Abstract
Executive function (EF), a set of neurocognitive processes, is central to students’ emotional and behavioral well-being. Despite students with emotional and behavioral disorders (EBD) being at risk for negative long-term outcomes, there is a paucity of EF research with students at risk for EBD in early elementary school, an important identification and intervention period. Thus, we conducted a cross-sectional latent profile analysis with a sample of 1,154 kindergarteners and first graders identified as at risk for EBD to determine whether unique EF profiles existed and the extent to which profile membership was related to distinct patterns of functioning. Results indicated a theoretically supported, three-profile solution of mildly, moderately, and clinically at-risk EF profiles. Differences between EF profiles were noted such that students with moderately and clinically at-risk EF profiles demonstrated more problematic behaviors, less social competence, and greater language difficulties. We discuss implications for early identification, intervention efforts, and future research.
Executive function (EF), a set of neurocognitive attention regulation processes (Miyake et al., 2000; Zelazo et al., 2016), not only is a significant predictor of students’ academic performance (Spiegel et al., 2021), but also is linked to their mental health and behavioral competence (Schoemaker et al., 2013; Zelazo, 2020). EF makes it possible for students to attend to tasks, ignore distractions, control impulses, remember and use information, and adapt to changing circumstances in both school and life (Miyake et al., 2000; Zelazo et al., 2016). For students identified with emotional and behavioral disorders (EBD), a limited ability to manage their impulses and behaviors is often associated with persistent negative outcomes (e.g., academic underachievement, school dropout, incarceration, unemployment; Bradley et al., 2008; Kauffman & Landrum, 2018), with growing evidence EF likely plays a role (e.g., Cumming et al., 2019; Mattison et al., 2006). From a dynamic-systems perspective, behavioral functioning involves the interplay between individual processes (e.g., EF) and environment (e.g., learning opportunities; Sameroff, 2020). Although there are linear generalities in behavioral development (e.g., greater ability to sit and attend in first grade than in kindergarten), students may exhibit very different patterns of growth due to this interplay (see Farmer et al., 2020). Yet, there is limited EF research with students identified at risk for EBD, especially in early elementary grades, which is a crucial period for prevention and intervention.
There may be more children and youths who are eligible or at risk for EBD than are identified, and EF is routinely overlooked during identification, referral, and instruction. Bruhn et al. (2014) estimated that up to 12% of children have a moderate EBD-related difficulty, and Forness et al. (2012) argued that one third are expected to experience an EBD prior to graduation. Still, fewer than 1% are identified and receive services for this classification (U.S. Department of Education, 2021), highlighting that the system may fail to adequately serve a large number of students with or at risk for EBD. Even students who do receive evidence-based supports in schools (e.g., behavior intervention plans, check-in/check-out) to address behavioral needs often make limited progress (Bradley et al., 2008; Kauffman & Landrum, 2018). Scholars posit that even well-validated and commonly used behavioral approaches (e.g., multi-tiered systems of support, functional behavior assessments) may be less influential if individual processes, such as EF, are not taken into account when considering both students' functioning in school and the type of supports that best align with their developmental needs (Chen et al., 2020; Farmer et al., 2020). Thus, to address the paucity of EF research with students at risk for EBD in early elementary grades and inform early identification and intervention efforts, the purpose of our study is to (a) identify the EF profiles of students at risk for EBD using a person-oriented approach and (b) determine the extent to which EF profile membership relates to distinct patterns of school functioning.
EF and Student-Related Competencies
Multiple theories have shaped EF conceptualization (Nigg, 2017), yet there is general agreement “EF” is an umbrella term for a set of top-down, neurocognitive processes that support goal-directed behaviors (Zelazo et al., 2016). The distinct but interrelated processes that make up EF include inhibitory control, working memory, and cognitive flexibility (Miyake et al., 2000), which develop throughout childhood and early adulthood (Best & Miller, 2010). Inhibitory control is the conscious controlling of a habitual or prepotent response, such as when a student refrains from aggression during an altercation or abstains from calling out in class. Working memory involves temporarily keeping information in mind and then using it, such as when a student remembers and uses rules to work through a social disagreement with a peer or to complete an assignment. Cognitive flexibility consists of shifting thoughts and behaviors, such as when a student imagines a friend's perspective or thinks of multiple ways to solve a problem. As such, EF is central to navigating the complexities and shifting demands of school and life.
Over 20 years of research has underscored EF's foundational role in not only students’ learning and academic achievement but also their behaviors, social interactions, and language abilities. EF is a well-known positive predictor of students’ long-term social and behavioral competence as well as mental well-being and physical health (Moffitt et al., 2011; Schoemaker et al., 2013; Zelazo, 2020). Those with EF difficulties are at risk of developing maladaptive behaviors (Schoemaker et al., 2013; Ogilvie et al., 2011), social difficulties (Tseng & Gua, 2013), and mental health problems (Zelazo, 2020). This pattern is evident as early as preschool (Schoemaker et al., 2013), with lasting implications for problematic behaviors in elementary school (Nelson et al., 2018) and for subsequent school dropout (Fitzpatrick et al., 2015). Furthermore, EF has been found to be associated with students’ language abilities (Kaushanskaya et al., 2017), which places those with challenges at increased risk for developing behavioral problems (Yew & O’Kearney, 2013). Yet, despite evidence students with or at risk for EBD display difficulties in behavioral competence, social interactions, and language skills (Bradley et al., 2008; Hollo et al., 2019), there is a dearth of EF research with these students (e.g., Cumming et al., 2019; Mattison et al., 2006), especially in early elementary school.
Given that early elementary school years comprise an active developmental stage for EF (Best & Miller, 2010), as well as a pivotal early identification and intervention period (Bradley et al., 2008), we wanted to examine the EF strengths and weaknesses of students identified as at risk for EBD using a person-centered approach. A person-centered approach highlights development as a bidirectional process of ongoing alignments and adjustments between individual processes (e.g., EF) and contextual experiences (e.g., relationships, instruction) that work together as a dynamic system (Farmer et al., 2020, 2021; Sameroff, 2020). For students with EBD, Farmer et al. (2021) proposed merging developmental-cascade and correlated-constraints perspectives when identifying patterns related to (a) emergence and (b) establishment or maintenance of emotional and behavioral problems. From a developmental-cascade perspective, difficulties in one domain are likely to contribute to challenges in other domains (Farmer et al., 2020, 2021), such that students with EF difficulties are likely to have behavioral and social challenges in school, placing them at potential risk for onset and escalation of behaviors leading to EBD identification.
From a correlated-constraints perspective, over time, difficulties in one domain (i.e., EF) may contribute and support problems in other domains (e.g., behavior, social interactions, language), such that a student's developmental system is organized to maintain problematic behaviors associated with EBD (Farmer et al., 2021). Identifying unique profiles of EF and how these relate to specific patterns of behaviors and functioning in the classroom, allows for a deeper understanding of varying levels of risk for EBD to inform identification and prevention and intervention efforts.
Methodologically, person-centered approaches rely on latent class or latent profile analysis (LPA) to identify groups that share unique patterns of characteristics (Farmer et al., 2020, 2021), which can provide insight into whether some groups of children have different outcomes than other groups. Previous studies using LCA or LPA offer valuable understanding about the behavioral and academic profiles of student subgroups with or at risk for EBD. For example, Gage (2013) found four profiles of students with EBD based on teacher reports of problem behaviors. Hollo et al. (2019) identified a relationship between language profiles and behavioral characteristics (e.g., externalizing, internalizing) among male students with EBD. King et al. (2019) found that three profiles identified academic subgroups among students at risk for EBD. These studies provide valuable information on patterns of behavioral manifestations, language-related needs, and academic difficulties of students with or at risk for EBD; however, researchers have not collected data on the EF processes that are suggested to underlie these outcomes.
Measuring EF to Identify EF Profiles of At-Risk Students
To gain insight into the EF profiles of students at risk for EBD in early elementary school and related patterns of school functioning, a comprehensive evaluation of EF skills is needed. Current EF measures are based largely on EF theoretical models (e.g., Miyake et al., 2000), with scholars assessing skills through either direct (i.e., performance based) or indirect (i.e., rating scale) measures. Researchers have viewed performance-based EF tasks (e.g., Head-Toes-Knees-Shoulders [HTKS]; Ponitz et al., 2009) that are administered directly to children under standardized conditions as the gold standard. Because these tasks tap into limited components of EF for a short time, some scholars argue they do not capture the multidimensional processes or behaviors often demonstrated in real-world situations (Fuhs et al., 2015). In contrast, rating scales or questionnaires gather various informant (e.g., parents, teachers) perceptions of children's behavioral and affective manifestations of EF (e.g., emotional regulation) across home and school settings (Gioia et al., 2015), yet they may be limited by informant bias (Duckworth & Yeager, 2015). Studies have found weak relationships between performance-based assessments and EF ratings scales (see Toplak et al., 2013, for a review), with scholars concluding these likely assess different underlying constructs, such that performance-based tasks tend to capture “cool” cognitive EF (e.g., inhibitory control; Zelazo & Carlson, 2012), whereas rating scales likely assess “contextualized” EF skills in everyday settings (Isquith et al., 2014). As such, Isquith et al. (2014) underscore the importance of using both performance-based assessments and rating scales to accurately capture students’ EF strengths and weaknesses.
Purpose
Despite a robust body of research on the EF of individuals with mental health and behavioral problems (Zelazo, 2020), few studies have been conducted with students at risk for EBD in early elementary school settings (e.g., Cumming et al., 2019), and to our knowledge, no prior studies have investigated EF using a person-centered approach with this population of students. To address this gap in the literature, and to provide researchers and practitioners with a comprehensive understanding of their diverse needs, we posed the following research questions:
Do students in kindergarten and first grade at risk for EBD demonstrate unique EF profiles based on teacher reports and performance-based EF measures? Is there a difference in performance on assessments of behavioral functioning, social-emotional competencies, and language skills among EF profiles? To what extent do EF profiles moderate relationships between student demographic covariates (i.e., gender, grade) and behavioral functioning, social-emotional competencies, and language skills?
We hypothesized students at risk for EBD would demonstrate unique at-risk EF profiles (i.e., EF difficulties) based on both teacher report and performance-based EF. We expected EF profiles would be associated with students’ functioning and adjustment in school, such that students with increasing levels of at-risk EF profiles (indicating EF difficulties) would demonstrate greater challenges with behavioral, social-emotional, and language skills. Last, given that male students tend to demonstrate higher levels of externalizing behavior problems than female students (Bertrand & Pan, 2013), and students in first grade tend to generally have more established EF skills than kindergarten students (Best & Miller, 2010), we anticipated EF profiles would moderate relationships between student demographic covariates and functioning.
Method
Sample and Participants
Data for the current study came from a 3-year randomized controlled trial examining the efficacy of the Social-Emotional Learning Foundations (SELF) curriculum (Daunic et al., 2021), which is a Tier 2 (i.e., targeted) literacy-embedded social-emotional program for kindergarten and first-grade students identified as at risk for EBD. To select eligible students, teachers used Stages 1 and 2 of the Systematic Screening for Behavior Disorders (SSBD-2; Walker & Severson, 1992), which uses a standardized process for screening students at risk for EBD. Stage 1 provides research-based definitions of externalizing (e.g., defiant) and internalizing (e.g., withdrawn) behavioral dimensions. After reading these definitions, teachers selected up to five students from their class rosters for each dimension based on their assessments of each student's behavioral characteristics. In Stage 2, teachers completed the Combined Frequency Index and Critical Events Index to nominate the top two or three students on each dimension (internalizing and externalizing) for possible involvement in the study. Research has demonstrated the psychometric properties of the SSBD-2, including test-retest reliability (ρ = .74–.88), internal consistency (α = .84–.86), interrater reliability (ρ = .72–.94), and convergent validity (see Glover & Albers, 2007).
Across 3 years, 304 teachers (163 kindergarten, 141 first grade) and 1,154 students (627 kindergarten, 527 first grade) participated in the project. For the current study, data from 1,141 participants were used in Step 1 (profile enumeration), which were later reduced to 1,071 in Step 3 (structural associations between covariates and distal outcomes) of the LPA due to missing data. Schools were selected for treatment or control condition via random assignment. Some districts did not provide student demographic information, resulting in demographic data being reported for 94% of the total sample. Most students were male (62%), with 64% identified as White, 20% as African American, 9% as Hispanic, and 7% as Other. Just over half were in kindergarten (54%), 3% were English language learners, and 20% received special education services. Students identified with a developmental delay had been excluded to eliminate a confounding variable in assessing treatment efficacy. All students came from one southeastern state and from 52 schools; 81% were eligible for free or reduced-price lunch.
Measures
Behavior Rating Inventory of Executive Function–Second Edition (BRIEF-2)
The BRIEF-2 Teacher Form assesses student emotion, behavior, and cognitive regulation (Gioia et al., 2015). Teachers completed items using a Likert-type scale from 1 (never) to 3 (almost always), with higher scores indicating EF challenges or dysfunction. The 63 items comprise nine scales intended to capture real-world manifestations of EF component skills: Inhibit, Self-Monitor, Shift, Emotional Control, Initiate, Working Memory, Plan/Organize, Task-Monitor, and Organization of Materials. These nine scales form three indices: Behavior Regulation, Emotion Regulation, and Cognitive Regulation. Yet, from a previous validation study of the BRIEF-2 with the current sample (Cumming et al., 2021), we found little support for the original three-index model. Instead, we found support for a modified 53-item model made up of nine scales and two indices: Cognitive Regulation Index and Behavior-Emotion Regulation Index. We noted changes in number of items and item-scale associations across six scales. Further, criterion-related validity analyses indicated positive relationships among indices and performance-based EF and internalizing and externalizing behaviors. Therefore, as a more parsimonious assessment of EF for at-risk students, we used the modified 53-item model for all analyses in the current study. Sample-based BRIEF-2 scales evidenced strong internal consistency (omega coefficients ranged from .88 to .99).
HTKS task
We administered the HTKS (Ponitz et al., 2009), a performance-based measure of overall EF (i.e., cognitive flexibility, working memory, inhibitory control) intended for use among students ages 4 to 6 years old. It is delivered in a game-like fashion. In the first 10 trials, children respond to the directions “Touch your toes” and “Touch your head,” followed by instructions to respond in the opposite manner (i.e., touch your head when asked to touch your toes). Children are then directed to touch their knees or shoulders, again followed by the opposite. In the second set of 10 trials, the assessors present students with all four possibilities (head, toes, knees, or shoulders) and ask them to respond in the opposite way. Each attempt is scored as 0 (incorrect), 1 (self-correct), or 2 (correct), with total scores ranging from 0 to 52 (six practices and 20 trials); higher scores indicate higher EF. In our sample, the HTKS showed strong internal consistency (α = .95). Previous psychometric evaluations have demonstrated adequate test-retest reliability (r = .74), interrater reliability (κ = .79), predictive validity across multiple behavioral and academic outcomes (d = .02–.87), and construct validity with related measures (rs = .31–.56) (McClelland et al., 2014; Ponitz et al., 2009).
Clinical Assessment of Behavior–Teacher Report (CAB-T)
To measure students’ school adjustment and behavioral functioning, teachers completed the CAB-T (Bracken & Keith, 2004). The CAB-T includes 70 items teachers rate from 1 (always or very frequently) to 5 (never) across four scales: Internalizing (e.g., “cries easily”), Externalizing (e.g., “is disruptive”), Social Skills (e.g., “annoys others”), and Competence (e.g., “difficulty following directions”). Lower scores indicate difficulties. Alphas for each scale were excellent (.88, .97, .93, and .93) for our sample. Previous studies have established validity (e.g., construct; Bracken & Keith, 2004).
Devereux Student Strengths Assessment (DESSA)
The DESSA (LeBuffe et al., 2009) was used to measure students’ social-emotional competencies and includes 72 items. Teachers rate items on a 5-point Likert-type scale from 0 (never) to 4 (very frequently), with higher scores reflective of strengths. The instrument includes eight scales: Personal Responsibility, Optimistic Thinking, Goal-Directed Behavior, Social Awareness, Decision Making, Relationship Skills, Self-Awareness, and Self-Management. In our sample, alphas varied from .88 to .94 across scales. Psychometric studies have established construct and criterion validity, test-retest reliability (r values > .90), and interrater reliability (r values > .78; LeBuffe et al., 2009).
Clinical Evaluation of Language Fundamentals, Fourth Edition (CELF-4)
The CELF-4 (Semel et al., 2003) is a direct assessment of students’ language content, structure, and use. The measure is comprised of 18 separate subtests, administered to students individually. We used raw scores from the Expressive Vocabulary (EV) and Understanding Spoken Paragraphs (USP) subtests in this study. In the EV subtest, students are presented with 27 pictures to identify verbally (e.g., “What is this?”). Sixteen questions are scored only as 2 (correct response) or 0 (incorrect response), but 11 questions can also receive a score of 1 (descriptive approximation of correct response). Total scores range from 0 to 54. The EV subtest includes a discontinue rule (i.e., administration stops after seven consecutive incorrect responses). For the USP subtest, students listen to three short stories consisting of seven or eight sentences and then verbally answer five questions about each story (e.g., “What did Marcus get when he went shopping?”). Answers are scored as 1 (correct) or 0 (incorrect), with total scores ranging from 0 to 15. Psychometric evidence for the CELF-4 subtests includes test-retest reliability (.60 to .90), interrater reliability (.90 to .98), and construct validity (Semel et al., 2003). Sample derived alphas were .85 (EV) and .79 (USP).
Data Analyses
To answer our research questions, we used LPA. Inspired by Morin et al. (2016), we considered HTKS (direct behavioral measure of EF; Ponitz et al., 2009) as the general indicator of the level effect of EF profiles (analogous to the general factor in Morin et al., 2016) and nine BRIEF-2 (Gioia et al., 2015) scales as specific indicators of the shape effect of EF profiles (similar to the six dimensions of the psychological health measure in Morin et al., 2016). Grade (0 = kindergarten, 1 = first grade) and gender (0 = male, 1 = female) were covariates. See Figure 1 for a conceptual model. Table S1, located in the supplemental material, contains a summary of model variables. Using pretest data only (control and treatment groups), we included scores from behavioral functioning (CAB-T; Bracken & Keith, 2004), social-emotional competencies (DESSA; LeBuffe et al., 2009), and language skills (CELF-4; Semel et al., 2003) as outcomes. Although all data were collected at the same time point, we use the term “outcomes” to highlight analysis of dependent variables after EF profiles were identified. Due to the complexity of our model and the use of continuous outcomes, we adopted the manual Bolck-Croon-Hagenaars (BCH) method (Bolck et al., 2004). This method is recommended for continuous outcomes (Asparouhov & Muthén, 2021) and is more resistant to shift in class membership (Nylund-Gibson et al., 2019). All analyses were conducted with Mplus 8.4 using robust maximum-likelihood estimator (deidentified data and codes are available upon request). To acknowledge the nesting structure of data, we specified CLUSTER = Teacher ID and TYPE = MIXTURE COMPLEX to account for dependence of observations.
Visual representation of the latent profile model with covariates and behavioral functioning, social-emotional competencies, and language skills outcomes. Note. C denotes the latent profiles enumerated from the executive function indicators (left-hand side of this figure). Grade and gender are covariates. Behavior measured by Clinical Assessment of Behavior scales: Internalizing, Externalizing, Social Skills, Competence. Social-emotional competencies measured by Devereux Student Strengths Assessment scales: Self-Awareness, Self-Management, Social Awareness, Relationship Skills, Goal-Directed Behavior, Personal Responsibility, Decision Making, Optimistic Thinking. Language skills measured by Clinical Evaluation of Language Fundamentals: Expressive Vocabulary, Understanding Spoken Paragraphs. HTKS = Head-Toes-Knees-Shoulders.
The BCH approach consists of three steps: class enumeration (Step 1), calculation of classification errors (Step 2), and estimation of structural relationships between variables (Step 3). In Step 1, we fit a series of unconditional latent profile models to identify a model that was empirically adequate and theoretically interpretable. As recommended (Nylund-Gibson & Choi, 2018), we began with one latent profile and used this as a baseline model. Then we consecutively fit models with one extra profile until a model presented inadequate fit. We followed Asparouhov and Muthén's (2012) guidelines by (a) fitting a latent profile model with varied random starting values to replicate the best loglikelihood value, and (b) based on the replicated loglikelihood, adding TECH11 to request the Lo-Mendell-Rubin (LMR) likelihood ratio test (LRT) that compares the analyzed latent profile model with a nested model with one fewer profile. It is worth noting that we did not request the TECH14 output for bootstrapped LRT described as the follow-up of TECH11 in Asparouhov and Muthén (2012); TECH14 is not available for a specification of TYPE = MIXTURE COMPLEX. Collectively, along with the relative fits (Akaike information criterion [AIC; Akaike, 1987], consistent AIC [CAIC; Bozdogan, 1987], Bayesian information criterion [BIC; Schwarz, 1978], sample-size-adjusted BIC [SABIC; Sclove, 1987], approximate weight of evidence [AWE; Banfield & Raftery, 1993]) and the entropy, we reviewed these statistics to assess model fit. Specifically, a statistically significant LRT indicates the analyzed model fits better than the model with one fewer profile, a smaller value on information criteria (IC) suggests better fit, and an entropy above .80 indicates acceptable classification quality (Weller et al., 2020). Following Nylund-Gibson and Choi (2018), we plotted values of IC for an intuitive comparison of change in IC across examined LPA models (see Supplemental Figure S1). Depending on where the “elbow” appears, this plot provides visual clues to determine the appropriate (range of) number of latent profiles.
In Step 2, classification errors were automatically calculated in Mplus and BCH weights saved using SAVE = BCHWEIGHTS. Last, in Step 3, we used these weights to fit the structural model with covariates and outcomes. Note: we did not preregister this research.
Missing Data
As shown in Supplemental Table S1, the percentages of missingness varied from 0.09% to 6.76% across variables, which was within the acceptable range of missingness of 5% to 10% (Lodder, 2013). We also examined the missing mechanism based on Jamshidian and Jalal's (2010) steps using MissMech package (Jamshidian et al., 2014) in R. The results of the nonparametric test were not statistically significant (p = .29), suggesting data were missing at random.
Results
Enumeration of EF Profiles
As shown in Table 1, we began our analysis with a baseline model with one latent profile (1-LPM) and then fitted additional models with varied numbers of latent profiles. The 1-LPM had the largest values on IC (AIC = 62677.02, CAIC = 62797.81, BIC = 62777.81, SABIC = 62714.28, AWE = 62807.81), with both 2- and 3-LPMs presenting larger decline in IC values. Starting from the 4-LPM, increasing the number of latent profiles led to a relatively smaller decrease. From the plot of IC (Supplemental Figure S1), there was a relatively clear “elbow” at 2-LPM, a nearly clear “diminishing point” (Nylund-Gibson & Choi, 2018) at 3-LPM, and a reduced decrease in IC when increasing the number of latent profiles from three to four or more. This suggested support for two- or three-profile models. We then reviewed the LMR that enables a comparison against the LPM with one fewer profile. As shown in Table 1, the results of LMR were statistically significant for LPMs of two to six latent profiles only (2-LPM to 6-LPM in Table 1). This indicated the two- to six-latent-profile models fit better than their one-fewer-profile model. Next, acceptable entropy was found on all LPMs except for the baseline model (i.e., no entropy is produced for 1-LPM), indicating adequate classification quality. Last, for each LPM, we reviewed model estimated indicator means and profile visualizations. Following recommendations by Nylund-Gibson and Choi (2018) and Weller et al. (2020), we considered the statistics of model fit, classification quality, and theory (e.g., Isquith et al., 2014), as well as previous LPA EF research with young children (Litkowski et al., 2020; Sasser et al., 2017). On the basis of these, we proceeded with the 3-LPM. Supplemental Table S2 contains model estimated means of EF indicators.
Summary of Fit Statistics of Latent Profile Models.
Note. AIC = Akaike information criterion; CAIC = consistent Akaike information criterion; BIC = Bayesian information criterion; SABIC = sample-size-adjusted Bayesian information criterion; AWE = approximate weight of evidence; LMR = Lo-Mendell-Rubin adjusted likelihood ratio test; −2LL = likelihood ratio test statistic.
Latent profile models are labeled as [number of profiles]-LP.
Prevalence of EF Profiles Based on 3-LPM
For descriptive statistics of auxiliary variables of the 3-LPM and visual comparison of profiles of the accepted model and model estimated means of EF indicators for each profile, see Table 2, Supplemental Table S2, and Figure 2, respectively. Overall, as HTKS score decreased in each profile (indicating lower EF performance), BRIEF-2 scale scores increased correspondingly (indicating greater EF difficulty). This relationship, according to Morin et al. (2016), indicated the potential of HTKS scores to determine the tendency of students to belong to a particular EF profile (level effect) and the added value of BRIEF-2 scale scores to demonstrate how they differ among the specific dimensions of EF (shape effect), as illustrated in Supplemental Figure S2. Based on BRIEF-2 scales Emotional Control, Initiate, and Working Memory, we named each profile according to its means and the clinical ranges suggested in the BRIEF-2 manual (Gioia et al., 2015) to maximize interpretability. Specifically, we (a) calculated the mean raw scores of the three BRIEF-2 scales for male and female students, (b) converted the scores using a t score scale (M = 50, SD = 10), and (c) compared the t score to the BRIEF-2 manual suggested clinical ranges: mildly elevated (60–64), potentially clinically elevated (65–69), and clinically elevated (70 or above). We were unable to convert raw scores to t scores for the remaining scales as these were modified BRIEF-2 scales (see Method section and Cumming et al., 2021, for details). For Profile 1, none of the three scales presented a score within clinical ranges. Hence, we named this profile as the “mildly at-risk EF” group. For Profile 2, one BRIEF-2 scale displayed a t score within the range of 60 and 64 (Emotional Control = 61) for male students. For female students, one BRIEF-2 scale displayed a mildly elevated t score (Initiate = 62), one scale presented a potentially clinically elevated t score (Working Memory = 65), and one scale showed a clinically elevated t score (Emotional Control = 72). Accordingly, we titled this profile as “moderately at-risk EF.” For the third profile, a clinically elevated t score (70 or above) was displayed on the majority of BRIEF-2 scales except for the Initiate scale for male students (Initiate = 69, potentially clinically elevated). We labeled this profile as “clinically at-risk EF.” Conversions are detailed in Table 3. As shown in Table 2, there were 301 students (26.38%) in the mildly at-risk EF profile, 437 (38.30%) in moderately at-risk EF profile, and 403 (35.32%) in clinically at-risk EF profile.
Visual comparison of profiles from the three-profile model. Note. Model-estimated mean scores of executive function profile indicators are plotted; the exact values are included in the table underneath the line plot. Of the 10 profile indicators, Inhibit, Self-Monitor, Shift, Emotional Control, Initiate, Working Memory, Plan/Organize, Task-Monitor, and Organization of Materials are Behavior Rating Inventory of Executive Function–Second Edition scales. HTKS = Head-Toes-Knees-Shoulders.
Descriptive Statistics of Covariates and Behavioral Functioning, Social-Emotional Competencies, and Language Skills Based on Three-Profile Model.
Note. EF = executive function; CAB = Clinical Assessment of Behavior; INT = Internalizing; EXT = Externalizing; SOC = Social Skills; COM = Competence; DESSA = Devereux Student Strengths Assessment; SEA = Self-Awareness; SEM = Self-Management; SOA = Social-Awareness; RS = Relationship Skills; GDB = Goal-Directed Behavior; PR = Personal Responsibility; DM = Decision Making; OT = Optimistic Thinking; CELF = Clinical Evaluation of Language Fundamentals; EV = Expressive Vocabulary; USP = Understanding Spoken Paragraphs; K = kindergarten; M = male; F = female.
Mean Raw Scores, T Scores, and Clinical Groups Based on Manual-Suggested Score Ranges.
Note. HTKS = Head-Toes-Knees-Shoulders; ME = mildly elevated (60–64); PE = potentially clinically elevated (65–69); CE = clinically elevated (70 or above) according to Behavior Rating Inventory of Executive Function–Second Edition manual.
Behavioral, Social-Emotional, and Language Skills Differences Between EF Profiles
To determine whether distinct differences in behavioral functioning, social-emotional competencies, and language skills emerged between EF profiles, we compared mean profile differences for each scale of the CAB, DESSA, and CELF (see Table 4). Further, we used Cohen's (1992) guidelines to interpret strength of the relationship (.10 = small, .30 = medium, .50 = large). Overall, all pairwise comparisons were statistically significant. For measures of problematic and adaptive behaviors, students in the mildly at-risk EF profile showed a higher mean score compared with the other two EF profiles. Those from the moderately at-risk EF profile had a higher mean score than students in the clinically at-risk profile. Across the four CAB scales, a medium-to-large effect size was found, with d ranging from 0.38 to 4.28.
Pairwise Comparisons of Means of Behavioral Functioning, Social-Emotional Competencies, and Language Skills.
Note. MD = mean difference; EF = executive function; CAB = Clinical Assessment of Behavior; DESSA = Devereux Student Strengths Assessment; CELF = Clinical Evaluation of Language Fundamentals.
*p < .05. **p < .01. ***p < .001.
For measures of social-emotional competencies and language skills, we observed similar elevations in mean differences. The mildly at-risk EF profile presented a significantly higher mean score than the moderately at-risk EF profile, which in turn showed a higher mean score than the clinically at-risk EF profile. Values for d varied from 0.51 to 3.27 (DESSA scales) and from 0.29 to 0.76 (CELF scales), which indicated a large effect size on profile differences of social-emotional competencies and a moderate-to-large effect on language skills.
Covariate Effects on Student Functioning Based on EF Profile
We included auxiliary variables (gender, grade, outcomes) in the 3-LPM and estimated the effects of gender and grade on outcomes. Before proceeding to model estimates of structural paths, we reviewed the prevalence of profile membership and compared it with the profile prevalence from the profile enumeration. This comparison did not signify any marked shift in profile prevalence, suggesting profile membership persisted over the BCH approach steps. As presented in Supplemental Table S3, for the mildly at-risk EF profile, gender had a significant effect on the Social Skills (β = 3.38, p = .02), Self-Management (β = 1.65, p = .01,), and Relationship Skills (β = 2.09, p = .02) scales, suggesting female students within this EF profile demonstrated stronger skills in these areas than male students. Within the same EF profile, grade had an effect on Social Skills (β = −3.44, p = .04), Social Awareness (β = –1.95, p = .02), Relationship Skills (β = –2.53, p = .03), and Decision Making (β = –2.09, p = .01) as well as on Expressive Vocabulary (β = 5.88, p < .05) and Understanding Spoken Paragraphs (β = 2.04, p < .05) scales. Overall, first graders scored lower than kindergarteners in the same profile on measures of behavioral and social-emotional competence but scored higher on language skills.
For the moderately at-risk EF profile, gender displayed significance on the Internalizing scale (β = −3.61, p = .01), indicating female students had more internalizing behaviors than male students in this profile. Grade had an effect on Internalizing (β = –2.94, p = .01), Self-Awareness (β = 1.49, p = .01), Expressive Vocabulary (β = 9.04, p < .05) and Understanding Spoken Paragraphs (β = 2.97, p < .05) scales. Overall, first graders scored lower than kindergarteners on internalizing behaviors but higher on self-awareness and language skills.
For students in the clinically at-risk EF profile, female students scored lower on the Internalizing scale (β = –4.41, p < .05) but higher on the Externalizing scale (β = 4.48, p = .03), suggesting female students demonstrated more internalizing but less externalizing behaviors than male students. Female students also scored lower on the Expressive Vocabulary scale (β = –2.51, p = .02). Yet, no significant effects of gender were found on social-emotional competencies. For grade, significance was found on Expressive Vocabulary (β = 8.29, p < .05) and Understanding Spoken Paragraphs (β = 3.37, p < .05), with first graders scoring higher on language skills. No effects of grade were found on behavioral or social-emotional measures.
Discussion
Given the (a) overall low identification rates and negative outcomes of students with EBD (Bradley et al., 2008; Bruhn et al., 2014), (b) EF's role in emotional and behavioral functioning (Zelazo, 2020), and (c) early elementary grades being an active EF developmental (Best & Miller, 2010) and prevention and intervention period (Bradley et al., 2008), our purpose was to determine whether kindergarteners and first graders at risk for EBD demonstrated unique EF profiles. Also, our intent was to determine whether profile membership was related to distinct patterns of students’ behavioral functioning, social-emotional competencies, and language skills. In alignment with Isquith et al.'s (2014) advocacy for both rating scales and performance-based measures of EF to provide a comprehensive picture of students’ EF strengths and weaknesses, we found scores on BRIEF-2 scales and HTKS served as significant indicators of distinctive EF profiles, such that three latent EF profiles emerged: mildly, moderately, and clinically at-risk. As expected, students at risk for EBD who were rated better by teachers on their use of EF skills in the classroom (BRIEF-2) and performed higher on a performance-based EF task (HTKS) than at-risk peers formed a mildly at-risk EF group. Students who demonstrated mildly to moderately elevated problematic EF scores across multiple areas formed a moderately at-risk group, whereas those who consistently were rated as demonstrating problematic contextualized EF skills in the classroom (e.g., trouble resisting impulses) and performed poorly on a performance-based EF task emerged as a critically at-risk EF group. These findings are compelling given that they align with the BRIEF-2 manual's (Gioia et al., 2015) guidelines for clinically elevated ranges. As such, students at risk for EBD in our sample demonstrated varying levels of EF difficulties based on heterogeneous EF groups, which aligns with robust research on relationships between EF and behavior and mental health (e.g., Zelazo, 2020), as well as emerging EBD research (e.g., Cumming et al., 2019).
Although this is the first known study to examine latent EF profiles of kindergarten and first-grade students at risk for EBD, our findings lend support to previous LPA studies. For example, in a study that used latent class growth analysis among preschool children from low-income families, Sasser et al. (2017) found a three-class solution, with low, moderate, and high EF trajectories. Litkowski et al. (2020) also found three EF classes of low, average, and high EF performance levels among kindergarten children based on direct assessments and teacher reports of EF. Yet, they also found two additional classes demonstrated discordant performance between teacher report and direct EF measures. The distinction with our findings is likely due to our sample including only students at risk for EBD, whereas Litkowski et al. used a nationally representative sample that likely included a wide range of behaviors and related EF abilities.
Results indicated student EF profiles were associated with similar patterns of functioning and adjustment in school but demonstrated distinct elevation levels. On average, students who placed in the mildly at-risk EF profile scored significantly better on all these skills than students who placed in EF profiles considered moderately and clinically at risk. In turn, students in the clinically at-risk EF profile demonstrated numerous pronounced areas of need, with effect sizes ranging from medium to large (Cohen, 1992). Thus, in our sample of kindergarteners and first graders at risk for EBD, students who fell within increasing at-risk EF profiles displayed more problematic behavioral, social, and language abilities. These findings align with a large literature base where researchers found children with EF challenges had greater difficulty suppressing a dominant response, which was predictive of aggression (Poland et al., 2016); curbing negative thoughts that led to anxiety and depression (Kertz et al., 2016); navigating challenging social situations (McQuade et al., 2013); and shifting from responding negatively (Schoemaker et al., 2013). Further, adding to the literature on language impairments and students with EBD (Hollo et al., 2019), our study is the first to our knowledge to connect distinct language skills to at-risk EF profiles with students at risk for EBD.
We found there were distinct behavioral functioning, social-emotional competencies, and language skills for female and male students, as well as kindergarteners and first graders, depending on whether they were in the mildly, moderately, or clinically at-risk EF profile. Overall, we found female students in the clinically at-risk EF profiles demonstrated fewer externalizing behaviors and more internalizing problems than male students in both the moderately and clinically at-risk EF profiles. Additionally, female students in the moderately and clinically at-risk EF profiles demonstrated lower expressive vocabulary than male students. These findings are in line with research demonstrating that female students tend to display fewer externalizing and greater internalizing behaviors compared with male peers (Bertrand & Pan, 2013), with evidence from our study suggesting EF likely plays an active role. It may well be that less established EF skills related to controlling and shifting away from ruminating thoughts may place female students at higher risk for depression and anxiety-related behaviors (e.g., Kertz et al., 2016) as well as limit their ability to express themselves.
Overall, first graders in the moderately at-risk EF profile demonstrated lower skills across multiple social-emotional competencies than kindergarteners, whereas those in the moderately and clinically at-risk EF profiles performed better on language skills than kindergarteners with the same profiles. Given that language and EF develop over time (e.g., Best & Miller, 2010), we had anticipated first graders would demonstrate better language skills than kindergarteners. Yet, differences in social-emotional competencies based on grade were mostly noted in the mildly at-risk EF profile. It may be students in this profile exhibit more heterogeneity across these skills, which suggests additional research is warranted. Overall, aligned with dynamic systems theory and a person-centered approach, grade and gender emerged as salient covariates.
Limitations and Future Directions
The current study has numerous strengths, including a large sample of students identified as at risk for EBD, the use of both teacher ratings and performance-based measures of EF, and data from well-validated measures of behavioral functioning, social-emotional competencies, and language skills. However, there are limitations to consider. First, LPA is an exploratory analytic approach; readers should use caution when interpretating results and their implications. Second, this first exploratory study with students at risk for EBD is restricted by our inclusion of participants from one state only, the majority of whom were receiving free or reduced lunch, thus limiting generalizability. Research teams should consider including participants from a diversity of geographic regions and socioeconomic backgrounds in future studies.
We recognize there are possible limitations with chosen measures. First, use of teacher reports of students’ behavioral functioning, although efficient, can increase the risk of teacher bias (Duckworth & Yeager, 2015). Thus, we recommend future studies incorporate classroom observations to corroborate findings. Although the HTKS is a well-validated measure, Gonzales et al. (2021) developed a revised version (HTKS-R) that demonstrated improved validity. Future studies should use the HTKS-R to assess whether different EF profiles emerge.
Future studies could also examine other potentially important factors. For instance, we did not include any academic measures in our model that investigators should consider, given the established link between EF and academics (Spiegel et al., 2021). We also did not collect information on family variables, such as parenting practices (e.g., warmth cognitive stimulation), or other ecological variables (e.g., classroom quality) that have been shown to shape EF development (Cumming, Poling, et al., 2022; Cumming, Zelazo, et al., 2022; Valcan et al., 2018). We recommend researchers include these as prime areas for examination in prospective studies.
We analyzed data from a single time point at pretreatment, which precluded an examination of profile trajectories over time; therefore, researchers should consider longitudinal data collection to understand the EF profiles of students as they age, given EF develops over time (Best & Miller, 2010). The use of longitudinal data could also uncover whether students move between profiles, which can be assessed through latent transition analysis (see Lanza & Rhoades, 2013). Understanding these possible developmental changes in EF and associated school functioning could inform the provision of prevention and intervention supports during the early elementary grades, which are particularly salient for students with clinically significant EF that places them at heightened risk for EBD. Last, we encourage researchers to develop targeted interventions focused on distinct EF profiles for early-elementary-grade students at risk for EBD.
Implications for Policy and Practice
Given its malleability and its foundational role in student learning, EF is a promising leverage point for improving the behavioral, social-emotional, and language functioning of kindergarten and first-grade students at risk for EBD.
Identifying how EF profiles among students at risk for EBD relate to behavioral and social functioning in the classroom has implications for early identification efforts and school-based programming for kindergarteners and first graders. For practitioners, we situate our findings within a tiered system of adaptive supports (TSAS; see Farmer et al., 2021), which merges developmental-cascades and correlated-constraints perspectives to guide the individualization of supports provided to students at risk for EBD within a multitiered systems approach. A core emphasis of TSAS is that tiered intervention programming should be tailored and adapted to students’ unique needs and strengths.
To identity distinct needs and strengths based on unique EF profiles, we recommend school professionals engage in initial screening and systematic progress monitoring that includes measures of EF to complement behavioral, social, and language measures. Specifically, both performance-based measures (e.g., HTKS; Gonzalez et al., 2021) and rating scales (e.g., BRIEF-2; Gioia et al., 2015) should be used to provide a holistic assessment of EF.
For students in the mildly at-risk profile, we suggest practitioners consider Tier 1 supports aimed at promoting student EF throughout the day. Based on a TSAS framework, Tier 1 involves universal and adaptable supports designed to promote everyday functioning (Farmer et al., 2020, 2021). For example, classwide activities that require children to recruit and practice EF skills, including circle games (e.g., Red Light, Green Light and Freeze Game; see Tominey & McClelland; 2011), physical exercise (e.g., running, jumping), computerized and noncomputerized EF training, and mindfulness, have been found to be effective at fostering EF (Cumming, Zelazo, et al., 2022; Takacs & Kassai, 2019). These activities can be implemented throughout the school day (e.g., mindfulness prior to a test) and as part of morning meeting (see Rimm-Kaufman & Chiu, 2007). Morning meeting is a classwide practice that increases the emotional support of the classroom, provides students opportunities to develop and practice social-emotional skills, and facilitates positive teacher–student and peer relationships, which have all been found to be essential to student EF development (see Cumming et al., 2020).
Students in the moderately at-risk EF profile would benefit from TSAS Tier 2 programming, which is meant to prevent the negative reorganization of the student's developmental system (Farmer et al., 2020, 2021). Specifically, practitioners can take an active role in building student EF while also strengthening other areas to prevent problematic behavioral, social-emotional, and language domains of functioning from escalating. For this tier, in addition to Tier 1, practitioners could incorporate a social-emotional learning program, such as Daunic et al.'s (2021) SELF program, which is a Tier 2 curriculum for student at risk for EBD. SELF has been found to not only have a positive effect on student EF but also enhance social competence and behavioral functioning. To address other domains of possible risk associated with EF difficulties, the SELF curriculum could be coordinated with programs that specifically foster behavioral and relational support, such as check-in/check-out (Wolfe et al., 2016).
Students in the clinically at-risk EF profile, who experience substantial difficulties across multiple domains, would benefit from TSAS Tier 3 supports. The goal of Tier 3 intervention is to positively reorganize the developmental system by ameliorating risks and fostering strengths in an intensive and systematic way over an extended period of time (Farmer et al., 2020, 2021). This would include multifaceted programming involving a variety of professionals working together in collaboration. For instance, in addition to programming that builds EF and social-emotional competencies (e.g., SELF; Daunic et al., 2021), students in the high-risk EF profile would benefit from targeted speech-language supports and individualized behavioral plans. Further, to complement school-based efforts, school professionals should consider partnering with caregivers to facilitate their use of home-based practices known to foster EF development, such as cognitive stimulation and positive discipline (Valcan et al., 2018).
Last, institutes of higher education and school districts should focus on preservice and in-service training to improve school professionals’ understanding of EF and its role in student outcomes, particularly for students at risk for EBD. Such knowledge has potential to inform the types of supports offered in schools and problem solving when students make limited progress.
Conclusion
Students with early-onset behavior problems are at risk for persistent and long-lasting negative outcomes (e.g., school dropout, incarceration; Kauffman & Landrum, 2018). Our study, using a person-oriented approach, is the first to shed light on how differential EF profiles relate to distinct behavioral, social-emotional, and language skills among kindergarteners and first graders at risk for EBD.
Supplemental Material
sj-docx-1-ecx-10.1177_00144029221135573 - Supplemental material for Executive Function Profiles of Kindergarteners and First Graders at Risk for Emotional and Behavioral Disorders
Supplemental material, sj-docx-1-ecx-10.1177_00144029221135573 for Executive Function Profiles of Kindergarteners and First Graders at Risk for Emotional and Behavioral Disorders by Michelle M. Cumming, Daniel V. Poling, Yuxi Qiu, Debra A. Prykanowski, Aniva Lumpkins, Ann P. Daunic, Nancy Corbett and Stephen W. Smith in Exceptional Children
Footnotes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
