Depressive Symptoms and Developmental Change in Mothers’ Emotion Scaffolding: Links to Children’s Self-Regulation and Academic Readiness

Abstract

This study examined whether the mechanism linking changes in Mexican-origin adolescent mothers’ depressive symptoms to children’s subsequent self-regulation and academic readiness was via their emotion scaffolding when their children were 2, 3, 4, and 5 years of age. Data included home interviews with adolescent mothers (N = 204), observations of mother-child interactions during a task that challenged children’s abilities to manage their emotional arousal, and assessments of children’s performance on measures of self-regulation and academic readiness. Adolescent mothers’ higher depressive symptoms at child age 2 years were associated with a greater decline in mothers’ emotion scaffolding from 2 to 5 years of age, which was subsequently linked to children’s lower self-regulation and academic readiness at age 5. Possible implications from this line of work for adolescent mothers and their children are discussed.

Children of adolescent mothers are at risk for delays in social-emotional and cognitive development (e.g., Borkowski et al., 2004), and mothers’ psychosocial functioning and mother-child interactions have been implicated as key factors that may inform such outcomes (e.g., Contreras, Narang, Ikhlas, & Teichman, 2002). The present study examined whether one mechanism linking adolescent mothers’ psychosocial functioning to children’s developmental outcomes was via mothers’ emotion scaffolding, given past work linking these maternal behaviors to social-emotional and academic outcomes (e.g., Hoffman, Crnic, & Baker, 2006). We studied these processes in Mexican-origin adolescent mothers, the ethnic group with the highest teen pregnancy rates in the U.S (National Vital Statistics Report, 2013). Given the complexity of maternal, child, and contextual factors that may contribute to adolescent mothers’ development, we examined whether there was developmental change in adolescent mothers’ emotion scaffolding across their children’s early development, and whether mothers’ depressive symptoms were linked to such change and, in turn, to children’s outcomes through changes in emotion scaffolding. We examined two key outcomes when children were 5 years of age, self-regulation and academic readiness, as these competences are linked to children’s success in elementary learning environments (e.g., Blair & Raver, 2015).

Emotion Scaffolding Behaviors in Adolescent Mothers

Parents’ scaffolding behaviors with their young children can generally be defined as the provision of support to assist children to achieve a subsequent level of functioning, and the fading of support as children’s competence grows (e.g., Hoffman et al., 2006). This type of support is believed to facilitate development by pushing children beyond what they can do on their own (Vygotsky, 1978). Emotion scaffolding, in particular, captures the provision of emotional support to assist children when circumstances challenge their ability to manage their emotional arousal. Effective emotion scaffolding can be distinguished from broader indices of positive parenting like parental sensitivity because emotion scaffolding not only includes sensitive awareness of children’s emotional arousal, but also parents’ abilities to reduce frustration and promote children’s sense of efficacy (Baker, Fenning, Crnic, Baker, & Blacher, 2007; Hoffman et al., 2006; Maslin-Cole & Spieker, 1990). Given the unique contributions of parents’ behaviors in the context of distress to children’s positive outcomes (Leerkes, Blankson, O’Brien, 2009), there is a need to understand how such behaviors develop in vulnerable mother-child dyads, including adolescent mothers and their children.

Although no study, to our knowledge, has explicitly observed emotion scaffolding in adolescent mothers, past work on broader dimensions of positive parenting have shown that, on average, adolescent mothers are at risk for displaying less optimal parenting (e.g., less sensitivity, responsiveness, positive affect and regard, praise, acceptance, and cognitive and behavioral support) in comparison to adult mothers (Berlin, Brady-Smith, & Brooks-Gunn, 2002; Culp, Osofsky, & O’Brien, 1996; East & Felice, 1996; Pomerleau, Scuccimarri, & Malcuit, 2003; Whitman, Borkowski, Keog, & Weed, 2001). Scholars argue that examining developmental change in parent behaviors is critical to our understanding of the parent-child relationship and its contribution to children’s development, offering information about how parents’ behaviors change over time, and potentially malleable targets and developmental periods for intervention (Ciciolla, Crnic, & West, 2013). To this end, we sought to fill a gap in previous research by examining developmental change in adolescent mothers’ emotion scaffolding in the context of a challenging teaching task during a critical period of children’s development from 2 to 5 years of age.

Cultural perspective.

In considering how Mexican-origin adolescent mothers’ emotion scaffolding behaviors may change across the early years of parenting, it is useful to consider both a cultural and developmental lens. Previous work suggests that non-cognitive attributes of intelligence (i.e., problem solving and interpersonal social skills; Okagaki & Sternberg, 1993) rank high among Latino parents’ orientations toward children’s development. Studies have revealed more physical guidance and emotional comfort and support among Latina mothers, relative to European American mothers in early childhood (e.g., Halgunseth, Ispa, & Rudy, 2006). Latino parents have been shown to emphasize dyadic co-regulation and support during early development, when children are viewed as more cognitively immature (Mosier & Rogoff, 2003) and to shift toward increased expectations of children’s self-regulation and compliance to rules at around six years of age (Halgunseth et al., 2006). Moreover, an understanding of mechanisms that facilitate self-regulation development in children of Mexican-origin adolescent mothers is especially warranted because regulation may mediate the link between the contextual risks faced by some minoritized populations (e.g., poverty) and children’s subsequent well-being (e.g., Li-Grining, 2012).

No study to our knowledge has examined emotion scaffolding over time among Latino parents. With respect to broader parenting behaviors, low-income Mexican American mothers have been shown to display steeper declines in directiveness behaviors compared to European American and African American mothers, and an increase in such behaviors just prior to kindergarten (Ispa et al., 2013). Although such work focused on other dimensions of parenting behaviors in parent-child contexts that may not have been marked with emotional arousal, they highlight potential developmental change and variability, and underscore the need for studying patterns of change in maternal behavior using within-ethnic-group designs.

Developmental perspective.

From a developmental lens, the limited work on how adult mothers’ behaviors change with their young children in teaching tasks suggest that broader maternal behaviors like sensitivity (Ciciolla et al., 2013) and cognitive assistance and questioning (Eisenberg et al., 2010) increase, possibly in response to children’s burgeoning emotion regulation capacities and more active role as communicative partners (Ciciolla et al., 2013). In contrast, we posit that adolescent mothers may fail to show increases in their scaffolding behaviors, and may even show a decrease in these behaviors across their children’s early development, due to developmental constraints on adolescent mothers’ emotion and self-regulation capacities (Azar, 2002) and limitations in their knowledge of parenting and children’s development (Bornstein et al., 2010; East & Felice, 1996). Given the normative tasks of adolescence (e.g., their own developing emotion regulation and problem-solving capacities; Steinberg, 2010), adolescent mothers may not yet have the foundational skills to be adequately emotionally aware and supportive of their children. Research on physiological changes during puberty support the notion that adolescence is marked by heightened emotional intensity and dysregulation (Dahl, 2004), and research on brain development suggests that maturity in emotional and self-regulatory capacities is ongoing during adolescence (Sebastian, Viding, Williams, & Blakemore, 2010; Steinberg; 2010), all of which are important to mothers’ emotion scaffolding with their children. Considering these developmental constraints in the context of Azar’s (2002) framework, which applies a social-cognitive perspective to changes that occur during parenthood, supports our hypothesis. Specifically, “learning” with respect to parenting behaviors is proposed to occur in the context of the parent-child social relationship; adolescent mothers’ suboptimal social-emotional and foundational capacities may interfere with the process of being flexible enough to gain effective parenting skills (Azar, 2002). Further, some have argued that, because of their developmental stage, adolescents have a tendency toward being idealistic, less self-reflective, and more egocentric, which may limit their understanding of their children’s needs (East & Felice, 1996). Adolescent mothers also may have less knowledge about children’s development than adult mothers (e.g., Bornstein et al., 2010), which may impede their use of behaviors that support their children’s learning. Thus, finding a decline in emotion scaffolding behaviors among adolescent mothers could support the notion that adolescents’ relatively immature developmental state and the other burdens and demands of adolescence undermine increases in their parenting behaviors.

The Role of Depressive Symptoms for Mothers’ Emotion Scaffolding

Theoretical accounts of parenting consider parents’ psychosocial characteristics and resources as important predictors of their parenting competence and children’s development (Belsky, 1984). Applying Belsky’s (1984) parenting model, Contreras and colleagues (2002) highlight depressive symptoms as a salient index of young Latina mothers’ psychological functioning. Indeed, adolescent mothers show heightened risk for depressive symptoms and simultaneous social-contextual challenges, compared to adult mothers and non-parenting adolescents (e.g., Lanzi, Bert, & Jacobs, 2009; Mollborn & Morningstar, 2009). Thus, adolescents’ depressive symptoms may be related to changes in their parenting behaviors. Among adult parents, depressive symptoms were positively associated with parents’ impatient use of directives in guiding children’s behaviors (Forehand, Lautenschlager, Faust & Graziano, 1986), and mothers’ use of three different forms of scaffolding (i.e., emotional, motivational, and technical) were lower for those with higher depressive symptoms when measured concurrently at child age 3 years (Hoffman et al, 2006). Interestingly, both depressive symptoms and scaffolding in the latter study were linked to children’s subsequent dysregulation at age 4. Although the study was limited to only two age points, it is an important first step toward understanding the mechanism by which mothers’ depressive symptoms may be related to children’s regulation skills.

Less work has focused specifically on the link between depressive symptoms and parenting behaviors in Latina adolescent mothers, but one study with a diverse sample (i.e., 43% Puerto Rican) found that depressive symptoms were related to fewer contingent responses by adolescent mothers during mother-child free play interactions (Leadbeater, Bishop, & Raver, 1996). It has been proposed that, given the reliance on family interdependence among Latino families, mothers’ depressive symptoms could have a particularly negative impact on their children’s outcomes (Valdez, Abegglen, & Hauser, 2013). In line with Hoffman et al.’s (2006) work with adult mothers, in the present study we examined indirect effects of depressive symptoms on children’s subsequent developmental outcomes through changes in emotion scaffolding behavior. We expected depressive symptoms to be negatively related to initial levels of emotion scaffolding, in line with past work on these relations (Hoffman et al., 2006). Furthermore, it may be that as the stressors and demands associated with adolescent motherhood accumulate, the negative impact of depressive symptoms is compounded, resulting in a more substantial decline in emotional scaffolding for mothers with more depressive symptoms.

Children’s Self-Regulation and Academic Readiness

The present study focused on children’s academic readiness and self-regulation at age 5 (i.e., a time when children are entering elementary school) given findings of poorer school readiness and social-emotional skills among children of adolescent mothers (Borkowski et al., 2004; Tang et al., 2016; Terry-Humen, Manlove, & Moore, 2005), and disparities among Latino children, including those of Mexican-origin, in their early academic performance relative to their White and African American peers (Crosnoe, 2007; Guerrero et al., 2013; Padilla, Cabrera, & West, 2017). Moreover, many children of adolescent mothers scored at borderline or mild levels of intellectual disability at age 3 (45%) and age 5 (26%), and nearly 40% of them met criteria for a learning disability or mild intellectual disability (Borkowski et al., 2002); furthermore, children of Mexican-origin adolescent mothers are at risk for developmental delays in early childhood (Jahromi, Umaña-Taylor, Updegraff, & Zeiders, 2016). In the present study, academic readiness reflected children’s pre-academic achievement (Woodcock, McGrew, & Mather, 2001) and self-regulation captured children’s effortful control (i.e., the ability to voluntarily control or modulate one’s attention or behavior; Rothbart & Bates, 2006) and executive function (i.e., mental forms of regulation; e.g., Zelazo, 2006). Together, these indices of children’s development have significant implications for subsequent social-emotional and cognitive functioning (e.g., Blair & Razza, 2007; Valiente et al., 2008) as they set the stage for children’s positive transition to, and further success in, the school setting (Raver, 2002).

Parents’ emotion scaffolding behaviors are believed to be critical to the development of children’s social-emotional and academic competencies (Calkins, 1994). Mothers’ engagement in emotion scaffolding behaviors during dyadic interactions are thought to support children in modulating their immediate emotional and behavioral responses, and to facilitate the further development of self-regulatory skills (Egeland, Pianta, & O’Brien, 1993). Cole, Michel, and Teti (1994) suggest that children whose parents provide them emotional support that is matched closely to their specific needs are more likely to internalize effective self-regulatory skills for the context of challenge. Moreover, scaffolding has been identified as one of the most important parental behaviors linked to individual differences in children’s executive functioning, a cognitive form of self-regulation linked closely to academic competence.

Specifically, in studies in which maternal scaffolding behaviors were observed during parent-child problem-solving tasks, mothers who engaged in more scaffolding behaviors had children with higher levels of executive function (Fay-Stammbach, Hawes, & Meredith, 2014). Thus, in line with prior findings and with the proposition that parents who engage in emotion scaffolding may model adaptive behaviors to their children, we expected that adolescent mothers with greater depressive symptoms would engage in fewer emotion scaffolding behaviors and have children who demonstrate poorer self-regulation (Halberstadt, Crisp, & Eaton, 1999). We also expected that poorer scaffolding behaviors would undermine children’s ability to engage in the academic behaviors associated with learning and problem solving (Fabes, Martin, Hanish, Anders, & Madden-Derdich, 2003). As part of these analyses, we conducted sensitivity analyses to contrast emotion scaffolding with other forms of scaffolding (i.e., motivational and technical) in the hypothesized links. Motivational scaffolding reflected parents’ ability to maintain the child’s attention and direction toward the end goal, while technical scaffolding included parents’ demonstrations and simplifications of the task (Maslin-Cole and Spieker, 1990). One might expect that technical and motivational scaffolding reflect external sources of support that are more task-specific, whereas emotion scaffolding (i.e., promoting children’s frustration regulation and sense of agency) could more readily be generalized by the child to other settings and other tasks, and thus carry over to children’s performance on self-regulation and academic measures. Interestingly, Hoffman et al. (2006) found that emotion scaffolding was the most consistent form of scaffolding across the multiple parent-child tasks used in their study, whereas technical and motivational scaffolding showed modest consistency across tasks, suggesting they may be more task specific. This contrast was addressed in sensitivity analyses.

Current Study

The current study examined emotion scaffolding among Mexican-origin adolescent mothers when their children were 2, 3, 4, and 5 years of age. We hypothesized that adolescent mothers’ emotion scaffolding behaviors may show no change or even decline over time, and that mothers’ higher depressive symptoms would be linked to greater declines in emotion scaffolding behaviors and, in turn, to children’s self-regulation and academic readiness. We explored variability by children’s gender, given that preschool-aged girls typically show higher self-regulation than preschool-aged boys (e.g., Kochanska & Knaack, 2003) and that parents discuss and explain emotions more frequently with girls (Putnam, Sanson, & Rothbart, 2002). In addition, we tested whether paths differed by adolescent mothers’ nativity given prior work suggesting that highly acculturated Latina adolescent mothers relied more heavily on verbal and nondirective teaching strategies than less acculturated adolescents (Teichman & Contreras-Grau, 2006) and that exposure to more formal sources of parenting information was greater among more acculturated parents (Bornstein et al., 2010).

We controlled for adolescents’ social competence, as this developmental asset may be related to mothers’ behaviors that promote children’s social functioning (e.g., Prinstein & La Greca, 1999). We also controlled for the availability of contextual supports and resources, namely family income, adolescents’ perceived social support, and adolescent-mother co-residence, as these factors have been directly linked to adolescent mothers’ parenting (East & Felice, 1996). We account for mothers’ school status, as this factor may be related to material and social resources (e.g., McDermott, 2018). Finally, to address the potential that child effects were linked to mothers’ scaffolding behaviors, we included measures of children’s birthweight, because low birthweight status has been shown to have a far-reaching effect on children’s developmental and educational outcomes (Resnick et al., 1999). We also control for children’s negative temperament in early development to account for the fact that children’s negative affect may be related to less optimal parenting (Lengua & Kovacs, 2005).

Method

Participants and Procedure

Data were from a longitudinal study that included 204 Mexican-origin adolescent mothers, their mother figures (e.g., biological mother, aunt, boyfriend’s mother), and a target child (Umaña-Taylor, Updegraff, Jahromi, & Zeiders, 2015). Data were collected when mothers were in their third trimester of pregnancy (Wave 1; W1), and when children were 10 months (W2), 2 years (W3), 3 years (W4), 4 years (W5), and 5 years (W6). Adolescent mothers’ emotion scaffolding was assessed in Waves 3, 4, 5, and 6. Adolescent mothers averaged 16.80 years of age (SD = 1.00) at W1 and 18.93 years of age (SD = 1.00) at W3, and 64% were U.S. born.

The majority of families participated across all six waves (i.e., 96.10% at W2, 87.75% at W3 and W4, 88.24% at W5, and 87.75% at W6). The percentage of missing data per variable ranged from 0% to 34%. Independent samples t-tests were conducted to test for potential mean level differences in control variables (i.e., adolescent mothers’ age, school status, and perceived social support; adolescent mother-mother figure coresidency; household size; and family income) between participants with complete (i.e., 172 individuals) versus incomplete data (i.e., 32 individuals). Results indicated that there were no significant differences on these variables at W1.

Participants were recruited from community agencies and high schools in a Southwestern metropolitan area. Eligibility criteria included that adolescents had to be of Mexican origin (i.e., self-identify with Mexican heritage), 15 to 18 years of age, currently pregnant, and not legally married. Interviews were conducted in participants’ preferred language, and most adolescent mothers completed interviews in English (range 61% – 69% from W1 to W6). Participants’ incentives consisted of monetary reimbursements, which increased from $35 for participation at W3, to $40 at W4, $50 at W5, and $60 at W6. Additionally, to support participant retention, we maintained close contact with families through birthday cards, newsletters twice per year, and tracking of contact information. All procedures were approved by the Institutional Review Board at Arizona State University for the Supporting Mexican-origin Adolescent Mothers and Infants project under approved protocol #1004005032.

Measures

Controls and moderators.

We controlled for several variables on the slope and intercept of adolescent mothers’ emotion scaffolding. Adolescent mother – mother figure coresidency was reported by mothers at W1, and coded as 1 = Live Together, 2 = Do Not Live Together (87% of adolescent mothers co-resided with their own mothers). Adolescent mothers reported their school status, coded as 0 = Dropped out of school and 1 = Attending school or Not attending because graduated or obtained a General Education Development (GED) diploma.

Adolescent mothers’ social competence was assessed using the 8-item subscale of the Developmental Assets Profile (DAP; Search Institute, 2004; e.g., “I plan ahead and make good choices”), which captures adolescents’ planning, decision making, cultural competencies, and social skills (α = .62). Adolescent mothers’ perceived social support was assessed with the Multidimensional Scale of Perceived Social Support (Canty-Mitchell & Zimet, 2000). The 12-item measure assesses perceived overall support from family, friends, and/or a significant other (e.g., “My family really tries to help me” and “I can talk about my problems with my friends”) on a scale from (1) Strongly disagree to (7) Strongly agree (α = .90). To assess family income, grandmothers’ (i.e., adolescent mothers’ own mothers) reports of family income at W1 were used, which was calculated by creating a sum of grandmothers’ income, additional funds contributed to the household by others (e.g., mothers), and public financial assistance (i.e., public assistance, food stamps). For analytic purposes, the original family income variable was rescaled for each participant by dividing each value by 1000. Higher scores indicated higher family income. In addition, mothers’ nativity (i.e., 0 = Mexico-born, 1 = U.S.-born) and children’s gender (1 = Female, 2 = Male) were tested as moderating variables. Adolescent mothers reported on their infants’ birthweight during a phone interview conducted approximately two weeks after birth. To assess temperamental negativity, the very short version of the Early Childhood Behavior Questionnaire was used (Putnam, Garstein, & Rothbart, 2006). This measure (α = .66) contains 12 items (e.g., “When told that it is time for bed or nap, how often did your child react with anger?”), with responses ranging from 1 (never) to 7 (always). An average score was computed across items, with higher scores indicating greater negative affectivity.

Adolescent mothers’ depressive symptoms.

At W3, the 20-item Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977) was used to assess mothers’ depressive symptoms during the past week (e.g., “My sleep was restless”). Responses ranged from (0) Rarely or none of the time to (3) Most of the time, and were coded so that higher mean scores indicated higher depressive symptoms (see Crockett, Randall, Shen, Russell, & Driscoll, 2005 for reliability and validity among Mexican-origin youth). Mean scores are reported in this study. A sum score of 16 or higher is (which corresponds to mean score of .80 or higher in our data) is widely accepted as a cutoff for risk of clinical levels of depressive symptoms (e.g., Beeghly et al., 2002). In the present study, 78 adolescent mothers (38%) met this criterion. Cronbach’s α in the current study was .90.

Adolescent mother’s scaffolding.

At W3, W4, W5, and W6, adolescent mothers’ emotion, technical, and motivational scaffolding behaviors were coded using a global rating of a 5-minute teaching task, in which adolescent mothers and their children played with a puzzle or Legos to make a model. This approach allowed us to use a similar task at each time point, while adjusting the difficulty level of the task at each age to a level that was deemed to be beyond children’s capacity to complete on their own without support. Adolescent mothers were instructed to teach their children how to complete the puzzle or build the model using the Legos. The task consisted of the following materials at each wave: a wooden puzzle of the alphabet (W3); Lego model image and large Lego pieces (W4); Lego model and small Lego pieces (W5); and a wooden puzzle that children attempted to complete without seeing the pieces, using only verbal guidance from their mother who could see the puzzle (W6).

The behaviors of mother-child dyads were videotaped for subsequent coding of adolescent mothers’ scaffolding, using the Maternal Scaffolding Coding System (Maslin-Cole & Spieker, 1990) according to the procedures of Hoffman et al. (2006). This coding paradigm captured the degree of adolescent mothers’ emotion scaffolding, which was defined as provision of the appropriate level of support for the child, aimed at making the experience positive and enjoyable for the child, and contributing to the child’s sense of accomplishment and efficacy. We utilized a 3-point version of the 5-point scale used in prior work (Hoffman et al., 2006) by combining the highest two (and lowest two) anchors on the scale because coders had difficulty distinguishing between a highest code (“Parent meets the child’s scaffolding needs almost the entire time but with a couple of noticeable missed opportunities”) and the second-highest code (“Parent meets the child’s scaffolding needs most of the time; there may be a rare instance in which the parent misses a minor opportunity for scaffolding”). The 3-point mothers’ emotion scaffolding scale reflected low (little to no acceptance/support of the child’s attempts at the task, praise for the child, contribution to child’s sense of accomplishment, sensitivity to their child’s emotional state, or mutual sharing of enjoyment; and deriving little enjoyment from the child successes), medium (moderate or partial levels of acceptance/support of the child’s attempts at a task, inconsistent praise or only conditionally adding to their sense of accomplishment, moderate sensitivity to their child’s emotional state, engaging in mutual enjoyment with responses that lack enthusiasm; and deriving some brief enjoyment from their child’s successes), or high emotion scaffolding (behaviors that were decidedly accepting and supportive of the child’s attempts at a task, frequent use of praise and strong contribution to the child’s sense of accomplishment; highly sensitive responses to the child’s emotional state; frequent shared enjoyment and demonstrations of vicarious enjoyment from the child’s successes).

In the subsequent sensitivity analyses, we contrasted emotion scaffolding with technical and motivational scaffolding. Technical scaffolding was defined as mothers’ abilities to structure the task effectively for children, including providing well-timed demonstrations, pointing out critical features, and filling in sub-steps. Motivational scaffolding was defined as mothers’ abilities to help children engage with the task and maintain attention and focus toward the end goal. Technical and motivational scaffolding were coded on the same 3-point scale (low, medium, high) as emotion scaffolding. The coding team consisted of five coders who trained until they reached an adequate level of reliability (> .75); mean ICC reliabilities for emotion, technical, and motivational scaffolding, respectively, for a total of 15% of videos coded at each wave by two independent coders were .77, .71, .70 (W3), .78, .79, .80 (W4), .69, .66, .65 (W5), and .70, .61, .71 (W6), indicating a “good” level reliability (Cicchetti, 1994).

Children’s self-regulation.

At W6, children’s self-regulation was measured using a composite of children’s performance on the gift bag task (Kochanska, Coy, & Murray, 2001) and the Dimension Change Card Sort task (DCCS; Zelazo, 2006). Once children had completed all activities for the home visit, and while the adolescent was in another room, the experimenter entered the room in which the child was seated and announced, “Guess what! This is a present for you!” A gift bag that contained a small gift along with tissue paper was placed in front of the child. The gift was not expected by the child. Tissue paper prevented the child from seeing the gift within the bag. Once the child turned his or her attention to the gift, the experimenter said, “But I want you to open your present with your mom. Wait here while I go to get her. Try not to touch it until I come back with your mom, OK?”. The experimenter left the room for three minutes. Children’s behaviors during the three-minute wait were video recorded and subsequently coded. After the three-minute wait, the experimenter returned with the child’s mother and allowed the child to open the gift bag. The child’s response was scored as follows, with higher scores reflecting better self-regulation: (1 = child pulls gift from the bag; 2 = child puts hand into the bag; 3 = child opens the bag to peek inside; 4 = child touches the bag without peeking; 5 = child neither touches nor peeks into the bag).

In the DCCS task (Zelazo, 2006) children were presented with cards that varied along the dimensions of shape and color to be sorted into piles according to explicitly-stated sets of sorting rules. Children were first asked to sort by shape (pre-switch), then by color (post-switch), and then by either shape or color depending on the presence of a border around the card (border). The percentage of correct trials within each set of trials (i.e., pre-switch, post-switch, and borders trials) was coded. Five out of six correct post-switch trials and nine out of 12 correct borders trials constituted a pass (Zelazo, 2006). An overall score was created as follows: 0 = failed the pre-switch phase; 1 = passed the pre-switch (shape) phase but failed the post-switch (color) phase; 2 = passed both the pre- and post-switch phases (shape and color) but failed the second post-switch phase (border); 3 = passed all three phases. The composite measure of self-regulation was created by standardizing and averaging the two individual indices of self-regulation (r = .05, p = .53). Conceptually, these measures are believed to tap meaningful dimensions of children’s self-regulation.

Children’s academic readiness.

At W6 (when children were 60 months old), we assessed children’s developmental functioning using the Woodcock Johnson III Tests of Achievement (WJ-III; Woodcock, McGrew, & Mather, 2001). Children whose primary language was Spanish were assessed using the Batería III Woodcock-Muñoz (Batería-III; Muñoz-Sandoval, Woodcock, McGrew, & Mather, 2005). Both tests are designed to be administered to respondents ranging in age from 2 to 90 years and have been reported to show high internal and test-retest reliability by their developers. Each test provides a standardized, normative score that enables a comparison of the respondent’s score against the national average for that respondent’s age (i.e., age-equivalence). Children were administered four subtests from the WJ-III and Bateria-III that were identified as appropriate for preschool-aged children: (a) Letter-Word Identification (Identificación de letras y palabras); (b) Applied Problems (Problemas aplicados); (c) Picture Vocabulary (Vocabulario sobre dibujos); (d) Passage Comprehension (Comprension de textos). On each test, children’s raw score (i.e., sum of correct responses) was converted to a standardized score called a W score using the WJ III Compuscore program provided by the test developers, and an average score was created across the four subtests.

Analytic Approach

Several steps were involved in the analytic approach. First, two nested models were compared to examine whether there was significant growth in adolescent mothers’ emotion scaffolding from W3 to W6. In the context of these analyses, “growth” could reflect an increase or decline in scaffolding over time. A no growth model was compared to a linear growth model, and a χ2 difference test was performed. If the χ2 difference test was significant, it suggested that there was significant growth (i.e., increase or decrease) over time in adolescent mothers’ emotion scaffolding (Grimm, Ram, & Estabrook, 2017).

Next, a linear growth model was used to test whether there were significant differences in mothers’ emotion scaffolding over time based on (a) children’s gender (i.e., boys or girls), or (b) mothers’ nativity (i.e., U.S.-born or Mexico-born). In particular, a series of four multigroup models that included gender or nativity as the grouping variable were compared using a χ2 difference test. In Model 1, all estimated parameters were invariant across groups (i.e., across boys and girls when gender was included as the grouping variable or across U.S.-born mothers and Mexico-born mothers when nativity was included as the grouping variable); in Model 2 latent variable means were allowed to vary freely across groups; in Model 3 latent variable means, variances, and covariances were allowed to vary freely across groups; and in Model 4 all of the estimated parameters were allowed to vary freely across groups. If any of the χ2 difference tests comparing models were significant it suggested that the second model (i.e., less constrained) was a better fitting model than the first model (i.e., more constrained), and it should be used for the next model comparison.

Additionally, the best fitting multigroup model from the prior step was used to test whether regression paths in the hypothesized model varied by children’s gender or adolescent mothers’ nativity. In particular, we added fourteen regression paths from the predictors (i.e., mothers’ age, family income, adolescent mother-grandmother coresidency, adolescent mothers’ school status, adolescent mothers’ perceived social support, number of individuals living in the home with adolescent mothers, and adolescent mothers’ depressive symptoms) to the intercept and slope of adolescent mothers’ emotion scaffolding, and four regression paths from the intercept and slope of adolescent mothers’ emotion scaffolding to children’s outcomes (i.e., children’s self-regulation and academic functioning). For each grouping variable (i.e., children’s gender or mothers’ nativity), we examined nested models, and tested the difference between them using a χ2 difference test. In this method, a first model allowed all estimates to be freely estimated across groups (i.e., an unconstrained model), and a second model constrained the estimates to be equal across groups (i.e., a fully constrained model). If the χ2 difference test between the nested models was significant, it suggested that there were significant differences in the model based on group. If the χ2 difference test between the models was not significant, it suggested that there were no significant differences across groups.

All models were tested via a structural equation modeling framework in Mplus version 7.2 (Muthén & Muthén, 2013). Three primary fit indices were used to examine overall model fit: the comparative fit index (CFI), the root-mean-square-error of approximation (RMSEA), and the standardized root-mean-square residual (SRMR). Model fit was considered to be good (acceptable) if the CFI was greater than or equal to .95 (.90), the RMSEA was less than or equal to .05 (.08), and the SRMR was less than or equal to .05 (.08; Hu & Bentler, 1999). Missing data were handled with full information maximum likelihood (Arbuckle, 1996).

Results

Tests of developmental change in emotion scaffolding.

Correlations, means, and standard deviations were computed for all study variables (see Table 1). Initially, a no growth model was compared to a linear growth model to examine whether there was significant change (i.e., significant increase or decline) in adolescent mothers’ emotion scaffolding over time. In the no growth model, paths from each observed emotion scaffolding variable to the latent intercept were constrained to be equal across waves, and in the linear growth model, paths from W3 to W6 of the observed emotion scaffolding variable to the latent slope were set equal to 0, 1, 2, and 3, respectively; therefore, data were centered at W3. Model fit indices from the no growth model demonstrated poor fit to the data: χ² (df = 11) = 19.21, p = .06; CFI = .87; RMSEA = .07 (90% C.I.: .00 - .11); SRMR = .10; however, the linear growth model (which reflected a decline in emotion scaffolding) resulted in good fit: χ² (df = 8) = 9.12, p = .33; CFI = .98; RMSEA = .03 (.00 - .10); SRMR = .06. Furthermore, when the no growth model was compared to the linear growth model, the χ2 difference test comparing these two models was significant [Δχ² (Δ df = 3) = 10.09, p = .02], suggesting that there was significant decline in adolescent mothers’ emotion scaffolding.

Table 1.

Means, Standard Deviations, and Correlations among Study Variables for Sample (N = 204).

	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1. W1 M School Status	--
2. W1 M Perceived Social Support	.05	--
3. W1 M-G Coresidency	−.15*	.01	--
4. W1 Family Income+	.23***	.10	.00	--
5. W2 C Birthweight	.04	.05	.06	.06	--
6. W3 C Negative Temperament	−.05	−.03	.16*	−.16*	−.21**	--
7. W3 M Social Competencies	.11	.32***	.14*	.08	.13	−.10	--
8. W3 M Depressive Symptoms	−.23***	−.31***	.02	−.21**	−.09	.24***	−.31***	--
9. W3 M Emotion Scaffolding	.02	.05	−.07	.19**	.13	−.16*	.09	−.04	--
10. W4 M Emotion Scaffolding	.19**	.01	−.12	.16*	−.01	−.16*	.11	−.15*	.32***	--
11. W5 M Emotion Scaffolding	.17*	.01	−.00	.08	−.03	−.01	.16*	−.27***	.21**	.37***	--
12. W6 M Emotion Scaffolding	.24***	.03	−.03	.18**	.06	−.19**	−.05	−.25***	.20**	.32***	.43***	--
13. W6 C Self-Regulation	.08	−.02	.04	.22**	−.00	−.06	−.04	−.07	.03	.18**	.11	.24***	--
14. W6 C Academic Functioning	.18**	.07	.04	.18**	.06	.02	−.01	−.07	.15*	.23***	.21**	.25***	.40***	--
Mean	.63	5.83	1.13	27.34	113.95	3.55	3.28	.85	2.59	2.69	2.42	2.42	2.64	390.18
Standard Deviation	.48	.77	.34	19.83	15.86	.80	.44	.58	1.03	1.01	.88	1.02	.65	16.67

Note. W = Wave, M = Mother, M-G = Mother-Grandmother, C = Child. Maternal school status coded: 0 = Dropped out of school, 1 = Attending school, graduated, or obtained a General Education Development (GED) diploma. Mother-grandmother coresidency coded: 1 = Live Together, 2 = Do Not Live Together.

⁺W3 Family Income was rescaled by dividing by 1000.

^*p < .05.

^**p < .01.

^***p < .001.

We also tested for the possibility of significant age-related changes in emotion scaffolding by testing a curvilinear change model. If the curvilinear change model was better fitting than the linear model, it would suggest that there were significant differences in the change between individual age periods. The χ²difference test comparing these two models was not significant [Δχ² (Δ df = 4) = 2.31, p = .68], suggesting there was not significant age-related change between age periods in mothers’ emotional scaffolding across time.

Tests of moderation by gender and nativity.

First, we tested whether there were differences in the change of adolescent mothers’ emotional scaffolding based on children’s gender (Table 2). Findings indicated that there were no significant gender differences. Next, we tested for differences by adolescent mothers’ nativity (Table 3). Results indicated that there were no significant nativity differences.

Table 2.

Nested Model Comparisons, Fit Statistics, and χ2 Difference Test for Gender Differences in the Developmental Change in Mothers’ Emotion Scaffolding.

Model	Models Compared	Constraints Removed	CFI	RMSEA	SRMR	χ2(df)	Δ χ2(Δ df)	p
1	--	--	.99	.01	.10	22.38 (22)	--	--
2	1 and 2	Means	.96	.04	.10	22.25 (20)	.13 (2)	0.94
3	1 and 3	Means, Variances, and Covariances	.93	.06	.09	21.42 (17)	.96 (5)	0.97
4	1 and 4	Means, Variances, Covariances, and Residual Variances	.92	.06	.09	20.98 (16)	1.41 (6)	0.97

Note. In Model 1, all estimated parameters were invariant across boys and girls; in Model 2 latent variable means were allowed to vary freely across boys and girls; in Model 3 latent variable means, variances, and covariances were allowed to vary freely across boys and girls; and in Model 4 variable means, variances, covariances, and residual variances were allowed to vary freely across boys and girls. If any of the chi square difference tests comparing models were significant it suggested that the second model (i.e., less constrained) was a better fitting model than the first model (i.e., more constrained), and it should be used for the next model comparison.

Table 3.

Nested Model Comparisons, Fit Statistics, and χ2 Difference Test for Nativity Differences in the Developmental Change in Mothers’ Emotion Scaffolding.

Model	Models Compared	Constraints Removed	CFI	RMSEA	SRMR	χ2(df)	Δ χ2(Δ df)	p
1	--	--	.91	.05	.12	27.52 (22)	--	--
2	1 and 2	Means	.90	.06	.13	26.58 (20)	.94 (2)	0.63
3	1 and 3	Means, Variances, and Covariances	.87	.07	.11	25.06 (17)	2.46 (5)	0.78
4	1 and 4	Means, Variances, Covariances, and Residual Variances	.87	.08	.12	20.98 (16)	6.54 (6)	0.37

Note. In Model 1, all estimated parameters were invariant across U.S.-born mothers and Mexico-born mothers; in Model 2 latent variable means were allowed to vary freely across U.S.-born mothers and Mexico-born mothers; in Model 3 latent variable means, variances, and covariances were allowed to vary freely across U.S.-born mothers and Mexico-born mothers; and in Model 4 variable means, variances, covariances, and residual variances were allowed to vary freely U.S.-born mothers and Mexico-born mothers. If any of the chi square difference tests comparing models were significant it suggested that the second model (i.e., less constrained) was a better fitting model than the first model (i.e., more constrained), and it should be used for the next model comparison.

Next, we tested whether any regression paths in the hypothesized model varied by children’s gender or adolescent mothers’ nativity. Given that prior model comparisons indicated no significant differences in means, variances, covariances, and residual variances across gender or nativity groups, these parameters were constrained to be equal in model comparisons. The first model allowed path estimates for each of the hypothesized associations to vary freely across groups (i.e., boys and girls for children’s gender; U.S.-born adolescent mothers and Mexico-born adolescent mothers for nativity) and the second model constrained path estimates to be equal across groups. The χ2 difference tests comparing the two nested models with children’s gender [Δ χ2 (Δ df = 8) = 6.56, p = .58] and adolescent mothers’ nativity [Δ χ2 (Δ df = 8) = 6.40, p = .60] were not significant.

Tests of the final hypothesized model.

Based on all prior results, which indicated that there were no significant gender or nativity differences, we examined the results from the initial linear growth model we conducted above, in which all participants were pooled together, and we tested the final hypothesized model. As mentioned previously, the linear growth model demonstrated good fit: χ² (df = 8) = 9.12, p = .33; CFI = .98; RMSEA = .03 (.00 - .10); SRMR = .06. Results indicated that the mean intercept of adolescent mothers’ emotion scaffolding at W3 was 2.64 (p = .00), and there was a trending annual change in emotion scaffolding from W3 to W6 of −.07, indicating that, on average, adolescent mothers’ emotion scaffolding decreased over time (p = .06; mean slope). There also was significant variability in adolescent mothers’ emotion scaffolding at W3 (ψ = .42, SE = .11; p = .00; variance of the intercept), and in adolescent mothers’ annual rate of change in emotion scaffolding (ψ = .05, SE = .03; p = .03; variance of the slope).

The final hypothesized model demonstrated good fit: χ² (df = 42) = 39.95, p = .56; CFI = 1.00; RMSEA = .00 (.00 - .04); SRMR = .04 (see Table 4 for all paths tested in the model, and Figure 1 for standardized, hypothesized paths only). Results indicated that, as expected, adolescent mothers’ higher depressive symptoms at W3 were associated with a greater decline in adolescent mothers’ emotion scaffolding from W3 to W6. Because the average slope factor for scaffolding was negative, decreases in the slope value represent a decline in these behaviors; therefore, higher depressive symptoms among adolescent mothers when their children were age 2 (W3) predicted a steeper decline in scaffolding (i.e., a smaller slope value) from W3 to W6 as compared to adolescent mothers who reported lower depressive symptoms at W3.

Table 4.

Unstandardized and Standardized Coefficients for Pathways in the Final Hypothesized Model (N = 204).

Paths	b	SE b	β	SE β	P
Hypothesized Paths
M Depressive Symptoms at W3 → Intercept of M Emotion Scaffolding at W3	.06	.14	.06	.13	.66
Intercept of M Emotion Scaffolding at W3 → C Self-Regulation at W6	.31	.13	.32	.14	.02
Intercept of M Emotion Scaffolding at W3 → C Academic Functioning	12.19	3.36	.48	.14	.00
M Depressive Symptoms at W3 → Change in M Emotion Scaffolding W3-W6	− .17	.07	−.42	.17	.01
Change in M Emotion Scaffolding W3-W6 → C Self-Regulation at W6	1.08	.53	.38	.18	.03
Change in M Emotion Scaffolding W3-W6 → C Academic Functioning at W6	32.97	14.22	.45	.17	.01
Direct Paths
M Depressive Symptoms at W3 → C Self-Regulation	.09	.12	.08	.11	.47
M Depressive Symptoms at W3 → C Academic Functioning	3.63	3.17	.13	.11	.25
Control Paths
C Birthweight → Intercept of M Emotion Scaffolding at W3	.00	.01	.07	.11	.55
C Negative Temperament → Intercept of M Emotion Scaffolding at W3	−.12	.10	−.14	.12	.25
M Social Competencies → Intercept of M Emotion Scaffolding at W3	.29	.19	.20	.13	.13
M School Status → Intercept of M Emotion Scaffolding at W3	−.01	.17	−.00	.12	.97
M Perceived Social Support → Intercept of M Emotion Scaffolding at W3	−.02	.11	−.02	.13	.87
M-G Coresidency → Intercept of M Emotion Scaffolding at W3	−.30	.23	−.16	.12	.19
Family Income → Intercept of M Emotion Scaffolding at W3	.01	.00	.21	.12	.07
C Birthweight → Change in M Emotion Scaffolding W3-W6	−.00	.00	−.10	.14	.51
C Negative Temperament → Change in M Emotion Scaffolding W3-W6	.02	.05	.09	.16	.59
M Social Competencies → Change in M Emotion Scaffolding W3-W6	−.16	.09	−.30	.16	.07
M School Status → Change in M Emotion Scaffolding W3-W6	.14	.08	.28	.16	.07
M Perceived Social Support → Change in M Emotion Scaffolding W3-W6	−.01	.05	−.03	.16	.87
M-G Coresidency → Change in M Emotion Scaffolding W3-W6	.15	.10	.22	.15	.14
Family Income → Change in M Emotion Scaffolding W3-W6	−.00	.00	−.07	.16	.68

Note. M = Mother, M-G = Mother-Grandmother, C = Child, W = Wave. Maternal school status coded: 0 = Dropped out of school, 1 = Attending school, graduated, or obtained a General Education Development (GED) diploma. Mother-grandmother coresidency coded: 1 = Live Together, 2 = Do Not Live Together.

Figure 1.

Predictors and Outcomes Associated with the Linear Developmental Change in Mothers’ Emotion Scaffolding.

Solids lines indicate significant paths and dashed lines indicate non-significant paths. Standardized coefficients are displayed. The following were included as controls predicting the intercept and slope of mothers’ emotion scaffolding in teaching task, but are not shown here for ease of illustration: children’s birthweight, children’s negative temperament, mothers’ social competencies, mothers’ school status, mothers’ perceived social support, mother-grandmother coresidency, and family income. All exogenous variables are mean-centered. Given that coefficients are standardized, coefficients of .20, .50, and .80 correspond to small, medium, and large effect sizes, respectively (Cohen, 1988). * p < .05. ** p < .01. *** p < .001.

In turn, initial levels and change in adolescent mothers’ emotion scaffolding was associated with children’s self-regulation and academic functioning at W6; that is, the more drastically emotion scaffolding decreased overtime, the lower children’s self-regulation and academic functioning were at W6. Finally, the direct effects from adolescent mothers’ depressive symptoms at W3 to children’s self-regulation and academic functioning at W6 were not significant.

To formally test for indirect effects, the RMediation web application was utilized to compute confidence intervals for the indirect effects (Tofighi & MacKinnon, 2011). Using this method, the indirect effect was significant if the confidence interval did not contain zero. Results indicated that there was a significant indirect effect between adolescent mothers’ depressive symptoms at W3 and children’s self-regulation at W6 through change in adolescent mothers’ emotion scaffolding from W3 to W6 [Indirect effect = −.16 (SE = .10), 95% CI = −.398, −.004]. In addition, there was a significant indirect effect between adolescent mothers’ depressive symptoms at W3 and children’s academic functioning at W6 through change in adolescent mothers’ emotion scaffolding from W3 to W6 [Indirect effect = −.19 (SE = .11), 95% CI = −.439, −.021].

Sensitivity analyses.

A series of additional analyses were conducted to further probe the associations examined. First, we tested whether results from our conditional models would change if we limited the analytic sample to adolescent mothers who did not meet the threshold for risk of clinical levels of depressive symptoms. We conducted all analyses without the 78 adolescent mothers in the sample who showed risk for clinically significant levels of depressive symptoms. The results did not change. Adolescent mothers who reported relatively higher levels of depressive symptoms at W3 showed a steeper decline in emotional scaffolding from W3 to W6 (b = −.44, p = .01), relative to those who reported lower levels of depressive symptoms. In turn, the decrease in adolescent mothers’ emotional scaffolding was associated with children’s poorer self-regulation (b = .61, p= .00) and academic functioning at W6 (b = .57, p = .00).

We also conducted additional analyses to explore divergent validity for emotion scaffolding, as compared to motivational and technical scaffolding. That is, we examined whether the effects tested in the current study were specific to emotion scaffolding and not motivational or technical scaffolding. The correlations between technical, motivational, and emotional scaffolding are as follows: .73, .67, .68 (W3), .71, .70, .69 (W4), .72, .71, .69 (W5), and .59, .69, .67 (W6). Technical and motivational scaffolding demonstrated significant linear decline over time (i.e., technical scaffolding: b = −.12, p = .000; motivational scaffolding: b = −.109, p = .002), but not significant variability (i.e., for technical scaffolding: b = −.01, p = .78 and for motivational scaffolding: b = −.02, p = .35). Next, we examined associations between these forms of scaffolding and adolescent mothers’ depressive symptoms and children’s outcomes. The model with motivational scaffolding would not converge due to lack of variability for this construct. The model with technical scaffolding revealed that this form of scaffolding was not significantly linked to maternal or child variables.

We next combined all three types of scaffolding into one measure. This scaffolding composite did not have significant variability (i.e., b = .001, p = .929). When we attempted to include this scaffolding composite in our model to assess links with mothers’ psychosocial wellbeing and child outcomes, the model would not converge due to lack of variability. These weakened results using the scaffolding composite variable support the notion that the effect may be primarily driven by emotion scaffolding.

Discussion

The present study was the first, to our knowledge, to examine changes in emotion scaffolding behaviors in adolescent mothers across their children’s early development. First, we found that, on average, adolescent mothers showed decreases in scaffolding behaviors from child age 2 to 5 years. Adolescent mothers’ higher depressive symptoms when their children were 2 years of age were associated with a greater decline in adolescent mothers’ emotion scaffolding from 2 to 5 years of age, which was subsequently linked to children’s lower self-regulation and academic readiness at age 5. Importantly, tests for indirect effects revealed that depressive symptoms were linked to children’s levels of both self-regulation and academic readiness through adolescent mothers’ change in emotion scaffolding. Together, these results indicate that Mexican-origin adolescent mothers’ early depressive symptoms could be associated with the development of their parenting skills and, subsequently, with key dimensions of children’s social-emotional and academic functioning.

Developmental Change in Emotion Scaffolding

Our study revealed that, in general, mothers decreased in their use of all forms of scaffolding behaviors from child age 2 to 5 years. Sensitivity analyses revealed that, in addition to the observed decrease in emotion scaffolding, there was also a linear decrease in technical and motivational scaffolding over time. It may be that adolescents’ less mature developmental stage may have undermined increases in these parenting behaviors. Because emotion scaffolding requires a deep awareness and understanding of children’s emotional states and regulatory capacities, the complexity of these parenting behaviors increases with child age and development. It is possible that, given adolescents’ ongoing development in the area of self-regulation (Steinberg, 2010), they may have difficulty keeping pace with their children’s increasing emotion scaffolding needs. Such a perspective would support the social-cognitive framework on parenting behavior (Azar, 2002), which argues that under-developed social-emotional foundational capacities may undermine parents’ learning or changes in their behaviors.

On the other hand, it might be that, in line with findings of other maternal teaching behaviors (i.e., directiveness) in samples of Latina mothers, Mexican-origin adolescent mothers in our sample showed similar declines in their behaviors that might reflect traditional Mexican cultural beliefs about children’s relative immaturity at younger ages (i.e., the need for more support at younger ages) and growing competence with age (i.e., the need for less support in later childhood; Ispa et al., 2013). However, given that the teaching task was designed to be too difficult for most children to complete on their own, it may not have been adaptive for adolescent mothers to reduce their scaffolding behaviors, which would explain the direction of associations between declining scaffolding and children’s developmental outcomes. These questions should be addressed in future research with Mexican-origin parents to explicitly understand how their beliefs and values regarding children’s development may predict their parent-child behaviors.

Finally, it is possible that simultaneous changes in children’s self-regulation, which were not measured in the present study, affected these findings. One possibility may be that mothers were better able to increase emotion scaffolding behaviors when their children showed developmental increases in self-regulation. Conversely, mothers may have struggled to maintain higher levels of emotion scaffolding if their children showed simultaneous lower levels of change in self-regulation. It will be important for future work to examine patterns of adolescent mothers’ change in scaffolding behaviors alongside patterns of change in children’s behaviors to understand how the interplay of these dynamic behaviors impacts children’s subsequent development. Future studies should also simultaneously measure patterns of change in other parent behaviors to further isolate the key behaviors that show changes in various contexts, and which behaviors are more versus less challenged over time in the context of adolescent development.

The Role of Depressive Symptoms for the Development of Emotion Scaffolding

It is notable that there was also significant variability in adolescent mothers’ emotion scaffolding at age 2 and in their change in scaffolding from age 2 to 5. Our examination of depressive symptoms as a predictor of individual differences in the initial level and changes in scaffolding revealed that adolescents who reported more depressive symptoms when their children were age 2 showed a greater decline in scaffolding behaviors between 2 and 5 years of age. Effect sizes were small to medium, consistent with prior findings on depressive symptoms and maternal behavior (e.g., Hoffman et al., 2006).

These findings underscore that depressive symptoms may be one aspect of the psychological risk factors linked to Latina adolescent mothers’ parenting behaviors (Contreras et al., 2002). Adolescent mothers with higher depressive symptoms showed steeper declines in scaffolding across child age 2 to 5 years, when children rely heavily on parental sources of coregulation in emotionally arousing situations, highlighting the far-reaching consequences of adolescent mothers’ early depressive symptoms. It is also possible that mothers with elevated depressive symptoms show less positivity and optimism in interactions with their children (Chang, Maydeu-Olivares, & D’Zurilla, 1997), which undermined their ability to demonstrate effective emotion scaffolding. Adolescents’ underdeveloped executive control capacities (Steinberg, Dahl, Keating, Kupfer, Masten, & Pine, 2006) could also challenge their ability to flexibly apply strategies to support their children’s developing regulatory needs across age. Finally, the findings may reflect adolescent mothers’ understanding of their children’s needs; adolescent mothers with more depressive symptoms may engage in fewer interactions with their children, in general, and be less attuned to their children’s development. Field et al. (2007) argue that depressive symptoms in motherhood lead to a state of “chronic emotional unavailability;” thus, it may be that depressed adolescents are less aware of or attuned to their children’s emotional arousal and regulatory capacities, leading to fewer scaffolding behaviors and less positive changes in the behaviors needed to match children’s developing capacities. Regardless of the mechanism that explains this association, the current findings identified Mexican-origin adolescent mothers’ depressive symptoms as one factor on which to intervene in hopes of ultimately improving their parenting capacities and, in turn, promoting better adjustment among their young children. It is important to highlight that our measure of depressive symptoms only indicates risk for clinical depression; thus, we do not know which of the adolescents in our sample suffered from clinical depression, nor do we know whether they had comorbid symptoms that may have influenced their parenting, or how their depressive symptoms changed over their child’s early development. Future work should aim to examine parallel development of adolescent mothers’ psychosocial wellbeing alongside emotion scaffolding behaviors to understand what maternal characteristics change alongside emotion scaffolding in early development.

Emotion Scaffolding and Children’s Subsequent Self-Regulation and Academic Readiness

Our study found that decline in adolescent mothers’ emotion scaffolding was associated with children’s poorer self-regulation and academic functioning at age 5; for adolescent mothers who demonstrated a greater decline in emotion scaffolding over time, children had lower self-regulation and academic functioning compared to children of adolescent mothers who showed a smaller decline. Our findings support work that has found children of mothers with higher depressive symptoms to have lower academic performance, poorer behavioral outcomes, and poorer social competence as compared to children of mothers with lower depressive symptoms (e.g., Valdez et al., 2013). Importantly, our work fills a gap in research linking maternal depressive symptoms at child age 2 with child outcomes at age 5 via developmental changes in adolescent mothers’ emotion scaffolding. Interestingly, sensitivity analyses revealed that these pathways were not significant for technical or motivational scaffolding. It may be that because emotion scaffolding reflects support for children’s burgeoning self-regulation and sense of efficacy it can be generalized to other contexts by the children, whereas technical and motivational scaffolding are task-specific strategies that do not as easily carry over to other contexts.

Although adolescent mothers decreased in all three forms of scaffolding, preliminary evidence from this study supports the notion that decline in emotion scaffolding plays a unique role in promoting children’s subsequent self-regulatory and academic development. A more drastic decline in such behaviors was linked to poorer outcomes in children at age 5. Adolescent mothers’ emotion scaffolding behaviors may have been linked to children’s own self-regulatory capacities in a number of ways. Adolescent mothers with greater scaffolding capacities may have helped their children to acknowledge and label their emotions; demonstrated or modeled the use of various emotion regulation strategies; and demonstrated the contingent association between use of strategies and subsequent soothing or relief. When faced with tasks requiring their independent performance at age 5, children whose mothers had used more emotion scaffolding in the teaching task across early development may have been better prepared to regulate their emotions and behaviors to succeed with the task.

Our finding that adolescent mothers’ emotion scaffolding showed even stronger relations to children’s performance on the academic assessment than to children’s self-regulation is consistent with past work that found stronger associations between mothers’ supportive behaviors and children’s cognitive relative to social-emotional outcomes (Fuligni et al., 2013). Future work is needed to elucidate these mechanisms, potentially linking maternal scaffolding to test achievement via children’s emotion regulation, problem-solving strategies, executive function, or other skills (Fabes et al., 2003; Fay-Stammbach, Hawes, & Meredith, 2014). The importance of understanding these mechanisms is critical, as past work with children of Mexican-origin adolescent mothers has demonstrated that, on average, they are just shy of one standard deviation below the normative mean on measures of academic achievement at kindergarten entry (Jahromi, Umaña-Taylor, Updegraff, & Zeiders, 2015). Thus, while not all children in our sample demonstrated developmental delays at age 5, a subset showed clear risk for delays. While multiple child, parental, and contextual factors are thought to play a role in predicting children’s developmental functioning, the results of this study suggest that one pathway is through adolescent mothers’ emotion scaffolding behaviors over time.

Limitations and Future Directions

Our study of change in adolescent mothers’ scaffolding is only a first step in this line of work, and limitations of this study offer important directions for future research. First, future studies are needed in which adolescent mothers’ own self-regulatory processes and depressive symptoms or other indices of well-being are directly measured across time, along with their scaffolding behaviors and their children’s self-regulation, using parallel process models, to enable an understanding of how these capacities co-develop over time. Future work should aim to examine intra-individual patterns of change in mothers’ scaffolding behaviors and a broader range of predictors of such behaviors to understand whether some mothers show a different pattern of change in these behaviors and what factors might be relevant to different patterns. In addition, it will be important to examine the nature of changes in children’s self-regulation over time which may be related to parents’ well-being.

Future studies, with larger samples, in different contexts (e.g., free play versus teaching; Ciciolla et al.,2013), and with different parenting behaviors are needed to accurately document changes in young mothers’ behaviors across their children’s early development. Our study aimed to design developmentally appropriate mother-child interactions that were equivalent over time, but we cannot confirm the validity of this design. Moreover, given that inter-rater reliabilities for observed scaffolding behaviors declined slightly in the final two waves of our study, it will be important for future work to explore whether differences exist in how such behaviors present across children’s and adolescent mothers’ development. Also, in the current study the assessment of emotion scaffolding included Waves 3-6, which overlapped with our outcomes at Wave 6. Although the end point of the slope is contemporaneous with the outcomes, it does not pose a threat to the temporal ordering because the rate of change is being determined largely by the pattern of what has happened in three years prior. However, it will be important for future work to test these notions empirically with a larger number of waves that do not overlap with outcomes.

Our study is also limited in its generalizability to populations beyond that which was studied (i.e., Mexican-origin 15- to 19-year-olds living in a Southwest metropolitan area); thus, future work should extend our findings to adolescent mothers in other ethnic and socio-contextual groups. It will also be important to examine scaffolding behaviors among other family members and caregivers of children in multi-generational families, including fathers and grandparents. Indeed, some work suggests that fathers may play a particularly important role in their children’s development of self-regulation (Stevenson & Crnic, 2013); thus, a study of adolescent fathers’ emotion scaffolding would be an important future direction.

Finally, although we examined the mechanism by which depressive symptoms were indirectly linked to children’s outcomes via an environmental factor (i.e., maternal behaviors), it is important to acknowledge that potential genetic effects may also be relevant to the link between adolescent mothers’ psychosocial functioning and that of their children (e.g., Rutter, 1990). To support this notion, there is empirical evidence that the link between parental depressive symptoms and children’s conduct problems may be a function of both family environment and genetic factors (e.g., Silberg, Maes, & Eaves, 2010). It will be important for future work to examine possible passive gene-environment correlations, wherein the genetic risk factors passed down to children and the parenting environment are related in a way that exacerbates children’s risk for subsequent maladaptive outcomes.

Conclusion

The current study makes an important contribution to knowledge about the development of parenting behaviors across children’s early development, particularly among vulnerable mother-child dyads. Using a longitudinal design with observations of Mexican-origin adolescent mothers and children at four age points during a key developmental period, this study tested a mechanism linking a maternal risk factor (depressive symptoms) to risk in child outcomes (self-regulation and academic achievement) by maternal scaffolding, which can be a target of intervention. Adolescent mothers’ depressive symptoms were found to be linked to their children’s self-regulation and academic readiness through adolescent mothers’ developmental change in emotion scaffolding such that depressive symptoms at age 2 were associated with a greater decline in adolescent mothers’ emotion scaffolding from 2 to 5 years of age, which was subsequently linked to children’s less self-regulation and academic readiness at age 5. Our study is a first step to understand the pattern of change in adolescent mothers’ emotion scaffolding behaviors and the unique contribution of these behaviors to aspects of children’s social-emotional and academic development.