Discussion

The major result of this research was that higher maternal depressive symptoms were found to predict an increased duration of child awakenings. An increased duration of child awakenings was not found to predict maternal depressive symptoms, and consistent findings were not identified in either direction for frequency of child awakenings. These results suggested that child awakenings were unlikely to contribute to maternal depressive symptoms but that maternal depressive symptoms could contribute to the duration of child awakenings across time. Previous research had not been able to fully test these hypotheses due to a lack of longitudinal data over multiple time points.

The effects of maternal depressive symptoms on children's sleep were thus different for duration of awakenings and frequency of awakenings. Maternal depressive symptoms when the child was 1 month of age significantly predicted a shift in the trajectory of duration of child awakenings over time such that the children displayed an overall worsening in sleep across the 3-year period, compared to children of mothers with fewer depressive symptoms. Thus, the effect of early maternal depressive symptoms on children's sleep appeared to be long lasting. In terms of the magnitude of the effect, the statistical model predicted that children of higher depressive symptom mothers (95th percentile) would be awake more than twice as long at 3 years of age than children of lower depressive symptom mothers (5th percentile). When examined at specific periods using cross-lag models, higher levels of maternal depressive symptoms at 15 and 24 months predicted greater duration of child awakenings at 24 and 36 months, respectively. In contrast, for frequency of awakenings, maternal depressive symptoms predicted an increased frequency of child awakenings only for the 15- to 24-month epoch. The two sleep variables thus demonstrated divergent results, highlighting their differences.

Although both frequency and duration of awakenings have been used to measure child sleep, these two symptoms have been considered to be influenced by differing processes. Frequency of awakenings has been thought to be most influenced by difficulties with sleep-maintaining processes, and/or factors that promote sleep-offset processes. In contrast, duration of awakenings has been thought to be most influenced by difficulties with (reinitiating) sleep-onset processes and/or factors that promote wake-maintaining processes. When characterizing sleep in young children, individual attention has been given to each of these processes (Gaylor et al., 2001). In addition, frequency and duration of awakenings have demonstrated divergent trajectories over the course of development (Klerman, Davis, Duffy, Dijk, & Kronauer, 2004).

The findings concerning duration of awakenings were strongest across the 15- to 24- and 24- to 36-month age epochs. These age periods have been described as times of increasing autonomy, self-awareness, and self-regulation for the child (DeHart et al., 2000; Lewis & Volkmar, 1990; Sroufe & Rutter, 1984), and could represent an important time of sleep consolidation (Moore & Ucko, 1957; Sadeh & Anders, 1993). Anders and colleagues have studied the self-regulation of internal states of arousal and falling asleep across early development, referring to an increasing ability to self-regulate during the nighttime hours as “self-soothing” (Anders & Keener, 1985; Goodlin-Jones, Burnham, Gaylor, & Anders, 2001). These researchers have described the process by which the child learns to self-sooth as involving a shift over the course of development from externally imposed calming and activating strategies to self-control strategies resulting, in part, from social interactions with the mother (Goodlin-Jones et al., 2001). Toddlers with mothers who had more depressive symptoms might have shown longer durations of awakenings because of diminished capacity to self-soothe or return to sleep. The more troubled sleep could be indicative of more general difficulties in terms of regulatory capacities in toddlers of parents with more depressive symptoms.

The longer duration of awakenings for children of mothers with more depressive symptoms could also be clinically meaningful. Although criteria for sleep disorders in young children have not been definitively established, a recent paper by Anders and Dahl (in press) has proposed that awakenings of 10 min or more for 3-year-old children would signify a sleep perturbation if the episodes occurred frequently enough. In the data reported here, the expected duration of awakenings for the 3-year-old children of highly depressed mothers would be more than 10 min, this would suggest that children of mothers with more depressive symptoms could be more likely to meet the criteria for sleep disorders.

Furthermore, the findings that maternal depressive symptoms at 1 month predicted a shift in the trajectory of child sleep and that maternal depressive symptoms at 15 and 24 months predicted more troubled sleep at 24 and 36 months, respectively, could be important for the development of interventions. Identifying and treating mothers with depressive symptoms could influence the children's sleep over long periods of time. Moreover, rather than treating children's sleep to reduce maternal depressive symptoms, as some previous studies have done, these findings suggest that treating maternal depressive symptoms directly might be beneficial both for reducing the maternal depressive symptoms and for improving the children's sleep. Certainly, the precise interventions depend, however, on the specific mechanisms involved in the associations between maternal depressive symptoms and children's sleep.

The prediction of duration of child awakenings from maternal depressive symptoms could result from a variety of influences. Children who had mothers with more depressive symptoms could have more difficulty with basic biological regulation of sleep and arousal, as such children have been shown to display a number of biological differences compared to children of mothers with less depressed symptoms, including lower vagal tone, right frontal EEG asymmetry, and higher norepinephrine levels (Field, 1995). Children with more depressed mothers could also have more difficulty with the regulation of affect, which could impact their ability to self-soothe. Dahl (1996) has described the close association between emotional and sleep regulation through which unregulated emotions interfere with sleep. Children with depressed mothers have been described as showing more irritability and difficulty self-soothing (Field, 1995; Zuckerman, Bauchner, Parker, & Cabral, 1990), which could contribute to increased problems settling to sleep.

Children who had mothers with more depressed symptoms might also have experienced less consistent maternal support of self-soothing strategies. Research concerning children with depressed mothers has described how depressed mothers behave in a more withdrawn or intrusive fashion with their children (Field, 1995; Weinberg & Tronick, 1998a). Such behaviors could interfere with the external calming strategies and thus not facilitate the development of internal self-soothing strategies. Goodlin-Jones and Anders (2001) have described a transactional sleep model that suggests that sleep and waking state organization in the first 3 years of life depend largely on the relationship context in which it develops. They have explained how transitions between sleep and waking occur many times each night, providing regular opportunities for homeostatic (hunger, temperature, dryness) and social/affective (separation, reunion, comfort, security) regulation. When the bedtime and nighttime interactions are consistent, predictable, and supportive, the mother and infant form a relationship that fosters biological and social regulation. Failure to develop consistent, predictable and supportive patterns of interaction surrounding sleep could lead to subsequent sleep–wake dysregulation. Dahl has also described how feelings of safety are essential for promoting sleep (Dahl, 1996). Erratic or nonsupportive parental behavior could also contribute to less feelings of safety and thus interfere with settling to sleep.

Depressed mothers, when compared with nondepressed mothers, have also been described as more inconsistent and ineffective in child management and discipline, trying to avoid conflict by accommodating to the child's demands (Cicchetti & Toth, 1995; Zahn-Waxler et al., 1990). Such behaviors could also be evident during interactions concerning sleep. Anders and colleagues have reported that self-soothers are more likely to have been put to bed awake than nonself-soothers (Goodlin-Jones et al., 2001). Depressed mothers could be less likely to put their children to bed awake, perhaps in an attempt to avoid conflict. When children awaken during the night, depressed parents could also have more difficulty implementing common sleep procedures such as controlled crying or systematic ignoring (Gelman & King, 2001; Green, 1990; Rickert & Johnson, 1988), thus contributing to sleep difficulties.

Research has also suggested that depressed mothers tend to reside in adverse circumstances, often experiencing more marital conflict and less support (Cicchetti & Toth, 1995; Downey & Coyne, 1990). Seifer et al. (1996) have reported that better overall family functioning measured by interviews and direct observations is related to fewer bedtime problems. Thus, the impact of maternal depressive symptoms on child sleep in terms of behavioral management of the sleep could be influenced by the social context. It is thus possible that the maternal depression could be a proxy or correlate for other factors associated with sleep problems in children such as temperament differences, infant health problems or family stress (Sadeh & Anders, 1993).

The finding that increased duration of child awakenings predicted decreased maternal depressive symptoms across the 15- to 24-month period, although contrary to the hypotheses, was not completely inexplicable. Sleep deprivation has been used as a treatment for major depression (Giedke & Schwarzler, 2002). Although much of the literature concerning maternal depressive symptoms and child sleep has focused on the negative aspects of awakenings for mothers, it is conceivable that the longer awakenings could have caused the mother with depressive symptoms to experience significant sleep deprivation. Such significant sleep deprivation could then have provided some treatment for the mother's depressive symptoms. Because the study did not include measurement of maternal sleep, such speculations could not be tested. Moreover, because this result was found in only one analysis, it could have been a chance finding.

Limitations of the research included the fact that the measures were not optimal. It would have been better to have conducted more comprehensive diagnostic assessments of the mothers and to have used more questions, sleep logs, actigraphy, videotaping, or EEG to measure sleep in the children. The weaker measures could have contributed to weaker findings. Although maternal depressive symptoms predicted the trajectory of duration of child awakenings at a p < .01 level, the prediction was only at the p < .05 level for the phase-specific findings. The smaller effect sizes might have resulted from the lack of optimal measures and thus could represent lower bound estimates. Another issue that could account for the smaller effect sizes is the time-lag period that was used. If a shorter time period had been used for the phase-specific models (e.g., 2–3 months), the effect sizes could have been stronger. Further research is needed concerning the optimal time periods for such investigations.

An additional limitation was the fact that mothers reported on both their own depressive symptoms and on the child sleep, so the reports of sleep could have been biased. There is some evidence in the data that such a bias did not account for the findings. If mothers with higher levels of depressive symptoms tended to report longer duration of child awakenings, maternal depressive symptoms and duration of child awakenings would show strong relations concurrently and across time, whether maternal depressive symptoms were being used to predict duration of child awakenings or visa versa. However, the results of this paper were that maternal depressive symptoms predicted child duration of awakenings across time but that child duration of awakenings did not predict maternal depressive symptoms across time. Moreover, the concurrent relations between maternal depressive symptoms and child sleep were not strong or consistent (see Table 2). Although these results suggested that bias was less likely in these analyses, bias could still be present, representing an important limitation.

There were additional limitations concerning the sample. Although every attempt was made to recruit a representative sample, the participants differed from the original sample (e.g., less likely to be Black or African American, more likely to have completed college, more likely to have a higher average income), and thus the results might not generalize to the entire population. The sample also did not show high levels of depressive symptoms and the children were not recruited based on having sleep problems; including higher risk families, mothers with diagnosed depressive symptoms, and children with diagnosed sleep problems could further clarify the predictive pathways.

Furthermore, the statistical models utilized were complex. Even though the samples were large, there could have been some instability in the estimates. For example, the standard error for the correlation between initial duration of child sleep and the duration of child sleep trajectory was large. An additional limitation of the causal modeling was that it could not identify third variable effects. For example, an unmeasured variable such as environmental noise or irregularity of day–night schedules could have accounted for the result that higher maternal depressive symptoms were found to predict longer duration of child awakenings.

Previous research has shown that children who show sleep problems in early childhood continue to show sleep problems later (Gregory & O'Connor, 2002; Zuckerman et al., 1987). Sleep difficulties could also shape affect, attention, cognitive and language development (Dahl, 1996; Dearing et al., 2001) and have been found to forecast behavior problems (Gregory & O'Connor, 2002). Thus, the effects of maternal depressive symptoms on child sleep could be significant. Moreover, the fact that early maternal depressive symptoms predicted child sleep problems over a long period of time suggests that addressing the predictive mechanisms could have an important and long-lasting impact on children's sleep. Further research is needed to determine the precise mechanisms by which maternal depressive symptoms predicted increased duration of child awakenings. Such information could clarify the best approaches for intervention strategies.