Mothers’ Teaching Strategies and Children’s Effortful Control: A Longitudinal Study

Abstract

Findings on the relation of maternal verbal teaching strategies to children’s effortful control (EC; i.e., self-regulation) are limited in quantity and somewhat inconsistent. In this study, children’s EC was assessed at 18, 30, and 42 months (ns = 255, 229, and 209, respectively) with adults’ reports and a behavioral measure. Mothers’ verbal teaching strategies were assessed while the mother and child worked on a task together. Children’s general vocabulary also was measured. In a structural panel model taking into account prior levels of constructs and correlations within time, as well as the relations of EC and teaching strategies to children’s vocabulary, socioeconomic status, age, and sex of the child, 18-month EC positively predicted mothers’ 30-month cognitive assistance and questioning strategies and negatively predicted 30-month maternal directive strategies. In addition, high 30-month EC predicted greater 42-month maternal cognitive assistance and fewer directive strategies. Thus, mothers’ teaching strategies were predicted by individual differences in self-regulatory skills, supporting potential evocative child effects on mothers’ teaching strategies.

The goal of this study was to examine within- and across-time bidirectional relations between young children’s dispositional self-regulatory skills (i.e., effortful control) and the quality of maternal verbal strategies during teaching interactions using longitudinal models that provide a relatively strong test of potential causal relations (i.e., by controlling for initial levels of constructs when predicting over time). Because of the relevance of individual differences in children’s self-regulatory capacities for many domains of functioning (see below), it is important to understand the role of potential environmental influences such as parenting in the development of self-regulation, as well as to examine if children’s self-regulation predicts parents’ verbal interactions with their children in a teaching context.

Effortful Control

Rothbart and Bates (2006) defined effortful control (EC) as “the efficiency of executive attention—including the ability to inhibit a dominant response and/or to activate a subdominant response, to plan, and to detect errors” (p. 129). EC is believed to regulate temperamental reactivity; it includes the ability to willfully deploy attention (attention focusing and shifting) and inhibit or activate behavior (inhibitory and activational control, respectively). EC appears to emerge in rudimentary forms in the first year of life. For example, 6- to 7-month-olds exhibit anticipatory looking, which is viewed as a rudimentary form of effortful attention (Sheese, Rothbart, Posner, White, & Fraundorf, 2008). In addition, Diamond (1991) found that around 12 months of age, infants were able to reach for a target not in their line of sight, demonstrating that they were able to coordinate reach and vision and attend to both. These abilities are believed to involve the execution of intentional behavior and planning and resistance to more automatic action tendencies (Diamond, 2002). Despite early evidence of EC, it improves substantially in the preschool years, especially around the third year of life (Gerardi-Caulton, 2000; Reed, Pien, & Rothbart, 1984; see reviews in Eisenberg, Spinrad, & Eggum, 2010; Rothbart & Bates, 2006). Moreover, after age 2, individual differences in EC become increasingly stable and measures of EC increasingly interrelate (Kochanska, Murray, & Harlan, 2000).

Numerous investigators have found that EC is related to and predicts a range of positive developmental outcomes across childhood. Specifically, greater levels of EC and/or self-regulation have been consistently negatively related to externalizing and internalizing problems (although relations to internalizing problems are somewhat less consistent if co-occurring externalizing problems are controlled; Eisenberg et al., 2009; see Eisenberg et al., 2010; Rothbart & Bates, 2006, for reviews). High EC also has been associated with relatively high levels of popularity, social skills, conscience, and sympathetic and prosocial tendencies (e.g., Eisenberg et al., 2007; Kochanska & Knaack, 2003; Spinrad et al., 2006), as well as academic performance (e.g., Duncan et al., 2007; McClelland et al., 2007; Valiente, Lemery-Chalfant, & Castro, 2007; Valiente, Lemery-Chalfant, Swanson, & Reiser, 2008; see Eisenberg, Sadovsky, & Spinrad, 2005, for a review).

EC and Teaching Strategies: Conceptual Views

Although EC is influenced by heredity (Rothbart & Bates, 2006), it also has been related to quality of parenting. For example, EC and observed self-regulation have been associated with relatively high maternal responsivity (Kochanska et al., 2000), positive guidance (e.g., praise, affection, positive affect, positive feedback, encouragement, and suggestions), and low negative control (e.g., scolding, punitive actions and verbalizations, physical and verbal control; Calkins, Smith, Gill, & Johnson, 1998; Eiden, Edwards, & Leonard, 2004; Lengua, Honorado, & Bush, 2007; Lunkenheimer et al., 2008; Olson, Bates, & Bayles, 1990; Putnam, Spritz, & Stifter, 2002). However, the vast majority of investigators examining the relations of parenting to EC and related self-regulatory constructs have measured general parenting style (e.g., warmth, support, punitiveness) rather than specific nondisciplinary parenting behaviors, such as in a teaching situation. Moreover, there has been relatively little research on the relations of maternal teaching behavior to children’s self-regulation across as well as within time.

There are conceptual reasons to expect the quality of parental teaching to affect children’s self-regulation and vice versa. Vygotsky (1978) argued that parents play a major role in the development of children’s self-regulation, in part by providing direction and guidance during teaching situations. Parents structure the way children approach, attend to, and deal with difficult tasks, and such behavior is believed to foster learning and affect regulation. However, various parental teaching strategies may have different effects, and investigators often have studied this issue from a Vygotskyian perspective. For example, when a parent gives explicit commands that solve a challenge, the child gains little information that can be applied to future learning opportunities and challenges. Nonetheless, directive parental strategies may act as a source of attentional regulation for the young child (Wertsch, 1979). Such external regulation may be appropriate for young children who do not have the skills to self-regulate; however, if used excessively after the early period of life, children may be ill-equipped to self-regulate in challenging contexts and may become overwhelmed by the demands of the situation.

In contrast, parents who verbally assist children to understand a task (sometimes called provision of cognitive assistance; e.g., Fagot & Gauvain, 1997) might help children to perform more competently on the task and develop effective strategies rather than becoming frustrated and possibly inattentive or dysregulated. Learning how to think about challenges as a consequence of parents’ provision of hints and information may give children general strategies for dealing with challenges in the future and opportunities to learn to effectively self-regulate their attention and performance-related behaviors.

Sigel, Stinson, and Kim (1993), like Vygotsky (1978), argued that parents’ discourse provides a source of intellectual stimulation, as it can place demands on children’s abilities to reason and understand. They further argued that parental distancing strategies—conceptualized by Sigel (1993) as behaviors or events that separate the individual cognitively from the immediate environment—are an important part of the process. Parents’ distancing strategies set cognitive demands that encourage children’s own thought processes. However, distancing strategies vary in their level of sophistication, from those that make minimal demands for the child to separate the self from the present situation (which involve minimal representation) to high-level demands that encourage the child to mentally go beyond the information given, make inferences, and use propositional and hypothetical thinking. The more sophisticated distancing strategies described by Sigel and colleagues would be expected to be used primarily with older children, but their argument that these parental strategies affect children’s attention, behavior, and cognition is consistent with Vygotsky’s view. Moreover, Sigel et al. (1993) noted that the form of the distancing strategy (e.g., statement, question, or imperative) might well be a key element of influence. For example, they suggested that distancing strategies in the form of questions (e.g., “What other way might you do this?”) demand that the child separate the self from the ongoing situation and seek different options—actions that often involve self-regulation.

Although both Vygotsky’s and Sigel’s frameworks provide conceptual reasons to expect parents’ verbal strategies during potential teaching contexts to affect children’s self-regulation and competence, they also provide reasons to expect children’s abilities, including their levels of regulated attention and behavior, to affect the quality of parental verbal strategies in teaching contexts. According to Vygotsky’s theory, adults’ efforts to teach children are the most effective if they are within the child’s zone of proximal development, defined as “the distance between the actual developmental levels as determined by independent problem solving and the level of potential development as determined through problem solving under adult guidance or in collaboration with more capable peers” (Vygotsky, 1978, p. 86). This zone differs for children of different ages (e.g., McNaughton & Leyland, 1999; Pratt, Kerig, Cowan, & Cowan, 1999), so parents must adjust their teaching behavior accordingly if it is to have a maximal effect. As Wertsch (1979) argued, there are several levels of young children’s transition from other- to self-regulation during problem solving; to effectively help children regulate, adults must use speech and gestures that are tied to the child’s understanding. Wertsch suggested that statements such as imperatives (e.g., “gimme a match” in a matching task) involve an explicit mention of an action and/or the object to be acted upon, whereas questions and hints do not involve an explicit mention of the desired actions. He also argued that parents tend to use more explicit directives (e.g., commands) when children are less attentive but can use more indirect types of verbal strategies (e.g., nonexplicit directives, questions, hints) when children’s self-regulation is more developed. If a mother is aware that her son or daughter has poor attentional or inhibitory control, she may be more likely to instruct the child using verbalizations that are easily followed, such as directive commands, rather than questioning or cognitive assistance. In other words, if the mother does not believe her child will be able to attend to a more indirect suggestion, formulate a plan independently, and execute the plan, she may be more likely to simply tell the child what to do.

Similarly, Pellegrini, McGillicuddy-DeLisi, Sigel, & Brody (1986) noted that children’s competence is associated with parents’ teaching behaviors, including their distancing verbalizations, and affects the ways in which parents teach them. Although Pellegrini et al. focused primarily on representational competence such as anticipatory imagery, memory, and general mental ability, children’s regulation (including regulated attention) is also expected to affect parents’ teaching strategies.

EC and Teaching Strategies: Empirical Findings

Empirical evidence is consistent with both pathways of influence: the effect of quality of parental teaching on children’s functioning and vice versa. However, few researchers who have examined this issue have had longitudinal data or controlled for prior levels of constructs in analyses (to better assess causality). In much of the research, the assumption has been that the quality of parental teaching fosters optimal outcomes for children, although the data usually are correlational.

In relevant research, parenting behaviors that undermine rather than promote toddlers’ autonomy (i.e., directive controlling behaviors) have been related to lower task persistence and competence (Frodi, Bridges, & Grolnick, 1985). In regard to more specific parental teaching strategies, a directive teaching style (e.g., direct commands) usually has been related to low levels of young children’s verbal and cognitive ability and school readiness, whereas mothers’ appropriate challenge, requests for attentional maintenance, or cognitive assistance has been associated with relatively high levels of these competencies (see Fagot & Gauvain, 1997; Gauvain, Fagot, Leve, & Kavanagh, 2002; Hess & McDevitt, 1984; Mulvaney, McCartney, Bub, & Marshall, 2006). Landry, Smith, Swank, and Miller-Loncar (2000) found that maternal directiveness at age 2 was positively related to children’s early cognitive and social skills, but the relation was negative for directiveness when children were 3.5 years old, suggesting moderation of the relation between parental use of directives and child functioning by age.

In contrast to directives, Fagot and Gauvain (1997) reported that mothers’ cognitive assistance on tasks with their children at 18 and 30 months predicted better cognitive task performance at age 5. Similarly, Smith, Landry, and Swank (2000) found that mothers’ verbal scaffolding—the provision of conceptual links between objects, persons, activities, or functions—with their 3-year-olds predicted the growth of children’s verbal and nonverbal skills over the next 3 years. Other investigators, however, have found relatively few relations of maternal verbal scaffolding to children’s social skills, social behavior with mother, or coordination of attention and word use with mother for typically developing children (Baker, Fenning, Crnic, Baker, & Blacher, 2007; Landry, Smith, Swank, & Guttentag, 2008).

Relatively few investigators have examined the relation of parental teaching strategies to young children’s self-regulation. Some research suggests that parents’ teaching behavior is related to children’s self-regulation, although the findings are not highly consistent. In contrast to the previously mentioned work by Frodi et al. (1985), Bomba, Goble, and Moran (1994) found no concurrent relation of temperamental persistence—a measure often viewed as reflecting EC—to parental teaching behavior, including directiveness. Similarly, one of the few studies of cognitively complex teaching strategies found that parental cognitive guidance (e.g., statements regarding how to complete the task in an efficient systematic or planful manner; explanations of goals and strategies; suggestions of general approaches) was not related to children’s concurrent EC, perhaps because cognitive guidance was combined with evaluation (attempt to check if the task was done correctly) and repair (any correction to a solution) in a single latent construct predicting EC (Gartstein & Fagot, 2003).

Relatively few investigators have examined relations of parental teaching strategies to children’s regulation in longitudinal studies. Landry et al. (2000) found that mothers’ strategies that provided children a choice on a task (e.g., questions, comments, suggestions) in the early preschool years (ages 2 and 3.5 years) predicted children’s cognitive goal-directed behavior (e.g., including persistence on a task, possibly tapping self-regulation) at 4.5 years. Hoffman, Crnic, and Baker (2006) found maternal scaffolding at age 3 to be negatively related to 4-year-olds’ dysregulation, operationalized as observed emotional experience across the duration of each laboratory task and an assessment of how much difficulty children had maintaining an optimal state of emotion regulation for completing the tasks. In their study, Hoffman et al. (2006) defined scaffolding as a mother’s ability to structure the task in such a way that it was within the child’s capabilities to successfully complete it with her support, to help the child become engaged with the task initially and to maintain enthusiasm, and to make the task a positive experience for the child. Thus, unlike a number of other studies, Hoffman et al.’s measures of scaffolding globally tapped an appropriate level of assistance rather than specific cognitive content and directiveness of parents’ statements.

In a longer longitudinal study, Supplee, Shaw, Hailstones, and Hartman (2004) found that maternal strategy instruction on a task when children were 42 months old—indirect (e.g., drawing attention to a strategy) or direct (e.g., “look at the picture to see where that piece fits on the barn”)—was positively related to children’s teacher-reported combined emotion regulation and low negative emotionality at age 6 to 7 years, whereas maternal active cognitive involvement (trying to involve the child in the task rather than doing it, often with questions) and use of advance organizers (using initial set-up instructions) were not. In another longitudinal study extending across multiple years, Olson et al. (1990) reported that the combination of teaching involving scaffolding and responsiveness when children were 13 months old was positively related to greater impulse control at age 6 when cognitive competence was controlled. Perhaps most relevant to the issue of cognitive scaffolding, Fagot and Gauvain (1997) found that maternal cognitive assistance during a challenging task was concurrently positively related to 30-month-olds’ task performance (as, surprisingly, were maternal behavioral directives) and low off-task play, which could be viewed as a measure of self-regulation. Moreover, they found that maternal cognitive assistance at 30 months was related to children’s cognitive and task performance at age 5. Thus, there is some evidence that the quality of maternal verbalizations is related to, and sometimes predicts across time, children’s performance on tasks that involve self-regulation; nonetheless, the direction of causal effects underlying these associations is unclear.

Indeed, in some of the same research there is evidence that children’s characteristics, including temperamentally based EC, may affect the quality and nature of parenting strategies—for example, via evocative genetic effects (Plomin, 1994). For instance, Fagot and Gauvain (1997) found that children’s difficult temperament at 18 months predicted higher levels of both mothers’ cognitive assistance and mothers’ directives at 30 months. In contrast, other findings suggest that mothers may be more likely to use relatively sophisticated cognitive scaffolding rather than directives if they perceive their children as attentive and likely to benefit from it. For example, Landry, Garner, Pirie, and Swank (1994) found that mothers of young children with Down syndrome were more likely to use directives and less likely to use suggestions than mothers of typically developing children, and the children with Down syndrome were more likely to comply with directive versus suggestive requests in more structured context. In addition, mothers’ use of directives has been negatively related to children’s language skills (for compatible findings, see Kermani & Brenner, 2001; Mulvaney et al., 2006; Pellegrini et al., 1986) Other investigators have found that mothers’ instructions become less explicit, concrete, and nonverbal, as well as more verbal, as children age (Heckhausen, 1999; also see Bellinger, 1979; Wood & Middleton, 1975), and mothers vary their teaching as a function of their child’s level of proximal development (McNaughton & Leyland, 1999) or success on tasks (e.g., less scaffolding after success than failure; Pratt et al., 1999). Thus, there is evidence consistent with the conclusion that parents’ teaching strategies are affected by their children’s competence, although very few of these studies have focused on individual differences in self-regulation.

The Present Study

To our knowledge, there is no research with a strong longitudinal design (the same measures assessed at multiple time points, especially with more than two assessments) that examines bidirectional relations between EC and parental teaching style. Nonetheless, there are compelling conceptual reasons to examine child driven as well as parent-driven associations. In the present study, we examined the relation of children’s EC to maternal teaching strategies at three time points (when children were 18, 30, and 42 months). EC was measured with a multimethod, multireporter approach. We focused on three types of maternal strategies: (a) one resembling scaffolding, called cognitive assistance strategies (verbalizations that help the child understand the task or complete more effectively); (b) one tapping a directive approach (explicit, direct verbal instructions about task-related behavior); and (c) one reflecting a more indirect approach, called questioning (verbal questions directed toward the child but not direct attempts to get the child to do something). Cognitive assistance taps mothers’ provision of task information; directives and questions reflect instructional styles (Gauvain et al., 2002). These or similar types of strategies frequently have been used in prior research (e.g., Fagot & Gauvain, 1997; Gauvain et al., 2002; Landry et al., 2000; also see above review), and they all have been viewed as involved in directing and maintaining children’s attention, at least if they are age appropriate (Landry et al., 2000; Wertsch, 1979).

Directive strategies are typically viewed as appropriate in the first years of life because children have difficulty focusing their attention and understanding what is expected of them (Landry et al., 2000; Wertsch, 1979). However, as already noted, such strategies are viewed as less likely to foster regulation and attention competence in older children (e.g., by age 3) and are more frequently used with children who are relatively low in self regulation and task competence (due to age or individual characteristics; Gauvain et al., 2002; Landry et al., 2000). Thus, we expected maternal use of directives at 30 months to predict lower EC at 42 months and/or vice versa. However, it also seemed possible that maternal directives at 18 months would be unrelated or even positively related to EC at 30 months.

In contrast, verbal strategies that foster thinking about objects and events were expected to foster attention and competence (e.g., Gauvain et al., 2002; Pellegrini et al., 1986; Supplee et al., 2004; Wertsch, 1979). Especially as young children move from the toddler to preschool years, maternal verbal strategies that help children learn ways to succeed on tasks rather than merely direct their behavior should provide children with a sense of efficacy and foster the skills needed to persist and focus on tasks rather than become frustrated. Thus, we expected questions and cognitive assistance to positively relate concurrently and longitudinally to EC across the preschool years, although it seemed possible that cognitive assistance would be less related to children’s EC at younger than at older ages. Moreover, we hypothesized that children with higher EC would elicit more cognitive assistance and questions from their mothers.

Furthermore, children with better language skills are more likely than their less verbal peers to benefit from socializers’ attempts to help them regulate emotions or behavior; in addition, language provides children with a tool for processing information and expressing their needs and wants in a constructive manner. Thus, children’s vocabulary was expected to positively predict EC and possibly more sophisticated maternal teaching strategies (e.g., cognitive assistance rather than directive strategies).

Because of the longitudinal nature of our data, we could assess whether parenting and EC predicted one another across time when at least partly taking into account the stability of the constructs, as well as controlling for children’s language skills, sex, age, and socioeconomic status (SES). To further strengthen the design of the study, we used three waves of data and multiple reporters and methods to assess EC; moreover, maternal teaching strategies were assessed with observational data. In addition to the aforementioned longitudinal relations between EC and maternal teaching strategies, we expected concurrent relations between these variables because we assumed they influence one another during the teaching session. Moreover, because SES and children’s language abilities might account for relations between parenting and EC (e.g., SES is related to parenting and EC: Parke & Buriel, 1998; Valiente et al., 2008; language is related to EC: Cole, Armstrong, & Perberton, 2010), we controlled for these constructs in the models. In addition, variability in children’s EC and mothers’ teaching due to age or sex of the child was controlled.

Method

Sample

Full-term, healthy infants with adult parents were recruited at birth at three hospitals in a southwestern city in the United States by hospital staff or a research assistant who approached parents in the hospitals (see Spinrad et al., 2007). If mothers were interested, they filled out a form so they could be contacted at a later time. Mothers completed at-home questionnaires and accompanied their child to a laboratory visit when the child was approximately 18, 30, and 42 months old (called T1, T2, and T3, respectively; nearly all children were tested within a month of the specified age). After dropping one family with only father-reported data, the ns at T1, T2, and T3 were 255, 229, and 209, respectively, including nine, 14, and 18 families who participated only by mail. The children at T1, T2, and T3 were 55%–56% male, and the mean ages in months were 17.78, 29.79, and 41.73 (SDs = .52, .68, and .69, respectively). Nonparental caregivers (ns = 176, 153, 151, for T1, T2, and T3, respectively) completed questionnaires with parental consent (nonparental caregivers and parents provided consent). Non-parental caregivers were relatives or nonrelatives providing care in the home, in another home, or at daycare; they were identified by the mother. More than one third of these caregivers remained the same from one assessment to the next (35%–36%), and 28% were the same at T1 and at T3.

At T1, T2, and T3, children were mostly Caucasian (and not Hispanic; 86.6%; 87%; 86%, respectively) or Hispanic (13%, 13%, and 12%). At T1, T2, and T3, 2.4%, 3.1%, and 3.5%, respectively, were African American; 0.8% to 1.2% were Asian; 2.2% to 3.0% were Native American; and 7% were of another race/ethnicity. At all three assessments, mean parental education was some college or a 2-year degree (range: eighth grade to graduate level), and mean annual family income was $45,000–$65,000 (range was < $15,000 to $100,000+).

One hundred seventy-nine families participated at all assessments; 55 participated at T1 but not T3. Mothers who attrited were younger at the child’s birth (M = 27.16 years) than were mothers still participating at T3 (M = 29.70), t(243) = 2.93, p < .01. No other demographic or key variables at T1 were related to attrition.

Procedure

At each assessment, families were mailed questionnaires that they mailed back or brought to the laboratory. Laboratory sessions were similarly structured at all assessments and lasted approximately 1.5 to 2 hr. As part of a series of tasks, mothers were observed interacting with their toddler during a challenging teaching task; children’s EC was assessed with a delay task (and videotaped for coding). In addition, four research assistants at each assessment rated children’s EC during the parts of the laboratory visit that they viewed. Caregivers completed questionnaires by mail.

Trained undergraduate or graduate research assistants coded the mothers’ behaviors during the teaching task and the EC task; reliabilities (computed for 25%–30% of participants) were calculated periodically to prevent drift. In the vast majority of instances, the coders were not people who rated the children in the laboratory, although in a small number of cases a person coding the videotapes might have been involved in the data collection in some capacity for a particular child. However, coding of videotapes typically occurred a considerable time—often 6 months to a year or more—after data collection, so it is unlikely that a coder would remember details about a given child.

Measures

At all assessments, mothers’ teaching strategies were observed during the teaching task, whereas children’s EC was assessed with a delay task as well as adults’ reports. Mothers also reported on their children’s expressive vocabulary at all assessments.

Maternal behavior and teaching strategies

At T1, T2, and T3, maternal directives, questioning, and cognitive assistance¹ were coded during a 3-min teaching task (adapted from Calkins & Johnson, 1998). Mothers were instructed to teach their child to complete a puzzle (T1, T2) or Lego model (T3) with whatever strategies they would use at home. Directives referred to explicit or direct behavioral instructions (e.g., “Put it there,” “Don’t do that”); questioning referred to questions directed toward the child (e.g., “What shape is this?” or “Where does this go?”; not attempts to get the child to do something, like “Can you put the piece here?”); and cognitive assistance referred to instances when the mother tried to help the child understand the task or complete it more effectively (e.g., “We are going to match the animals with their pictures”; “Try turning the pieces around”). Behaviors were rated 1 = none to 3 = repeatedly at 10-s intervals and then averaged across intervals. Mothers’ verbal strategies could be coded for more than one category (as in Gauvain et al., 2002). Interrater reliability, assessed with intraclass correlations (ICCs) at T1 to T3, was as follows: directives (.89, .74, .92), questioning (.83, .89, .76), and cognitive assistance (.76, .73, .62). Analogous interrater correlations at T1, T2, and T3 were .89, .75, and .93 for directives; .84, .90, and .88 for questioning; and .76, .79, and .82 for cognitive assistance, respectively.

Effortful control

At T1 and T2, mothers and caregivers rated children’s EC from 1 = never to 7 = always with three 12-item scales of the Early Childhood Behavior Questionnaire (ECBQ; Putnam, Gartstein, & Rothbart, 2006): (a) attention-focusing—the ability to concentrate on a task (αs for T1 and T2 = .76 and .81 for mothers and .79 and .85 for caregivers); (b) attention shifting—the ability to move attention from one activity to another (αs at T1 and T2 = .69 and .73 for mothers and .76 and .71 for caregivers); and (c) inhibitory control—the ability to voluntarily control behavior (αs at T1 and T2 = .81 and .88 for mothers and .90 and .88 for caregivers).

At T3, mothers and caregivers rated (from 1 = never to 7 = always) aspects of children’s EC with subscales from the Children’s Behavior Questionnaire (CBQ; Rothbart, Ahadi, & Hershey, 1994), assessing the same constructs as the ECBQ (some items differ from the ECBQ): attention focusing (14 items; αs = .77 and .74 for mothers and caregivers, respectively), attention shifting (12 items; αs = .67 and .80 for mothers and caregivers, respectively), and inhibitory control (13 items; αs = .77 and .82 for mothers and caregivers, respectively). ECBQ and CBQ sub-scales were correlated at each assessment from .23 to .51 (ps < .01 for mothers), and .34 to .68 (ps < .01 for caregivers), and were averaged within reporter to create an EC composite.

At T1, T2, and T3 we used the snack delay task as an observed measure of EC (Kochanska et al., 2000). In this task, children were asked to wait until a bell was rung to get a piece of candy from under a clear cup (T1 and T2) or to eat candy that was on their tongue (T3). At T1 and T2, there were four trials (with delays of 10, 20, 30, and then 15 s) and three trials at T3 (with delays of 20, 40, and 30 s to make the task more age appropriate). Children were rated (on a scale from 1 = no restraint to 4 = extreme restraint) for self-restraint on each trial, and scores were averaged across trials (ICCs = .92–.99). Although other behavioral measures of EC were collected, none was available across all three assessments, and it is more difficult to draw conclusions from panel structural equation models if the measures change over time than if they are comparable across time.

At the end of each lab visit, four researchers (graduate student, computer and audiovisual technicians, and experimenter) completed an adapted version of the Infant Behavior Record (IBR; Stifter & Corey, 2001) that captures children’s behaviors across the entire visit. Attention and persistence items (1 item each) were rated (1 = consistently off task or lacks persistence to 5 = continued absorption in toy/activity/person or consistently persistent; attention α = .74, .84, and .84, and persistence α = .76, .81, and .73 at T1, T2, and T3, respectively). Attention and persistence were averaged across items (rs = .57–.83) and reporters (rs = .44–.68).

Vocabulary

At each assessment mothers completed vocabulary checklists (Short Form [SF] Level II at T1 and T2; Level IV at T3), both of which are part of Macarthur Communicative Development Inventories (CDI; Fenson et al., 2000; Dale, Price, Bishop, & Plomin, 2003). The CDI SF Level II (Parallel Version A was used) contains a 100-word vocabulary production checklist. The Level IV checklist contains a list of 48 words. A child’s score was the total number of words spoken in either English or Spanish.

SES

SES (the mean of standardized maternal and paternal educational levels and family income) was reported by mothers. Because T1 SES was correlated > .93 with SES at later ages, only T1 SES was used to reduce the complexity of the models.

Results

After descriptive results, correlations between maternal strategies and EC are summarized; then longitudinal structural equation models are presented.

Descriptive Results

Means, standard deviations, and sex differences for the key variables are presented in Table 1; correlations among measures of EC are presented in Table 2. Following the recommendation of Curran, West, and Finch (1996), no variables needed to be transformed due to excessive skew or kurtosis. Children’s age at the laboratory visit was concurrently negatively correlated with mothers’ cognitive assistance at T2, r(216) = −.15, p < .05. Age within an assessment did not vary greatly and was not significantly correlated with any other measures in the study.

Table 1.

Means and Standard Deviations of Study Variables at Time 1 (T1), Time 2 (T2), and Time 3 (T3)

Variable	Time 1			Time 2			Time 3
	N	M	SD	N	M	SD	N	M	SD
Teaching strategies
Directives	246	1.52	0.23	216	1.31	0.20	192	1.36	0.20
Questioning	246	1.32	0.20	216	1.37	0.21	192	1.40	0.20
Cogn. assistance	246	1.15	0.14	216	1.17	0.15	192	1.40	0.21
Effortful control
IBR	247	3.24	0.71	216	3.55b	0.85	192	3.46d	0.82
Mother report	234	3.91	0.57	220	4.29c	0.63	205	4.34	0.57
Caregiver report	157	4.42	0.78	143	4.68	0.73	149	4.61	0.67
Snack delaya	235	2.33	0.96	213	3.29	0.98	192	3.34e	0.90

Note. IBR = Infant Behavior Record; Cogn. = cognitive.

^aFor this task, observed data points were treated as missing values if experimenters’ instructions were very unclear and/or if the child had no understanding of the task (based on coders’ ratings). This occurred for six data points across three assessments.

^bSex difference: t(214) = −2.28, p < .05, Ms = 3.69 for girls, 3.43 for boys.

^cSex difference: t(218) = −2.35, p < .05, Ms = 4.40 for girls, 4.20 for boys.

^dSex difference: t(189.56) = −2.73, p < .01, Ms = 3.63 for girls, 3.32 for boys.

^eSex difference: t(184.02) = −3.71, p < .01, Ms = 3.59 for girls, 3.13 for boys.

Table 2.

Correlations Among EC Measures at Time 1 (T1), Time 2 (T2), and Time 3 (T3)

Variable	1	2	3	4	5	6	7	8	9	10	11	12
1. IBR T1	—
2. Mother T1	.10	—
3. Caregiver T1	.14	.17*	—
4. Delay T1	.04	.06	.22**	—
5. IBR T2	.21**	.01	.06	.09	—
6. Mother T2	.17*	.54**	.24**	.02	.18**	—
7. Caregiver T2	−.01	.15	.49**	.03	.04	.24**	—
8. Delay T2	.21**	.09	.22**	.02	.36**	.24**	.23**	—
9. IBR T3	.19*	.07	.17	−.06	.53**	.12	.09	.31**	—
10. Mother T3	.07	.38**	.23*	.05	.32**	.61**	.26**	.33**	.24**	—
11. Caregiver T3	.17*	.15	.38**	.07	.13	.28**	.51**	.13	.05	.30**	—
12. Delay T3	.20**	.01	.11	−.03	.20**	.18*	.08	.17*	.35**	.23**	−.03	—

Note. IBR = Infant Behavior Record.

^*p < .05.

^**p < .01.

Repeated measures general linear models indicated that mothers’ strategy use differed across age, Greenhouse-Geisser F(1.94, 359.55) = 79.35, F(1.96, 361.72) = 8.06, F(1.71, 316.17) = 138.50; ps < .001, partial η² = .30, .04, and .43 for directives, questions, and cognitive assistance, respectively. Trend analyses indicated that mothers’ strategy use changed in a linear manner with age, Fs(1, 185) = 80.47, 13.84, and 194.04, ps < .001, partial η² = .30, .07, and .51 for directives, questions, and cognitive assistance, respectively. Significant quadratic trends also were present for directives and cognitive assistance, Fs(1, 185) = 77.85 and 59.58, ps < .01, partial η² = .30 and .24, respectively. Pairwise comparisons indicated that T1, T2, and T3 directives all differed significantly (ps < .014 or lower), with T1 being the highest mean, followed by T3 (Ms = 1.52, 1.32, and 1.36 for T1 to T3). T1 maternal questioning was considerably lower than T2 or T3 questioning (ps < .001); T2 and T3 did not differ significantly from one another (Ms = 1.32, 1.38, and 1.39 for T1 to T3). Maternal cognitive assistance was substantially higher at T3 than at T1 or T2 ( ps < .001), whereas T1 was near-significantly lower than T2 (p = .057; Ms = 1.15, 1.18, and 1.40 for T1 to T3). Means sometimes differed slightly from those in Table 1 due to dropping data from families with data at only one assessment in a pairwise comparison.

Within and across time, we found no significant correlations between maternal strategies and children’s vocabulary. Within and across time, we found significant, positive correlations between vocabulary and mothers’ reports of EC (see Table 3). T2 vocabulary was positively related to T2 and T3 IBR reports of attention and persistence and EC during snack delay. T3 vocabulary also was positively related to T2 snack delay. Caregiver-reported EC was infrequently related to children’s vocabulary.

Table 3.

Correlations Between Vocabulary and Effortful Control (EC) Measures at Time 1 (T1), Time 2 (T2) and Time 3 (T3)

Index of EC	Vocabulary T1	Vocabulary T2	Vocabulary T3
IBR T1	.05	.05	.04
Mother T1	.16*	.22**	.15*
Caregiver T1	.06	.01	.06
Delay T1	.03	.04	.03
IBR T2	.12†	.25**	.12
Mother T2	.22**	.24**	.24**
Caregiver T2	.18*	.04	.09
Delay T2	.05	.27**	.18*
IBR T3	.04	.22**	.11
Mother T3	.16*	.20**	.12†
Caregiver T3	.08	.06	.07
Delay T3	.13†	.23**	.12

Note. IBR = Infant Behavior Record.

^†p < .10.

^*p < .05.

^**p < .01.

Correlations Among Measures of Maternal Strategies

Directives were significantly, positively related across time, as was questioning (see Table 4). Cognitive assistance at T1 and T2 were positively correlated, but T3 cognitive assistance was unrelated to earlier cognitive assistance. Within each time, modest negative correlations were found between questioning and directives at T1, T2, and T3 (at p < .06). Cognitive assistance at T1 was weakly positively related to T1 directives; T2 cognitive assistance was modestly positively related with T1 and T2 questioning; and T3 cognitive assistance was negatively related to T2 directives.

Table 4.

Correlations Among Maternal Teaching Strategies Measures at Time 1 (T1), Time 2 (T2), and Time 3 (T3)

Variable	1	2	3	4	5	6	7	8	9
1. Directives T1	—
2. Questioning T1	−.17**	—
3. Cogn. Assistance T1	.13*	−.03	—
4. Directives T2	.41**	−.17*	.00	—
5. Questioning T2	−.08	.35**	.07	−.24**	—
6. Cogn. Assistance T2	.12	.19**	.33**	.04	.16*	—
7. Directives T3	.32**	−.13	−.02	.40**	−.08	−.14	—
8. Questioning T3	.01	.19*	.13	−.05	.20**	.01	−.14	—
9. Cogn. Assistance T3	−.14	.08	.12	−.24**	.10	.10	−.02	.06	—

Note. Cogn. = cognitive.

^*p < .05.

^**p < .01.

Correlations Between EC and Maternal Teaching Strategies

Mothers’ questioning and cognitive assistance were sometimes positively related and directives were sometimes negatively related to observed and reported EC within time and sometimes across time (see Table 5). There were some across-time correlations from earlier EC to later maternal strategies as well as vice versa. Sex differences in these relations were found no more frequently than expected by chance.

Table 5.

Correlations Between Maternal Teaching Strategies and Effortful Control (EC) Measures at Time 1 (T1), Time 2 (T2) and Time 3 (T3)

Index of EC	Teaching strategy
	D T1	Q T1	C T1	D T2	Q T2	C T2	D T3	Q T3	C T3
IBR T1	.04	.15*	.22**	−.18**	.21**	.22**	−.02	.08	.12
Mother T1	−.03	.14*	.13	−.01	.06	.11	−.01	.06	.04
Caregiver T1	−.06	.15	.08	−.03	.10	.19*	−.08	−.04	.13
Delay T1	−.03	.08	.07	−.05	.05	.08	−.03	.04	−.05
IBR T2	.02	.17*	.17*	−.13	.29**	.15*	−.07	.07	.15*
Mother T2	−.17*	.19**	.17*	−.07	.12	.06	−.10	.07	.13
Caregiver T2	−.02	.03	.17*	−.06	−.01	.18*	−.12	.00	.06
Delay T2	−.05	.11	.12	−.17*	.21**	.19**	−.17*	.17*	.14
IBR T3	−.06	.10	.16*	−.20**	.20**	.12	−.20**	.00	.18*
Mother T3	−.04	.17*	.21**	−.07	.10	.10	−.11	.06	.22**
Caregiver T3	−.05	.10	.16	.01	.00	.15	−.11	−.05	.09
Delay T3	.03	.08	.14*	−.06	.15*	.07	−.07	.11	.17*

Note. IBR = Infant Behavior Record; D = directives; Q = questioning; C = cognitive assistance.

^*p < .05.

^**p < .01.

Confirmatory Factor Analyses (CFAs) and Structural Equation Models (SEMs)

The CFAs and SEMs were calculated using Mplus (Version 4.21; Muthén & Muthén, 1998–2007); full information maximum likelihood was used to handle missing data. Separate CFAs were initially computed for cognitive assistance and for the measures related to degree of maternal directiveness (i.e., directives and questioning in a single latent construct). Consistent with conceptual expectations and the modest within-time negative correlations between directives and questioning, we initially tried to use directives and questioning as two indicators of a latent construct. However, the models with questioning and directives on the same latent construct would not converge.

Therefore, three separate sets of CFAs and SEMs were run: one for mothers’ cognitive assistance, one for directives, and one for questioning. Each CFA model contained two constructs (mothers’ teaching strategies and children’s EC) at three time points. In all three models, IBR ratings (averaged across raters), mothers’ and caregivers’ reports of EC, and the delay task were used as indicators of children’s EC. Measurement errors of reported EC were correlated across time (within reporter) when indicated by the modification indices. Model fit was assessed with chi-squares, comparative fit indices (CFI), root-mean-square error of approximation (RMSEA), and standardized root-mean-square residual (SRMR).

In each CFA, we examined the invariance of loadings for indicators of caregiver-reported, mother-reported, and observed EC across time. Loadings are commonly constrained (when possible) to show that the constructs are equivalent across time; constraints are kept if they do not produce a significant reduction in fit (using chi-square difference tests). In the final constrained CFA model for maternal cognitive assistance, all loadings that were estimated could be constrained across time. In the constrained CFA models for directives and questioning, all constraints were implemented, except caregiver-rated EC at T1 could not be set equal to the analogous indicator of EC loadings at older ages. These CFAs fit well, χ²(77) = 86.05, ns, CFI = .980, RMSEA = .021, 90% CI [.00, .042], SRMR = .067, for cognitive assistance, with all estimated variables loading at p < .001; χ²(76) = 84.82, ns, CFI = .981, RMSEA = .021, 90% CI [.00, .042], SRMR = .063, for questions, with all estimated variables loading at p < .005 or better; χ²(76) = 97.36, p = .05, CFI = .957, RMSEA = .033, 90% CI [.00, .050], SRMR = .065, for directives, with all estimated variables loading at p < .014 or better.

In CFAs, only correlations among the latent constructs for EC and maternal strategies were tested. In the SEMs, we tested autoregressive paths for EC and maternal strategies, as well as concurrent relations between these constructs and bidirectional cross-lagged paths. In addition, covariates were added to the models. In the SEMs, we kept the same loading constraints used in the CFAs and added constraints on (a) the autoregressive paths for EC (or maternal teaching strategies) from T1 to T2 and from T2 to T3; and (b) correlations between the latent constructs for EC and strategies at each time point (because we had no reason to expect developmental differences in correlations between EC and teaching within each time). We did not constrain the cross-lagged paths because we expected that there might be developmental change in these paths. In addition, children’s vocabulary (assessed at each time) tended to correlate with children’s EC (but not mothers’ strategies) and was used as a covariate of EC at each time. Child sex and age at the laboratory assessment were used as covariates for all constructs at all times; findings for the covariates are described in text and are not included in the figures. In addition, SES at T1 was used as a covariate when it was significantly or near significantly correlated with a variable (see Figures 1, 2, and 3 for those paths; the marginal paths are nonsignificant in the figures). Finally, Mplus automatically correlated the scores for vocabulary with sex, age, and SES (these correlations are not included in the figures). Chi-square difference tests were computed to test if the constraints listed above could be implemented without significantly reducing the fit of the models. Chi-square difference tests comparing the constrained models (see below) with the unconstrained models were nonsignificant for the directive, questioning, and cognitive assistance models, Δχ²(9, 9, 9) = 10.90, 12.35, and 11.30, ns; thus, the constrained models (as described above) were kept.

Figure 1.

Loadings of study variables and longitudinal relations between mothers’ cognitive assistance teaching strategies and children’s effortful control (EC): structural equation model. The numbers above parentheses are unstandardized coefficient estimates; the numbers in parentheses are completely standardized coefficient estimates. All the model-estimated loadings for the indicators were significant at least at p < .01. Measurement errors were fixed at (1 − α)SD² (.005, .005, and .008 for cognitive indicators at T1, T2, and T3, respectively). Measurement errors for vocabulary checklist indicators were fixed at 0. Measurement errors were correlated within reporters for mothers’, caregivers’ and IBR reports (between T1–T2, T2–T3, and T1–T3 for the first two, and between T2–T3 for the third). Solid lines represent significant paths or correlations; dashed lines indicate nonsignificant paths or correlations. Age and sex were also used as covariates of EC and maternal teaching strategies but are not shown in the figure. C = cognitive assistance. * p < .05. ** p < .01. *** p < .001.

Figure 2.

Loadings of study variables and longitudinal relations between mothers’ directive teaching strategies and children’s effortful control (EC): structural equation model. Numbers before parentheses are unstandardized coefficient estimates; the numbers in parentheses are completely standardized coefficient estimates. All the model-estimated loadings for the indicators were significant at least at p < .01. Measurement errors were fixed at (1 − α)SD² (.006, .010, and .003 at T1, T2, and T3, respectively). Measurement errors for vocabulary checklist indicators were fixed at 0. Measurement errors were correlated for mothers’, caregivers’ and IBR reports (between T1–T2, T2–T3, and T1–T3 for the first two, and between T2–T3 for the third). Solid lines represent significant paths, dashed lines indicate nonsignificant paths or correlations. Age and sex were also used as covariates of EC and maternal teaching strategies but are not shown in the figure. D = directive. * p < .05. ** p < .01. *** p < .001.

Figure 3.

Loadings of study variables and longitudinal relations between mothers’ questioning teaching strategies and children’s EC: structural equation model. Numbers before parentheses are unstandardized coefficient estimates; the numbers in parentheses are completely standardized coefficient estimates. All the model-estimated loadings for the indicators were significant at least at p < .01. Measurement errors were fixed at (1 − α)SD² (.006, .004, and .005 at T1, T2, and T3, respectively). Measurement errors for vocabulary checklist indicators were fixed at 0. Measurement errors were correlated for mothers’, caregivers’ and IBR reports (between T1–T2, T2–T3, and T1–T3 for the first two, and between T2–T3 for the third). Solid lines represent significant paths or correlations; dashed lines indicate nonsignificant paths. Age and sex were also used as covariates of EC and maternal teaching strategies but are not shown in the figure. Q = questioning. * p < .05. ** p < .01. *** p < .001.

The three final constrained SEM models fit the data adequately: (a) for cognitive assistance, χ²(187) = 195.133, ns, CFI = .989, RMSEA = .013, 90% CI [.00, .030], SRMR = .064; (b) for directive strategies, χ²(186) = 217.748, ns, CFI = .958, RMSEA = .025, 90% CI [.000, .038], SRMR = .065; and (c) for questioning strategies, χ²(185) = 214.254, ns, CFI = .960, RMSEA = .024, 90% CI [.00, .038], SRMR = .066. See Figures 1 to 3 for unstandardized and completely standardized estimates of the loadings and path coefficients. In all three models, all factor loadings were significant and all autoregressive paths for EC were significant, indicating longitudinal stability of the construct. The autoregressive paths for maternal teaching were significant except for the model for cognitive assistance (despite a significant zero-order correlation across time from T1 to T2). In the model for cognitive assistance, EC at T1 (18 months) positively predicted cognitive assistance at T2 (30 months), and EC at T2 positively predicted maternal cognitive assistance at T3 (42 months). In the models for maternal directives, T1 EC negatively predicted T2 use of directives and T2 EC negatively predicted T3 mother directives. In the model for questioning, T1 EC positively predicted T2 maternal use of questioning. In all three models, T2 vocabulary was positively related to T2 EC (but T3 vocabulary did not relate to T3 EC); T1 vocabulary related to T1 EC only in the cognitive assistance model. SES was significantly positively associated with T1 EC in all three models; SES positively predicted T2 EC in the questioning and cognitive assistance models (and was near significant in the directives model); it also was positively related to T1 maternal questioning and negatively related to T1 directives in the respective models.

Although this is not indicated in the figures, sex and age also were related to T1 EC in all three models; girls and older children were higher in EC (unstandardized coefficients for sex ranged from .16 to .19, ps < .02 or better; unstandardized coefficients for age ranged from .13 to .14, ps < .02 or lower). In addition, sex and age were associated with T2 maternal cognitive assistance (unstandardized coefficients = −.06 and −.04, ps < .02 and .003); mothers used more cognitive assistance with boys and younger children at 30 months of age.

Discussion

Using a three-wave longitudinal model with autoregressive and bidirectional cross-lagged paths, we found evidence of not only some concurrent relations between children’s EC and mothers’ teaching strategies (especially for cognitive and questioning strategies) but also predictive relations across time. EC consistently predicted maternal teaching strategies across time rather than vice versa.

Measures of maternal cognitive assistance were positively correlated with some indices of children’s EC at all three assessments, and the correlation between EC and maternal cognitive assistance in the model was significant within time at 18 and at 42 months. Moreover, 18-month (T1) EC positively predicted 30-month (T2) maternal cognitive assistance in the SEM, and 30-month EC predicted frequent use of such strategies at 42 months (T3). Thus, the data are consistent with the conclusion that mothers’ use of cognitive assistance was higher across time if their children were well-regulated. Specifically, mothers tended to use more verbalizations that were likely to help the child understand the task or complete it effectively when children were better at regulating their attention and behavior. Judging from the path coefficients and their significance, the across time path was more substantial from T1 to T2 than from T2 to T3. Moreover, because mothers’ cognitive assistance and children’s EC were related within T3 even when accounting for stability of the constructs, the data suggest (but do not prove) that they affected one another concurrently (although the direction of effect is unclear), likely during the task session.

Somewhat similar findings were obtained for maternal directives, at least in terms of across-time relations. The measures of maternal directives were relatively infrequently negatively related to indices of EC in the zero-order correlations and the maternal directives construct was uncorrelated with EC within time at all three assessments in the SEM. However, there were some across-time correlations and in the SEM, T1 EC predicted lower T2 maternal directives, and T2 EC predicted lower T3 maternal directives. Thus, when children were well-regulated, their mothers were less likely to use directives a year later, even when controlling for prior levels of the constructs. Mothers of well-regulated children probably felt less need to use statements that were directive with their children. This finding, combined with the lack of within-time correlations in the model, suggests that mothers modulate their use of directives in accordance with their prior knowledge of their child’s regulatory capacities and do not vary their directives substantially as a function of the child’s regulation while doing the task.

In zero-order correlations, indices of maternal questioning were somewhat more frequently positively related to indices of EC than were directives, especially at T1 and T2, and high EC was associated with high maternal questioning within all three assessments in the SEM. Of most interest, analogous to the pattern for cognitive assistance, high EC at 18 months predicted high levels of maternal questions at 30 months. Thus, although it is likely that mothers varied their questioning as a function of their child’s EC when working on a given task (i.e., concurrently), mothers of 30-month-olds appeared to use more questions if their children were well-regulated a year prior. Perhaps the path from 30-month maternal questioning to 42-month EC was not significant because the level of maternal questioning did not change between T2 and T3, a finding suggesting stability in the relation between the two variables from T2 to T3. The fact that two of the four zero-order correlations between T2 maternal questioning and the four T3 measures of EC were significant, whereas the path from T2 questioning to T3 EC was not, is consistent with this explanation.

Thus, in all three models, there was evidence of child-driven effects. Effects were found at 18 months even though toddlers’ EC is not very well developed, suggesting that even emerging levels of EC can have an effect on adults in the child’s environment. These predictive relations were found when controlling for the level of EC and maternal behavior at the earlier assessment(s) and the relation between EC and maternal behavior at 18 months, a pattern that provides stronger (albeit not conclusive) evidence of a causal relation than simple across-time correlations. In addition, we controlled for the relations between vocabulary and EC in the models, as well as the effects of age, sex, and SES. To our knowledge, no one has used such a stringent panel model to test the relations of maternal teaching strategies to children’s functioning across time.

Kochanska, Tjebkes, and Forman (1998) found that young children who were oriented toward their mother during a teaching task—and likely more regulated—were those who also exhibited committed (willing, enthusiastic) compliance (which was related to self-regulation during a temptation task in the same study). Children with regulated attention, especially if they are compliant, likely require less direction than less regulated children and are good candidates for relatively subtle maternal techniques such as questioning and providing cognitive assistance. Because children with low EC have difficulty regulating attention, this deficit may evoke relatively controlling (directive) maternal verbal strategies and reduce mothers’ tendencies to promote children’s autonomy through the use of less directive strategies. The apparently evocative effects of EC are consistent with findings that mothers of children with cognitive deficits (e.g., Down syndrome) use more directive teaching methods (Landry et al., 1994) and with findings linking children’s age to maternal use of directives (e.g., Bellinger, 1979; Heckhausen, 1999).

The lack of any prediction of EC from parental teaching behaviors was surprising, given a number of significant correlations between the indices of maternal strategies and EC a year or two later. It is also surprising given findings suggesting that a warm, sensitive rather than negative parenting style predicts EC across time, even when controlling for earlier levels of EC (e.g., Eisenberg, Zhou, et al., 2005; Spinrad et al., 2007; Valiente et al., 2006), and other findings of relations between scaffolding-type parenting and positive outcomes for children (Fagot & Gauvain, 1997; Hoffman et al., 2006; Supplee et al., 2004). In contrast to mothers’ teaching strategies, mothers’ general style of parenting, including their sensitivity/warmth, may be more dependent on fairly stable and global parental beliefs and values rather than on their children’s competencies, with the consequence that paths from parenting behavior to children’s EC are more likely for measures of general parenting style than teaching strategies. Moreover, maternal parenting goals probably vary across different contexts (Bugental & Grusec, 2006). Because mothers likely have the goal of teaching their children specific ideas or skills in a teaching context and eliciting optimal performance, they may be more reactive to the capabilities of their children, including their EC, than in many other situations.

The high stability of children’s EC over time may have partly precluded prediction of EC from maternal teaching strategies in the models. However, 18-month maternal warmth/sensitivity was found to predict 30-month EC in the same sample, although 30-month maternal warmth/sensitivity did not provide additional prediction of 42-month EC (Eisenberg et al., in press; Spinrad et al., 2007). Thus, although the stability of EC probably reduced the probability of finding prediction from maternal teaching strategies in our study, there was enough unexplained variance in 30-month EC for a path from 18-month maternal strategies to 30-month EC to be significant. Moreover, controlling for SES, which was positively related to maternal questioning and children’s EC as well as negatively related to maternal directives, did not account for the lack of parent-driven findings because the findings were similar when SES was not controlled.

Of course, we also cannot rule out the possibility that maternal teaching strategies, especially simpler strategies, have effects on children’s EC prior to 18 months that are stable over time; moreover, effects of maternal teaching strategies could emerge after children are 42 months old. It is also possible that other strategies have a greater effect on children’s self-regulation. Perhaps maternal teaching strategies have more of an effect on children’s regulation in a specific context (e.g., a teaching situation) than across time; such effects may account for the within-time relations in the models between children’s EC and mothers’ strategies, even when controlling for prior levels of these variables (e.g., for relations of EC with T2 and T3 questioning and T3 cognitive assistance). Moreover, it is quite possible that maternal teaching strategies have effects on children’s cognitive competence over time but that the pattern of relations with maternal teaching strategies differs for children’s dispositionally based regulatory skills and their cognitive or social skills. Maternal teaching strategies may also have a stronger relation to children’s executive functioning skills involved in self-regulation than to skills contributing to delay of gratification; our measure of EC did not directly assess executive functioning (although adults’ reports of EC and the delay task would be expected to partly reflect children’s executive functioning). Nonetheless, our findings support the conclusion that mothers of young children adapt their teaching strategies to their child’s level of effortful control, suggesting that child characteristics play an important role in determining a child’s zone of proximal development and mothers’ attempts to teach therein.

It is of interest that maternal questioning and directive strategies were stable over time, whereas cognitive assistance strategies were not. Teaching strategies can be differentiated in regard to their form (as indicated by directives and questioning) or informational content (e.g., for cognitive assistance; see Gauvain et al., 2002). Perhaps the form of mothers’ strategies is more stable than the content, and mothers’ parenting views and personalities contribute more to the former. However, another possibility is that the information provided in a maternal strategy is more likely than the form of a strategy to vary as a function of mothers’ perceptions of their child’s capacities (even above and beyond their age). Discontinuity in mothers’ cognitive assistance might be especially marked during periods when children change rapidly in terms of relatively sophisticated regulatory and other task-related skills. Clearly, there is substantial change in children’s EC from ages 2 to 4.5 years (see Eisenberg et al., 2010; Rothbart & Bates, 2006).

In summary, our results suggest that the level of children’s EC may contribute to mothers’ teaching activities. Thus, mothers seem to be more reactive to their children’s abilities to attend and regulate their behavior than the reverse. Strengths of the study are its multimethod, multireporter, longitudinal design. Limitations are that EC could not be assessed in precisely the same way at each age (although the measures were similar and indicators of EC usually could be constrained across time). In addition, some of the measures of EC loaded fairly weakly, albeit significantly, on the EC latent construct; and it appears, on the basis of the coefficients and p values, that some of the across-time paths were modest (as were many of the significant zero-order correlations), albeit significant. In addition, for a small number of children, one of coders of the observational data also may have been one of the four raters of children’s persistence and attention (on the IBR) in the laboratory many months earlier, which could cause a slight degree of shared method variance across EC and measures of parenting. Moreover, our findings may not generalize to primarily lower SES samples or other cultures. In general, however, our findings suggest that it is important to consider the effects of children’s regulatory skills on mothers’ teaching strategies. The pattern of data also suggests that interventions with parents (and perhaps teachers) of young children might focus on making adults aware of their perceptions of children’s skills and if their teaching behavior is varying as a function of their child’s regulation rather than his or her cognitive and language competence. In future work, it would be beneficial to test if the direction of predictive relations varies across teaching and less structured mother-child interactions.