergarten or first grade in the fall of that year and their parents, and they were recruited while the family was residing in one of three emergency shelters during the summers of 2008 and 2009. These shelters perennially house the majority of homeless children each year who attend school in the district where they are located. Eligible families had to be in residence at the shelter for at least 3 days, have a child expected to enter kindergarten or first grade the following fall, speak English, and be free of disabilities that would preclude participation. Two children were ineligible because of severe disabilities. Only one child per family was tested. The overall participation rate was 72% of all families believed to be eligible, with 6.8% of families actively declining participation; the remaining did not come for appointments, could not be scheduled before they left the shelter, or never came in direct contact with research staff. Of participating children, 78 (56.5%) were female and 60 (43.5%) were male. The mean age was 5.77 (SD = .58, range 4.83 to 6.92). Child ethnic-racial composition included 63.8% African American, 16.7% multiracial, 5.8% American Indian, 4.3% Caucasian, 1.4% Asian, and 6.5% Other, which is typical of these shelters. Most of the children lived in a single-caregiver household (73%) headed by a mother. Average length of shelter stay at the time of assessment (reported by parents) was 32.93 days (SD = 44.90).

Procedures

Parents and children were assessed separately for approximately 60 minutes on the shelter premises and then participated in a joint session focused on parent-child interaction that is not part of the present study. Children completed a battery of EF tasks and three standardized intelligence subtests, and parents completed questionnaires about the family. Children received small gifts and parents received gift cards for participating. After the school year began, teachers were contacted to complete a questionnaire about each child’s adaptive functioning. Despite continued mobility in many cases, 83% of the children were located in schools, and outcome measures were obtained for 80% of the original sample, with a very high teacher response rate (97%). Teachers received a gift card as a thank-you for returning the questionnaire.

Research in an emergency shelter context requires flexibility and sensitivity in design and implementation, including attention to the ongoing stresses of homelessness, cultural competence, and dealing with the inevitability that families will be unable to make appointments for a variety of reasons. Importantly, close community partnerships and collaboration with shelter providers and educators contributed to the study design, execution, and interpretation of findings. Community partners informed our appreciation of the context of the project and helped us to fine-tune our procedures and attend to the well-being of the families we encountered. Furthermore, given the study’s goals of furthering a larger program of research aimed at reducing educational disparities, partnerships with providers are necessary to guide research questions and recommendations toward plausible and practical ends and means. Parents gave permission during the consenting process for us to contact schools to request information on school adjustment. All procedures for this study were approved by the Institutional Review Board at the University of Minnesota.

Measures

Measures used for the study were well validated or had shown good psychometric properties when used with children staying in emergency shelters in past research and pilot testing.

EF

EF was measured with a battery of six tasks that emphasize inhibitory control, set shifting, and delay of gratification. The battery included four cool EF tasks. In Simon Says, the researcher stated and demonstrated a series of actions, half of which are preceded by the phrase “Simon says” (Kochanska, Murray, & Coy, 1997). The child was instructed to do all the actions preceded by “Simon says” (activation trials) and refrain from doing actions that do not begin with “Simon says” (inhibition trials). Child behavior was video-recorded and later coded by a team of raters (K = .94). Scores were based on the percentage of correct inhibition trials. No scores were computed for cases in which the child failed all activation trials or failed to inhibit correctly in any of the practice trials. These children were presumed to misunderstand the demands of the task. The second EF task, the Dimensional Change Card Sort (DCCS; Zelazo, 2006), assessed cognitive flexibility. Children were instructed to sort the test cards first by color, then by shape. Scores for the DCCS reflected the number of correct sorts out of six following the rule switch. In the Peg Tapping task (Diamond & Taylor, 1996), children were presented with two rules: Tap the table with a wooden dowel twice when the experimenter tapped once, and tap once when the experimenter tapped twice. Scores were based on percentage of correct taps (K based on 1/3 of sample = .97). In the Computerized Pointing Stroop (Berger, Jones, Rothbart, & Posner, 2000), children saw two different animals on a computer screen and heard an animal sound corresponding to one of the animals. In compatible trials, children were instructed to point to the animal that makes the sound they heard. Then, in the incompatible trials, children were instructed to point to the animal that does not make the sound they heard. Scores were based on percentage of correct responses during incompatible trials.

In addition, two tasks involved inhibition in the presence of an anticipated reward to assess EF in hot or reward-salient conditions. In the Dinky Toys task (Kochanska et al., 1997), children were shown a box full of small toys, such as plastic cars or spinning tops, and were told to ask for, but not to touch, the toy they wanted. Scores were a composite based on worst transgression (5-point scale), frequency of transgressions, and latency to worst transgression. Child behavior was video-recorded and later coded by a team of trained raters (K = .76 for worst transgression code; 93% agreement for frequency within 1; latency agreement within 2 seconds = 95%). The final EF task was Gift Delay (Kochanska et al., 1997), which has two distinct parts. In Part 1, children are positioned facing away from the experimenter while he or she noisily wraps a gift for 1 minute and are instructed not to peek. In Part 2, the experimenter leaves the room for 3 minutes, having instructed the child not to peek at the gift while he or she is gone. Scores on each part of this task were based on a code for worst transgression as well as frequency of and latency to worst transgression. Interrater agreement was calculated on the basis of one third of the cases by a team of raters (K = 1.0 for both tasks on code; 95% and 100%, respectively, agreement within 1 for frequency; 96% and 91% agreement within 2 seconds for latency). For these three hot EF scores, we formed composites by averaging z scores for code, frequency, and latency. Scores were truncated to z scores of ±3.0 before compositing.

Based on results presented below, we computed an EF composite by averaging z scores from six EF task scores (α_c = .71). The Gift Delay Part 2 score was not included because it did not load with the other tasks in the measurement model (see below).

Intelligence

Scores from three measures were used to estimate the general factor of IQ: the Peabody Picture Vocabulary Test, Fourth Edition (PPVT-4; Dunn & Dunn, 2007), a standardized assessment of receptive vocabulary, and two nonverbal subscales (Block Design and Matrix Reasoning) of the Wechsler Preschool and Primary Scales of Intelligence, Third Edition (WPPSI-III; Wechsler, 2002). We formed a composite by averaging z scores based on the scaled scores from the two nonverbal subscales (r = .29 in this sample) and then averaging the z score of that score with the z score based on PPVT-4 standardized scores (r = .36).

School outcomes

Measures of school adaptation in five specific areas were drawn from the teacher version of the MacArthur Health and Behavior Questionnaire (HBQ-T; Armstrong, Goldstein, & The MacArthur Working Group on Outcome Assessment, 2003). Academic competence was assessed with the five-item HBQ-T Academic Competence subscale (e.g., “How would you evaluate this child’s current school performance in reading-related skills?” α_c = .96 for the present sample). Correlations of this score with standardized assessments conducted by the school district for a subgroup of children with available test scores indicated good construct validity for this index of achievement (r = .64 to .76; see Supplemental Table S1 at http://edr.sagepub.com/supplemental). Peer acceptance was assessed with the eight-item HBQ-T Peer Acceptance/Rejection subscale (e.g., “has lots of friends at school”; α_c = .91). Prosocial behavior was assessed with the 20-item HBQ-T Prosocial Behavior subscale (e.g., “comforts a child who is crying or upset”; α_c = .95). Inattention-impulsivity symptoms were assessed with the HBQ-T ADHD Symptoms Scale (α = .90), comprising the Inattention subscale (e.g., “can’t concentrate; can’t pay attention for long”; 6 items, α = .90) and the Impulsivity subscale (e.g., “impulsive or acts without thinking”; 9 items, α = .92). Aggressive-defiant symptoms were assessed with a composite of the three HBQ-T subscales (α = .94): Oppositional Defiant Behavior (e.g., “defiant, talks back to adults”; 9 items, α = .91), Conduct Problems (e.g., “steals; takes things that don’t belong to him/her”; 11 items, α = .86), and Overt Hostility (e.g., “kicks, bites, or hits other children”; 4 items, α = .84).

Strategy of Analysis

Analysis proceeded in two main stages. First, a confirmatory factor analysis (CFA) of the seven EF scores and three IQ scores was conducted to establish whether it was feasible to assess the construct of EF, or two components of EF (cool, hot), as distinct from IQ, despite expected shared variance among these tasks. The CFA was conducted with the lavaan package (Version 0.4-10; Rosseel, 2012) in R (Version 2.14.0; R Development Core Team, 2011). Second, we tested the predictive validity of the EF composite in a series of hierarchical regressions of the five school adjustment criteria on a planned set of models to test hypotheses and partition the response variance. Site differences were examined prior to regression analyses, and no significant differences were found in any of the school adjustment variables.

The models had the following predictors: (a) control variables (gender, age), (b) IQ composite added to (a), and (c) EF composite added to (b). This a priori ordering of models explicitly tested the variance that could be attributed to EF (Model 3) after gender, age, and IQ were controlled. Planned follow-up regressions were conducted, reversing the evaluation order of Models 2 and 3 (adding EF prior to IQ), to provide a more complete picture of the correlated predictors. This strategy allowed us to partition the total variance in each school outcome domain into variance predicted by EF uniquely, by IQ uniquely, and shared by EF and IQ.

Analyses were first conducted with listwise deletion of cases that had missing values (n = 110 for all regressions except Academic Competence, where n = 108). The same analyses were also conducted with multiply imputed data (20 data sets were imputed on the basis of the larger data set) with statistical pooling of individual results (Little & Rubin, 2002). Imputation was conducted with the mice package (Version 2.11; van Buuren & Groothuis-Oudshoorn, 2011) in R (Version 2.14.0; R Development Core Team, 2011). Regressions based on imputed data corroborated the listwise deletion findings with highly similar beta weights and significance test results. For simplicity, the nonimputed results are presented below (imputed results for the correlations and key regression models are included in supplemental materials at http://edr.sagepub.com/supplemental). After constructing the models, we examined the residuals for normality and for any presence of heteroscedasticity. We examined Cook’s distance for each model to see whether there were any influential cases. Only externalizing showed non-normality and heteroscedasticity, which were corrected with a square root transformation. For all other outcomes, the residual analysis indicated consistency with assumptions of traditional regression (normality and homoscedasticity).

Results

Descriptive data for the EF, IQ, and HBQ measures are provided in Table 1. CFA results are presented next, followed by the regression results.

Table 1

Descriptive Data for IQ, EF, and HBQ-T Scores

Measure	M	SD
WPPSI-III
Block Design	7.23	3.01
Matrix Reasoning	7.72	2.86
PPVT-4	88.10	12.79
Animal Stroop (0–1)	.73	.25
Simon Says (0–100)	44.80	37.42
DCCS (0–6)	4.38	2.41
Peg Tapping (0–100)	68.38	31.01
Dinky Toys
Worst transgression (0-4)	2.07	1.72
Frequency of transgressions	1.78	2.97
Latency to first transgression	49.51	44.89
Gift Delay
Worst transgression (0–2)	1.36	.90
Frequency of transgressions	1.16	2.27
Latency to first transgression	45.84	21.61
HBQ-T
Oppositional Defiant (0–2)	.42	.47
Conduct Problems (0–2)	.25	.34
Overt Hostility (0–2)	.42	.49
Inattention (0–2)	.73	.60
Impulsivity (0–2)	.72	.56
Peer Acceptance (1–4)	3.29	.61
Prosocial Behavior (0–2)	.99	.50
Academic Competence (1–5)	2.49	.91

Note. The range of possible values is shown in parentheses. EF = executive function; HBQ-T = teacher version of the MacArthur Health and Behavior Questionnaire (Armstrong, Goldstein, & the MacArthur Working Group on Outcome Assessment, 2003); WPPSI-III = Wechsler Preschool and Primary Scales of Intelligence, Third Edition (Wechsler, 2002); PPVT-4 = Picture Vocabulary Test, Fourth Edition (Dunn & Dunn, 2007); DCCS = Dimensional Change Card Sort (Zelazo, 2006).

The CFA analysis tested the plausibility of the two-factor model of the cognitive scores, delineating EF and IQ as separate factors, and also compared the two-factor model against a one-factor and a three-factor model. The one-factor model had both the EF and IQ variables loading on a single factor. The three-factor model had separate EF and IQ factors, with EF further divided into hot EF (Dinky Toys, Gift Delay Parts 1 and 2) and cool EF factors (Simon Says, DCCS, Stroop, Peg Tapping). Age was a control variable in all analyses. Models were compared with the Akaike information criterion (AIC; Akaike, 1973). The three-factor model could not be estimated because of lack of convergence. Results indicated the two-factor model had good absolute fit (comparative fit index = 0.966; Tucker-Lewis index = 0.955; root mean square error of approximation = 0.039; standardized root mean square residual = 0.049), and its relative fit (AIC = 7081.97) was better than the one-factor model (AIC = 7087.06; AIC difference = 5.09) according to the criteria of Burnham and Anderson (2002). Factor loadings and the factor correlation for the two-factor CFA are shown in Table 2. Thus, separate composite scores were computed for EF and IQ. The score for Gift Delay 2 contributed little or nothing to the EF factor. Therefore, only z scores for the remaining six EF task scores were averaged to form the EF composite for the regression analyses.

Table 2

Confirmatory Factor Analysis Loadings and Factor Correlation for the Two-Factor Model of IQ and EF Variables

Factor and Indicator	λ
IQ Measures
WPPSI-III Block Design scaled score	.563
PPVT-4 standard score	.501
WPPSI-III Matrix Reasoning scaled score	.556
EF measures
Stroop	.785
Simon Says	.758
Peg Tapping	.670
Dimensional Change Card Sort	.591
Gift Delay Part 1	.474
Dinky Toys	.348
Gift Delay Part 2	.027
Factor correlation
Correlation of EF with IQ	.775

Note. CFI = 0.966; TLI = 0.955; RMSEA = 0.039; SRMR = 0.049. EF = executive function; WPPSI-III = Wechsler Preschool and Primary Scales of Intelligence, Third Edition (Wechsler, 2002); PPVT-4 = Picture Vocabulary Test, Fourth Edition (Dunn & Dunn, 2007).

Correlations among the regression variables are shown in Table 3. Results of the primary regressions are shown in Table 4, and Table 5 summarizes the findings on the shared and unique variance contributed to each outcome for EF and IQ after controlling for gender and age (corresponding to Figure 1).

Table 3

Intercorrelations of the Regression Variables

	Age	Gender	IQ	EF	Peer	PS	Aca	In/Im
Age
Gender	.04
IQ	.16	−.12
EF	.41***	−.15	.44***
Peer	.10	−.04	.07	.25**
Prosocial	−.07	−.20*	.24**	.29***	.56***
Aca	−.09	−.11	.49***	.39***	.31***	.49***
In/Im	.10	.30***	−.26**	−.36***	−.45***	−.60***	−.43***
Agg	.10	.25**	−.25**	−.21*	−.53***	−.68***	−.35***	.72***

Note. EF = executive function; Peer = peer acceptance; PS = prosocial behavior; Aca = academic competence; In/Im = inattention-impulsivity; Agg = aggressive-defiant behavior. Positive gender correlations indicate that boys had higher scores.

p < .05. **p < .01. ***p < .001.

Table 4

Results of Hierarchical Regressions to Test Value-Added Models of EF in Relation to IQ for Teacher-Rated School Adjustment

	School Adjustment Outcome
Step	Academic Competence	Peer Acceptance	Prosocial Behavior	Inattention-Impulsivity	Aggressive-Defiant Behavior
1. Gender	−.194	.068	−.169	.319**^a	.200**^a
Age	−.136	.081	−.057	.089	.50
ΔR²	.020	.009	.034	.094**	.072*
2. Gender	−.072	.080	−.140	.290**^a	.182*^a
Age	−.269*	.065	−.096	.128	.064
IQ	.495***	.058	.145**	−.148	−.068
ΔR²	.255***	.008	.075**	.064**	.029
3. Gender	.006	.118	−.109	.250**^a	.164*^a
Age	−.458**	−.046	−.186*	.247**	.117
IQ	.375***	−.027	.075	−.057	−.027
EF	.313***	.199**	.163**	−.212***	−.096*
ΔR²	.079***	.076**	.075**	.106***	.045***
Total R²	.353***	.094*	.183***	.264***	.146**
Planned follow-up entering EF in Step 2
2. Gender	−.033	.120	−.115	.254** ^a	.166*^a
Age	−.468**	−.045	−.188*	.248**	.118
EF	.469***	.188**	.195***	−.237***	−.107**
ΔR²	.211***	.083**	.133***	.162***	.070**

Note. Estimated beta weights are shown. EF = executive function.

Boys have higher scores.

p < .05. **p < .01. ***p < .001.

Table 5

Partitioned Variance for EF and IQ as Predictors of School Adjustment

	Academic Competence	Peer Acceptance	Prosocial Behavior	Inattention-Impulsivity	Aggressive-Defiant Behavior
A. EF uniquely	.079	.076	.075	.106	.045
B. IQ uniquely	.123	.001	.016	.008	.004
C. EF-IQ jointly	.132	.007	.059	.056	.025
Total R² EF+IQ	.334	.084	.150	.170	.074

Note. R² values are shown (after gender and age are controlled). EF = executive function; A = unique variance accounted for by EF, B = unique variance accounted for by IQ, C = variance accounted for by EF and IQ jointly. See Figure 1 for a visual representation of A, B, and C.

Figure 1.

Diagram of shared and unique variance of executive function (EF) and IQ related to an outcome. Total R² and partitioned variance predicted by (A) EF uniquely, (B) IQ uniquely, and (C) IQ-EF jointly for each outcome.

EF scores accounted for unique variance in school outcomes for all five domains tested, whereas IQ accounted for unique variance only for academic achievement. Results support the partly unique and partly shared effect of EF as a marker of risk or resilience for early school problems or success. Results for academic achievement indicate that EF and IQ have unique and shared variance. Results for peer acceptance suggest that only EF is a significant predictor. IQ does predict prosocial behavior, but the variance is shared entirely with EF, which also accounts for additional unique variance in prosocial behavior. Teacher ratings of inattention-impulsivity symptoms were uniquely related to EF scores only; with EF controlled, IQ did not contribute to the prediction of this outcome. This was also the case for aggressive-defiant behavior; only EF skills predicted the teacher ratings.

Conclusions

EF skills offer a broad indicator of risk or resilience in the early school years among children experiencing homelessness. Results suggest that EF composite measures may index fundamental skills important for learning and classroom adjustment in these children as they also navigate the numerous challenges associated with homelessness. Children who had better EF skills did better in kindergarten or first grade with respect to all five school adjustment criteria, with better academic achievement, peer acceptance, and prosocial behavior and fewer problems of impulsivity, inattention, aggression, or noncompliance. Findings corroborate and extend the results from the preliminary study by Obradović (2010) based on a smaller sample of children from one of the same shelters. Results demonstrate the feasibility of measuring EF in the shelter context using relatively brief and easy-to-administer methods. EF showed expected associations with age as well as IQ scores: Older children tended to have higher EF scores, as did children who performed better on IQ measures. Even so, individual differences in EF predicted unique variance in school adjustment beyond age and IQ.

Future research should describe the developmental processes that contribute to shared and unique aspects of EF and IQ task performance. Understanding the common and distinctive causal pathways of risk and protection will be particularly important for understanding differences in cognitive functioning among high-risk children. Extreme adversity and stress, particularly in a context of few resources and limited protections, could interfere with general as well as specific aspects of cognitive development. There are general neurocognitive processes shared by many tasks, including most IQ and EF tests (Blair et al., 2011; Blair & Raver, 2012; Shonkoff, 2011). Performance on any kind of novel cognitive task requires EF, and especially in childhood, EF skills may be fundamental for performance on any IQ assessment battery, given that these tests require children to pay attention, remember the instructions, switch from task to task, and carry out the expected response.

In the current study, there was evidence for differentiation of EF and IQ as well as evidence of covariance and shared predictive validity consistent with the possibility of common underlying processes. There was little support for a three-factor model that differentiated cool and hot EF, although the lower loadings of the latter tasks may be congruent with the possibility of emerging neurocognitive specialization. Further study with different or additional hot EF measures, larger sample sizes, and longitudinal data would be informative in this regard.

With respect to results suggesting differentiation and unique predictive validity for EF in relation to IQ, the EF tasks specifically called for inhibition of a prepotent response, delay of gratification, and cognitive flexibility. These skills may have specific relevance for children in frequently changing and challenging circumstances, including children experiencing the turmoil of homelessness and high mobility. These children must contend with high levels of novelty and inconsistency as they transition from school to school or from home to home. EF skills also may be specifically important for success in the complex social world of school for these children under stress, as they try to cope with new social contexts, manage their emotions, and understand new teachers, school cultures, and rules.

Much remains to be elucidated when it comes to the origins, nature, and meaning of variability in EF skills among children experiencing adversities such as homelessness. Numerous investigations are under way to understand development and individual differences in EF as well as the processes by which EF can be promoted and protected in the face of adversity. Additionally, investigators are focused on the processes by which EF protects learning and other adaptive behaviors. Factors across multiple levels of child function, relationships, and context appear to contribute to the development of EF and other skills. Among high-risk children, including those in shelter, implicated factors include parenting behavior (e.g., Herbers, 2011; Herbers et al., 2011), stress and biological reactivity (e.g., Cutuli, 2011; Shonkoff, 2011), and early childhood education (Diamond & Lee, 2011). It will be important to identify the similarities and differences in EF processes for HHM children compared to other populations of children who are not currently homeless but who share the risk factors of high mobility, trauma exposures, or sociodemographic disadvantage. It will be important to systematically improve the theory and assessment of EF in relation to IQ, school readiness, and behavioral symptoms. Functional imaging of neural activity may be helpful in delineating the neural networks engaged by different measures of higher cognitive function, including measures of EF and IQ, and this in turn may help hone the concepts and methods for assessing, understanding, and promoting these crucial cognitive skills. Longitudinal studies of EF are needed in high-risk children and especially children who experience homelessness, even though it is challenging to follow highly mobile families (NRC, 2010). Intervention studies to promote school success by improving EF skills could be particularly important for testing hypotheses about the role of EF in resilience (Masten, 2011).

This study had a number of strengths, including assessments in settings with high ecological validity, high participation rates by homeless families and classroom teachers, multiple measures of EF and IQ, multiple shelters representing the majority of homeless children entering the school district, and sensitive, successful data collection with an important and understudied population. There were also notable limitations, including a moderate sample size, short-term assessment of school outcomes, and a mixture of children with diverse school and preschool experiences. The correlational study design also precluded conclusions about causality. Results indicate that children in emergency shelter who show higher levels of EF do better in multiple ways when they enter kindergarten or first grade. However, the current design cannot speak to whether these differences resulted in any way from EF, predated homelessness, or reflect other causal processes that may influence school outcomes as well as EF. Moreover, results of this study may not generalize to other populations of mobile children, including homeless children who have not resided in shelters. Additionally, without a broader range of residentially stable and unstable families from varying adversity backgrounds, it is not possible to determine the extent to which EF skills may be a particularly important protective factor for school success among families who are residentially mobile or in other ways turbulent. It also is unclear why EF skills are better or worse among different children living in emergency shelters, although evidence from our ongoing studies implicate stress and parenting processes as important predictors of individual differences. Good EF exhibited during the crisis of family homelessness may in itself reflect antecedent protective processes that are preserving EF development in the midst of adversity, such as might be afforded by effective parenting or genetic variations in sensitivity to the context. Finally, to keep the participant burden low, we did not assess all aspects of school readiness or conduct comprehensive assessments of mental and physical health disorders that might be important for understanding EF skills and their significance for school.

Implications

Limitations notwithstanding, results of this study suggest that further research on EF may be important for addressing achievement disparities in children who experience homelessness as well as other very disadvantaged, mobile children. The adversities that precipitate homelessness, in addition to the experience of homelessness itself, may disrupt the development or deployment of EF skills, with lasting consequences for learning and school success. On the other hand, the plasticity of EF skills may offer a window of opportunity for effective interventions to promote school success through targeting change in these cognitive control processes (Blair & Raver, 2012; Diamond & Lee, 2011). EF appears to be a key mediator linking risk and protective processes to school success and, as such, may play a central role in the efforts to bridge behavioral neuroscience and education (Blakemore & Bunge, 2012).

EF skills may be an important target for screening and preventive interventions to promote school success in homeless and highly mobile children. Growing evidence indicates that EF skills are malleable, with randomized intervention trials showing promise (Diamond & Lee, 2011; Raver et al., 2011). Many of these interventions are school based and emphasize activities that require students to use EF skills, such as sustained attention, working memory, inhibitory control, and cognitive flexibility. Another promising avenue for intervention is positive parenting practices, another “natural context” that supports the development of child EF with subsequent implications for school success (Herbers, 2011; Herbers et al., 2011). Such interventions may be particularly beneficial for higher-risk and lower-EF children. Evidence on the rapid change with age in EF task performance in the preschool years, the sensitivity of EF to early stress and environmental insults, and indications that EF skills respond to educational interventions, both in preschool and the early school years, all suggest that there is considerable plasticity in EF, especially during childhood. Thus, early interventions targeting EF skills in high-risk populations, either directly or indirectly, may be an important strategy for closing the marked achievement gaps observed for HHM and other very disadvantaged children (Cutuli et al., in press; Obradović et al., 2009).

Clearly, there is much to learn about the processes by which individual and contextual differences influence the development of EF as well as the processes by which EF contributes to learning and school adjustment. Ultimately, randomized control trials targeting change in EF will be important for testing causal hypotheses about the role of EF in school readiness and success, including the resilience of children who experience homelessness.

Family homelessness and residential instability are likely to continue in the United States, even after the most recent surge related to the Great Recession begins to ebb. Thus, it is imperative that the needs of homeless children be recognized and addressed by all stakeholders in their future. Early childhood education programs may represent a crucial front line and EF skills a key target for ameliorating risk and promoting resilience in homeless and highly mobile children.

Footnotes

Notes

Authors

Ann S. Masten is a Distinguished McKnight University Professor in the Institute of Child Development, University of Minnesota, 51 East River Road, Minneapolis MN 55455; amasten@umn.edu . Her research focuses on competence and resilience in development, including school success in homeless and highly mobile children.

Janette E. Herbers is a clinical postdoctoral appointee at the CARES Institute of the University of Medicine and Dentistry of New Jersey and a research associate at University of Minnesota, Institute of Child Development, 51 East River Road, Minneapolis, MN 55455; herbe064@umn.edu . Her research focuses on risk and resilience among children and families facing poverty and trauma.

Christopher David Desjardins is a doctoral candidate at University of Minnesota Twin Cities, Educational Psychology, 56 East River Road, Minneapolis, MN 5545; desja004@umn.edu . His research interest include longitudinal, multilevel, and latent variable modeling; statistical modeling for unusual, nonstandard data; and Bayesian statistics.

J. J. Cutuli is a research associate at University of Pennsylvania, School of Social Policy and Practice, 3701 Locust Walk, Philadelphia, PA 19104; jcutuli@upenn.edu . His research focuses on the processes of resilience and of risk in development.

Christopher M. McCormick is a doctoral candidate at the University of Minnesota Twin Cities, Institute of Child Development, 51 East River Parkway, Minneapolis, MN 55455; mccor217@umn.edu . His research is broadly focused on understanding processes of risk and resilience in children and families experiencing adversity.

Julianna K. Sapienza is a graduate student at University of Minnesota Twin Cities, Institute of Child Development, 51 East River Parkway, Minneapolis, MN 55455; jksapienza@gmail.com . Her focus is on translational and intervention research on executive function skills in high-risk children.

Jeffrey D. Long is a professor in the Department of Psychiatry, Carver College of Medicine, University of Iowa, 1-328 Medical Education Bldg., Iowa City, IA, 52242-1000; jeffrey-long@uiowa.edu . His research focuses on the analysis of longitudinal data.

Philip David Zelazo is the Nancy M. and John E. Lindahl Professor in the Institute of Child Development, University of Minnesota Twin Cities, 51 East River Road, Minneapolis, MN 55455; zelazo@umn.edu . His research focuses on the development and neural bases of executive function.

References

Akaike

(1973). Information theory and an extension of the maximum likelihood principle. In Petrov

B. N.

Csaki

(Eds.), Second international symposium on information theory (pp. 267–281). Budapest, Hungary: Akademiai Kiado.

Armstrong

J. M.

Goldstein

L. H.

, & the MacArthur Working Group on Outcome Assessment. (2003). Manual for the MacArthur Health and Behavior Questionnaire (HBQ 1.0). MacArthur Foundation Research Network on Psychopathology and Development (David J. Kupfer, Chair), University of Pittsburgh.

Berger

Jones

Rothbart

M. K.

Posner

M. I.

(2000). Computerized games to study the development of attention in childhood. Behavioral Research Methods and Instrumentation, 32, 297–303.

Best

J. R.

Miller

P. H.

(2010). A developmental perspective on executive function. Child Development, 81(6), 1641–1660.

Blair

(2002). School readiness: Integrating cognition and emotion in a neurobiological conceptualization of children’s functioning at school entry. American Psychologist, 57(2), 111–127.

Blair

(2006). How similar are fluid cognition and general intelligence? A developmental neuroscience perspective on fluid cognition as an aspect of human cognitive ability. Behavioral and Brain Sciences, 29(2), 109–125.

Blair

Granger

D. A.

Willoughby

Mills-Koonce

Cox

Greenberg

M. T.

, . . . the FLP Investigators. (2011). Salivary cortisol mediates effects of poverty and parenting on executive functions in early childhood. Child Development, 82, 1970–1984. doi:10.11111/j.1467-8624.2011.01643.x

Blair

Raver

C. C.

(2012). Individual development and evolution: Experiential canalization of self-regulation. Developmental Psychology, 48, 647–657. doi:10.1037/a0026472

Blair

Razza

R. P.

(2007). Relating effortful control, executive functioning, and false belief understanding to emerging math and literacy ability in kindergarten. Child Development, 78, 647–663.

10.

Blakemore

S.-J.

Bunge

S. A.

(2012). Editorial: At the nexus of neuroscience and education. Developmental Cognitive Neuroscience, 2S, S1–5.

11.

Brock

L. L.

Rimm-Kaufmann

S. E.

Nathanson

Grimm

K. J.

(2009). The contributions of “hot” and “cool” executive function to children’s academic achievement, learning-related behaviors, and engagement in kindergarten. Early Childhood Research Quarterly, 24(3), 337–349.

12.

Buckner

J. C.

(2008). Understanding the impact of homelessness on children: Challenges and future research directions. American Behavioral Scientist, 51, 721–736.

13.

Buckner

J. C.

Mezzacappa

Beardslee

W. R.

(2003). Characteristics of resilient youths living in poverty: The role of self-regulatory processes. Development and Psychopathology, 15, 139–162.

14.

Buckner

J. C.

Mezzacappa

Beardslee

W. R.

(2009). Self-regulation and its relations to adaptive functioning in low income youths. American Journal of Orthopsychiatry, 79, 19–30.

15.

Burnham

K. P.

Anderson

D. R.

(2002). Model selection and multimodel inference: A practical information-theoretic approach (2nd ed.). New York, NY: Springer-Verlag.

16.

Cutuli

J. J.

(2011). Context, cortisol, and executive functions among children experiencing homelessness (Doctoral dissertation). University of Minnesota, Minneapolis and St. Paul.

17.

Cutuli

J. J.

Desjardins

C. D.

Herbers

J. E.

Long

J. D.

Heistad

Chan

C.-K.

. . . Masten

A. S.

(in press). Academic achievement trajectories of homeless and highly mobile students: Resilience in the context of chronic and acute risk. Child Development.

18.

Diamond

Lee

(2011). Interventions shown to aid executive function development in children 4 to 12 years old. Science, 333, 959–964. doi:10.1126/science.1204529

19.

Diamond

Taylor

(1996). Development of an aspect of executive control: Development of the abilities to remember what I said and to “do as I say, not as I do.” Developmental Psychobiology, 29, 315–334.

20.

Dunn

L. M.

Dunn

D. M.

(2007). Peabody Picture Vocabulary Test (4th ed.). Minneapolis, MN: NCS Pearson.

21.

Garon

Bryson

S. E.

Smith

I. M.

(2008). Executive function in preschoolers: A review using an integrative framework. Psychological Bulletin, 134(1), 31–60.

22.

Herbers

(2011). Parent-child relationships in young homeless families: Co-regulation as a predictor of child self-regulation and school adjustment (Doctoral dissertation). University of Minnesota, Minneapolis and St. Paul.

23.

Herbers

J. E.

Cutuli

J. J.

Lafavor

T. L.

Vrieze

Leibel

Obradović

Masten

A. S.

(2011). Direct and indirect effects of parenting on academic functioning of young homeless children. Early Education and Development, 22, 77–104.

24.

Herbers

J. E.

Cutuli

J. J.

Supkoff

L. M.

Heistad

Chann

C.-K.

Hinz

Masten

A. S.

(in press). Early reading skills and academic achievement trajectories of students facing poverty, homelessness, and high residential mobility. Educational Researcher.

25.

Hughes

Ensor

(2011). Individual differences in growth in executive function across the transition to school predict externalizing and internalizing behaviors and self-perceived academic success at 6 years of age. Journal of Experimental Child Psychology, 108(3), 663–676.

26.

Johnson

M. H.

(2011). Interactive specialization: A domain-general framework for human functional brain development? Developmental Cognitive Neuroscience, 1, 7–21.

27.

Kochanska

Murray

K. T.

Coy

K. C.

(1997). Inhibitory control as a contributor to conscience in childhood: From toddler to early school age. Child Development, 68, 263–277.

28.

Little

R. J. A.

Rubin

D. B.

(2002). Statistical analysis with missing data (2nd ed.). New York, NY: Wiley.

29.

Masten

A. S.

(2011). Resilience in children threatened by extreme adversity: Frameworks for research, practice, and translational synergy. Development and Psychopathology, 23, 141–154.

30.

Masten

A. S.

Heistad

Cutuli

J. J.

Herbers

J. E.

Obradović

Chan

C.-K. L.

. . . Long

J. D.

(2008, Summer). School success in motion: Protective factors for academic achievement in homeless and highly mobile children in Minneapolis. CURA Reporter, 38(2), 3–12.

31.

Masten

A. S.

Miliotis

Graham-Bermann

Ramirez

Neemann

(1993). Children in homeless families: Risks to mental health and development. Journal of Consulting and Clinical Psychology, 61, 335–343.

32.

Miller

P. M.

(2011). A critical analysis of the research on student homelessness. Review of Educational Research, 81(3), 308–337. doi:10.3102/0034654311415120

33.

Miyake

Friedman

N. P.

Emerson

M. J.

Witzki

A. H.

Howerter

Wager

T. D.

(2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100.

34.

National Research Council and Institute of Medicine. (2010). Student mobility: Exploring the impact of frequent moves on achievement. Summary of a workshop. Washington, DC: National Academies.

35.

Obradović

(2010). Effortful control and adaptive functioning of homeless children: Variable-focused and person-focused analyses. Journal of Applied Developmental Psychology, 31, 109–117. doi:10.1016/j.appdev.2009.09.004

36.

Obradović

Long

J. D.

Cutuli

J. J.

Chan

C.-K.

Hinz

Heistad

Masten

A. S.

(2009). Academic achievement of homeless and highly mobile children in an urban school district: Longitudinal evidence on risk, growth, and resilience. Development and Psychopathology, 21, 493–518.

37.

R Development Core Team. (2011). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org/

38.

Raver

C. C.

Li-Grining

Bub

Jones

S. M.

Zhai

Pressler

(2011). CSRP’s impact on low-income preschoolers’ preacademic skills: Self-regulation as a mediating mechanism. Child Development, 82(1), 362–378.

39.

Riggs

N. R.

Jahromi

L. B.

Razza

R. P.

Dillworth-Bart

J. E.

Mueller

(2006). Executive function and the promotion of social-emotional competence. Journal of Applied Developmental Psychology, 27, 300–309.

40.

Rosseel

(2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1–36. Retrieved from http://www.jstatsoft.org/v48/i02/

41.

Samuels

Shinn

Buckner

J. C.

(2010). Homeless children: Update on research, policy, programs, and opportunities. Report prepared for the Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services.

42.

Sapienza

J. K.

Masten

A. S.

(2011). Understanding and promoting resilience in children and youth. Current Opinion in Psychiatry, 24, 267–273.

43.

Shonkoff

J. P.

(2011). Protecting brains, not simply stimulating minds. Science, 333, 982–983.

44.

van Buuren

Groothuis-Oudshoorn

. (2011). mice: Multivariate imputation by chained equations in R. Journal of Statistical Software, 45(3), 1–67.

45.

Wechsler

(2002). WPPSI-III administration and scoring manual. San Antonio, TX: Psychological Corporation.

46.

Willoughby

M. T.

Blair

C. B.

Wirth

R. J.

Greenberg

(2010). The measurement of executive function at age 3 years: Psychometric properties and criterion validity of a new battery of tasks. Psychological Assessment, 22, 306–317.

47.

Willoughby

M. T.

Kupersmidt

Voegler-Lee

Bryant

(2011). Contributions of hot and cool self-regulation to preschool disruptive behavior and academic achievement. Developmental Neuropsychology, 36(2), 162–180.

48.

Zelazo

P. D.

(2006). The Dimensional Change Card Sort (DCCS): A method of assessing executive function in children. Nature Protocols, 1, 297–301.

49.

Zelazo

P. D.

Bauer

P. J.

(Eds.). (in press). National Institutes of Health Toolbox Cognition Battery (NIH Toolbox CB): Validation for children between 3 and 15 years. Monographs of the Society for Research in Child Development.

50.

Zelazo

P. D.

Müller

(2002). Executive functions in typical and atypical development. In Goswami

(Ed.), Handbook of childhood cognitive development (pp. 445–469). Oxford, UK: Blackwell.