Currently, empirical findings of children's reading development indicate that reading achievement is multiply determined, predicted by such multiple interrelated factors as children's developmental characteristics (e.g., age, attention, and behavior), home environment, and classroom quality (e.g., Rayner et al., 2001; Connor et al., 2005; Petrill et al., 2007). Thus, ecological effects are important to understanding children's reading development. As such, our study was guided by an ecological developmental approach.

Bronfenbrenner and Morris (2006) contended that an ecological developmental approach is served by the four constructs triggering growth and change in academic competence, namely person, process, context, and time. The present study focused on three constructs: person, process, and context, all of which directly influence children's development. Person indicates child characteristics such as prior reading skills and family SES. Process represents the activities or interactions in which children are engaged in. Context indicates the environment and its associated learning opportunities (Ponitz et al., 2009). In this study, children's engagement represented process; classroom quality (quality of teacher–child interactions) was used as an indicator of context and children's SES and prior literacy skills were used as indicators of person. Time is captured indirectly by the longitudinal design of the study.

The ecological developmental approach emphasizes the importance of interrelationships among person, process, and context when studying children's development. Rather than examining individual contributions of each construct, the ecological framework suggests that researchers also attend to the interactive effects among these constructs (Bronfenbrenner, 1979). In fact, researchers have attempted to examine the interactive process between process and context in predicting children's learning. Specifically, context (teacher–child interactions) promotes child achievement indirectly through positive association with process (children's engagement; e.g., Guthrie et al., 2001; Greenwood et al., 2002; Perry et al., 2002; Ponitz et al., 2009). In the current study, we aimed to expand the existing literature by examining the individual influences of process (children's engagement) and context (teacher–child interactions) and how they interact. We also controlled for key person variables: children's SES and prior literacy skills. A brief review of classroom quality and children's engagement associated with child achievement follows.

The global quality of classroom is typically differentiated in terms of structural (i.e., teacher education; group sizes) versus process features (teacher–child interactions; Mashburn et al., 2008). In the present study, we restrict our focus on process features – teacher–child interactions most proximal to child learning. Teacher–child interactions indicate how teachers and students talk together, listen to each other, and generally interact verbally and non-verbally. These interactions, which take place on a daily basis, are thought to be the primary mechanism through which children learn (Hamre and Pianta, 2005, 2007; Mashburn et al., 2008). Two broad dimensions of teacher–child interactions emerge: teachers’ emotional support and instructional support, both of which have been linked to children's development theoretically and empirically. In general, emotional support encompasses the teachers’ sensitivity and responsiveness toward specific children, emotional warmth, and negativity (NICHD-ECCRN, 2002; Pianta et al., 2008a). Theories of motivation suggest that sensitive, responsive, and positive interactions between teachers and children could make children perceive teachers as more supportive and hence children are more motivated to learn (Deci and Ryan, 1985; Connell and Wellborn, 1991; Eccles, 1993). Likewise, children who have more motivation to learn generally show stronger academic achievement (Roeser et al., 2000; Gregory and Weinstein, 2004). Empirical evidence has suggested that teachers’ emotional support is a potentially important factor for predicting children's reading growth over time (e.g., NICHD-ECCRN, 2000; Rivik et al., 2000; Pianta et al., 2002; Connor et al., 2005).

Teachers’ instructional support refers to teacher's use of the classroom activities to effectively support children's academic development (Hamre and Pianta, 2005; Pianta et al., 2008b). The classrooms with higher levels of instructional support are characterized by frequent conversations, modeling of conceptual thinking, frequent feedback loops, and explicit discussion of language and literacy (e.g., Taylor et al., 1986; Snow et al., 1998; Makin, 2003; Meehan et al., 2003; Hamre and Pianta, 2005; Justice et al., 2008; Pianta and Hamre, 2009). These instructional inputs are important contributors to children's literacy achievement (e.g., NICHD-ECCRN, 2000; Justice et al., 2008).

In addition to the relation between classroom quality as indexed by teacher–child interactions and students’ skills, several studies have reported a link between teacher–child interactions and children's engagement (Ryan and Patrick, 2001; Assor et al., 2002; Hughes and Kwok, 2007). For instance, Ryan and Patrick (2001) found that the social environment of eighth-grade classrooms was significantly associated with changes in engagement when students moved from seventh to eighth grade. Hughes and Kwok (2007) recently also reported that the quality of teacher–child interactions was positively correlated with first-grade children's engagement. The documented relationship between teacher–child interactions and higher level of child engagement was not unexpected. Many researchers propose that children who receive high-quality support from teachers are more engaged in that they study harder, attend more to teachers, and confirm to classroom rules and routines (e.g., Wentzel, 1999; Ridley et al., 2000; Little and Kobak, 2003).

Researchers have operationalized engagement as a multidimensional concept in terms of behavioral, cognitive, and motivational engagement (Linnenbrink and Pintrich, 2003). Given that many studies supported the significant association between behavioral engagement and academic achievement for elementary, middle, and high school students (e.g., Skinner et al., 1990; Marks, 2000), this study conceptualized engagement as observable behavior. Although behavior engagement has been defined differently by different studies, most often behavioral engagement has been defined as involvement in learning activities and includes behaviors such as attention, persistence, and concentration (e.g., Birch and Ladd, 1997; Fredericks et al., 2004). Among these behaviors, attention which represents the extent to which children are observed to attend to learning activities, appeared to be critically important in promoting children's learning as a consistent, significant relationship was reported between the level of attention and children's reading or math achievement (e.g., Peterson et al., 1984; Wasson et al., 1990). In addition, recent research has placed an emphasis on another child behavior related to engagement – self-reliance, which “reflects the degree to which the child displays autonomy, self-regulation, and personal initiative in the classroom” (Rimm-Kaufman et al., 2002, p. 459). Children who are self-reliant are more likely to exhibit behavior that promotes learning and demonstrate more reading skills (Rimm-Kaufman et al., 2002; Ponitz et al., 2009). Thus, the current study used the level of children's attention and self-reliance to represent the construct of children's engagement.

Educational researchers have suggested that, compared with those who are not engaged, engaged children “are thought to be more intensively and extensively involved – behaviorally, intellectually, and emotionally – in their learning” (Bangert-Drowns and Pyke, 2001, p. 23). From this perspective, children's engagement contributes to their social and cognitive achievement (Newmann, 1992). In fact, the importance of student engagement is supported in empirical work. For example, Singh et al. (2002) found that eighth-grade students’ engagement, defined as academic time spent in math and science, was a strong predictor of their math and science achievement. In terms of reading skills, Campbell et al. (1997) reported a significant association between engagement and reading achievement in a national sample of elementary school students. Wigfield et al. (2008) also found a significant association between fourth-grade students’ reading engagement and their reading achievement. Similarly, middle school students with higher level of reading engagement had higher reading achievement than those students who were lowly engaged (Guthrie and Wigfield, 2000). The studies referenced above have supported the view that engaged learners are motivated to use cognitive strategies and interact in a classroom community (Guthrie, 1996). These behaviors that children displayed are all associated with their achievement.

The importance of children's engagement in reading achievement reflects social-interactionist theories of how children acquire reading. Social-interactionist theories view reading acquisition as occurring within the “social context of discourse, in the miniaturized culture that governs the communicative interaction of children and adults” (Bruner, 1981, p. 175). Such a perspective stresses the role of child's active involvement in the literacy acquisition process in fostering growth.

Although in the previous sections we have considered classroom quality and children's engagement separately, it is possible that children's engagement may mediate or interact with the quality of teacher instruction in the classroom to impact achievement (Guthrie et al., 2001; Greenwood et al., 2002; Perry et al., 2002; Ponitz et al., 2009). By way of illustration, consider a case in which children have difficulty in being engaged in learning. In such instances, teachers might interact with them in sensitive, responsive, and positive ways. Such children would be more likely to feel more support from teachers and hence be more engaged in learning. High levels of children's engagement may improve their academic achievement. Therefore, social interactions with teachers within the school setting may serve as a protective factor for children who are weakly engaged in learning.

In fact, some empirical studies have shown that children's engagement mediated the effects of classroom instruction on children's reading skills (e.g., Wigfield et al., 2008; Ponitz et al., 2009). For example, Ponitz et al. (2009) found that kindergartener's behavior engagement significantly mediated the relation between classroom quality and their reading achievement. Wigfield et al. (2008) also found that reading instruction improved fourth-grade children's reading achievement through increasing engaged reading. Such findings are important because they help explain how classroom quality increased children's reading skills and highlighted the importance of children's engagement. However, to our knowledge, only these two studies have attempted to determine whether the effects of classroom quality on children's learning were attributable to children's engagement in instruction. Therefore, the primary focus of the present study was to examine to what extent the effects of classroom quality on children's reading skills were mediated by the children's level of engagement. The current study focused on third-grade children because academic achievement during middle childhood (from Grade 1 to Grade 5) is critical to a successful developmental trajectory through this period and into adolescent (Eccles et al., 2006). Furthermore, much of high-stakes accountability assessment starts in third grade (Pianta et al., 2008a). It seems imperative to understand the underlying mechanism of classroom quality and children's engagement contributing to reading achievement in third grade.

The findings reviewed above provide evidence of a growing body of literature suggesting the importance of both classroom quality and children's engagement as sources of influence on student achievement. In the present study, our goal was to examine the relations among children's engagement, classroom quality, and students’ third-grade reading, controlling for family SES, and students’ Grade 1 reading ability. Our conceptual model (see Figure 1) allows us to examine the influences of classroom quality and student engagement simultaneously, while investigating the unique contribution of each to students’ reading skills. Specially, we asked three research questions to guide our investigations:

Children included in this study took part in the Phase III study of the larger ongoing longitudinal NICHD Study of Early Child Care and Youth Development (SECCYD; please see http://secc.rti.org/ for additional information about the NICHD study). Children participating in NICHD SECCYD are followed and their developments are assessed at frequent intervals from birth through adolescence.

Of the students and classrooms, the present study included 1,364 students and their teachers. Given that NICHD SECCYD is a longitudinal study, some students and their teachers did not provide complete data on all variables in this study. We used the full information maximum likelihood method in AMOS 18.0 (Arbuckle, 2006) to deal with the missing data. Of all the participants, 24% of the children were ethnic minority group and 11% of the mothers had less than a high school education. Mean family income was 3.6 times the poverty threshold. Almost half of the children were male (49%). The children participating in NICHD SECCYD came from 10 locations in the United States (Little Rock, AR; Irvine, CA; Lawrence, KS; Boston, MA; Philadelphia, PA; Pittsburgh, PA; Charlottesville, VA; Morganton, NC; Seattle, WA; and Madison, WI). For each classroom, only one child was selected to participate in study. If more than one child within the same classroom participating in the study, one child was randomly selected for classroom-level data analysis.

Structural equation modeling (SEM) was used as our data analytic strategy. We employed several criteria to assess the fit for our model including x² which is ideally non-significant, indicating a good fit of the model. However, the value of x² was sensitive to large sample size (Marsh, 1994). Thus other statistics were also used to assess the model fit including the Tucker Lewis index (TLI), comparative fit index (CFI), and the root mean squared error of approximation (RMSEA). A cut-off value of TLI and CFI should be 0.90 or greater, indicating a close fit (Hu and Bentler, 1999). The value of RMSEA should be around 0.05, representing a close fit (Brown and Cudeck, 1993). Last, the Sobel test (Preacher and Leonardelli, 2001) was used to test whether indirect effects were significant or not because the Sobel test was considered as the strongest compared with other 14 methods for testing mediation effects (MacKinnon et al., 2002).

We hypothesized that classroom quality would impact children's reading outcomes indirectly through children's engagement. We tested this hypothesis first by comparing two models. The first model did not include direct paths from Grade 3 classroom quality to students’ reading achievement when they are in Grade 3 (see Figure 2). The second model included the direct paths from Grade 3 classroom quality to students’ reading achievement when they are in Grade 3. Both models fit the data reasonably well. The first model represented in Figure 2 has a x² of 284.22, df = 34, CFI = 0.96, TLI = 0.92, and RMSEA = 0.07. The second model has x² of 281.61, df = 33, CFI = 0.96, TLI = 0.92, and RMSEA = 0.07. As the first model was nested within the second model, we thus compared their relative fit by performing a x² difference test. Results showed that the first model did not fit significantly worse than the second model (x² difference = 2.56, df = 1, p = 0.11), suggesting classroom quality indirectly impacted student reading outcomes when they were in Grade 3. The first model represented in Figure 2 was used in the present study. Parameter estimates for both models are summarized in Table 3. The path model (the first model) is provided in Figure 2. In both instances, children's SES was a significant predictor of third grade engagement and reading outcomes. First-grade reading skills significantly predicted third-grade engagement and reading skills.

First, we examined the mediated effect of Grade 3 classroom quality through Grade 3 children's engagement as they affect the reading skills of third-graders (classroom quality → children's engagement → children's reading outcomes). Sobel test results demonstrated that the indirect effect from classroom quality to children's reading achievement via children's engagement was significant (unstandardized path coefficient = 0.79, p < 0.001). Multiplying the two relevant paths (see Figure 2) provided the size of the indirect effect of classroom quality on Grade 3 reading skills via children's engagement: 0.49 × 0.10 = 0.05.

Second, we examined the direct effect of Grade 3 classroom quality on children's engagement. As the direct path of Grade 3 classroom quality to Grade 3 reading achievement is not significant, we did not report the direct effect of Grade 3 classroom quality on children's reading achievement. Our results showed that classrooms with high-quality had children who were more engaged (standardized path coefficient = 0.49, p < 0.01, see Table 3 and Figure 2), after controlling for children's SES and their reading skills when they are in Grade 1.

We found that Grade 3 engagement was a significant predictor of their reading achievement (standardized path coefficient = 0.10, p < 0.001, see Table 3 and Figure 2), after controlling for children's SES and their reading skills when they are in Grade 1. This result indicated that students who were more engaged tended to demonstrate higher reading skills. In addition, we also found evidence concerning the reciprocal relation between children's engagement and their reading skills. As shown in Table 3 and Figure 2, Grade 1 reading skill significantly and directly predicted Grade 3 engagement (standardized path coefficient = 0.14, p < 0.001).

A number of limitations and areas for future investigation warrant note. First, similar to any correlational study, we cannot assume that the observed relations among classroom quality, children's engagement, and children's reading achievement are causal in nature. In addition, our analysis did not test the reciprocity effects of classroom quality and children's engagement on children's reading skills. Future investigation should focus on experimental studies to identify any causal and reciprocity effects. Second, the results of this study are limited by children's engagement measures in that the metric used only assessed children's behavioral engagement. However, there are other types of children's engagement, namely cognitive engagement (e.g., strategy use, metacognition) and motivational engagement (e.g., interest, value, affect). Another necessary component to represent the multidimensional nature of children's engagement is disaffection, reflecting the core behaviors of disengagement such as lack of initiation or effort (Skinner et al., 2008). Given the importance of cognitive and motivational engagement and disaffection in student learning (Pintrich and Schrauben, 1992; Eccles et al., 1998; Skinner et al., 2008), future studies should broaden to include children's cognitive and motivational engagement and disaffection. Third, the measure of classroom quality in the current study was also potentially biased, in the sense that it used quality ratings of teacher–child interactions, based on classroom observation conducted at one point in time. Future investigation may employ multiple methods to capture classroom quality such as observation, self-reported survey, and quantitative analysis from videos.

Additionally, amounts, types, and quality of reading instruction were not included in the model. It is possible that both quality of the classroom environment and amounts and types of instruction both contribute, directly and indirectly, to children's reading achievement in third grade. There is accumulating evidence that appropriate and evidence-based reading achievement contributes to reading skill gains (NRP, 2000). Thus, research efforts designed to study types and quality of reading instruction that influence children's engagement and reading skills are important future lines of inquiry.