Early Pubertal Timing Mediates the Association Between Low Socioeconomic Status and Poor Attention and Executive Functioning in a Diverse Community Sample of Adolescents

Abstract

Low socioeconomic status (SES) may be associated with earlier pubertal timing and impaired attention and executive function (EF) in youth; however, whether pubertal timing mediates the relation between socioeconomic status and attention or executive functioning remains unclear. Structural equation models tested concurrent and prospective relations between SES, pubertal timing, and attention and executive functioning measures in a gender and racially diverse sample of adolescents (N=281, 45.6% male, 50.5% White/Caucasian, 46.3% Black/African American, 3.2% Biracial/other, and 44.5% low SES; complete data were not available on some measures). Youth from low socioeconomic status families experienced earlier pubertal timing, and this accelerated development was associated with worse performance on attention and executive functioning tasks, both concurrently and longitudinally. These findings highlight a pathway by which youth from low socioeconomic backgrounds may develop worse attention and executive functioning abilities during adolescence.

Introduction

There is substantial evidence that socioeconomic status has a striking effect on cognitive functioning in childhood and adolescence (Duncan & Magnuson, 2012), such that youth from low socioeconomic status backgrounds show markedly lower scores on standardized assessments of cognitive functioning than their same-aged peers from high socioeconomic status backgrounds. In fact, there is a 20-point (1.3 standard deviations) difference between low socioeconomic status and high socioeconomic status children on the Weschler Intelligence Scale for Children – Fifth Edition (Wechsler, 2014). These differences in cognitive functioning are evident in childhood and persist across development (Hackman, Gallop, Evans, & Farah, 2015) and are particularly evident on measures of executive functioning (Hackman et al., 2015). In fact, one study found that coming from a low socioeconomic status family predicted lower executive functioning scores in youth at age 2, and the difference in scores between low and high socioeconomic status youth nearly tripled by age 16 (von Stumm & Plomin, 2015). However, questions remain about the mechanisms of this association. The current study investigated pubertal timing as a potential mediator of the relation between low socioeconomic status and attention and executive functioning weaknesses in adolescence.

Executive functioning (EF) is a term used to describe higher-order cognitive abilities that enable individuals to orient towards the future, practice self-control, and engage in flexible, goal-directed behavior (Logue & Gould, 2014). There is broad agreement that executive functioning includes the ability to inhibit a behavior, hold and update information in working memory, and switch between tasks (Miyake & Friedman, 2012), and that these abilities are coordinated by activity in the prefrontal cortex (PFC; Best & Miller, 2010). However, there are a wide range of other functions that are primarily reliant on executive functioning, such as organization, planning, and interference control, that are not always featured in specific theories of executive functioning (Diamond, 2013; Miyake & Friedman, 2012). Deficits in executive functioning in childhood are associated with worse academic and occupational outcomes across adolescence and into young adulthood (Miller, Nevado-Montenegro, & Hinshaw, 2012) and have been shown to be associated with most types of psychopathology (Snyder, Miyake, & Hankin, 2015). In fact, it has been proposed that executive functioning deficits may represent a transdiagnostic mechanism of risk for emotional, behavioral, and psychotic disorders (Nolen-Hoeksema & Watkins, 2011). Therefore, it is essential to elucidate trajectories of risk for impairments in executive functioning in youth.

Attention often is referred to as the “gatekeeper” of executive functioning because performance on executive functioning tasks often requires the capacity to both sustain attention over extended periods of time as well as the ability to selectively attend to relevant, rather than irrelevant, stimuli (Awh, Vogel, & Oh, 2006). For example, performance on a classic measure of verbal auditory working memory (Digit Span Backwards) requires participants to sustain their attention for enough time to hear the information before they can hold it in mind and manipulate it. However, similar to executive functioning, there is considerable disagreement on precisely what processes are included within the domain of attention. Although there is broad agreement that attention includes simpler cognitive functions, such as alerting and orienting, it is unclear the degree to which more complex executive attention functions, such as sustained, switching, selective, and divided attention, overlap with executive functioning abilities (Miyake & Friedman, 2012). There is strong evidence that, although sustained attention likely overlaps considerably with EF and poor EF may impact sustained attention, it also is dependent on a separate neural network and is closely associated with other cognitive functions (Unsworth et al., 2010). Constructs such as selective attention and switching attention are much more intimately related to well-known aspects of executive function, such as inhibition and cognitive flexibility (Diamond, 2013; McCabe, Roediger, McDaniel, Balota, & Hambrick, 2010). Further, there is strong empirical evidence that performance on executive functioning tasks is closely related to performance on tasks of executive attention (Kane, Bleckley, Conway, & Engle, 2001; Magimairaj & Montgomery, 2013) and is coordinated by activity in the PFC (Ottowitz et al., 2002), whereas alerting, orienting and sustaining attention, on the other hand, are differentially coordinated by subsets of larger neural attention networks (Mezzacappa, 2004; Raz & Buhle, 2006).

In addition to difficulties in attention and executive functioning, SES predicts variation in the onset of pubertal development, otherwise known as pubertal timing. Researchers have theorized that cumulative stress in childhood, in the form of maternal psychopathology, family conflict, family structure, or abuse and adversity (Belsky et al., 2007; Brooks-Gunn, 1988; Ellis & Garber, 2000a) signal scarcity or instability in the early environment and cause youth to reach pubertal maturation earlier, reproduce earlier, and, ultimately, increase genetic fitness via a higher likelihood of passing on their genes (Belsky, Steinberg, & Draper, 1991). Individuals in low socioeconomic status households are exposed to higher rates of these stressors than individuals in high socioeconomic status households (Repetti, Taylor, & Seeman, 2002). As a result, some studies have evaluated socioeconomic status as an independent predictor of pubertal timing. These studies found that, consistent with expectations, youth from low socioeconomic status families have earlier pubertal timing than youth from high socioeconomic status families (Deardorff, Abrams, Ekwaru, & Rehkopf, 2014; Sun, Mensah, Azzopardi, Patton, & Wake, 2017).

Several lines of research suggest that early pubertal timing, in addition to SES, also may be associated with attention/executive functioning difficulties during adolescence, although this hypothesis has yet to be tested directly. Multiple pathways exist that may link pubertal timing with attention/executive functioning. First, sex steroids, which increase across puberty, appear to contribute to a reorganization of the prefrontal cortex (Juraska & Willing, 2017). This process involves a proliferation of synapses at the onset of puberty, followed by a pruning of synapses, which is thought to be important for increased efficiency (Blakemore & Choudhury, 2006). Importantly, it has been posited that this increased efficiency is achieved by balancing the quantity of excitatory and inhib itory neurons, and there is evidence that excitatory neurons are selectively pruned, and inhibitory neurons are spared (Selemon, 2013). Therefore, youth who enter puberty earlier than their peers may, at least temporarily, have worse executive functioning abilities as a consequence of this synaptogenesis and transient imbalance of excitatory and inhibitory neurons compared to their peers. Alternatively, early pubertal timing may confer risk for executive functioning impairments via psychosocial factors, like substance use, stress and depression. Youth with early pubertal timing are more likely to engage in risky behavior, including drug and alcohol use, perhaps due to an exaggerated mismatch between neural emotional and executive functioning systems (Biehl, Natsuaki, & Ge, 2007; Stumper, Olino, Abramson, & Alloy, 2019). Drug and alcohol abuse disrupts adolescent frontal cortical development and the maturation of executive functioning across adolescence (Crews, He, & Hodge, 2007). Similarly, early-maturing youth have been shown to experience more stress than their on-time or late-maturing peers (Conley & Rudolph, 2009; Hamilton et al., 2014), and stress can affect both the function and structure of the PFC (Arnsten & Shansky, 2004; McEwen & Morrison, 2013). Early pubertal timing also has been linked with depressive symptoms (Hamilton et al., 2014), which, in turn, is associated with worse executive functioning (Mac Giollabhui et al., 2019). Thus, early-maturing youth are more likely to use drugs and alcohol and experience more stressors and depressive symptoms than their same aged peers, which may, in turn, disrupt the development of the neural systems that underlie executive functioning development, leading to impairments in executive functioning abilities. However, there has been little empirical investigation into a potential relation between pubertal timing and executive functioning impairments. This may be due to the belief that pubertal processes are not related to executive functioning (Luna, 2009). Furthermore, if these youth do evince impairments in EF, it is also unclear whether they go on to “catch up” to their peers, or if these impairments persist across development.

Current Study

The aims of the current study were two-fold: to test whether pubertal timing was associated with performance on attention and executive functioning tasks, both cross-sectionally and longitudinally, and to test whether pubertal timing mediated the relation between socioeconomic status and performance on these tasks during adolescence. Attention was assessed via two sustained auditory tasks (Digit Span Forwards/Test of Everyday Attention (for Children)’s Score!) and aspects of executive functioning also were assessed (Working Memory: Digit Span Backwards; Cognitive Flexibility: Test of Everyday Attention (for Children)’s Creature Counting (Timing/Accuracy); Selective Attention: Test of Everyday Attention (for Children)’s Sky Search). We hypothesized that earlier pubertal timing would be associated with poorer performance on executive functioning tasks, both concurrently and longitudinally. We also hypothesized that low socioeconomic status youth would have earlier pubertal timing and, in turn, poorer performance on executive functioning tasks, with early pubertal timing mediating the longitudinal association between low socioeconomic status and attention and executive functioning. These hypotheses were tested in a gender and racially diverse community sample of adolescents drawn from an urban area in the US, with nearly half of the sample classified as low SES. Additionally, childhood stress was included as a covariate in the present analyses because of its established association with socioeconomic status (Repetti et al., 2002), pubertal timing (Ellis & Garber, 2000a), and executive functioning (Pechtel & Pizzagalli, 2011).

Methods

Participants and Procedure

Participants were drawn from the Adolescent Cognition and Emotion (ACE) project at Temple University, a large, public university located in an urban setting in the United States. A community sample of 639 adolescents and their mothers or primary female caregivers were recruited from the Philadelphia area. Most youth recruited were 12 or 13 years old, although a small portion of youth recruited at baseline were aged 14–16. Recruitment involved both mailings and follow-up calls to families with children attending Philadelphia area public and private middle schools (68% of the total sample) and advertisement in local newspapers (32% of the sample). Inclusion criteria included sufficient competence with the English language to complete the assessments. Additionally, adolescents had to identify as either Caucasian/White, African American/Black, or biracial. Individuals who identified as members of other racial or ethnic groups were excluded, as the investigation of differences in the etiology of depression comparing Caucasian/White and African American/Black youth was one of the aims of Project ACE. All demographic information was self-reported during the first visit of the study. Exclusion criteria also included a history of severe psychiatric illness or developmental disorders (see Alloy et al., 2012 for further information). Informed written consent was obtained from mothers and written assent from adolescents at the first study visit.

At baseline and annually thereafter, participants completed a battery of assessments, including self-report measures and behavioral tasks assessing executive functioning and attention. However, the annual visits were completed in two sessions, and the two outcome measures used in the present study (Test of Everyday Attention for Children [TEA-Ch] and Digit Span) were given at different sessions. Participants frequently attended just one of the two sessions. As a result, the Ns for statistical models conducted in the present study differ slightly. Data used in the current study were drawn from the participants’ first study visit and two subsequent annual visits, when youth were 13, 14, and 15 years old, on average.

Samples were determined by selecting participants who had data on outcome measures at baseline and at least one annual follow-up (a robust maximum likelihood estimator was used in the present analyses). This resulted in a different sample size for each outcome measure used. The two samples were largely overlapping, such that 205 participants were in both samples, 76 additional participants were unique to the TEA-Ch sample, and 38 additional participants were unique to the Digit Span sample. For analyses using the TEA-Ch, the sample consisted of a subsample of 281 adolescents (M_age at baseline = 12.96 years, SD = .79; M_age at T2 = 14.26 years, SD = .85; M_age at T3 = 15.09 years, SD = .79). This sample was 45.6% male, 50.5% White/Caucasian, 46.3% Black/African American, 3.2% Biracial/other, and 44.5% were categorized as low SES. Within this sample, 71% of youth (N=200) completed the assessment at age 14, and 77% (N=216) completed the assessment at age 15. For analyses using the Digit Span, the sample consisted of a subsample of 243 adolescents (M_age at baseline = 13.01 years, SD = .79; M_age at T2 = 14.09 years, SD = .81; M_age at T3 = 15.08 years, SD = .84). This sample was 46.5% male, 47.3% White/Caucasian, 49.0% Black/African American, 3.7% Biracial/other, and 46.5% were categorized as low SES. Within this sample, 74% (N=179) completed the assessment at age 14, and 58% (N=141) completed the assessment at age 15.

The TEA-Ch sample included in analyses did not differ from the complete sample of Project ACE participants who were excluded from analyses on the basis of: sex (χ²(1, N=636) = .34 p = .56), race (χ²(2, N=636) = 1.55 p = .46), baseline Selective Attention (t(566) = −.22, p = .83), baseline Sustained Attention (t(577) = −1.14, p = .26), Digit Span Forward (t(464) = −.11, p = .92), Digit Span Backward (t(462) = −1.22, p = .31, or childhood life events (t(613) = 1.64, p = .11). Participants included in the TEA-Ch sample had later pubertal timing (t(625) = 2.36, p = .02) and higher scores on Switching Attention (timing: t(530) = −2.10, p = .04; accuracy: t(573) = −4.32, p < .001). Participants included in the TEA-Ch sample also differed on the basis of SES; they were less likely to be eligible for school lunch (χ²(1, N=607) = 4.01, p = .045), had higher income (t(604) = −3.47, p=.001), and had higher maternal education (t(615) = −3.29, p=.001) than those excluded.

The Digit Span sample did not differ from participants who were excluded on the basis of sex (χ²(1, N=636) = .02 p = .88), race (χ²(2, N=636) = .08 p = .96), free lunch status (χ²(1, N=607) = .78, p = .41, baseline Selective Attention (t(566) = −.15, p = .88), baseline Switching Attention (timing: t(530) = −1.71, p = .09; accuracy: t(573) = −1.54, p = .13), baseline Sustained Attention (t(577) = −1.46, p = .15), Digit Span Forward (t(464) = −.36, p = .72), Digit Span Backward (t(462) = −1.62, p = .11, or childhood life events (t(613) = −.57, p = .57). Participants included in the Digit Span sample had later pubertal timing (t(625) = 2.64, p = .01), higher income (t(604) = −3.20, p=.001), and maternal education (t(615) = −3.11, p=.002) than those excluded.

Measures

Pubertal Timing.

The Pubertal Development Scale (PDS; Petersen, Crockett, Richards, & Boxer, 1988) is a self-report questionnaire designed to assess pubertal development. The questions ask about growth in height, body hair, skin change, breast (females) or voice (males) change, and facial hair (males) or menstruation (females). All questions aside from menstruation are rated on a 4-point scale (1 = no development, 2 = development has barely begun, 3 = development is definitely underway, 4 = development is complete). Menstruation is scored as 1 = “I have not yet begun to menstruate” or 4 = “I have begun to menstruate.” Item scores are averaged, and the scale yields a final score ranging from 1–4 (less to more pubertally developed). Consistent with past research assessing pubertal timing (e.g., Alloy, Hamilton, Hamlat, & Abramson, 2016; Dorn, Dahl, Woodward, & Biro, 2006), timing scores were obtained by regressing PDS total score on age. Timing scores were computed separately for males and females. The residual was used as a continuous measure of pubertal timing. Both mothers and adolescents completed the five-item questionnaire at baseline, and an average of the timing scores computed from both adolescents’ and mothers’ reports was used in analyses (correlation between mother and adolescent report: r = .84, p < .001). The PDS has acceptable psychometric properties (average alpha of .77 for just five items) and good convergent validity (correlations of .61-.67 with physician ratings) (Petersen et al, 1988). Internal consistency in this sample based on 5 items was adequate (Child Report: TEA-Ch Sample: girls = 0.69, boys = .72; Digit Span Sample: girls = .65, boys = .78; Parent Report: TEA-Ch Sample: girls = 0.70, boys = .75; Digit Span Sample: girls = .70, boys = .76).

Test of Everyday Attention for Children/Test of Everyday Attention.

Three subtests of the Test of Everyday Attention for Children (TEA-Ch) and the Test of Everyday Attention (TEA) were administered, which were the age- and gender-normed behavioral assessments of selective, sustained and switching attention (Manly et al., 2001; Robertson, Ward, Ridgeway, & Nimmo-Smith, 1994). Scaled scores, for which a score of 10 is indicative of performance in the 50th percentile (SD = 3), were reported; higher scores are indicative of superior performance. The TEA was designed and normed to assess adults, whereas the TEA-Ch was adapted from the TEA to assess attention in youth ages 6 to 16 years. The TEA-Ch was administered to youth up to age 16; the TEA was administered to youth older than 16 years. The TEA and TEA-Ch have demonstrated adequate to good reliability in healthy adult and child samples, respectively (Manly et al., 2001; Robertson, Ward, Ridgeway, & Nimmo-Smith, 2001). See Mac Giollabhui et al. (2019) for more detailed explanations of each subtest. For both measures, Sky Search was administered to assess selective attention. The Sky Search subtest is a nonlinguistic measure of selective attention, in which participants were asked to identify cases as quickly as possible in which identical stimuli are paired together on a page. In the TEA, the total score reflects the total number of accurately identified cases; in the TEA-Ch, the total score reflects the total number of accurately identified cases, controlling for psychomotor speed. For the TEA-Ch, sustained attention was assessed using the Score! subtest. Score! is a 10-item counting measure in which between 9 and 15 tones (345 ms) were presented, interspersed by silent intervals of variable duration (500–5,000 ms), and participants were asked to count the number of tones. The TEA used a comparable seven -item task. Sustained attention performance was assessed by the number of tones correctly identified. Because scaled scores are not provided for the TEA, raw scores were used for both the TEA and TEA-Ch; raw scores on the TEA were scaled to match the 10-item TEA-Ch measure. Attentional switching was assessed in the TEA-Ch using the Creature Counting subtest and was assessed in the TEA using the Elevator Counting subtest. Both tests measure the temporary slowing that is associated with temporarily switching from one mental set to another. Two total scores are determined based on i) the speed with which participants successfully switch response sets on all successful trials, and ii) the accuracy with which participants completed all trials.

Digit Span.

The Digit Span subtest of the Wechsler Intelligence Scale for Children – Fourth Edition (WISC-IV; Wechsler, 2003) and the Wechsler Adult Intelligence Scale – Fourth Edition (WAIS-IV; Wechsler, 2008) was administered. The WISC-IV was validated for use with youth ages six to 16, and the WAIS-IV was validated for use with adults over 16 years old. Youth were given the WISC-IV until age 16; they were given the WAIS-IV when they were over 16. The Digit Span measures auditory verbal working memory. In both the WISC-IV and the WAIS-IV, participants listen to and then repeat a series of numbers read aloud by the experimenter (Digit Span Forward). This task is intended to measure short-term attention/memory abilities. Participants subsequently listen to a series of numbers presented by the experimenter and then repeat them aloud in reverse order (Digit Span Backward), a task thought to measure the maintenance and manipulation of information in one’s working memory. A total score is used for both tasks, which reflects the total number of times the participant repeats the series of numbers correctly. The Digit Span has demonstrated good reliability in the WISC-IV and WAIS-IV (Wechsler, 2003; Wechsler, 2008). Digit Span Forward is used as a measure of simple auditory attention in the current study, and Digit Span Backward is used as a measure of working memory in the current study. Scaled scores, for which a score of 10 is indicative of performance in the 50th percentile (SD = 3), were reported where higher scores are indicative of superior performance.

Childhood life events.

The CLES-PR measures the occurrence of stressful life events during childhood, as reported by a parent (Crossfield, Alloy, Gibb & Abramson, 2002). At baseline, parents respond “yes” or “no” for 50 childhood events deemed to be moderately-to-majorly stressful, including items of physical and sexual abuse, bereavement, poor school performance, achievement failures and negative emotional feedback. A total score was calculated by summing all affirmative answers, with higher scores indicating greater experiences of stress. The CLES was included as a covariate in the present analyses. Internal consistency for this measure was good in both the TEA-Ch and Digit Span samples (α = .72 in both samples).

Socioeconomic status.

Socioeconomic status was operationalized as a latent variable comprised of maternal reports of family income, maternal education, and eligibility for the National School Lunch program. Mothers reported on income (1 = $0–14,000, 2 = $15–29,999, 3 = $30–44,999, 4 = $45–59,999, 5 = $60–74,999, 6 = $75–89,999, 7 = $90,000 and over), educational attainment (1 = Less than 8th grade completed, 2 = 8th grade completed, 3 = some high school, 4 = high school diploma, 5 = some college, 6 = Associate’s degree, 7 = Bachelor’s degree, 8 = some graduate school, 9 = Master’s degree, 10 = professional degree), and eligibility for school lunch (re-coded so that 0 = eligible, 1 = not eligible) via self-report at the baseline visit. All variables loaded significantly onto the latent factor across models (p < .001); average loading across models for income was 88.17 (SD = 1.17), for maternal education was 61.33 (SD = 2.87), and for eligibility for free school lunch was 70.5 (SD = 3.15). Higher scores on this variable indicate higher socioeconomic status.

Results

Descriptive statistics and bivariate correlations for the main study variables are presented in Table 1 for the 281 participants who were present at baseline and at least one follow up in the TEA-Ch sample. Descriptive statistics and bivariate correlations are provided for all attention and executive functioning variables at all timepoints in Supplementary Table 1.

Table 1:

Descriptive statistics and bivariate correlations for baseline variables

Variables	1	2	3	4	5	6	7	8	9	10	11	12
1: Age	-
2: School Lunch Eligibility	0.09
3: Income Level	0.10	0.61**	-
4: Mother’s Education	−0.06	0.41**	0.53*	-
5: Childhood Stress	0.05	0.01	−0.05	0.14*	-
6: Pubertal Timing	0.85***	0.11*	−0.15*	−0.09	0.02	-
7: Selective Attention	−0.07	0.02	−0.01	−0.03	−0.01	−0.04	-
8: Switching Attention (Timing)	−0.08	−0.17**	0.15*	0.06	−0.17**	−0.18**	0.17**	-
9: Switching Attention (Accuracy)	−0.01	−0.16**	0.15*	0.08	−0.04	−0.03	−0.10	0.08	-
10: Sustained Attention	0.05	−0.07	0.11	0.11	−0.06	0.04	0.05	0.16**	0.26**	-
11: Digit Span Forward	−0.24***	−0.17**	0.17**	0.18**	−0.06	−0.21**	0.05	0.21***	0.01	0.12*	-
12: Digit Span Backward	−0.17**	−0.19**	0.22**	0.21**	−0.01	−0.18**	0.06	0.19**	0.19**	0.12*	0.39***	-
Mean	12.46	.45	4.15	6.11	9.9	−0.09	10.02	8.29	10.14	9.00	10.15	9.66
(SD)	.79	.50	2.07	2.02	4.56	0.87	2.52	2.55	2.71	1.57	3.12	2.75

Note: Age=Age at baseline.

^*= p<.05;

^**= p<.01;

^***= p<.001

Data Analysis

All analyses were conducted in Mplus (Version 7.4). Missing data were handled using a Full Information Maximum Likelihood and using a robust estimator. Path analysis within a structural equation model framework examined the concurrent and prospective associations between age 13 pubertal timing and (i) age 13 attention/executive functioning as well as (ii) prospective changes in attention/executive functioning, controlling for socioeconomic status and childhood stress. In addition, statistical mediation was tested based on 5000 bootstrapped samples assessing whether pubertal timing mediated the association of socioeconomic status with age 13 and age 14/15 attention/executive functioning. All pathways are graphically presented in Figure 1. Models analyzed assumed that the relationship between specific variables were equivalent across all timepoints, although models in which pathways are unconstrained are presented as supplementary information (see Supplementary Table 2). For example, we assumed that the association of socioeconomic status with attention or executive functioning was equivalent across all three ages, and likewise, we assumed that the association of pubertal timing with changes in future attention/executive functioning was equivalent at age 14 and age 15. We made this assumption because attention and executive functioning variables were age-standardized at each time-point and because of the absence of strong theory or empirical data to suggest otherwise. Models predicting Sustained Attention controlled for age and gender, because the scores on this measure were not standardized by age or gender. Model fit was estimated using chi-square estimate of goodness of fit, comparative fit index (CFI), and root-mean-square error of approximation (RMSEA). The Chi-Square test of model fit was reported according to convention, but not interpreted given its limited utility in large samples (Chen, 2007; Cheung & Rensvold, 2002). For the CFI, “good” fit is indicated by a value >.90 and “excellent” fit by a value >.95. A RMSEA statistic between .05 and .10 is indicative of “good” fit and a value <.05 is indicative of “excellent” fit (Schermelleh-Engel et al., 2003).

Figure 1:

Structural equation model examining concurrent and prospective associations of socioeconomic status, pubertal timing, and childhood stress with measures of executive functioning (EF) or attention over three years.

Note: Dashed arrows indicate that pathways were constrained to be equal; all models controlled for early childhood stress; models predicting Sustained Attention also controlled for age and gender; socio-economic status is a composite of eligibility for the National School Lunch Program, maternal education, and income.

Model Fit

Model fit statistics are provided in Table 2 for the six models examining each of the cognitive functioning outcomes. Based on the Comparative Fit Index and Root mean Square Error of Approximation (90% CI), good to excellent model fit was observed for all models.

Table 2.

Structural equation model examining concurrent and prospective associations of socio -economic status (SES), pubertal timing, and childhood stress with measures of attention/executive functioning (EF) over three years.

	Selective Attention	Switching Attention (Timing)	Switching Attention (Accuracy)	Sustained Attention	Digit Span Forward Ψ	Digit Span Backward Ψ
Indices of Model Fit
Chi-Squared Test of Model Fit	34.35*	22.96	25.26*	47.88	26.29*	38.24*
Comparative Fit Index	.93	.98	.97	.995	.98	.94
Root Mean Square Error of Approximation (90% CI)	.06(.03–.09)	.04(.00–.07)	.04(.00–.07)	.02(.00–.05)	.05(.00–.08)	.07(.04–.10)
Pathways
Age 13/14 Attention/EF predicting Age 14/15 Attention/EF†	.31***	.34***	.22***	.29***	.65***	.44***
Age 13 Pubertal Timing predicting Age 13 Attention/EF	−.03	−.15*	.01	−.01	−.20**	−.18**
Age 13 Pubertal Timing predicting Age 14/15 Attention/EF †	−.10*	−.09*	−.13**	−.05	−.02	−.05
SES predicting Age 13/14/15 Attention/EF †	.05	.15***	.18***	.05	.12**	.19***
Childhood Stress predicting Age 13/14/15 Attention/EF †	−.05	−.14***	−.07*	−.07	−.04	−.07
SES predicting Age 13 Pubertal Timing	−.16*	−.16*	−.16*	−.06	−.15*	−.15*
Childhood Stress predicting Age 13 Pubertal Timing	.02	.02	.02	−.01	.06	.06
SES predicting Childhood Stress	−.01	−.02	−.01	−.01	−.06	−.06
Indirect Pathways
SES predicting Age 13 Attention/EF via Age 13 Pubertal Timing	.01(−.01,.05)	.03(.004,.08)	−.01(−.03,.03)	.001(−.03,.04)	.05(.005,.13)	.04(.004,.09)
SES predicting decline in Age 14/15 Attention/EF via Age 13 Pubertal Timing	.02(.002,.07)	.02(.001.06)	.03(.01,.08)	.01(−.01,.03)	.004(−.01,.03)	.01(−.002,.05)

Note:

^†= Pathways were constrained to be equal.

^Ψ= Based on smaller sample of 243 participants. Parameters meeting criteria for statistical significance are presented in bold. Models predicting Sustained Attention control for Age and Gender. Socio-economic status is operationalized as a composite of maternal education, income, and eligibility for the National School Lunch program.

^*= p<.05;

^**= p<.01;

^***= p<.001

Concurrent and Prospective Associations between Pubertal Timing, Socioeconomic status, Childhood Stress, and Attention/Executive Functioning

The structural equation model estimated for each of the six attention and executive functioning variables are represented in Figure 1. Small to large correlations were observed in the auto-regressive associations of attention/executive functioning across timepoints (see Table 2). Earlier pubertal timing at age 13 was associated with poorer performance on tasks assessing attention (Digit Span Forward), switching attention (timing) and executive functioning (Digit Span Backward) at age 13. Earlier pubertal timing at baseline also was associated with declines in performance on measures of selective attention and switching attention (timing and accuracy) at age 14/15. Low socioeconomic status was associated with earlier pubertal timing at age 13 across all models as well as poorer performance on switching attention (timing and accuracy), attention (Digit Span Forward) and executive functioning (Digit Span Backward) at age 14/15. Childhood Stress independently was associated with worse switching attention (timing and accuracy).

Indirect Effects: Socioeconomic Status on Attention/Executive Functioning via Pubertal Timing

Given the associations observed between SES, pubertal timing, and attention/executive functioning measures, additional analyses were conducted using 5000 bootstrapped samples to test whether low socioeconomic status predicted worse attention or executive functioning indirectly via earlier pubertal timing (see Table 2). Earlier pubertal timing significantly mediated the relation between low SES and age 13 switching attention (timing; b=.03, 95% CI=.004, .08), attention (Digit Span Forward; b=.05, 95% CI=.005, .013), and executive functioning (Digit Span Backward; b=.04, 95% CI=.004, .09). Additionally, earlier pubertal timing significantly mediated the relation between low SES and age 14/15 selective attention (b=.02, 95% CI=.002, .07), and switching attention (timing; b=.02, 95% CI=.001, .06; accuracy; b=.03, 95% CI=.01, .08); in the case of switching attention (timing), the estimate was run with 10,000 bootstrapped samples, because it was unclear whether the lower bound of the confidence interval crossed zero when using 5,000 bootstrapped samples.

Alternate Model Analyses

Two alternative structural equation modelling approaches were run to test the proposed hypotheses. First, growth curve models were fit to the data, such that the intercept and slopes for the six measures of attention/executive functioning were estimated and regressed on age 13 pubertal timing and socioeconomic status. Model fit for all models was good to excellent, but technical problems (linear dependencies) prevented interpretation for four of the six models. In the case of two models that were interpretable, a similar pattern of results was observed when compared to results reported in this manuscript, such that low socioeconomic status was associated with earlier pubertal development and both were associated with worse performance at baseline (i.e., the intercept of the Digit Span Forward/Backward, but not the slope). Second, alternative models were estimated to examine whether data for attention and executive functioning measures loaded on (i) two latent variables: executive functioning (Digit Span Backward, Selective Attention, Switching Attention timing and accuracy) and auditory attention (Score! and Digit Span Forward) and (ii) a general latent factor underpinning attention and executive functioning variables. A general latent factor model utilizing all attention/executive functioning variables was generated at baseline that fit the data well; low socioeconomic status was associated with earlier pubertal development and both were associated with lower scores on the latent factor underpinning attention/executive functioning. Unfortunately, longitudinal models estimating latent factors underpinning attention/executive functioning at all three timepoints could not be generated. In addition, models were run that allowed all constrained pathways to vary. The model fit and results of these models closely mirrored that of the results presented in this manuscript (see Supplementary Table 2).

Discussion

There is compelling evidence that youth from low socioeconomic backgrounds perform worse on tests measuring a broad range of cognitive functions, including attention and executive functioning, but questions remain about the mechanisms underlying this relationship. The present study investigated pubertal timing as a mediator of this relation in light of its potential associations with both socioeconomic status and executive functioning. To address this question, analyses tested whether SES, childhood stress, and pubertal timing were associated with impairments in performance on attention and executive functioning measures concurrently and longitudinally, and whether the association between socioeconomic status and attention or executive functioning was mediated by pubertal timing. The overall pattern of results offered support for a developmental pathway from low socioeconomic status to early pubertal timing to impairments in both attention and executive functioning both concurrently and prospectively.

These results were counter to the hypothesis that low SES and early pubertal timing would predict impairments in executive functioning but not less complex attention tasks. This hypothesis was derived from the literature on brain development during adolescence, and the belief that low SES and early pubertal timing may disrupt the normative development of the prefrontal cortex (Hackman, Gallop, Evans, & Farah, 2015, Juraska & Willing, 2017), which plays an important role in coordinating executive functioning abilities (Ottowitz et al., 2002). However, low SES and early pubertal timing predicted worse performance across a range of tasks assessing aspects of executive functioning and attention abilities in the current sample. Therefore, these results suggest that the impact of low socioeconomic status and early pubertal timing is broader than just the prefrontal cortex, or that disruption to prefrontal cortex development confers risk for impairment in both sets of abilities, perhaps due to the significant overlap between them (Unsworth et al., 2010).

An important finding from the current study is the relation between pubertal timing and measures of “cold” executive functioning (the ability to engage executive functioning in the absence of emotional stimuli). The extant literature has focused on the relation between pubertal processes and “hot” executive functioning abilities (the ability to engage executive functions in the context of emotional stimuli; Steinberg, 2008). This may be due to the belief that pubertal processes do not influence the development of neural areas that underpin executive functioning (e.g., the PFC) and that “hot” executive functioning abilities suffer because of the pubertally-induced maturity of neural networks responsible for emotionality and reward sensitivity (Luna, 2009). However, the current study offers support for early pubertal timing as a risk factor for impairments in some facets of “cold” executive functioning during adolescence. The mechanisms of this effect are not addressed in the present study, but increased psychosocial stress (McEwen & Morrison, 2013), propensity for alcohol use among early-maturing youth (Crews et al., 2007), depressive symptoms (Mac Giollabhui et al., 2019) and alterations in neurological development as a result of off-time maturation (Juraska & Willing, 2017) are plausible candidates.

These findings also add to our understanding of the mechanisms of the association between low socioeconomic status and impairments in executive functioning. There is evidence that family characteristics, like a single parent household and quality of family relationships, and stress more broadly, mediate the relation between socioeconomic status and executive functioning among youth (Sarsour et al., 2011; Ursache & Noble, 2016). Family characteristics and stress also predict an earlier onset of pubertal development (Belsky et al., 1991; Ellis & Garber, 2000b; Ellis, McFadyen-Ketchum, Dodge, Pettit, & Bates, 1999; Wierson, Long, & Forehand, 1993). Perhaps early pubertal timing and executive functioning impairments are both consequences of the stress associated with growing up in a low socioeconomic status family. Consistent with this idea, Joos and colleagues (2018) proposed that pubertal timing is not a consequence of stress or an antecedent of negative outcomes, but rather a reflection of allostatic load and biology’s attempt to maximize the likelihood of reproduction and, therefore, genetic fitness (Joos, Wodzinski, Wadsworth, & Dorn, 2018). According to this theory, negative consequences of early pubertal timing, in this case, worse attention and executive functioning, also can be conceptualized as allostatic load. Thus, results from the current study are in line with this theory and may reflect just one pathway of many from childhood stress to negative outcomes. However, it is important to note that early childhood stress was related to poorer switching attention (cognitive flexibility) in the current sample, but early childhood stress was not related to pubertal timing or socioeconomic status. There is robust evidence for relations between these variables in the extant literature, and it is unclear why they weren’t evident in this sample (Repetti et al., 2002; Ellis & Garber, 2000a). This may be due to differences in the measure used to assess childhood stress, as the measure used in the present stud y captured events that occurred any time from birth through age 12/13, when it was given.

Further, these results suggest that low socioeconomic status youth are at risk for impairments in attention and executive functioning both concurrently and 1 and 2 years later, and this relation was mediated by early pubertal timing. Adolescence is understood to be a critical period for neural development, and the present study highlights the potentially deleterious effects of low socioeconomic status and pubertal timing on this development. It remains unclear whether early maturing youth “catch up” to their peers with regard to performance on these tasks, but these findings offer evidence that these impairments persist at least through mid-adolescence. This is consistent with past work that shows that impairments in executive functioning among low SES youth persist from childhood into adolescence (Hackman et al., 2015), and may even increase in magnitude across development (von Stumm & Plomin, 2015).

In light of evidence that both early pubertal timing and executive functioning deficits are risk factors for a number of psychological disorders (Graber, Lewinsohn, Seeley, & Brooks-Gunn, 1997; Nolen-Hoeksema & Watkins, 2011), poor executive functioning may be a potential mechanism by which early pubertal timing confers risk for psychopathology. In addition to psychological disorders, impairments in attention and executive functioning also are associated with academic and occupational outcomes (Miller, Nevado-Montenegro, & Hinshaw, 2012), highlighting the importance of this developmental pathway as a potential target for prevention and intervention efforts among youth.

Use of a prospective longitudinal design, a sample diverse in SES, race, and sex, and repeated behavioral assessments of attention/executive functioning are strengths of the current study. However, the results should be interpreted in light of the study’s limitations. First, pubertal timing was assessed relatively late in development (age 12–13). Although there was variability in the development of youth in the current sample, it is unclear whether the same effects would be observed earlier in development. It would be useful to replicate these findings in a younger sample. Second, although the TEA-Ch is a validated measure of attentional functioning, it was not normed in a large sample of US adolescents. Therefore, it is possible that the normed scores used in the present study do not match age and gender norms of US adolescents. Concerns about the use of this measure are attenuated by the use of the Digit Span as well, with which we found a similar pattern of results.

Importantly, the latent variable assessing socioeconomic status is comprised of self-reports that were completed concurrently with the baseline assessments of executive functioning, attention, and pubertal timing. The models tested implicitly assume that our measurement of SES applies retrospectively to the adolescents’ histories. The sample in the current study was recruited largely from North Philadelphia, an urban area with high rates of poverty and limited social mobility (Venkatesh, 1994). Thus, it is unlikely that a substantial number of these families experienced significant changes to their financial situations. However, some families may have experienced different financial conditions in previous years. Findings should be interpreted in light of this limitation. Finally, whereas important aspects of executive functioning were examined (working memory/switching attention/selective attention), others were not (inhibition/planning/updating working memory). Future work should investigate whether early pubertal timing predicts impairments in other facets of executive functioning as well as global cognitive functioning. Finally, there was a high rate of attrition in the current study. The current study samples did not differ from those excluded on the basis of most demographic characteristics and baseline measures. However, the TEA-Ch sample had higher scores on switching attention than those excluded, and participants in both the TEA-Ch sample and the Digit Span sample had later pubertal timing, higher maternal education and income, and were less likely to be eligible for school lunch than those excluded. This overall pattern of more economically advantaged families being more likely to attend follow ups is not surprising, and perhaps to be expected. Low SES families have more demands on their time and resources, such as increased demands placed on caregivers in single parent households or parents working multiple jobs often with unpredictable schedules. Although it is understandable that such barriers would make it challenging for low SES families to attend follow up sessions, this may lead to biased results. Further, although results in the current study suggest that low SES youth have earlier pubertal timing and poorer performance on switching attention, it is possible that the biases in the samples due to attrition may have, in turn, biased the results.

Conclusion

The present study demonstrated that low socioeconomic status youth had impairments in attention and executive functioning abilities concurrently and prospectively, and this relation was mediated by earlier pubertal timing. These findings demonstrate the potentially deleterious effects of low socioeconomic status and early pubertal timing on attention and executive functioning development during adolescence and highlight a subset of youth who are particularly vulnerable to developing impairments in these abilities. Given the link between executive functioning impairments and negative outcomes in adolescence and young adulthood, these findings have implications for prevention and intervention efforts. However, more work is needed to understand the mechanisms of these associations to provide more specific targets for intervention.

Biography

Allison Stumper is a doctoral student in Lauren B. Alloy’s Mood and Cognition Lab at Temple University. Her research interests include how biological and environmental factors interact to confer risk for adolescent depression, with an emphasis on the role of biological changes as a result of the pubertal transition.

Naoise Mac Giollabhui is a graduate student in Temple University’s clinical psychology program. He is working in Dr. Alloy’s Mood and Cognition Lab at Temple University and is interested in how cognitive biases and neuropsychological functioning differ in those who have experienced depression. He also is interested in how immune functioning is implicated in the etiology of depression and, in particular, whether immune dy sfunction is related to weaknesses in neuropsychological functioning observed in depression.

Lyn Y. Abramson is the Sigmund Freud Professor of Psychology at the University of Wisconsin-Madison. She received her doctorate in Clinical Psychology from the University of Pennsylvania. Her major research interests include the developmental, cognitive, motivational, and cultural determinants of information processing about the self and the effects of early psychological, physical, and sexual maltreatment on the development of cognitive styles and vulnerability to depression in adulthood.

Lauren B. Alloy is Laura H. Carnell Professor and Joseph Wolpe Distinguished Faculty in Psychology at Temple University. She received her doctorate in Experimental and Clinical Psychology from the University of Pennsylvania. Her major research interests include cognitive, psychosocial, developmental, and, neurobiological processes in the onset and course of depression and bipolar disorder.