Anxious and depressive symptoms are subsumed under the overarching category of internalizing problems (Zahn-Waxler et al., 2000). They affect ~20% of youth at some point from childhood to adolescence, and have been associated with a series of negative outcomes such as academic failure, substance abuse, psychopathology, difficulties with social relationships, and suicide (e.g., Lewinsohn et al., 1993; Fergusson and Woodward, 2002). A review by Kovacs and Devlin (1998) suggested that anxiety disorders typically surface sometime during early to middle childhood (e.g., 6–11 years) and show moderate stability in childhood and adolescence (Gullone et al., 2001). In contrast, depressive disorders typically appear during early to middle adolescence (e.g., 11–15 years). Cole et al. (1997) further differentiated this progression with findings that anxiety symptoms and depressive symptoms around third grade were virtually indistinguishable in a normative sample. By sixth grade, each construct was distinct but highly related. Clark and Watson's tripartite model (1991) posits that the high co-occurrence for anxiety and depression emanates from a shared underlying component—negative affective temperament. Research has substantiated this claim in that difficult temperament and behavioral inhibition have been associated with risk for internalizing disorders (e.g., Biederman et al., 1990).

Recent evidence supports a competency-based model in the development of psychopathology asserting that a number of stage-salient skills or self-regulatory processes must be negotiated for competent development—for example, affect regulation, impulse control, social competence, and managing school transitions (Bosquet and Egeland, 2006). Failure to achieve competence in regulating one or more developmental processes may lead to internalizing disorders. At the same time, the development of internalizing disorders may interfere with attainment of competency. The authors examined this model by studying the etiology and course of anxiety symptoms from infancy through adolescence. They found that anxiety symptoms reflected moderate stability over the period. Heightened negative reactivity in infancy predicted emotion dysregulation in preschool which predicted anxiety symptoms in childhood. In turn, failure to regulate behaviors and emotion in childhood social interactions predicted anxiety symptoms in preadolescence, and anxiety symptoms in preadolescence predicted social incompetence in adolescence.

Buck and Dix's (2012) examination between development of depressive symptoms and peer relationships in early adolescence offers insight into why depressive symptoms may increase as children approach adolescence. Their findings suggested that youth who were less capable at regulating responses to physical and social stresses later in childhood were more likely to develop more depressive symptoms in adolescence.

In middle childhood, common sleep problems across various sub-categories of anxiety disorders include sleep onset latency, insomnia, nightmares, and nocturnal panic attacks (Gregory and O'Connor, 2002; Alfano et al., 2006; Kushnir and Sadeh, 2011; Kushnir et al., 2014) Overlap exists between the neurological processes required to self-regulate emotions and sleep. An increasing body of literature has concluded that the neurological circuitry for regulating emotions lies within the prefrontal cortex and the amygdala (e.g., Bechara et al., 1999). Yoo et al. (2007) found that sleep deprived individuals exhibited over 60% greater magnitude of amygdala reactivity in response to aversive stimuli compared to controls. Functional connectivity between the amygdala and the medial prefrontal cortex (mPFC) also showed significant reduction. Similar patterns of neuronal activity have been associated with anxiety disorders and depressive symptomatology (Davidson, 2002; Kent and Rauch, 2003).

Research has demonstrated that children who experience difficulty regulating their anxiety and fear at bedtime may also experience problems with falling asleep, potentially leading to bedtime resistance. Excessive parental involvement in managing bedtime resistance may adversely affect children's capacity to self-regulate both their sleep and emotion with risk for later developmental problems in both areas (Tikotzky and Sadeh, 2009). Mindell and Barrett (2002) reported increased levels of trait anxiety in 5–11 year old children with three or more nightmares per week. The ability to return to sleep without parental intervention after a nightmare requires the capacity to self-sooth and regulate. These findings demonstrate the importance of self-regulation in children's ability to manage the overlap between sleep and emotions. Evidence exists for developmental differences in relations between sleep and anxiety and sleep and depressive symptoms (Alfano et al., 2009). In prepubertal children, sleep and anxiety appear to be more closely related whereas sleep and depressive symptoms are more closely related in adolescents.

Social competence requires regulation of behaviors and emotions to forge friendships, build cooperation, and willingly participate in group activities (Bodrova and Leong, 2006). The developing ability to regulate attention control and inhibit responses, along with growth in cognitive processes such as working memory and planning functions, increase the child's ability to not only manage response strategies but to attend more constructively to social cues in engagement with family, teachers, and peers (Bronson, 2000; Blair, 2003). Achieving milestones on the path to social competence necessitates a series of transactions and dependencies involving motor, representational, and linguistic systems. Development of sensorimotor skills and growth in psychological processes such as attention, memory, language, and executive function alter arousal and response tendencies from infancy through middle childhood—all of which underpin and promote regulation of behavior and emotion in support of successful social interaction.

Children who demonstrate poor self-regulation are more likely to be rejected and ostracized by peers (Bierman et al., 2015). Consequently, the interplay of poor behavioral control and peer rejection creates a dynamic process highlighting the effects of individual differences in social-emotional functioning as early as elementary school (Miller-Johnson et al., 2002). Evidence has shown that social incompetence early in middle childhood predicted increased internalizing symptoms at the end of middle childhood (Gazelle and Ladd, 2003). Burt et al. (2008) reported similar associations during this period using a design capable of establishing directionality. Social competence in middle childhood showed negative longitudinal links to internalizing symptoms in adolescence in a normative sample; reverse associations and gender differences were not found. The authors concluded that findings were broadly consistent with developmental competency-based models of internalizing symptoms with important implications for positive adjustment in the future.

Developing social competence involves the interaction and integration of multiple regulatory systems that govern brain/behavior relations. Maturational processes within the brain and nervous system continue to develop into adolescence (Gunnar and Vazquez, 2006), and regions of the prefrontal cortex, associated with behavior and emotion regulation, are among the slowest maturing brain systems. Numerous parallels and overlap exist with sleep-related neurobiologic systems and the neurobehavioral systems that regulate affect, attention, learning and memory, and language (Dahl, 1996). Sleep is thought to optimize functioning of the prefrontal cortex (PFC), and research shows that sleep deprivation interferes with the cognitive modulation of impulses, drives, and emotions governed by PFC (Horne, 1993). Dahl suggests that waking synchronization of multiple PFC-governed behaviors is particularly sensitive to sleep loss, and sleep loss may interfere with the complex processing required for effective social interaction.

During the preschool years and beyond, girls demonstrate better social competence due to lower prevalence of externalizing behaviors, better effortful control and language development, and less physical aggression (Zahn-Waxler et al., 1995; Kochanska et al., 2000; Broidy et al., 2003). Evidence suggests such behaviors are relatively stable across development and thus contribute to stability in levels of social competency over time (Masten, 2005; Burt et al., 2008).

Two studies of children's sleep problems examined relations between sleep and social competence. Vaughn et al. (2015) found that sleep duration was positively correlated with social competence and receptive vocabulary. Lemola et al. (2011) reported that sleep quantity and quality were related to optimism and self-esteem, but no significant associations surfaced between sleep and social competence.

Pubertal onset has been closely linked to changes in affective domains such as increased emotional intensity, sensation seeking, risk-taking, sleep/arousal regulation, and affective disorders (Herman-Giddens, 2006). It appears likely that affective changes, typically ascribed to adolescence and the teenage years, may occur somewhat earlier as puberty may occur somewhat earlier (Carskadon, 2011). Thus, we use the term “preadolescence” to isolate this period of rapid and extensive change when children typically transition from elementary to middle school. In light of these emerging trends, it is important to consider relations between sleep, affective functioning, and children's adjustment immediately prior to the teen years.

Biological and psychosocial changes that take place during this time manifest themselves through increased stress responsiveness and emotional reactivity (Dahl and Lewin, 2002; Nelson et al., 2005; Steinberg, 2008). Although this period may represent unique opportunities for learning and developing social and emotional skill, it also presents a time of considerable risk. In accordance with previous sections of this paper that reviewed the literature on children's psychological functioning and sleep problems across middle childhood, children with feelings of maladjustment in preadolescence may be more likely to report lower levels of self-esteem, feeling less connected and competent in school, and more depressive symptoms. A dramatic increase in the extent of time spent with peers coincides with heightened emotional reactivity, depression, sensation-seeking, risky behaviors, and a greater need to align with peer group norms (Gardner and Steinberg, 2005; Steinberg, 2008). Evidence has shown considerable overlap among these constructs such that biological and environmental risk factors may explain why some children develop maladaptive behaviors leading to poor emotional and social adjustment. Insufficient sleep, more anxious and depressive symptoms, and poor social competence have shown unique associations with poor social and emotional adjustment during the transition to adolescence. Less clear, however, are the interactions among these domains across middle childhood and their cumulative effects on preadolescent functioning.

Children's feelings of competence and well-being are the product of the dynamic interplay of academic and social experiences that have taken place over a lengthy period of childhood (Parker et al., 1995; Ladd and Burgess, 2001). Developmental differences are evident in children's perception of academic competence (Stipek and Daniels, 1988; Wigfield et al., 1997). In early elementary school, young children typically overestimate their academic abilities as demonstrated by lack of convergence with teacher ratings of performance (Stipek and Iver, 1989). By second or third grade, however, closer alignment begins to exist between children's perspectives and objective indices of performance (Wigfield et al., 1997). By the time children reach the end of middle childhood, their feelings about themselves as students have had sufficient time to form. The adaptive or maladaptive feelings that children develop at the end of middle childhood begin to form the basis for coping with broader academic and psychological adjustments associated with impending school transitions.

By adolescence, coping styles and strategies to regulate emotions have likely influenced by both biological and environmental factors. Emotion regulation is based on the premise that “emotion is a multifaceted phenomenon involving physiological arousal, neurological activation, cognitive appraisal, attention processes, and response tendencies” (Thompson, 1994, p. 30). It reflects a developmentally changing phenomenon due to its reliance on biological and neurological processes. Similar to temperament models, individual differences illuminate why emotional arousal and selection of emotion response strategies lack uniformity and consistency across individuals (Thompson et al., 2011). Development of sensorimotor skills and growth in psychological processes such as attention, memory, language and executive function alter arousal, and response strategies from infancy through middle childhood (Harman et al., 1997; Zelazo and Cunningham, 2007).

Emotional dysregulation has been related to poor social functioning, internalizing and externalizing behavior problems, and maladaptive risk-taking (Aldao et al., 2016). Because sleep deprivation has been associated with emotional processing, it is not surprising that some correlational studies have demonstrated decrements in mood, increased irritability, heightened sensitivity to negative stimuli, and ineffective coping strategies after experimental sleep restriction. Less clear, though, is the cumulative effect of interactions between sleep and affect on emotion regulation across middle childhood.

As children approach adolescence, personal responsibility for judgment and decision-making increases while adult oversight of both academic and social behaviors gradually diminishes over time. Deciding to engage in maladaptive risk-taking behaviors stems from a breadth of factors including temperament, neurobehavioral disinhibition, social competencies, parenting, and peer behaviors (Feldstein and Miller, 2006). Risk-taking in the preadolescent period has been referred to as lack of attention to personal safety, stealing, cheating, property destruction, and substance abuse. A recent study by Thomas et al. (2015) has linked sleep disturbances (a factor in PFC performance) to maladaptive risk-taking, largely mediated by working memory processes. O'Brien and Mindell (2005) also found evidence for insufficient sleep and increased risk-taking behavior in adolescence. Tarter et al. (2004) have referred to neurobehavioral disinhibition as a composite of executive cognitive functioning, emotional regulation, and behavioral control with each function governed by the prefrontal cortex (PFC).

Longitudinal evidence across this period of middle childhood in normative samples (Fredriksen et al., 2004; Gregory et al., 2005; Alfano et al., 2009) suggests that sleep problems precede symptoms of anxiety and depression with relations beginning as early as school entry. These studies tested mostly unidirectional associations between domains and did not follow children across middle childhood. More recently, Kelly and El-Sheikh (2013) identified reciprocal relations between sleep and internalizing behaviors in children ages 8–13. Using a cross-lagged research design capable of identifying bidirectionality between domains, they found bidirectional effects with sleep having stronger effects on behavioral adjustment (e.g., internalizing and externalizing behaviors) than the reverse associations. In a prior study using a similar sample, El-Sheikh et al. (2012) reported similar associations with relations more pronounced for girls. Using a longitudinal cross-lagged model, Gregory et al. (2009) reported that sleep problems at 8 years predicted symptoms of depression 2 years later but reverse associations were not found. Wang et al. (2016) most recently found that emotional difficulties at age 5 differentiated normal sleepers from troubled sleepers over almost a decade, but the groups did not differ in emotional functioning in late adolescence. Even though gender was not significantly related to the initial level of sleep problems, girls with more early anxious and depressed symptoms were more likely over time than boys to be troubled sleepers. These studies lend insight into relations between sleep and affect during middle childhood and illustrate the importance of studying boys and girls separately. The limited concordance in findings, however, leaves open questions about directionality over extended periods of development, pathways of risk for emotional and social adjustment, and the potential for gender differences.

The current study investigates the series of transactions that take place for sleep, anxiety and depressive symptoms, and social competence in a large community sample of children from preschool to fifth grade using a four-wave design capable of uncovering the direction of effects (Selig and Little, 2012) among these domains and their cumulative effects on emotional and social adjustment at the end of middle childhood (e.g., preadolescence). We chose a 6-year interval—between 4 ½ and 10 ½ years-of-age—because it covers all of middle childhood as well as two important transitionary periods—the start of formal schooling during which time children learn to adapt to new academic and social demands—and the transition to adolescence, a sensitive period of physical, cognitive, emotional, and social development. We defined sleep problems as night awakenings that disrupt sleep (including having nightmares, talks or walks in sleep, or wets the bed), sleeping less or more than typically developing children, difficulty initiating or maintaining sleep, and daytime overtiredness. We use the term preadolescent adjustment to encompass multiple aspects of emotional and social development that have been associated with sleep problems and affective functioning such as the ability to control emotional reactivity, levels of depressive symptoms, feelings about school, and engagement in maladaptive risk-taking behaviors. Our study used multiple informants, controlled for covariates associated with sleep and affect, and we examined effects separately by gender. Several studies have suggested that biological changes to sleep during pubertal development may result in different findings for boys and girls (Angold et al., 1998; Herman-Giddens, 2006; Sadeh et al., 2009; Carskadon, 2011).

Figure 1 reflects a conceptual model that hypothesizes bidirectional relations among domains across middle childhood to predict emotional and social adjustment during the transition to adolescence. This model depicts the interplay of multiple domains that may result in effects deemed as direct and unidirectional, direct and bidirectional, or indirect through various pathways, with each successive stage considered a partial product of previous stages and the current stage. Similar to developmental cascade models (Masten and Cicchetti, 2010), this model suggests that transactions that occur among domains over time may have adaptive or maladaptive cumulative effects on later functioning. Because cascade models typically focus on an entire developmental period, they have the potential to uncover periods of risk that may inform preventative intervention.

Finally, to rule out the possibility that any relations over time are due to shared risk factors (Coulombe et al., 2010) we included controls for maternal depressive symptoms (Warren et al., 2006; Seifer, 2011; El-Sheikh et al., 2012; Weinraub et al., 2012), difficult child temperament (Van Tassel, 1985; Sadeh et al., 1994; Atkinson et al., 1995; Owens-Stively et al., 1997; El-Sheikh and Buckhalt, 2005; Weinraub et al., 2012), marital instability (El-Sheikh and Kelly, 2011; Mannering et al., 2011), lax parental control (Owens-Stively et al., 1997; Noble et al., 2012), and lower socio-economic status (Grandner et al., 2013; Singh and Kenney, 2013).

We addressed five specific questions: (1) how are domains at adjacent time points from preschool to fifth grade related, (2) what is the magnitude of these relations and the pathways of influence to preadolescent functioning, (3) does early functioning directly predict later functioning over and above their indirect effects, (4) are there cross domain indirect effects from transactions across middle childhood, and (5) are there differences in these relations for boys and girls? Because research has focused limited attention on relations among all of the domains of interest and spanned the entire period of middle childhood, this study is exploratory.

With a large and diverse sample of families from 10 different locations across the United States, the National Institute of Child Health and Development Study of Early Child Care and Youth Development (NICHD-SECCYD) publicly available longitudinal data set was ideal for testing our model. Details of study recruitment and protocols are described at http://www.icpsr.umich.edu. NICHD-SECCYD researchers recruited 1,364 families in 1991 through hospital visits with mothers giving birth (n = 8,986) during selected 24-h intervals in 10 locations in the United States. Written informed consent was obtained from all participants—parents and teachers. Proper assents were also obtained from the children at each age. The sample used in this report consisted of families in which mothers reported children's sleep problems, beginning at 54 months old, the initial time point for analysis (n = 1,057), and excluded those families where the informant for sleep was listed as “other” (n = 7). The sample was diverse, including 24% ethnic minority children, 11% mothers who had not completed high school, and 14% single mothers. At the start of the study, mothers had an average of 14.4 years of education (SD = 2.49 years); 51.7% of children were boys. The present report focused on children beginning at 54 months-of-age (M = 4.5, SD = 0.23) with assessments proceeding in 2-year, equal intervals (Grade 1, Grade 3, Grade 5) to provide a sufficient time lag for effects to take place. In Grade 1, children were 6.5 years of age; in Grade 3, children were 8.5 years of age, and in Grade 5 children were 10.5 years of age. There were no differences in ages between the boys and the girls.

Children and their mothers came to the university laboratories when children were 54 months old and when children were in Grades 1, 3, and 5. At these visits, standardized assessments were administered to the children and mothers. Teachers completed questionnaires regarding the study child's academic and social development from kindergarten through Grade 5. Children completed questionnaires about their social and emotional functioning in Grade 5.

Table 1 depicts the timing of measurement for each predictor, outcome, and control variable used in the study.

Predictor variables were measured repeatedly at the four time points listed above.

The primary analyses used structural equation modeling (SEM) in Amos™ 19.0 (Arbuckle, 2010) to establish the relations and pattern of effects among the variables in the conceptual model Bentler (1988). Given the longitudinal nature of the data, some participants had missing values. We used full information maximum likelihood estimation (FIML) to address missing data in order to avoid omitting participants. FIML produces less biased estimates than either mean substitution or list wise case deletion (Graham, 2009; Acock, 2012).

Data analyses proceeded in seven stages and followed procedures similar to Little (2013) and Masten et al. (1995) for testing longitudinal cross-lagged panel models. For the entire sample, first we tested a model that examined stability between repeated measures (autoregressive model) from one assessment period to the next to examine rank order among participants over time while controlling for across and within-time covariation (Model 1). Second, we added cross-lagged paths between adjacent time points to test for bidirectional relations and the magnitude of effects among the domains across four waves of data (Model 2). Third, to test whether early effects of sleep problems, anxious-depressed symptoms, and social competence in middle childhood would exert direct effects on these same domains later in middle childhood (above their indirect effects), we added direct paths to Model 2 from each 54 month and Grade 1 domain to each domain at Grade 5 (Model 3). We removed (trimmed) all non-significant paths to improve parsimony. Then, we added a set of variables associated with preadolescent emotional and social adjustment to Model 3 to uncover the direct and indirect effects across middle childhood for sleep problems, anxious-depressed symptoms, and social competence (Model 4). Next, we added covariates to Model 4 using the semipartial control method (Little, 2013) to examine their predictive influence on the domains and the potential for moderating effects, and to determine whether relations among domains changed after inclusion of covariates (Little et al., 2007; Model 5). Then, we conducted a multiple group analysis based on Model 4 to examine whether relations differed for boys and girls before adding covariates (Model 6). Next, we added covariates to Model 6 (using the semipartial control method) to examine the influence of covariates in the gender model (Model 7). Last, we assessed the cross-domain overall indirect effects on outcomes across middle childhood and on preadolescent adjustment.

We used multiple criteria to examine how well the data fit the model. First, we assessed the chi-square index. A non-significant chi-square reflects a good fit between the observed covariances and the covariances implied by a fitted model. As Kline (2011) notes, chi-square is sensitive to sample size and may reflect statistical significance even though a model fits the data well when sample sizes are large, as in the present study. We also examined the root mean square of approximation (RMSEA; Steiger and Lind, 1980), the Tucker-Lewis index (TLI; Tucker and Lewis, 1973), and the comparative fit indices (CFI; Bentler, 1990). The RMSEA represents the magnitude of the discrepancies between the implied and actual covariances. A value of zero indicates a perfect fit and a value of 0.05 is an indication of a close fit (Browne and Cudeck, 1992; Hu and Bentler, 1999). The TLI and CFI represent comparisons of the current model to a null model, estimating the amount of total information explained by the current model. Values >0.90 and 0.95, respectively, indicate a good fit (Hu and Bentler, 1999).

For the entire sample, average sleep problem scores gradually and significantly declined from 54 months-of-age to Grade 3 (M = 1.41, SD = 1.51; M = 1.00, SD = 1.27; M = 0.87, SD = 1.25, respectively; p < 0.001), but tended to remain relatively stable between Grades 3 and 5 (M = 0.81, SD = 1.24 for Grade 5; p < 0.10). Mothers rated nightmares as the most frequent sleep problem at each time point. At 54 months and Grade 1, mothers considered enuresis and sleeping less than other children as the next most prevalent sleep problems. At Grades 3 and 5, talking and walking in sleep, sleeping less than other children, and overtiredness ranked highest after nightmares. Mean scores for children's anxious-depressed symptoms increased from 54 months-of-age to Grade 1 (M = 52.08, SD = 4.26, M = 52.77, SD = 4.84, respectively, p < 0.001), and social competence declined from Grade 1 to Grade 3 (M = 15.34, SD = 0.3.62; M = 14.91, SD = 3.87, respectively; p < 0.01).

Table 2 presents a comparison of descriptive statistics for boys and for girls for each variable. With the exception of sleep problems at 54 months, there were no sex differences on sleep. As expected for a community sample, a considerable number of mothers reported no sleep problems for their child at each assessment. Thus, the mean values for sleep problems reflected a “floor effect” that might imply that sleep problems were not a problem for most children. To address this potential interpretation, we categorized sleep problems into two groups with “0” = mothers rated their child as having no sleep problems, or “1” = mothers rated their child as having one or more sleep problems “sometimes or often,” we found that at 54 months, mothers reported that 71% of boys and 61% of girls experienced one or more sleep problems sometimes or often times. In Grade 1, this percentage declined to 54% for boys and 51% for girls. By third grade, the percentage dropped to 49% for boys and 45% for girls. By fifth grade, 41% of boys and 43% of girls continued to experience one or more sleep problems sometimes or often. Thus, almost 50% of boys and girls experienced one or more sleep problems at some point across middle childhood.

There were no differences between boys' and girls' anxious-depressed symptoms at any assessment. Teachers consistently rated girls' as more socially competent compared to boys at all assessments. Boys' exhibited more risk-taking behaviors and felt less competent in school compared to girls. Mothers reported more emotional reactivity in girls compared to boys in fifth grade. No differences surfaced between boys' and girls' self-reported depressive symptoms, or emotional reactivity in the classroom in fifth grade. We found no significant differences for control variables between boys and girls with the exception of marital quality. Boys' mothers reported less satisfaction with the quality of their martial relationships compared to girls' mothers.

Tables 3, 4 present bivariate correlations for each measure for boys and girls, respectively. For both boys and girls, all repeated measures were significantly and moderately to highly correlated over age. Also, concurrent associations between model variables indicated that more sleep problems were associated with higher rates of anxious-depressed symptoms and less effective interactions with peers at all assessments. Girls' sleep problems and anxious-depressed symptoms were significantly related to preadolescent adjustment more often than boys. Boys' social competence was significantly related to preadolescent adjustment more often than girls.

All models provided good overall fit to the data as shown in Table 5. Table 6 presents the direct, overall indirect, and total effects for our final model for the entire sample (Model 5) and for boys and for girls (Model 7) with covariate effects. Because effects extended across four waves of data, we employed the joint significance test (MacKinnon et al., 2002; Haller et al., 2010) to assess the significance of each mediational chain across middle childhood. With this method, a mediational chain is deemed significant if each path in the chain is statistically significant. Our final model reflected only significant structural pathways thus suggesting significance for our overall indirect effects. Figures 2, 3 depict the significant pathways for relations among domains for boys and for girls, respectively, after controlling for covariates and moderating effects. Standardized regression coefficients are shown for each significant pathway with unstandardized regression coefficients shown in parentheses. According to Cohen (1977), cutoff values for small, medium, and large standardized path coefficients are 0.10, 0.30, and 0.50, respectively. When path coefficients were significant both for boys and for girls, tests of critical ratios revealed no significant differences in estimates by gender.

Results demonstrated several consistent findings for both boys and girls. First, all autoregressive effects were significant indicating continuity over time. Second, cross-lagged results revealed a pathway in which more preschool sleep problems predicted higher rates of anxious-depressed symptoms during middle childhood which in turn predicted more maternal-reported preadolescent emotional reactivity. The direct and indirect effects of anxious-depressed symptoms across middle childhood on preadolescent emotional reactivity were stronger than the effects of preschool sleep. Last, more preschool sleep problems directly predicted higher rates of maternal-reported preadolescent risk-taking 6 years later. For control variables, mothers' reported depressive symptoms were associated with anxious-depressed symptoms, social competence, and several adjustment variables for boys and girls.

For boys, higher rates of preschool anxious-depressed symptoms directly predicted more sleep problems 6 years later in preadolescence. Social competence across middle childhood predicted all preadolescent adjustment domains with the exception of maternal-reported emotional reactivity. Lower levels of teacher-reported preschool social competence directly predicted more preadolescent maternal-reported anxious-depressed symptoms, lower self-reported school competence, and more maternal-reported risk-taking. Social competence across middle childhood was negatively associated with teacher-reported emotional reactivity and self-reported risk-taking. The effects of preschool sleep problems on first grade social competence just missed attainment of statistical significance (p = 0.051). Boys' negative affective temperament was positively associated with sleep problems and anxious-depressed symptoms. We found evidence for moderating effects between sleep and anxious-depressed symptoms. The strength of the effect of preschool sleep problems on first grade anxious-depressed symptoms increased as boys exhibited increasingly higher levels of negative affective temperament.

For girls, cross-lagged results revealed several pathways involving sleep problems predicting later functioning. First, more preschool sleep problems predicted lower teacher-reported first grade social competence which in turn predicted more maternal-reported preadolescent anxious-depressed symptoms. The indirect effect of preschool sleep on preadolescent anxious-depressed symptoms was small in magnitude. Second, results also revealed a pathway in which more preschool sleep problems predicted more maternal reported anxious-depressed symptoms across middle childhood which in turn predicted higher rates of self-reported depressive symptoms and maternal-reported risk-taking. Direct and indirect effects of anxious-depressed symptoms on depressive symptoms and risk-taking were stronger in magnitude than the indirect effects of preschool sleep. Third, more sleep problems across middle childhood predicted lower self-reported school competence and higher rates of teacher-reported emotional reactivity in preadolescence. Last, social competence in middle childhood negatively predicted two self-reported preadolescent adjustment domains—school competence and risk-taking.

Reliance on maternal report rather than more objectives measures of polysomnography or actigraphy represented a considerable limitation in this study and introduced the potential for shared reporter variance to influence some of our findings; however, our analyses controlled for mothers' report of their own depressive symptoms and children's temperament, two factors found to predict children's sleep problems and anxious-depressed symptoms. The use of child and teacher reports as outcome measures and teachers' reports for measures of social competence also help extend these findings beyond the limits of maternal report (Podsakoff et al., 2003). Nevertheless, our findings using maternal reports are similar to those of Kelly and El-Sheikh (2013) using more objective measures of sleep. Thus, it may be reasonable to consider the importance of these findings despite the potential for shared rater bias. Our large sample (n = 1,057) enabled us to conduct complex and rigorous analyses using four waves of data with a developmental cascade model, but cost and time factors limited the collection of more objective assessments of sleep with this large sample (Becker et al., 2015).

These findings suggest that elevated levels of sleep problems and deficits in affective functioning as children begin formal schooling may have implications for children's school success and optimal development. For girls, early and middle childhood sleep problems influenced adjustment at the end of middle childhood, more so than boys, suggesting there may be developmental differences in relations between sleep and pubertal development. Thus, parents and clinicians may want to pay closer attention to girls' sleep in relation to emotional behavior and social adjustment throughout middle childhood. For boys, because mothers reported more sleep problems in boys with higher levels of negative affectivity and also reported more depressive symptoms in themselves, researchers may want to examine these interrelations more closely for families with boys.

In their review of the efficacy of behavioral treatments for bedtime problems and night wakings in early childhood, Mindell et al. (2006) found that over 90% of behavioral interventions produced reliable and durable improvements in children's sleep problems. These findings are encouraging because they suggest that behavioral treatments, including preventive parent education, may alter the course of early maladaptive patterns that may lead to future psychosocial disorders in typically developing children.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.