Longitudinal Relations Among Household Chaos, SES and Effortful Control in the Prediction of Language Skills in Early Childhood

Abstract

Previous research has shown that children’s home environment plays an important role in children’s early language skills. Yet, few researchers have examined the unique role of family-level factors (SES, household chaos) on children’s learning, or focused on the longitudinal processes that might explain their relations to children’s early language skills. The goal of this study was to investigate the longitudinal relations from family socioeconomic status (SES), household chaos, and children’s effortful control (EC) to children’s language skills during early childhood, controlling for stability of the constructs. At 30 months (T1), mothers reported family SES and children’s vocabulary, and their own linguistic input was assessed during a free-play session with their child. At 30, 42, and 54 months (T1, T2, and T3), household chaos was reported by mothers, and children’s EC was rated by mothers and nonparental caregivers and observed during a gift delay task. At T3, children’s expressive and receptive language were measured with a standard assessment. Path analyses indicated that higher SES predicted higher levels of EC at T2 and language skills at T3, and greater levels of household chaos at T2 predicted poorer EC and language skills a year later, even when controlling for stability of the constructs. Results indicated that T2 EC partially mediated the relations between SES and T3 language skills. Findings from this study can be used to identify key factors for early learning and perhaps inform programs designed to support families and young children.

There is wide variation in the mastery of basic cognitive and socioemotional competencies at school entry; children who lack fundamental school readiness skills generally struggle early on with learning and achievement, and their problems often persist throughout the school years (Zauche, Thul, Mahoney, & Stapel-Wax, 2016). Language is a critical foundation that contributes to school readiness because the abilities to understand and communicate with language support transactions within the social and physical environments that promote learning. In fact, early linguistic skills predicted improvements in reading ability and other domains of academic achievement during the early and later elementary years (Nelson, Welsh, Trup & Greenberg, 2011). Therefore, it is important to identify predictors of language development during early childhood, including mediators of predictors, in order to support successful transitions and performance in the first years of school and beyond.

Broad measures of socioeconomic risk have been consistently negatively related to early language development and other school readiness skills (Evans & Rosenbaum, 2008; Fernald, Marchman, & Weisleder, 2013; Hoff, 2003), but examination of more specific aspects of children’s environments is warranted. To this end, researchers have demonstrated that environmental chaos—including noise and disorganization within the home—is a consistent predictor of diminished language and literacy skills in children (Maxwell & Evans, 2000; Vernon-Feagans et al., 2012).

A cornerstone of developmental research is the aim not only to describe developmental phenomena, but to understand the processes through which they emerge, including the specific pathways from risk to children’s outcomes (Bailey, Duncan, Watts, Clements, & Sarama, 2018). Effortful control (EC) has increasingly emerged as an important contributor to the development of children’s cognitive, socioemotional, and behavioral competencies (Eisenberg, Valiente, & Eggum, 2010; Kopystynska, Spinrad, Seay, & Eisenberg, 2016; Nigg, 2017), including language and literacy abilities (Lonigan, Allan, & Phillips, 2017; McClelland et al., 2007). Moreover, children’s EC (or the broader construct of self-regulation) has been shown to account for some of the relations between family risk factors and child outcomes (Bernier, Beauchamp, Carlson, & Lalonde, 2015; Evans & Rosenbaum, 2008; Garrett-Peters et al., 2019; Ispa, Su-Russell, Palermo, & Carlo, 2017). Thus, the purpose of this study was to explore the longitudinal and unique relations of two family-level characteristics - namely, household chaos and socioeconomic status (SES)— to early language skills, as well as to examine the potential role of children’s EC in mediating these relations across early childhood. Few researchers have examined the unique role of these constructs in understanding children’s language skills, particularly using panel models in which the stability of the constructs is controlled. Moreover, researchers have not systematically examined these processes in early childhood, when children’s EC and language skills are rapidly developing.

Household Chaos, SES, and Language

Researchers have measured many aspects of environmental chaos in children’s lives, including community noise, residential crowding, and household disorganization. Household chaos, as conceptualized in this and other studies, refers to aspects of the environment that represent temporal and structural disorganization, commotion, and disorder within the home (Bronfenbrenner & Evans, 2000). Some of these characteristics are captured by a lack of regularity in household routines and schedules, high levels of ambient noise and other background stimuli, high levels of traffic within the home, an unkempt physical environment, and/or a general sense of disorganization in daily experiences (Bronfenbrenner & Evans, 2000; Vernon-Feagans, Garrett-Peters, Willoughby, Mills-Koonce, & The Family Life Project Key Investigators, 2012; Whitesell, Teti, Crosby, & Kim, 2015). Families reporting their households to be chaotic often perceive them to be “hectic, unstructured, unpredictable, and, at times, simply out of control” (Evans, Gonnella, Marcynyszyn, Gentile, & Salpekar, 2005, p. 560). Despite the variety of indices of chaos examined, converging evidence suggests that individuals exposed to chronic environmental chaos exhibit a number of maladaptive outcomes across social-emotional, cognitive, and physical domains (Bradley, 2015; Evans & Wachs, 2010).

One of the most robust research findings is that exposure to aspects of chaos relates to deficits in children’s learning. In both laboratory and naturalistic studies, noise exposure has been associated with deficits in language, pre-literacy and literacy skills, speech perception, attention, and memory (Maxwell & Evans, 2000). A number of studies have found chaos in particular to relate to language development even when considering other risk factors (Johnson, Martin, Brooks-Gunn, & Petrill, 2008; Pike, Iervolino, Eley, Price, & Plomin, 2006; Vernon-Feagans et al., 2012). Household chaos also has been related to broader academic outcomes, such that children who experience more household chaos have lower academic competencies (Garrett-Peters et al., 2016) and more disengagement from school (Garrett-Peters et al., 2019).

It has been argued that in the presence of chaos, developmentally-facilitative transactions between the child and their social and physical microcontexts may be disrupted or attenuated due to lack of duration, regularity, or intensity (Bronfenbrenner & Evans, 2000; Evans et al., 2005). In line with research on elevated physiological stress responses and “cognitive overload” in adults and children exposed to environmental stress (Saegert, 1981), it has been suggested that children cope with excessive noise and other stimulation (i.e., social, visual, temporal movement) by filtering out unwanted stimuli. With chronic exposure, they may begin to overgeneralize and apply this strategy indiscriminately, screening out developmentally-relevant cues and information along with that which is irrelevant or excessive (Maxwell & Evans, 2000). Thus, chaotic home environments may disrupt the vital proximal transactions that support children’s learning (Bronfenbrenner & Evans, 2000). It is also possible that timing of the exposure to chaos (i.e., in infancy, toddlerhood, or preschool period) may differentially predict children’s learning, although to our knowledge, this issue has not been examined.

As cautioned by Evans and colleagues (Evans, Li, & Whipple, 2013), examination of a single risk factor in isolation could overestimate its effects when overlapping factors exist. Notably, the families of those exposed to chaos often also struggle with economic hardship, which is an additional risk factor for deficits in language development at school entry and persistent difficulties in academic achievement (Fernald et al., 2013; Hoff, 2003). Low-income children have more language difficulties and slower rates of vocabulary growth than their more economically-advantaged peers (Burchinal, Roberts, Zeisel, Hennon, & Hooper, 2006; Hart & Risley, 1995). Sociodemographic risk factors for poor language development include poverty, large family size, and low parental education or IQ, and less sophisticated language utilized with children (Hoff, 2003; McClelland et al., 2007; Neuman, Kaefer, & Pinkham, 2018).

Chaos and socioeconomic status (SES) are often related, as there are a multitude of factors that may render the lives of socioeconomically disadvantaged children more chaotic than those in more affluent families. Low-income families often experience a number of stressors and negative life events, a lack of resources, and family and residential instability (Can & Ginsburg-Block, 2016); they tend to live in more noisy, crowded residences with substandard housing quality (Evans, Eckenrode, & Marcynszyn, 2010); and they often have inconsistent childcare arrangements and variable work schedules that can interfere with family routines (Evans, 2006). However, the correlations between SES and measures of household chaos are generally modest (Dumas et al., 2005; Evans et al., 2005; Pike et al., 2006), or sometimes nonexistent (Coldwell, Pike, & Dunn, 2006). Moreover, chaos relates to developmental outcomes across socioeconomic class (Dumas et al., 2005; Evans, 2006; Pike et al., 2006; Wachs & Evans, 2010). It is therefore important to understand the unique role of both SES and household chaos on children’s learning.

The Mediating Role of Effortful Control on the Relations of Household Chaos and SES on Language Skills

Another main goal of this study was to examine children’s EC as both a direct predictor and a potential mediator of the relations between family-level characteristics and the development of language abilities in children. EC has been defined as “a dispositional trait-level representation that represents the tendency to be able to employ top-down control to self-regulate” (Nigg, 2017, p. 363). EC and its underlying neural bases develop significantly over early childhood (Jones, Rothbart, & Posner, 2003; Kochanska, Murray, & Harlan, 2000), although regulatory capacities, overall, continue to increase throughout childhood (Williams, Ponesse, Schachar, Logan, & Tannock, 1999). In general, children who do not possess age- EC skills are at risk for social, emotional, and cognitive difficulties (Kopystynska et al., 2016; Ladd, Birch, & Buhs, 1999; McClelland et al., 2007; Olson, Sameroff, Kerr, Lopez, & Wellman, 2005; Spinrad et al., 2007; Valiente, Lemery-Chalfant, & Swanson, 2010).

Importantly, EC may support children’s ability to capitalize upon the learning opportunities afforded by the environment. In order to learn, children must selectively orient toward a stimulus (physical or social), event, or task, sustain attention, and inhibit attention to distractions. Along with attentional regulation, appropriate behavioral regulation is necessary in order to initiate actions with stimuli or tasks, to inhibit potentially competing action tendencies (e.g., wandering off to play with something else), and to modulate emotional reactions (e.g., during a challenging task). There is mounting evidence that EC is concurrently and longitudinally associated with the development of children’s learning and academic achievement (Eisenberg, et al., 2010; Kopystynska et al., 2016; Merz et al., 2014). For example, EC has been positively related to various cognitive abilities in preschool and elementary school children (Allan & Lonigan, 2011; Liew, McTigue, Barrois, & Hughes, 2008; NICHD ECCRN, 2003; Valiente et al., 2010). Moreover, researchers found that performance on a behavioral regulation task (i.e., tapping inhibitory control, as well as attention and working memory) positively predicted later literacy, vocabulary, and math skills in preschoolers and kindergarteners (Lonigan et al., 2017; McClelland et al., 2007; Ponitz, McClelland, Matthews, & Morrison, 2009).

Additionally, family-level risk—and, specifically, environmental chaos—is associated with deficits in children’s EC across childhood as well as diminished growth in executive functioning skills from 2 to 4 years of age (Dumas et al., 2005; Hardaway, Wilson, Shaw, & Dishion, 2012; Hughes & Ensor, 2009; Raver, Blair, & Garrett-Peters, 2015). School-aged children and adolescents living in chaotic homes have exhibited impaired performance on laboratory tasks and have lower levels of adult-rated EC (Evans et al., 2005; Fuller-Rowell, Evans, Paul, & Curtis, 2015; Garrett-Peters et al., 2019; Valiente, Lemery-Chalfant, & Reiser, 2007). The development of EC may be compromised under conditions of persistent overstimulation because children may become less able or willing to flexibly deploy, allocate, and shift their attentional resources to appropriate sources of learning (Evans et al., 2005; Wachs & Evans, 2010). Likewise, it is difficult for children to develop appropriate behavioral regulation when experiences are unpredictable and non-contingent (Grolnick & Farkas, 2002).

Similar to those exposed to chaos, young children facing economic hardship also tend to have greater difficulties with EC. For children living in poverty, high levels of stressors experienced in their daily lives may disrupt the developing neurobiological systems that underlie stress reactivity and regulation (Blair & Raver, 2012; Blair, Raver, Granger, Mills-Koonce, & Hibel, 2011). Other investigators have demonstrated that economic stress may impact children’s developing EC through its impact on parenting. They note that low income has been related to poor EC independently of other measures of cumulative family risk in high-risk samples (Lengua et al., 2014), so it is important to consider these processes in lower-risk samples as well.

In addition to direct effects of EC on children’s language, we hypothesized that children’s EC would mediate the negative relations between family-level risk factors and language across early childhood. A number of investigators have found that deficits in self-regulatory skills or EC account for some of the negative effects of poverty and cumulative family risk on children’s cognitive and behavioral outcomes (Doan, Fuller-Rowell, & Evans, 2012; Evans & Rosenbaum, 2008; Kim & Brody, 2005). Moreover, in a longitudinal study with low-income families, household chaos negatively predicted children’s inhibitory control (a component of EC), which in turn related to greater externalizing problems (Hardaway et al., 2012). In one recent study, household disorganization predicted children’s outcomes through its relations with positive parenting and children’s regulation (Garrett-Peters et al., 2019). However, this study did not control for stability in the constructs and considered a composite of household disorganization across the first five years of life; therefore, it is unknown whether household chaos at specific developmental periods (i.e., toddlerhood, preschool period) predicts children’s regulation or learning in early development. More work on the specific roles of both income and household chaos in early development, when both EC and language skills are rapidly emerging, is needed.

The Current Study

The goal of the current study was to examine the longitudinal associations among household chaos, family SES, and young children’s language abilities, and the potentially mediating role of children’s EC when predicting language ability, using path modeling. Because of the critical importance of foundational language skills for reading and other domains of learning (Nelson et al., 2011) and the rapid growth in vocabulary across the first years of life (Farkas & Beron, 2004), it is important to focus attention on the development of these competencies during early childhood, prior to school entry. We expected to replicate findings that children living in more chaotic homes, and those with lower family SES, would exhibit deficits in both their language and EC during early childhood. We expected that both household chaos and SES would uniquely predict these outcomes.

We hypothesized that both SES and chaos would directly predict later language skills and EC across time. Further, we expected EC to mediate the relations between family-risk factors and language. That is, low SES and high chaos were expected to predict relatively low EC, and in turn, low EC was expected to predict lower language skills. The current work is novel in its focus on two family-level factors simultaneously and for the use of a cross-lagged panel model design—a design that provides a more causal interpretation than could be provided in prior studies on the topic.

Because children’s oral language abilities may be influenced by both prior language skills and linguistic input from caregivers (Hart & Risley, 1995; Hoff, 2003), and because mothers experiencing sociocontextual risk (i.e., low SES) typically engage in less frequent and less complex speech with their young children (Evans, Hart, & Maxwell, 1999; Hart & Risley, 1995) and those in chaotic homes appear less verbally responsive to their children (Bradley & Caldwell, 1984), we included initial levels of children’s vocabulary and maternal speech in the models. Additionally, we included child sex as a covariate in the models because sex differences (generally favoring girls) have been found in some of the study constructs, including self-regulation (Else-Quest, Hyde, Goldsmith, & Van Hulle, 2006; Moilanen, Shaw, Dishion, Gardner, & Wilson, 2010) and vocabulary (Fenson et al., 1994; Huttenlocher, Haight, Bryk, Seltzer, & Lyons, 1991). We also examined children’s ethnicity/race and non-parental caregiver stability as covariates in the models because these factors may account for children’s language skills in the preschool period.

Method

Participants

The study was approved by the Institutional Review Board at Arizona State University (protocol number 0610001212; Emotion, Regulatory Processes and Social Functioning). Children and families who participated in this study were part of a longitudinal investigation of early emotional development and were recruited at birth from three hospitals in a large city in the southwestern United States. Families visited a university laboratory and completed questionnaires when children were 18, 30, 42, 54, and 72 months of age. Because the third year of life generally brings increased stability in EC (Kochanska et al., 2000), data for the current study were drawn from the assessments at 30, 42, and 54 months of age (T1, T2, and T3, respectively, in this study).

At the initial assessment (at 18 months; not included in the current study), the sample consisted of 256 children and their mothers. At T1, 230 families participated (including 14 who participated only by mail; 128 boys, 102 girls; ages 27.2 – 32.0 months, M = 29.8 months, SD = .65). At T2, 210 families participated (18 only by mail; 117 boys, 93 girls; ages 39.17 – 44.20 months, M = 41.75 months, SD = .65). Finally, the T3 assessment included 192 children and mothers (including 24 only by mail; 108 boys, 84 girls; 52.97 – 57.20 months, M = 53.89, SD = .80). The full sample for the current study was comprised of 236 families (132 boys, 104 girls) with data from T1, T2, and/or T3, with 183 families participating across all three time points.

At T1, the majority (84%) of children were Caucasian, with African Americans (6%), Native Americans (5%), and Asians (3%) also included in the sample (1% of children were identified by parents as “more than one race,” “other” or “unknown”). Additionally, 77% of children were non-Hispanic, with 23% of Hispanic ethnicity. Race and ethnicity percentages were similar at T2 and T3. Annual family income was reported on a 7-point scale and ranged from less than $15,000 to over $100,000, with a median income level of $45,000–60,000. Parents’ education ranged from 8^th grade to the graduate level; the highest level of each parent’s education was rated on a 7-point scale (1 = grade school; 2 = some high school; 3 = high school graduate; 4 = some college or 2-year college; 5 = 4-year college graduate; 6 = Masters degree; 7 = Ph.D. or M. D.); median formal education completed by both mothers and fathers was 4 on the scale. At T1, 59% of all mothers were employed; 61% of mothers at T2 and 66% of mothers at T3 were employed (at each time point, 78–80% of working mothers were employed full-time). The majority (81%) of parents were married and had been married from less than one year to 26 years (M = 6.9 years, SD = 3.9). Approximately 71% of the children had siblings at T1 (M = 1.21, SD = 1.18, range = 0 – 8), and 44% of all children were firstborns.

At each time point, families provided contact information for another adult who knew their child well (e.g., a non-parental caregiver). These caregivers were sent questionnaires that they returned via mail (ns = 152, 151, and 145 at T1, T2, and T3, respectively). Caregivers were often not the same adult across time points (53% of caregivers changed across the study). Those who returned questionnaires described their role as a relative providing care in their own home (T1 = 45%; T2 = 49%; T3 = 34%) or the child’s home (T1 = 23%; T2 = 19%; T3 = 14%); a nonrelative providing care in their own home (T1 = 19%; T2 = 16%; T3 = 14%) or the child’s home (T1 = 7%; T2 = 6%; T3 = 7%); or a childcare provider at a day care center or preschool (T1 = 7%; T2 = 9%; T3 = 30%).

Demographic characteristics of families who began the study at 18 months and continued participation until T3 (n = 185) were compared to those of families lost due to attrition by the last time point (n = 71). Families no longer participating at T3 had marginally lower annual income (M = 3.70; 3 = between $30K and 45K; 4 = $45K to 60K) and significantly lower maternal education (M = 4.03; 4 = some college; 5 = 4-year college degree) than those who remained in the study (Ms = 4.21 and 4.37), ts(226, 238) = 1.93 and 2.17, ps < .06 and .04, respectively. Additional analyses were conducted to compare demographic characteristics and study variables assessed at T1 between families who had data from all three time points in the current study (n = 183) and those who only participated at one or two of these times (n = 53). Children with incomplete data had significantly higher scores on T1 caregiver reports of EC (M = 5.04) and marginally lower gift delay scores (a method of assessing EC) at T1 (M = 2.73) than those with complete data (Ms = 4.63 and 3.24), ts(150, 212) = 2.67 and −1.90, ps < .01 and .06, respectively.

Procedures

At each assessment point, mothers completed a mailed packet of questionnaires prior to the laboratory visit and additional questionnaires during the visit. The lab visits lasted approximately 1.5 to 2 hours and included a mother-child free play session and a delay task in which children waited for a prize. The visits were videotaped and maternal speech during the free play and children’s EC during the delay task were later coded. At the T3 visit, a standard assessment was used to evaluate children’s expressive and receptive vocabulary. After each laboratory visit, mothers were asked to provide permission for questionnaires to be sent to children’s caregivers. Children received a small toy at the end of the laboratory visits, and mothers and caregivers received a modest payment after returning each set of questionnaires.

Measures

Children’s vocabulary.

At T1, children’s expressive vocabulary (i.e., spoken words) was reported by mothers using the short form of the MacArthur Communicative Development Inventory (CDI-Short Form; Fenson et al., 2000). This measure has been shown to have excellent reliability and correlates with other indices of vocabulary (see Fenson et al., 2000). Because some families were bilingual, the form was adapted so that mothers indicated the words spoken by their child in either English and/or Spanish on a checklist of 100 words. Vocabulary scores were computed by summing the number of words spoken in either language.

Language.

At T3, children’s receptive and expressive language were assessed by a trained experimenter, using the Wechsler Preschool and Primary Scale of Intelligence (WPPSI-III; Weschler, 1991; Sattler & Dumont, 2004). The expressive scale included 25 items, and children were asked to define each given word (e.g., “What is a dog?”). Open-ended responses were scored according to a standard checklist. The receptive scale included 38 items, and children were asked to identify the picture (four pictures were presented per item) that corresponded with a given word (e.g., “Show me the easel.”). At the beginning of each scale, children had to provide two correct responses in order to proceed with the remaining words and testing for each scale was discontinued after five consecutive incorrect responses. The total raw scores on the expressive and receptive scales (possible ranges of 0 – 43 and 0 – 38, respectively) were each converted into scaled scores (0 – 19) based on chronological age (Sattler & Dumont, 2004). Expressive and receptive scaled scores have excellent internal reliability scores (see also Lichtenberger & Kaufman, 2004). Expressive and receptive language scores were substantially correlated, r(167) = .49, p < .001; thus, as with other studies (Bandstra et al., 2011; Snowling, Duff, Nash, & Hulme, 2016), scores were averaged to create a composite of children’s language ability at T3.

Maternal mean length utterance (MLU).

At T1, mothers’ MLU was assessed during a 3-minute free play task with children. During the free play session, mothers and children were given a basket of toys and encouraged to play “as they would at home.” Maternal speech was transcribed verbatim from videotape. Of the 216 transcripts of families who participated in the T1 lab visit, six were excluded from further coding because mothers spoke primarily in Spanish, and maternal speech during one additional free play session was not fully audible. The remaining 209 transcripts were segmented into utterances and coded for the use of grammatical morphemes, following the conventions described in Brown (1973). Maternal MLU was calculated by dividing the total number of morphemes by the total number of utterances. The intraclass correlation (ICC) computed for 20% of the sample transcripts coded by two independent coders was .85.

Socioeconomic status (SES).

At T1, mothers reported on mothers’ and fathers’ educational levels and family income. The highest level of each parent’s education was rated on a 7-point scale (1 = grade school; 2 = some high school; 3 = high school graduate; 4 = some college or 2-year college; 5 = 4-year college graduate; 6 = Masters degree; 7 = Ph.D. or M. D.). Annual family income also was reported on a 7-point scale (1 = <$15K; 2 = $15K to 30K; 3 = $30K to 45K; 4 = $45 to 60K; 5 = $60 to 75K; 6 = $75 to100K; 7 = >$100K). Mothers’ and fathers’ highest educational levels and annual family income level were significantly correlated, rs(215–219) = .52 to .62, ps < .001. These three variables were each standardized into z-scores and then averaged to create an index of SES.

Household chaos.

At each assessment, the degree of household chaos was reported by mothers using the Confusion, Hubbub, and Order Scale (CHAOS; Matheny, Wachs, Ludwig, & Phillips,1995). The CHAOS consists of 15 items (e.g., “You can’t hear yourself think in our home,” and “No matter how hard we try, we always seem to be running late”). Observations of various indices of environmental chaos within the home environment have been found to correlate with the CHAOS scale (Matheny et al., 1995), and mothers’ and fathers’ reports on this measure have been moderately correlated (i.e., r = .52). CHAOS items were rated on a true/false scale (1 = true; 2 = false) and averaged (with appropriate items reversed) to create a total score, with higher scores indicating greater levels of chaos within the home. Cronbach’s alphas for this scale were .83, .80, and .79 at T1, T2, and T3, respectively.

Effortful control (EC).

Children’s EC was measured with the Early Childhood Behavior Questionnaire (ECBQ; Putnam, Gartstein, & Rothbart, 2006) at T1 and the Childhood Behavior Questionnaire (CBQ, Rothbart, Ahadi, Hershey, & Fisher, 2001) at T2 and T3. Mothers and caregivers rated items tapping EC from the attention focusing, attention shifting, and inhibitory control subscales (1 = never, 7 = always). The attentional focusing subscale included 12 items (ECBQ) or 14 items (CBQ) that assessed children’s ability to concentrate on a task (e.g., “When engaged in play with his/her favorite toy, how often did your child play for more than 10 minutes?”), αs = .81, .77, and .77 for mothers and .85, .74, and .72 for caregivers, at T1, T2, and T3, respectively. The attention shifting subscale consisted of 12 items (for both the ECBQ and the CBQ) that assessed children’s ability to alter the focus of their attention from one activity to another (e.g., “During everyday activities, how often did your child seem able to easily shift attention from one activity to another?”) αs = .73, .67, and .73 for mothers and .71, .80, and .72 for caregivers, at each age. The inhibitory subscale included 12 items (ECBQ) or 13 items (CBQ) reflecting children’s ability to control their behavior (e.g., “When told ‘no’, how often did your child stop an activity quickly?”), αs = .88, .77, and .80 for mothers and .88, .82, and .83 for caregivers, at each age. Composite scores of reported EC were created by averaging the subscale scores for attention focusing, attention shifting, and inhibitory control, separately for mothers and caregivers, respectively, at T1, rs(218–221) = .30 to .36, ps < .01, and rs(141–143) = .45 to .53, ps < .01; at T2, rs(203) = .23 to .51, ps < .01 and rs(147–148) = .41 to .68, ps < .01; and at T3, rs(186) = .21 to .56, ps < .01 and rs(143–144) = .39 to .64, ps < .01.

Children’s EC also was assessed at each age during a delay paradigm (Kochanska et al., 2000). Children were presented with a gift bag and were told that it contained a prize for them. Children were left alone in the room, with the gift bag directly in front of them on the table and were asked not to touch or look in the bag until the experimenter returned with the bow (2 minute delay). Children’s level of restraint was coded on a 5-point scale (1 = child pulls box from bag, 2 = child puts hand into bag, 3 = child peeks in bag, 4 = child touches bag but does not peek, 5 = child does not touch bag). ICCs computed for 25% of the sample were .96, .95, and .98 at T1, T2, and T3, respectively.

Due to a relatively small sample size and a large number of constructs across three time points, we decided to create a composite score of observed and reported EC by standardizing and averaging mothers’ and caregivers’ reports on EC and observed EC during the gift delay task. At all time points, mothers’ and caregivers’ reports on EC were significantly correlated, rs = .18, .25, and .31, ps < .05, .01., and .001 for T1, T2, and T3 respectively. Further, mothers’ reports were significantly correlated with children’s observed restraint at all time points, rs = .16, .27, and .18, ps < .05, .01, .05. Although caregivers’ reports were correlated with observed delay at T2, r =.24, p < .01, it was at the trend level at T1, r =.16 and was nonsignificant at T3. To maintain the same measurement over time, we used the composite score of EC in all data analyses.

Results

Preliminary Analyses

Descriptive statistics.

Descriptive statistics for the study variables can be found in Table 1. Given the assumption of univariate normality underlying model estimation procedures, skew and kurtosis of the study variables were examined; all variables were distributed normally (Curran, West, & Finch, 1996).

Table 1.

Correlations, Means, and Standard Deviations of Study Variables at T1, T2, and T3

Variable	1	2	3	4	5	6	7	8	9	10
1. T1 vocab	---	.12	.16*	-.04	.29**	.02	.18*	−.01	.14	.30***
2. T1 MLU		---	.25**	−.03	.14*	−.01	.10	−.01	.15	.25**
3. T1 SES			---	−.19**	.21**	−.21**	.24**	−.10	.24**	.45***
4. T1 chaos				---	−.27**	.79***	−.22**	.75***	−.22**	−.25**
5. T1 EC					---	−.24**	.62**	−.20**	.51**	.26**
6. T2 chaos						---	−.33**	.73***	−.35**	−.28***
7. T2 EC							---	−.24**	.62**	.37**
8. T3 chaos								---	−.23**	−.18*
9. T3 EC									---	.42**
10. T3 language										---
M	72.88	4.23	−.05	1.27	−.03	1.27	−.01	1.26	−.002	11.25
SD	22.90	.87	.84	.22	.72	.21	.78	.20	.76	2.81

Note. Vocab = vocabulary; MLU = maternal mean length utterance; SES = socioeconomic status; EC = effortful control.

^*p < .05.

^**p < .01.

^***p < .001.

Zero-order correlations among the study variables.

Correlations for the study variables can be found in Table 1. Maternal reports of household chaos and children’s EC showed stability over time. Children’s early vocabulary level at T1 was positively and at least marginally related to SES, children’s EC at T1 and T2, and to children’s T3 language skills but was unrelated to mothers’ reports of chaos. Maternal MLU was positively related to children’s EC at T1 and to language skills at T3.

In examining the main constructs of interest, SES was negatively related to household chaos at T1 and T2 and was positively related to children’s EC at all ages. Moreover, children from higher-SES families had higher language scores at T3. Household chaos was negatively related to children’s EC within and across time. Chaos at each age was negatively related to children’s language at T3. Finally, children’s EC at all time points were positively related to the T3 language skills.

Differences by child sex, ethnicity, and caregiver stability.

We tested whether children’s sex, children’s race/ethnicity (i.e., White non-Hispanic or not, ns = 142 and 68 respectively), and nonparental caregivers’ stability (i.e., the caregiver was the same individual across the study or not, ns = 70 and 78 respectively) were related to children’s T3 language. T-tests indicated that children’s T3 language differed for boys and girls, but it was unrelated to ethnicity or caregiver stability, ps > .05. Specifically, girls’ T3 language (M = 11.71, SD = 2.99) was higher than boys’ T3 language (M = 10.85, SD = 2.59), t(166) = 2.00, p = .047. Because each of these variables could contribute to children’s language skills (and may have been related to other variables in the model), we included children’s sex, children’s race/ethnicity, and caregivers’ stability as control variables in the hypothesized model.

Primary Analyses

Hypothesized model.

Longitudinal path analyses controlling for stability in the constructs over time were computed to evaluate the main study hypotheses. The path analysis was conducted with eleven variables, including early vocabulary, mothers’ MLU, SES, children’s sex, children’s race/ethnicity, chaos and EC at T1, T2, and T3, and language at T3. The model specified within-time correlations and included autoregressive paths for chaos and EC (i.e., T1 to T2; T2 to T3) and structural paths from a) T1 Child sex to T3 language, b) T1 Race/ethnicity to T3 language, c) T1 SES to chaos and EC at T2, and T3 language, d) T1 chaos to T2 EC, e) T1 vocabulary to T3 language, f) T1 maternal MLU to T3 language, g) T2 chaos to T3 EC and language, and h) T2 EC to T3 language. Models were tested using Mplus Version 7.4 (Muthen & Muthen, 1998-2019). To test model fit, we examined the chi-square, the comparative fit index (CFI), root mean square error of approximation (RMSEA) and the standardized root mean square residual (SRMR). The model fit the data adequately (Hu & Bentler, 1999), χ²(26) = 63.277, p < .001; CFI = .94; RMSEA = .07 (90% CI = .05 - .10); SRMR = 0.04. The modification indices in Mplus suggested that we add an additional stability path from T1 chaos to T3 chaos, which improved the model fit: χ²(25) = 34.48, p = .09; CFI = .99; RMSEA = .04 (90% CI = .00 - .07); SRMR = 0.03. As portrayed in Figure 1, all autoregressive paths in the model were positive and significant. In terms of hypothesized relations among the constructs, the path from T1 SES to T2 chaos was non-significant, but SES positively predicted children’s T2 EC and children’s language at T3. Although T1 chaos did not predict T2 EC after controlling for stability in the constructs, the path from T2 chaos to T3 EC was negative and significant. Chaos at T2 predicted T3 language. T1 vocabulary and T2 EC predicted children’s language at T3, but children’s sex, children’s race/ethnicity, and maternal MLU was unrelated to children’s language at T3 after accounting for the other constructs.

Figure 1.

Unstandardized estimates are presented first; completely standardized estimates are presented in parentheses. Solid lines are significant, p < .05; dashed lines are non-significant. Child sex is coded 1 = boy, 2 = girl; Race/ethnicity is coded 1 = White non-Hispanic, 2 = ethnic minorities; SES = socio-economic status; Chaos = household chaos; EC = effortful control; MLU = maternal mean length utterance.

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

In terms of concurrent relations among the constructs, SES was negatively related to chaos and positively related to children’s EC, children’s vocabulary, and maternal MLU within T1; moreover, EC was positively correlated with vocabulary and negatively correlated with chaos within time at T1. Chaos and EC were negatively related within time at T2. EC and language were positively related at T3. Children’s sex was related to children’s vocabulary and EC at T1 (favoring girls). Children’s race/ethnicity was related to SES and maternal MLU at T1 (favoring White non-Hispanic).

The test of mediation of the effect of SES on language at T3 by children’s EC at T2 yielded a significant indirect effect, β = .04, 95% CI [.01, .09]. Because the direct effect of SES on language at T3 was still significant after taking children’s EC at T2 into account, children’s EC at T2 was a partial mediator of the relation of T1 SES to T3 language.

Discussion

Our research showed that EC longitudinally mediated the relations between SES and children’s language skills in early development. Although causal conclusions can not be drawn from our data, the longitudinal design with stability paths controlled allowed us to test the predictive relations of family-level environmental factors to children’s language skills more stringently than in prior studies. Further, we showed that household chaos by 42 months of age was uniquely and negatively predictive of young children’s developmental outcomes. As important factors of the family environment, household chaos and SES independently and longitudinally predicted children’s later language skills.

Given that children in socioeconomically disadvantaged families often have more chaotic households (Whitesell et al., 2015), one aim of this investigation was to examine and tease apart the effects of these family-level risk factors. As hypothesized, household chaos negatively predicted children’s later language abilities, even after accounting for family SES and a number of other relevant influences (i.e., child sex, race/ethnicity, early vocabulary and maternal speech, EC). These findings are consistent with other investigations demonstrating similar links between aspects of environmental chaos and poorer language, pre-literacy, and literacy skills in preschoolers and academic readiness in older children (Burchinal et al., 2018; Garrett-Peters et al., 2019; Maxwell & Evans, 2000; Merz et al., 2014; Pike et al., 2006), as well as those in which chaos and SES have had independent effects on children’s development (Dumas et al., 2005; Pike et al., 2006). It is important to note, however, that a parent’s own low level of EC could contribute to household chaos and/or genetically influence children’s EC. Although this possibility was not examined in the current study, see Bridgett and colleagues (2015) for a review of this issue.

We also explored the possibility that household chaos might be a manifestation of family socioeconomic risk within children’s most immediate environment, thus mediating its effects. Although related concurrently, SES did not predict chaos over time, suggesting that in this sample, household chaos was not simply a more proximal function of SES; nor did it not account for the impact of SES on later language. However, the lack of cross-time prediction of chaos from SES was likely due to the high level of stability in mothers’ reports of chaos (with correlations > .70), which has been found in other work using both survey and observational measures of chaos (Deater-Deckard, Chen, Wang, & Bell, 2012). Recognizing that the duration of exposure to adversity can increase the severity of developmental consequences, these high levels of stability are also a cause for concern. For some children, household chaos may be less tied to transient family circumstances and more of an enduring feature of their daily lives, and these children may be most vulnerable to its effects. Research illustrating the link between cumulative risk and elevated allostatic load (Evans, 2004) suggests that chronic exposure to household chaos may have important cumulative effects on physical, cognitive, and psychosocial functioning.

Another important goal of this investigation was to better understand the mechanisms through which family-level risk might impact children’s development. Thus, we examined longitudinally the role of EC in the relations between SES, household chaos, and language across early childhood while controlling for stability of the constructs over time. Contrary to expectations, chaos failed to predict children’s T2 EC when controlling the stability of EC, which may in part be due to the considerable stability in EC across this period. However, higher levels of chaos within the home predicted lower EC from T2 to T3, even when controlling for earlier EC. Multiple studies have shown that children experiencing environmental chaos have poorer attentional and regulatory skills (Evans et al., 2005; Garrett-Peters et al., 2019; Vernon-Feagans, Willoughby, & Garrett-Peters, 2016). However, one novel feature of the current study is that we examined unique relations of household chaos at multiple time points in early development (rather than using a composite across the first years of life).

In addition to the stability in EC over time, another reason for the lack of prediction of 42 month EC from 30 month chaos may be the timing of exposure to chaos. Although the measure of household chaos used in this study demonstrated high stability across time, it may also be the case that there are particular developmental points at which its relations with EC are more pronounced. For example, as children progress through the early childhood years, they often encounter more rules and greater expectations for compliant and socially-acceptable behavior, particularly if they are in an early educational environment. The difficulties in effortful control associated with exposure to chaos may become more evident to parents and caregivers closer to the preschool years than during the toddler years because EC is not fully developed earlier and lapses may be seen as more normative. Alternatively, it is possible that the predictive relations between chaos and EC are apparent only at later ages due to a dosage effect. Exposure to chaos may have cumulative effects such that only with chronicity does it begin to reveal significant relations with children’s regulatory abilities. Both of these explanations would parallel evidence that the timing and duration of exposure to economic hardship influence the strength of its relation with developmental outcomes (Bradley & Corwyn, 2002).

Our findings also showed that T2 chaos predicted lower language skills at T3. If chaotic environments impair children’s capacity to flexibly and volitionally regulate their attention and behaviors, they may have difficulty focusing and engaging with developmentally-facilitative stimuli and may miss critical opportunities for learning. Of particular relevance, if children indiscriminately “tune out” external stimuli (Evans & Lepore, 1993), this may also include speech, which may limit the amount of linguistic input they receive and hinder their developing language skills.

In the current study, EC mediated the longitudinal relation between SES and language skills across three years. SES at T1 was positively related to change in EC at T2, in accordance with previous studies that found the positive relations between economic hardship and difficulties in young children’s EC (Blair & Raver, 2012; Lengua et al., 2014). EC at T2 was positively related to children’s language abilities a year later, even when including other predictors of language in the model and early vocabulary. These findings are in concordance with a number of investigations demonstrating significant associations, often predictive, between aspects of self-regulation and young children’s vocabulary, literacy, and math skills (McClelland et al., 2007; NICHD ECCRN, 2003; Ponitz et al., 2009); although, our study is unique in the use of three-waves of longitudinal data across the early years and controlling for stability in the constructs.

The current study had multiple strengths that should be highlighted. We utilized a longitudinal design with observational and multi-reporter data to assess the relations among household chaos, SES, children’s EC, and their language, including disentangling the roles of chaos and SES and examining potential mediating processes. The use of three-wave, autoregressive panel models allowed for more rigorous analyses and tests of mediation that revealed unique effects of the predictors after taking into account the stability in constructs over time (Cole & Maxwell, 2003).

Despite considerable strengths, some limitations of the study must be acknowledged. Importantly, household chaos was assessed only by maternal report, and the extent to which adult perceptions mirror children’s “felt experience” of chaos is unclear. Maternal ratings may not accurately represent the level of environmental chaos to which children were actually exposed, although prior research has found significant associations between ratings on the CHAOS measure and observations within the home (Matheny et al., 1995). There may also be unmeasured individual differences that contributed to mothers’ perceptions of chaos, such as aspects of their personality, mental health, self-regulatory abilities, or tolerance for environmental stressors. For example, Valiente and colleagues (2007) found that parents’ ratings of their own EC were negatively related to their reports of chaos within the home. Combining parental ratings of chaos with more objectively-measured indices (e.g., Vernon-Feagans et al., 2012), as well as examining whether distinct dimensions of chaos relate uniquely to different child outcomes (Hardaway et al., 2012), would strengthen the conclusions that can be drawn from such investigations. Additionally, mothers reported on both chaos and children’s EC, which may raise concerns about shared reporter variance. However, the EC composite was comprised of both observational and adult-reported measures, and caregivers often different individuals reporting across time, lending greater confidence in the finding that greater levels of chaos reported within the home are predictive of lower EC in children. Another limitation was that we did not measure children’s language at T1 and T2 to control for language’s stability across waves. The approximate variable utilized was vocabulary at T1. Finally, as with many longitudinal studies, there was moderate attrition across the course of the study. Although expected, researchers should make every effort to reduce attrition as well as examine whether missing data may be attributed to family-level factors, children’s EC, or children’s learning.

Overall, the findings of this study highlight important predictors of language development prior to school entry and suggest potential avenues for prevention and intervention. Given the detrimental effects of household chaos on preschool-aged children, regardless of their socioeconomic standing, efforts should be made to reduce the level of chaos in all families’ lives. This includes identifying the factors that prevent parents from structuring the household effectively and the supports that they need in order to accomplish this, and these likely differ across families of different backgrounds. The broader goal of limiting exposure to environmental chaos should also be extended to childcare, educational, and other institutional contexts in which young children have regular or prolonged experience. Classroom chaos has been shown to predict elementary students’ school motivation; thus, there is a clear need to examine how chaos in other environments may influence children’s learning (Berger et al., 2017). Moreover, the effects of EC highlighted in this study call for programs aimed at improving children’s early learning through a focus on building and strengthening effortful control. Specifically, this study indicates that EC directly predicts children’s language skills, even when taking into account a number of other relevant factors. Thus, developing effective early interventions to promote children’s effortful control may be an important way to improve language skills and eventual academic achievement, particularly for children exposed to sociocontextual risks.