Exploring the relationship between student approaches to learning and reading achievement at the school level

Abstract

Using the Early Childhood Longitudinal Study, Kindergarten Cohort of 1998 (ECLS-K, 98), a nationally representative sample of kindergarteners in the United States, we investigated the relationship between approaches to learning and spring reading achievement with particular emphasis on classroom and school-level differences. We employed hierarchical linear modelling to investigate classroom- and school-level influences on children’s spring reading achievement and their teacher-rated approaches to learning. Results indicate that classroom-level variables are particularly important predictors of spring reading in kindergarten and also relate to the strength of the relationship between teacher-rated approaches to learning and spring reading achievement. Specifically, in classrooms with a high frequency of reading activities, there is a stronger relationship between teachers’ ratings of approaches to learning and spring reading achievement than in classrooms with a low frequency of reading activities. Also, the relation between teachers’ ratings of approaches to learning and achievement appears to be stronger in schools with high enrolment, as compared to schools with low enrolment. Findings suggest that the frequency of reading activities in the classroom is not only important for students’ reading achievement, but that these activities are also related to the important relationship between approaches to learning and reading achievement.

Keywords

Reading kindergarten classroom practice school approaches to learning

Introduction

Children’s engagement in learning, including their persistence in tasks, attentiveness and eagerness, is indicative of later school achievement (George and Greenfield, 2005). Further, early attitudes to learning are important predictors of children’s achievement throughout elementary school (Li-Grining et al. 2010; Welsh et al. 2010; Xue and Meisels, 2004). Taken together, students’ orientations toward engagement in academic tasks have been termed approaches to learning (Biggs, 1987; Fantuzzo et al. 2004).

The learning context has been identified as key to determining how students’ approaches to learning, or academically oriented attitudes and behaviours, may manifest (e.g. Biggs, 1987; Denton and West, 2002). Despite the fact that these contextual factors have been identified as key to students’ academic achievement, the ways in which classroom and school factors influence students’ approaches to learning have been under-examined in the literature. To explore how young children’s approaches to learning are related to their reading achievement across classroom and school contexts, we used a nationally representative sample of kindergarten students in the United States (ECLS-K: 1998–99) to answer our research questions. Specifically, due to the importance of contextual influences for students’ approaches to learning and achievement (Dart et al. 2000; Denton and West, 2002), we examined how school composition and the frequency of instructional practices used in the classroom relate to the relationship between approaches to learning and reading achievement. We also examined the more general influence of school composition and classroom practices on the between-school variation in spring reading achievement for kindergarten students.

To explore these relations, we employed hierarchical linear modelling (HLM; Raudenbush and Bryk, 2002) to examine both individual-level and context-level influences on spring reading achievement and the relationship between approaches to learning and spring reading achievement in kindergarten. See Figure 1 for the full conceptual model that guided the current study.

Figure 1.

Conceptual model.

Approaches to learning

Approaches to learning have been defined as a constellation of students’ attitudes and behaviours, including attentiveness, eagerness, flexibility, persistence, independence and organization (Claessens et al. 2008), which are indicative of students’ engagement with academic tasks. Fantuzzo et al. (2004) identified four dimensions underlying the construct of approaches to learning: (a) competence motivation, (b) attention/persistence, (c) strategy/flexibility and (d) attitude toward learning. The competence motivation dimension reflects students’ curiosity about and motivation for academic tasks; the strategy/flexibility dimension refers to children’s problem solving capacities; and finally, the attitude toward learning dimension refers to students’ engagement with adults and peers in task completion.

In assessing approaches to learning, Xue and Miesels (2004), asked teachers to report the frequency with which students’ demonstrated behaviours in each of the following categories, (a) attentiveness, (b) task persistence, (c) eagerness to learn, (d) learning independence, (e) flexibility and (f) organization (Xue and Meisels, 2004), on the Social Rating Scale (SRS). Similarly, the United States’ Head Start Child Development Early Learning Framework (2010) defines approaches to learning as students’ initiative, curiosity, persistence, attentiveness and cooperation (US Department of Health and Human Services (DHHS), 2010). Thus, approaches to learning can be conceptualized as a child’s fit within the instructional, organization and social context of the classroom.

There is a growing interest in the relationship between approaches to learning and young children’s achievement (Chen and McNamee, 2011; Li-Grining et al. 2010; McDermott, 1984; McDermott and Beitman, 1984). Specifically, the National Task Force on School Readiness (1991) cited approaches to learning as a key factor influencing student achievement and learning throughout the course of schooling. In the Head Start Child Development and Early Learning Framework (US Department of Health and Human Services, 2010), approaches to learning were highlighted as an important and unique domain in which children should develop, independent from academic and social domains. Further, approaches to learning have been determined to be a particularly important factor to consider in young students, due to their malleability and capacity to be developed within the classroom context (Barnett et al. 1996; George and Greenfield, 2005).

Approaches to learning have also been identified as an important predictor of achievement in early schooling (DiPerna et al. 2002; Schaefer and McDermott, 1999), a relationship that appears to persist throughout the elementary school years. Claessens et al. (2008) found that teacher-rated approaches to learning measured in the autumn of students’ kindergarten year predicted their achievement in fifth grade. DiPerna et al. (2007) similarly found that approaches to learning in kindergarten predicted growth in maths achievement from kindergarten through to third grade. Children who displayed higher approaches to learning in kindergarten had steeper growth in their maths achievement scores over the first four years of elementary school. Li-Grining et al. (2010) replicated these findings through to fifth grade, and demonstrated the same relationship in reading as well as maths. A recent study by Welsh et al. (2010) similarly showed that approaches to learning skills during pre-kindergarten predicted reading and maths achievement in kindergarten, suggesting the importance of these skills for achievement in a formal schooling context (Welsh et al. 2010).

Ready et al. (2005), examining the relation between approaches to learning and reading achievement in kindergarten, determined that girls entered kindergarten with higher teacher-rated attentiveness and task persistence, and this, in large part, accounted for the fact that girls had higher classroom reading scores at the beginning of the school year and greater growth in reading skills from the beginning to the end of kindergarten. Li-Grining et al. (2010) found that variability in approaches to learning resulted in differences in young children’s ability to cope with the demands of formal schooling, and resulted in variations in learning trajectories and reading achievement. Thus, prior research has established that approaches to learning are related to both initial reading achievement and the growth of reading scores over time.

Similarly, children’s attention, another central dimension of approaches to learning, is necessary for learners to be able to focus on relevant stimuli in the environment, ignore irrelevant distractions and to persist in tasks, even in the face of challenge (Fantuzzo et al. 2004). Longitudinal studies following children with attentional issues find that these students are more likely to have reading problems and lower achievement several years later (Rabiner and Coie, 2000; Spira et al. 2005). Thus, in considering possible difficulties in reading achievement, it is essential to consider these aspects of approaches to learning (Biggs, 1987; Denton and West, 2002) as affecting achievement within this domain. More work is needed to understand the particular relation between approaches to learning and reading achievement and how these relations may vary across school contexts.

Assessing approaches to learning

In the current literature, young children’s approaches to learning are based on adult ratings of their behaviours and attitudes (George and Greenfield, 2005). Teachers are typically asked to rate approaches to learning-related items either as part of another measure or via a stand-alone instrument (George and Greenfield, 2005). However, teacher ratings of this variable are not without bias. Studies have shown that teachers may rate children’s knowledge and attitudes differently depending on their student and family characteristics (i.e. family socioeconomic status, race, gender; Auwater and Arguete, 2008; Lynch, 2010). As such, child demographic variables are important individual-level factors to consider when examining the relationship between teacher-rated approaches to learning and student achievement. The current study controls for socioeconomic status (SES) at the individual student level and minority status and overall SES at the school level to account for these potentially biased ratings.

Kindergarten in the US

Before reviewing the current literature on school-level contributions to achievement, it is important first to situate the context of kindergarten in the United States. Although more common presently, full-day kindergarten (as opposed to half-day or part-day) is not universal. In 2003, 65% of all kindergarteners in the United States were enrolled in full-day programmes and the rate of full-day kindergarten varied by region of the country, race/ethnicity of students and income of families (Wirt et al. 2004). In 2010, only nine states required that all children enrol in full-day programmes and only 16 states required that students attend kindergarten at all (USDE, 2010b).

Children in low-income communities and those from minority households are more likely to be enrolled in full-day kindergarten programmes compared to other students (Lee et al. 2006; Wirt et al. 2004). Federal programmes, such as Title I, a provision of the Elementary and Secondary Education Act of 1965, provide additional funds to school districts with a high percentage of low-income students to supplement school day programmes, which is one reason why more low-income students attend full-day kindergarten (USDE, 2010b).

Classroom and school context

Students’ early approaches to learning have consistently been shown to be related to students’ reading achievement (Chen and McNamee, 2011; Finn and Pannozzo, 2004; Xue and Miesels, 2004). It is therefore essential to examine whether the relationship between these variables differs across schools, and if so, which contextual factors explain the variation in this relationship. Xue and Meisels (2004) used multi-level modelling to examine teacher-level predictors of students’ approaches to learning in kindergarten in the Early Childhood Longitudinal Study, Kindergarten Cohort (ECLS-K). The researchers found that teachers’ instruction in reading, specifically using an integrated language arts approach as opposed to phonics instruction, predicted children’s approaches to learning. Additionally, teachers’ experience, average class reading ability and being enrolled in a full-day kindergarten classroom were significant in predicting children’s approaches to learning in the full ECLS-K sample. This study stresses the importance of children’s approaches to learning attitudes upon entry to kindergarten and also highlights critical within-classroom practices and student-level factors that are related to variations in approaches to learning.

The current study considered the frequency of reading instruction rather than the overall instructional approach teachers used in the classroom. There were a variety of reasons why we chose to examine classroom instruction in this way. First, by using the frequency of instructional practices, our variables and subsequent factor had more variation than if we had chosen to use a dichotomous presence/absence indicator of particular instructional practices. Second, using an instructional approach variable would not have allowed us to discern if the practice happened once a week versus daily, as it is likely the frequency of implementation has a differential influence on children’s achievement. By using frequency of instructional practices, we were able to examine not only whether certain practices were used within the classroom, but also the frequency with which they were used. Finally, extant research suggests that the frequency of certain classroom practices is associated with approaches to learning in kindergarten and we wanted to investigate these relationships in our model. By assessing the frequency of instructional practices, as compared to a dichotomous rating of presence or absence of these same practices, we believed our variable was able more accurately to reflect students’ experiences in the classroom.

In one study, examining the frequency of classroom activities and approaches to learning, Chen and McNamee (2011) found that daily activities, such as reading, drawing, playing games and dancing, are related to approaches to learning in pre-school children. The researchers stress the importance of providing children with a variety of activities to enhance their approaches to learning. They also posit that teachers play a critical role in enhancing approaches to learning by supporting children and adjusting activities to keep their interest (Chen and McNamee, 2011), pointing to the need to examine classroom-level factors in order fully to understand students’ approaches to learning. Both Chen and McNamee (2011) and Xue and Miesels (2004) illustrate the importance of instructional-level factors in predicting approaches to learning; they do not, however, simultaneously examine the relationship between classroom instruction, approaches to learning and reading achievement.

As such, it is important to understand what mechanisms within the school support the relationship between approaches to learning and child outcomes. As demonstrated in prior research, demographic (e.g. school SES), and process-level variables (e.g. frequency of instructional practice) are important to consider in these types of analyses and will be considered in the present investigation, in conjunction with one another. A hierarchical framework is most appropriate for examining the influence of both these classroom- and school-level factors, while taking student-level characteristics into account. Considering the myriad of factors impacting on the relation between approaches to learning and students’ reading achievement, there is a need to examine the nature of this relationship in more nuanced ways. Beyond establishing the existence of such a relationship, we were interested in investigating how this relation might vary across class and school contexts. Multi-level methods were used to better understand the differences in the relations between approaches to learning and reading achievement.

Structural indicators of the school day and its relationship to approaches to learning have been investigated in prior studies. Finn and Pannozzo (2004) found that the length of the programme (full-day or half-day), the number of aides in the classroom and school enrolment were all related to children’s spring kindergarten approaches to learning. Children in full-day programmes demonstrated higher teacher reported approaches to learning. Further, larger schools contained children with lower teacher-rated approaches to learning overall (Finn and Pannozzo, 2004). This study demonstrates the overall importance of classroom and school-level variables for approaches to learning, but does not consider the effect approaches to learning may have on reading achievement.

Yan and Lin (2005) used the same data set employed for the current study to demonstrate that children who are enrolled in full-day programmes make more gains in reading than children in half-day programmes, controlling for a variety of demographic variables, including race, family composition, poverty status and parent education. Similarly, Baskett et al. (2005) documented the increases in students’ achievement associated with full-day kindergarten attendance. These studies illustrate the importance of classroom- and school-level variables in explaining the variation in reading achievement in kindergarten, but do not include approaches to learning. The current study builds on prior research by examining predictors of both reading achievement and the relation between approaches to learning and reading achievement.

Demographic school-level characteristics may also contribute to the between-school variance that is shown in the relationship between approaches to learning and student achievement. Lee et al. (2006) demonstrated that school-level demographic differences, including school location and size, contribute to differential achievement of students within schools during kindergarten. For example, although full-day kindergarten appears to be beneficial for children’s spring reading achievement, and those children who attend full-day kindergarten tend to show the most growth in their reading scores, students in low-income areas, who are more at risk of school failure, were more likely to have full-day kindergarten offered in their school. Therefore the growth demonstrated may be a product of these children starting behind and making stronger gains in reading, and these gains may be a function of their initial disadvantage. This is why school-level variables are essential to understand the complete picture of why students’ approaches to learning may vary across schools.

It seems possible that there may be a different relationship between students’ approaches to learning and their achievement, depending on school composition. For example, at an “at-risk” school, with a higher proportion of low-income students, students with weaker approaches to learning may perform much worse on a measures of reading achievement than students with similarly negative approaches to learning who attend more advantaged schools. The specific characteristics of schools typically associated with “at-risk” included a low aggregate SES of the students and the enrolment of a high percentage of minority students (Curley, 2005). Additionally, following from the Lee et al. (2006) study, we were interested in examining the influence of contextual level variables, such as the availability of full- or half-day kindergarten, and the size of the school, on the relation between students’ approaches to learning and reading achievement, particularly as factors related to school-readiness, such as approaches to learning, may be particularly important for the success of at-risk children (George and Greenfield, 2005).

Although the time spent on instruction in full-day kindergarten is not double that of half-day kindergarten, there is certainly more time available for teachers to work with students on academic skills. The benefit of this extra instructional time may lead to increases in end-of-year achievement. This is evidenced by McCoach et al. (2006), who found that kindergarten and first grade reading achievement was predicted by the amount of instruction present in the kindergarten classroom. Similarly, smaller schools may allow more time for one-on-one instruction and children may feel less overwhelmed in a smaller school context, especially if they have limited prior experience of formal classroom work. Magnuson et al. (2007) characterized classroom contexts as differing not only in the time devoted to instruction (frequency of behaviours), but also in the type of instruction offered to students, pointing to the need to examine the effects of both structural characteristics of classes and classroom practices on students’ achievement. In the current study we were able to examine these characteristics in two ways: by creating a factor that represented the frequency of reading activities in each classroom and by using HLM to examine variation between classrooms. Studies that have used HLM find that practices within the classroom may compensate for disadvantages students face outside of school (Bennett et al. 2005). For example, full-day kindergarten has been found to be particularly beneficial in producing gains in reading comprehension and vocabulary, as compared to half-day kindergarten (Fusaro and Royce, 1995). Although aspects of the model of interest in the current study have been explored previously, we sought to assess the extent to which frequency of instructional practices affects both reading achievement in general and the relationship between approaches to learning and reading achievement – the combination of which has not been adequately addressed in prior research.

Instruction, family background and reading achievement

Finally, teacher-specific instructional practices are also related to children’s reading outcomes in kindergarten. Magnuson et al. (2007) found that although family background is predictive of achievement, children who are in classrooms with a rigorous instructional climate during kindergarten exhibit greater gains in reading, as compared to those in classrooms with lower levels of instruction. Classrooms associated with a high-level instructional climate were small in size, allowing for more individualized support, more teacher-student interactions and fewer teacher-directed activities (Magnuson et al. 2007). Conversely, McCoach et al. (2006) found that the amount of reading instruction that kindergarteners received weakly predicted their reading ability at the end of kindergarten and first grade. However, family background characteristics, including family income and race, were much stronger predictors of reading at the end of first grade (McCoach et al., 2006). It is clear that classroom instructional factors alone do not account for the variations in students’ reading achievement. Family background and approaches to learning are also important to consider, but have not been included in prior research examining how reading achievement varies at both individual and school level.

Additionally, the relation between approaches to learning and reading achievement may be particularly pertinent to examine within low-income, high-poverty schools, wherein students may exhibit more behavioral needs in the classroom and heightened instances of reading difficulties, making a predisposition to engage in learning-related activities particularly important (Bulotsky-Shearer and Fantuzzo, 2011). Therefore, this study will add an important piece to the research on reading achievement in young children by examining how both classroom-level and school-level factors not only influence reading achievement, but also how these factors influence the relationship between approaches to learning and reading achievement. Specifically, this study will address the following research questions:

How do children’s approaches to learning relate to reading achievement in the spring and to what extent does this relationship vary across schools?

How do frequency of reading activities in the classroom, half- or full-day kindergarten attendance and school-level demographic factors affect children’s reading achievement in the spring of kindergarten?

How do frequency of reading activities in the classroom, half- or full-day kindergarten attendance and school-level demographic factors affect the variation in the relationship between children’s approaches to learning and children’s reading achievement in the spring?

Method

The sample was drawn from the first two waves of the 1998–1999 Early Childhood Longitudinal Study, Kindergarten Cohort, (ECLS-K), which included assessments during the autumn of kindergarten when students have just begun formal schooling and the spring of kindergarten, approximately six to seven months following the autumnal assessment. Only students with complete data for both time points were included in these analyses. This is a subsample of the larger ECLS-K data set. The final subsample used in our analyses included 3,034 students in 184 schools out of the original subsample of 3,309 students in 200 schools. Several schools were missing data on one of the school-level factors of interest and HLM requires no missing data at level 2. Thus, the sample was further reduced due to this. More information on these missing schools and the students within them is supplied in the results section. Appropriate sampling weights were used at both student level and school level in all the analyses to account for the complex sampling structure of this data set.

The ECLS-K data set is the first nationally representative data set of its kind to sample kindergarten students and follow them through middle school. Children from the ECLS-K are representative of children entering kindergarten in the autumn of 1998. This includes children in public and private schools and those from a variety of ethnic, racial and socioeconomic backgrounds. Children were followed from kindergarten through to the spring of eighth grade in 2007. This paper only examines data from the first two data collection waves, occurring in the autumn of 1998 and the spring of 1999 (USDE, 2012)

Outcome variable

Due to our interest in students’ reading achievement, we selected a standardized measure of reading achievement in the spring of students’ kindergarten year as the outcome. The Common Core Standards for Kindergarten foundational reading skills include print understanding, understanding of spoken words, syllables and sounds, and phonics-based decoding skills (National Governors Association, 2010). Similarly, Ponitz et al. (2009) adopted a definition of reading achievement in kindergarten, consistent with the National Reading Panel (2000), including both phonological and communicative language skills.

Accordingly, the topics for the reading assessment in the ECLS-K include: Basic Skills; Vocabulary; Initial Understanding; Developing Interpretation; Personal Reflection and Response; and Demonstrating a Critical Stance. The first topic, Basic Skills, covers children’s print knowledge and identification of letters. This topic represented 40% of the questions on the reading assessment. The Initial Understanding topic required children to describe the main point of a passage and represented 10% of the questions on the assessment. The Developing Interpretation topic asked students to provide a deeper understanding of what they had read and represented 25% of the questions on this assessment. Personal Reflection and Response asked children to relate what they had read to their own lives. This topic represented 10% of the questions on this assessment. Demonstrating a Critical Stance asked students to reflect on what they had read objectively and represented 5% of the questions on this assessment. The final 10% of questions on this assessment covered children’s vocabulary knowledge (USDE NCES, 2002).

Because children were not tested on every item in the reading assessment (they were routed to the appropriate level of difficulty in the assessment based on a routing section), item response theory (IRT) was used to predict children’s scores as if they were given all of the test assessment items. Students’ performance on individual items is not available so we chose to use the IRT scale score provided in the ECLS-K data file due to the interpretability of these scores and their use in several of the studies reviewed in the introduction (e.g. Ready et al. 2005; Xue and Meisels, 2004; Yan and Lin, 2005). We transformed these scores into z-scores in order to have a standardized metric for reference.

Child variables

The means and standard deviations for the variables included in the current study are presented in Table 1. The first predictor variable was teacher assessments of children’s approaches to learning. The teacher-rated scales consisted of six items that asked respondents to rate the attentiveness, persistence, eagerness to learn, learning independence, flexibility and organization of the child (Claessens et al. 2008). Specifically, the approaches to learning scale consisted of a composite score of teachers’ ratings of six items. Individual items are not provided in the ECLS-K data set so we could not calculate reliability for our specific sub-sample; however, overall reliability for this measure is quite high (α = 0.89; USDE, 2010a). Teachers were asked how frequently each child: (1) keeps belongings organized; (2) shows eagerness to learn new things; (3) works independently; (4) easily adapts to changes in routine; (5) persists in completing tasks; and (6) pays attention well (USDE, 2010a). Teachers rated each learning-related attitude or behaviour on a scale of 1 to 4, ranging from “never” to “very often”, for each student. These items were asked for during both the autumn and spring data collection periods. Teacher ratings for autumn and spring assessments were correlated at 0.72. Due to the high correlation, and in order to get a broader picture of the net effect of students’ approaches to learning on achievement, we averaged the autumn and spring scores. These scores were then converted to z-scores for ease of interpretation.

Table 1.

Means and standard deviations of variables used in HLM analyses

Mean (SD)
Student level variables (n = 3034)
Spring reading IRT score	31.66 (10.01)
Teacher-rated ATL (avg)	3.06 (0.64)
Natural log of autumn reading IRT score	3.04 (0.34)
SES indicator	0.02 (0.75)
Time between testing (days)	182.80 (19.32)
School level variables (n = 184)
Aggregate school SES	0.36 (0.68)
High minority enrolment	0.29 (0.45)
Full-day kindergarten^a	0.55
Total school enrolment	374.74 (219)
Aggregate teacher ATL	−0.014 (0.44)
Reading activities	−0.074 (0.89)

Note: Means and standard deviations are presented in the original metrics of the variables. HLM, hierarchical linear modelling; SES, socioeconomic status; ATL, Approaches to Learning.

Proportion of schools, not mean of variable.

We also included a series of student-level control variables into our model. First, we controlled for children’s autumn reading achievement and the time between assessments because this varied significantly across schools, and it certainly seemed plausible that students who had been in the classroom for longer periods of time would demonstrate higher reading abilities. The autumn reading achievement test was also an IRT scale score that measured the same aspects of reading achievement as the spring assessment. After examining the distribution of this variable, we transformed the scores using a natural log transformation to correct for non-normality and then standardized this variable. The time between assessments was entered in the model and represents the number of days that elapsed between the autumn and spring assessments. In order to control for family background, the continuous measure of individual student’s family SES was included in the model and was standardized.

Classroom and school variables

School-level variables represent signifiers of the “at-risk” status of the school as well as other school-level characteristics of interest. We included the aggregate SES of students in the school because this reflects the average SES of the students at the school who were sampled. Due to the high correlation between receipt of Title I funding and school level SES, only aggregate SES was included as a school-level predictor. SES was calculated based on parent reports of income and the number of household residents. We also included a school-level variable to indicate whether the school had over 40% of minority student enrolment, as previous research has shown that race influences both teacher ratings of approaches to learning and reading achievement (Auwater and Arguete, 2008; Lynch, 2010). This variable was categorical and the cut-off level of 40% of minority enrolment was provided in the ECLS-K data set. This variable was used to identify schools with minority enrolment in the top third of schools, or high minority schools.

Availability of full-day kindergarten was included as a school-level variable and was dichotomously coded into 1 = “full-day” and 0 = “other”. The “other” category included those schools that only offered half-day kindergarten as well as schools that were mixed, that is, they offered both full- and half-day kindergarten. The percentage of schools offering the mixed option was quite small (14 schools; 7% of the total sample). School size was included as a continuous variable that measured total student enrolment and was standardized as a z-score.

Finally, we created an instructional-level factor that captured the frequency of reading activities in the classroom using principal axis factoring in SPSS version 20. The individual items included in the factor asked teachers to indicate how often they taught different concepts associated with reading, including how often teachers spent time on vocabulary, on talking about the contents of a book and on teaching children about rhyming. The individual items and their factor loadings are presented in Table 2. We first created this factor at the student level and aggregated to the school level. Two of the original response options (“taught at a higher grade level” and “children should already know”) did not fit along a frequency continuum and were therefore re-coded as “not taught” under the assumption that if teachers believed children should already know something or that it should be taught only to older students they would not teach it in their class. Although these response options describe conceptually different reasons for teachers electing not to teach certain material, they both represent reasons for choosing not to teach these particular concepts and therefore represent the frequency of “not taught”. This also served to make the factor more interpretable as a measure of the frequency of reading activities in the classroom given the other response options available to teachers. The final response choices were: 1 = “not taught”, 2 = “once a month or less”, 3 = “2–3 times a month”, 4 = “1–2 times a week”, 5 = “3–4 times a week” and 6 = “daily”.

Table 2.

Factor matrix for frequency of reading activities

Loading
Rhyming words and families	0.485
Reading multi-syllable words	0.571
Common prepositions	0.557
Identify main idea of story	0.611
Make predictions based on text	0.530
Use cues for comprehension	0.612
Communicate ideas orally	0.425
Use capitalization/punctuation	0.606
Write complete sentences	0.656
Story has beginning/middle/end	0.655
Vocabulary	0.457
Reading aloud fluently	0.530

Note. 37.19% of variance explained by factor; Cronbach’s α = 0.83.

We forced the variables onto one factor based on the high correlations between the individual items (r’s ranging from 0.13 to 0.59 with most in the 0.3–0.4 range and all significant at p < 0.01). In addition, previous studies had collapsed items into one factor in order to examine teacher’s overall reading instruction (e.g. McCoach et al., 2006). Several variables from the original scale were eliminated because they did not load above our criterion of 0.4, which is a common cut-off value used in factor analysis (Pett et al. 2003). We used the factor regression scores as the predictors of frequency of reading instruction in the classroom. Due to the negative skew of the factor regression scores, these scores were squared, which resulted in an approximately normal distribution. The final scale was reliable (α = 0.83). Finally, the factor scores were standardized to z-scores and aggregated to the school level.

We used HLM to conduct these analyses because we are interested in both individual-level predictors of children’s reading achievement in kindergarten as well as the classroom- and school-level variables that may contribute to this achievement. HLM uniquely allows the variance in students’ reading achievement and the relation between approaches to learning and reading achievement to be examined at both the individual and school levels. Examining these relations at the individual level allows us to take into account and control for individual differences; the school level allows us to examine how contextual influences matter above and beyond individual-level variation. All students included in the current sample were purposely selected from the same class due to the small number of students sampled at each site; therefore, nesting at the classroom and school levels could not be considered separately because only one teacher was nested within each school. However, because students all shared the same teacher, it seems reasonable to assume that school- and classroom-level effects would similarly affect students.

Results

The multi-level analyses in this paper used the HLM 7.0 program. However, before we conducted our analyses, we examined the correlations between our predictors and outcome variables to test for multicollinearity between our variables. Results for the student- and school-level predictors are reported in Tables 3 and 4, respectively.

Table 3.

Correlations among student-level variables

Spring reading	Teacher ATL	Log autumn reading	Gap testing	SES
Spring reading	1.00
Teacher ATL	0.422**	1.00
Log autumn read	0.793**	0.446**	1.00
Gap testing	0.056*	0.001	−0.032	1.00
SES	0.367**	0.257**	0.409**	0.032	1.00

Correlation is significant at the 0.05 level; **Correlation is significant at the 0.01 level. SES, socioeconomic status. ATL, Approaches to Learning.

Table 4.

Correlations among school-level variables

School SES	High minority	Full day	School size	Teacher ATL	Lit. activities
School SES	1.00
High minority	−0.370**	1.00
Full day	−0.112	0.188*	1.00
School size	−0.093	0.126	−0.007	1.00
Teacher ATL	0.262**	−0.164*	−0.094	0.076	1.00
Lit. activities	0.018	0.262**	−0.320**	0.129	−0.003	1.00

Correlation is significant at the 0.05 level; **Correlation is significant at the 0.01 level. ATL, Approaches to Learning.

Almost all of the individual-level predictor variables were correlated significantly with the outcome of interest in the expected direction. That is, average teacher-rated approaches to learning were associated with higher kindergarten reading achievement scores in the spring. Higher SES, higher autumn reading achievement and a longer gap between autumn and spring testing were also associated with higher spring reading achievement. Finally, most of the control variables were also significantly correlated with the other variables in the model in the expected direction, but the correlations between variables did not indicate that multicollinearity was a problem.

Similarly, no multicollinearity issues were evident when we examined the correlations between the school-level variables. These correlations between school-level variables were also in the expected direction. School-level SES and high minority enrolment were negatively correlated, indicating that low aggregate SES was associated with a school having a higher percentage of minority students. High school-level SES was associated with higher teacher-rated approaches to learning at the school level, as indicated by the moderate and positive correlation between these variables. The availability of full-day kindergarten was associated with a higher percentage of minority students in a school. In addition, a higher percentage of minority student enrolment was associated with lower teacher-rated approaches to learning. Finally, somewhat surprisingly, frequency of reading activities was positively associated with high-minority enrolment and negatively associated with full-day kindergarten enrolment.

Missing data

Sixteen schools were lost when we imported the data into HLM due to some individuals missing data on the frequency of reading activities factor. There were some significant differences between the schools we lost and the schools we retained for our variables of interest, including average teacher-rated approaches to learning and the availability of full-day kindergarten. Schools that were excluded were less likely to offer full-day kindergarten and had higher mean teacher-rated approaches to learning at the school level.

In addition to the loss of school-level data, the data for all students that attended the excluded schools were excluded from the final analysis. Students with missing data differed significantly for several of our variables of interest, including spring reading, teacher-rated approaches to learning, autumn reading achievement and individual SES. Students that were excluded had higher means for all four of these variables. Therefore, the data for the current study are not considered to be nationally representative. In terms of the missing data, the students who attend these schools and the schools themselves seem to be somewhat more advantaged as compared to the overall sample. However, because we were specifically interested in school-level risk factors it is not so concerning that we lost some of the more advantaged schools.

Fully unconditional model

Prior to conducting HLM analyses, it is advisable to first fit a fully unconditional model (Raudenbush and Bryk, 2002). This model does not specify any predictors and allows for assessment of whether there is significant variance in the outcome of interest between the level-2 groups, in this case, schools. It also partitions the total variance in spring kindergarten reading achievement into the within-group, or individual student variance, and the between-group, or school variance. Finally, the fully unconditional model provides an initial rating of the reliability of our dependent variable as well as of the initial variance within and between schools. For the current model, the reliability for spring reading achievement was λ = 0.832, indicating that the observed scores for spring reading achievement were reliably estimated. The parameter variance of τ₀₀ = 0.257 is statistically significant, indicating that there is a significant amount of variance in students’ average spring reading achievement between schools.

Using the fully unconditional model, we calculated the intraclass correlation (ICC) which describes the proportion of total variance that is between groups. The ICC calculation indicates that 25% of the variance in spring kindergarten reading achievement is between schools. This indicates that HLM is an appropriate analysis strategy for pursuing our current set of research questions because students’ spring reading achievement does vary between schools.

Within-school model

The purpose of the within-school model is to use student-level predictors to model the outcome of interest, in this case, kindergarten reading achievement measured in the spring. No level 2 or group-level predictors are entered in the model at this step. The within-school model allows an exploration of individual-level differences while controlling for school-level differences (Raudenbush and Bryk, 2002). This is an important first step in modelling between-school variables, as we can look at whether there is between-school variance in any of the relationships between predictors and the outcome of interest, while controlling for individual-level variance.

To explore our within-school model, we entered the independent variables one at a time to evaluate how the model would change with each new predictor. We entered teacher ratings of approaches to learning group-mean centred with random effects, which allows for the relationship between approaches to learning and spring reading achievement to vary across schools. The reliability for the slope of the relationship between teacher ratings of approaches to learning was λ = 0.138, and the slope varied significantly across schools (χ²₁₈₃ = 214.69, p = 0.05). The significant variation in slope indicates that the relationship between approaches to learning and spring reading achievement was different in different schools, and that there are probably school-level factors that influence the strength of the relationship. Therefore, we left this variable in the model, group-mean centred with random effects. Next, we entered our control variables into the model: time between testing, standardized SES and autumn reading achievement. All control variables were entered one at a time, grand-mean centred with fixed effects. This is because we were not interested in how these predictors varied with achievement across schools and only wanted to control for the possible influence that they may have on spring reading achievement. The fixed and random effects for the full set of predictors and covariates included in the final within-school model are presented in Table 5.

Table 5.

Fixed and random effects for the final within-school model

Fixed effects	Coefficient	Std. error	t-ratio	p-Value
Spring reading achievement	0.022	0.023	0.974	0.331
Teacher ATL	0.105	0.015	6.804	<0.001
Autumn reading achievement (log)	0.733	0.015	48.590	<0.001
Time between assessments (days)	0.003	0.001	3.333	<0.001
Student SES	0.044	0.015	3.063	0.002
Random effects	Std. dev.	Variance	χ²
Intercept	0.231	0.053	701.39	<0.001
Teacher ATL	0.062	0.004	214.69	0.049

ATL, Approaches to Learning; SES, socioeconomic status.

The reliability for spring reading achievement in the final model was λ = 0.73, indicating that the observed score for spring reading achievement was still reliably estimated between schools. The reliability for the slope of the relationship between teacher-rated approaches to learning and spring achievement was λ = 0.14. There is no firm cut-off point when determining whether the reliability of the slope is adequate, but Raudenbush and Bryk (2002) advise that reliability above λ = 0.1 may be an appropriate reference point. The reliabilities in this model were lower than those found for the fully unconditional model, as expected. This decrease in reliability can be attributed to the addition of grand-mean centre predictors to the model that explain some of the variance in the mean levels of students’ spring reading achievement. As such, some of the differences across schools at level 2 are being controlled.

The intercept for spring reading achievement in the fixed effects table can be interpreted as the average spring reading achievement (in standard deviations) after controlling for parent-rated approaches to learning, SES, time between assessments and autumn reading achievement. Teacher-rated approaches to learning significantly predicted the spring reading achievement after controlling for the other predictors in the model. That is, a one standard deviation increase in teacher-rated approaches to learning was associated with a 0.11 standard deviation increase in spring reading achievement, indicating that students with higher approaches to learning scored higher on the measure of spring reading achievement.

Although the control variables are not of primary interest to the research questions we posed, their effects on spring reading achievement were consistent with prior research. On average, a one standard deviation increase in log of fall reading achievement was associated with a 0.73 standard deviation increase in spring kindergarten reading achievement, indicating that for every standard deviation increase in the autumn reading score, we would expect almost three quarters of a standard deviation increase in spring reading scores. Perhaps not surprisingly, this indicates that higher initial reading achievement in the autumn is associated with higher reading achievement in the spring. A one standard deviation increase in SES was associated with an increase of 0.04 standard deviation in the spring reading achievement, that is, for every one standard deviation increase in SES we would expect spring reading scores to increase by 0.04 standard deviation, indicating that spring reading achievement increases as SES increases. A one day increase in the gap between autumn and spring testing was associated with a 0.003 standard deviation increase in the spring reading achievement, indicating that one day longer between autumn and spring testing was associated with a very small improvement in spring reading achievement scores.

The variance of the slope of the relationship between teacher-rated approaches to learning and spring reading achievement was significant, indicating that the relationship between these variables varies significantly between schools. Further, it is important to note that the variance of the intercept is still significant in the random effects portion of the model. This indicates that spring reading achievement still significantly varies between schools. Taken together, these results suggest that it would be informative to add school-level predictors into the model to attempt to explain the between-school differences in reading achievement and the relation between teacher-rated approaches to learning and spring reading achievement.

Of final interest in this model was the amount of individual-level variance that the current within-school model was able to account for. Our calculation showed that our within-school model explained about 62% of the individual-level variance in spring reading achievement. This indicated that autumn reading achievement, SES and the time between autumn and spring assessments accounted for a relatively large portion of the individual variance in students’ spring reading achievement.

Between-school model

The purpose of a between-school model is to use school-level predictors to model the variation between schools for an outcome of interest, in this case average kindergarten spring reading achievement (Raudenbush and Bryk, 2002). It also offers a way for school-level predictors to model the relationship between an outcome of interest and an individual-level predictor, in this case the relationship between teacher-rated approaches to learning and kindergarten spring reading achievement. After building our within-school model, we found that teacher-rated approaches to learning not only predicted children’s kindergarten spring reading achievement (controlling for other variables in our model), but that the nature of this relationship varied between schools. Therefore, we decided to run a between-school model to understand what school-level variables contribute to this variation. This model addresses our last two research questions, which focus on classroom- and school-level differences and how they affect spring kindergarten reading achievement and the relation between approaches to learning and spring reading achievement.

To explore our between-school model, we began with the within-school model discussed above. We next added school-level predictors of aggregate teacher-rated approaches to learning, whether the kindergarten programme was full-day or half-day, whether the school had a high proportion of minority students, aggregate SES, frequency of reading instruction and school size. These variables were added one by one to see which variables predicted both average kindergarten spring reading achievement and the strength of the relationship between kindergarten spring reading achievement and teacher-rated approaches to learning. Decisions about centring were uniform and followed the standards recommended by HLM analysts; all dichotomous variables were entered un-centred and all continuous variables were entered mean-centred (Enders and Tofigi, 2007).

Average spring kindergarten reading achievement

First, we entered the school-level aggregate of teacher-rated approaches to learning as a predictor of average spring reading achievement as a control variable because this student-level variable was group-mean centred in our within-school model at level 1. This variable was non-significant once entered into the model, but was left in the final model to serve as a control. We next added aggregate school-level SES, high-minority enrolment, availability of full-day kindergarten and frequency of classroom reading activities into the model as predictors of average spring reading achievement because these variables were central to our research questions at the school level. When only these four predictors were included in the model, the availability of full-day kindergarten was a significant predictor of spring reading achievement in the expected direction. That is, children in schools with full-day kindergarten had higher mean reading achievement. However, after adding the factor representing the frequency of reading activities into the model, this variable became non-significant. Only the factor measuring classroom frequency of reading activities was a significant predictor of the average spring reading achievement. Regardless of significance, all of these variables were kept in the model as controls (see Table 6).

Table 6.

Fixed and random effects for the final between-school model

Coefficient	Standard Error	t-statistic	p-Value
Estimated effects
Intercept
Spring reading achievement	−0.007	0.037	−0.185	0.854
High minority enrolment	−0.007	0.055	−0.132	0.895
Aggregate school SES	0.034	0.047	0.797	0.426
Aggregate teacher-rated ATL	−0.002	0.057	−0.041	0.964
Full-day kindergarten	0.072	0.047	1.534	0.127
Reading activities	0.072	0.029	2.509	0.013
Teacher ATL
Intercept	0.134	0.024	5.521	<0.001
High minority enrolment	−0.015	0.034	−0.430	0.668
Aggregate school SES	0.015	0.029	0.589	0.557
Full-day kindergarten	−0.037	0.026	−1.566	0.119
School size	−0.015	0.018	−2.538	0.012
Reading activities	0.050	0.015	2.747	0.007
Time between assessment (days)
Intercept	0.004	0.001	4.182	<0.001
Student SES
Intercept	0.042	0.018	2.334	0.020
Autumn reading achievement (Log)
Intercept	0.727	0.023	31.691	<0.001
Parameter	Parameter variance	d.f.	χ ²	p-Value
The chi-square table
Intercept	0.047	178	644.02	<0.001
Slope	0.002	178	195.10	0.181

SES, socioeconomic status; ATL, Approaches to Learning; d.f., degrees of freedom.

Teacher-rated approaches to learning

The variables entered as possible predictors of the strength of the relationship between approaches to learning and spring reading achievement included: high-minority enrolment, school size, school SES, availability of full-day kindergarten and frequency of reading activities. Although high-minority enrolment, availability of full-day kindergarten and school SES were not significant predictors of the relationship between teacher-rated approaches to learning and spring reading achievement, these were left in the model as controls, and because they were central to our research question. After developing our final between-schools model we were able to completely explain the variance in teacher-rated approaches to learning across schools, with school size and frequency of reading activities both significantly predicting the variance in the relationship between teacher-rated approaches to learning and spring reading achievement across schools.

Based on the results of this final between-schools model, we were able to conclude several things about spring kindergarten reading achievement. The estimated average of the school averages in spring kindergarten reading achievement scores (standardized) for schools with less than 40% minority enrolment, average school SES at schools offering half-day or mixed kindergarten with average teacher-rated approaches to learning and average frequency of reading activities was −0.007. A one standard deviation increase in frequency of reading activities at the school was associated with an increase of 0.072 standard deviation in the spring kindergarten reading achievement, controlling for the other predictors in the model. That is, schools with a higher frequency of reading activities had higher mean reading achievement scores in the spring. Although not significant, the effects of the other variables included as predictors of average spring reading achievement are mostly in the expected direction.

Several variables were significant predictors of the strength of the relationship between approaches to learning and spring reading achievement. For each standard deviation increase in the frequency of reading activities at the school level, there was an increase in the effect of teachers’ ratings of approaches to learning on spring kindergarten reading achievement by 0.05 (see Figure 2). This indicates that the relationship between teacher-rated approaches to learning and spring reading achievement was stronger for schools that had a higher frequency of reading activities. Additionally, controlling for whether a school offered a full- or half-day kindergarten programme was not significant in predicting the strength of the relationship between approaches to learning and spring reading achievement, so the number of hours in school does not seem to be influencing these results. This indicated that the frequency with which reading instruction occurs in the classroom, regardless of the length of the school day, affects the relation between approaches to learning and spring reading achievement. Classes in which reading instruction was more frequent showed a stronger relationship between approaches to learning and spring reading achievement.

Figure 2.

The strength of the relationship between approaches to learning and spring reading achievement in schools with varying frequencies of reading instruction in kindergarten classrooms

For every one standard deviation increase in school size, the effect of teacher-rated approaches to learning on spring kindergarten reading achievement was decreased by 0.039 (see Figure 3). This indicates that the relationship between teacher-rated approaches to learning and spring reading achievement was weaker in larger schools. Conversely, schools that are smaller in size showed a stronger relationship between approaches to learning and spring reading achievement.

Figure 3.

The strength of the relationship between approaches to learning and spring reading achievement in schools of varying size

In order to calculate the between-group variance that was explained by the model, we calculated the percentage change in the variance component of spring reading achievement and teacher rating of approaches to learning slope. Based on the results presented in the table, we explained about 13% of the variance in spring kindergarten reading achievement.

We also completely explained the variation in teacher-rated approaches to learning between schools. That is, the school-level predictors entered into the model accounted for the variance between schools in the differing strength of the relationship between approaches to learning and spring reading achievement.

Discussion

In this study, we investigated three specific research questions:

How do children’s approaches to learning relate to reading achievement in the spring and to what extent does this relationship vary across schools?

In response to these specific enquiries, first, the within-school model revealed that both spring reading achievement and the relation between approaches to learning and spring reading achievement varied between schools. Second, in the between-school model, the school-level predictors of aggregate teacher-rated approaches to learning (i.e. whether the kindergarten programme was full-day or half-day, whether the school had a high proportion of minority students, aggregate SES, frequency of reading instruction and school size, frequency of reading instruction) were found to have a positive effect on spring reading achievement, controlling for individual characteristics and the other school-level factors in the model. Finally, a between-school model explaining the differences in the relation between students’ approaches to learning and spring reading achievement across schools found that the frequency with which teachers engaged in reading activities significantly explained part of the between-school variance in this relationship.

With no predictors entered into the model, spring reading achievement varied significantly across schools. This is not entirely surprising given that the data were a random sample of students across the US. Important to note is that even after taking individual-level predictors into account in the within-school model, spring reading achievement still varied significantly across schools. This indicates that while individual student-level factors certainly contribute significantly to explaining the variance in spring reading achievement scores, there are still classroom- and school-level factors that must be considered.

The between-schools model allowed us to examine some of these possible factors. This model examined the school-level variables that predict students’ spring kindergarten reading achievement as well as predictors of the relationship between teacher-rated approaches to learning and spring kindergarten reading achievement. Frequency of reading instruction and school size were able to account for the school-level variation in the relationship between students’ approaches to learning and spring reading achievement, accounting for the difference in the strength of this relationship between schools. Thus, both classroom- and school-level factors have been shown to play an important role in explaining the variance across schools in the relation between teacher-rated approaches to learning and reading achievement toward the end of the school year.

Although statistically significant, the percentage of explained variance in overall spring reading achievement was low, but we were able to explain the difference between schools in the variance of approaches to learning and spring reading achievement. The variance explained accounts for the variance in spring reading achievement overall as well as the variance in the relation between approaches to learning and spring reading achievement. Considering we were able to completely explain the between-school variance in the relationship between approaches to learning and reading achievement, the percentage of variance explained probably remained somewhat small because we were unable to account for all the factors that influence students’ reading achievement between schools. Further, because we were able to explain a high percentage of the variance in spring reading achievement at the individual level, we would expect the proportion of variance explained at the school level to be low.

Time spent on reading activities in the classroom is important for students’ academic achievement (McCoach et al. 2006), as well as for strengthening the relationship between their attitudes about learning and their achievement (Chen and McNamee, 2011). The factor we created accounted for the frequency of reading behaviours teachers perform when in the classroom and is one way to account for reading instruction within the kindergarten classroom. The relationship between reading activities and children’s academic achievement is consistent with the literature on reading activities in kindergarten (Chen and McNamee, 2011). It is promising that reading behaviours also influence the relationship between approaches to learning and spring reading in addition to their direct effect on reading achievement. Although Xue and Meisels (2004) predicted approaches to learning from teacher reading behaviours, the relationships presented in this paper are a key and novel finding. Our findings illustrate not only the importance of approaches to learning, but also the value of examining how the teaching practices that kindergarten teachers use may strengthen the relationship between approaches to learning and achievement.

Frequency of instruction within the classroom specifically accounted for a large proportion of the variance in the relationship between approaches to learning and reading achievement across schools. This echoes research on young children’s approaches to learning which demonstrates the importance of these approaches to learning on children’s achievement in elementary school (George and Greenfield, 2005; Li-Grining et al. 2010). Research on approaches to learning needs to be examined further as it seems to be such an important factor influencing children’s achievement. More studies examining ways to develop and support young children’s approaches to learning are necessary in order fully to understand what can help young children’s academic achievement.

It was somewhat surprising that the only school-level demographic variable that emerged as a significant predictor of the relationship between teacher-rated approaches to learning and spring reading achievement was the size of the school. Although it does make sense that the relationship between approaches to learning and spring reading achievement may be stronger in smaller schools, we were not expecting this to be the only significant demographic predictor. It would be informative in future studies to unpack this variable further. That is, it seems likely that this relationship is stronger in smaller schools due to the closer relationships between teachers and students. Perhaps this finding can be explained by the overall collegiality of smaller schools. Prior research by Koth et al. (2008) indicates that both smaller class size and smaller school size positively predict perceptions of school climate.

Magnuson et al. (2007) suggest that differences in class size produce differences in the quality of instruction, with smaller classes allowing more devoted attention to students and fewer teacher-directed activities. Smaller classes are likely to occur in smaller schools. Future studies should examine both school and class size to see if these account for the differences observed in the current study. An interesting area for future research is an examination of the instructional processes through which teachers capitalize on students’ approaches to learning in the classroom as related to both school and class size, as well as school structure.

Finally, even though previous research has demonstrated the importance of school-level variables in explaining academic achievement (Lee et al. 2006), the current model found that frequency of instructional practices was the only significant predictor of differences between schools when added to the model. Although our percentage of explained variance is small, school-level demographic variables were not predictive of achievement in reading once reading behaviours were taken into account. Demographic factors significantly predicted achievement before frequency of instructional practices was added to the model, indicating that these factors are still important to consider when examining achievement differences across schools. However, teacher practices in the classroom may be more important than the demographic characteristics of schools in predicting achievement. This finding is promising for improving student achievement through teacher training and support (Chen and McNamee, 2011). The importance of teacher practices in the classroom is also consistent with understandings of approaches to learning as mutable and developed through instruction and classroom interactions (Bulotsky-Shearer, and Fantuzzo, 2011). Further, this finding demonstrates the importance of providing high-quality instructional experiences, particularly to students in at-risk schools and to students with potentially limited approaches to learning. Although not included in our model, future research should investigate teachers’ instructional practices in the classroom, not only with regard to reading activities, but also geared toward the development of students’ approaches to learning.

Limitations

There are several limitations to this study. First of all, the factor that we created based on the frequency of reading activities in the classroom only accounts for the number of times certain concepts are taught; it does not capture the concepts of reading and the depth to which they were taught. Although it was beyond the scope of this project and this data set to capture reading concepts, future research should explore the influence of depth of concepts on predicting the relationship between approaches to learning and reading achievement in young children. Further, we are interested in the reasoning behind teachers’ choice of instructional activities for their classrooms as an area for further investigation. Our frequency of reading activities in the classroom factor asked teachers to report the frequency of instructional activities on a scale that allowed teachers to report content areas as “not taught,” because they expected this content to be covered in either the preceding or subsequent school year. However, more research is needed to understand why teachers elect to engage in certain instructional activities and not others, and how classroom characteristics, such as class size, may influence this decision.

The ECLS-K was designed to be a nationally representative sample of kindergarten students in the United States. However, due to some missing data on predictors of interest to our model, we lost several schools from our sample. As such, the results reported in the paper are not representative of the kindergarten population in the United States as a whole. The final sample of schools in this study was more likely to offer full-day kindergarten and had lower mean levels of teacher-rated approaches to learning. Additionally, the students who attended the schools in our final sample were somewhat more disadvantaged in terms of SES as compared to the full sample.

Future directions

In examining the percentage of variance explained by our model for average spring reading achievement, it is clear that there is still a significant amount of between-school variance that remains to be explained. Although we included one instructional level factor as a possible explanation of the between-school differences in average reading achievement, there may be other instructional-level or school-level factors that were not included in our current model that could explain more of the variance in reading achievement at the end of kindergarten.

It is possible that other variables, that we were unable to include in the current study, may affect kindergarten reading achievement. For example, Magnuson et al. (2007) found that those children who experienced preschool type programmes the year prior to kindergarten demonstrated higher literacy skills compared to children who spent the year before kindergarten being cared for by either their parents or another family member. Future studies should examine how preschool experiences may influence both reading achievement in kindergarten and also how these early schooling experiences may influence the relationship between approaches to learning and reading achievement.

We examined many school-level contextual variables; however, we were only able to include one classroom-level variable. There are certainly other instructional practices and school-level factors that might explain children’s spring reading scores and the relationship between spring reading and approaches to learning. The current data set makes it a little difficult to get at these underlying factors, especially those that are teacher-level characteristics. We had to aggregate all teacher-level variables to the school level and therefore could not take variability between teachers into account. It could also be that there are important classroom-level variables that were not collected for the ECLS-K that could account for the differences in these types of schools.

In addition to giving further consideration to classroom practice as differentiating the relation between approaches to learning and reading achievement, it may be beneficial to assess more extensively students’ approaches to learning. Specifically, in this study, a single teacher-report measure of approaches to learning was included in the models and this measure was averaged between teachers’ autumn and spring ratings. Further work should consider direct observational measures of students’ approaches to learning or ratings collecting from several sources (George and Greenfield, 2005).

Finally, it is important to note that the results of this study are situated within the context of kindergarten classrooms in the United States. Although we can make no claims as to how these results may transfer to other school contexts in other countries, it seems likely that classroom- and school-level factors influence students’ reading achievement during early schooling experiences across contexts. As such, we believe that the results of the current study can be applied to schooling in other countries, although replication in those unique contexts would certainly be important for future research.

Conclusion

Students’ approaches to learning in kindergarten are related to spring reading achievement and frequency of instruction, and school size are two variables that can account for these differences. It is gratifying that instructional-level factors significantly predicted this relationship while many of the school-level risk factors did not. School size and other demographic factors at the school level may be difficult to alter; however, frequency of reading instruction in the classroom is something that schools and teachers can control. This variable had the strongest effect on enhancing the strength of the relation between approaches to learning and reading achievement. Therefore, these practices may be particularly important to target in future intervention work. The frequency of reading instruction in the classroom is essential in influencing student achievement. Further, because approaches to learning are so essential to children’s later engagement in school, it is promising that the individual level effect holds after all of the individual and school-level risk factors are controlled for in the model.

In examining the relation between individual-level approaches to learning and reading achievement at the end of kindergarten, in addition to the influence of classroom and school-level factors on the difference in this relation between schools, several important implications for instruction can be highlighted. First, consistent with earlier research, this study identified differences in spring reading achievement across schools, even when individual demographic characteristics were controlled for. With the addition of class- and school-level factors, however, we were able to explain the between-school variance in spring reading achievement, pointing to the importance of considering not only student-level but also classroom and school level characteristics as affecting students’ learning. Significantly, teachers’ instructional practices, assessed in terms of time spent on reading activities, predicted students’ spring reading achievement, whereas other demographic features did not. This finding illustrates the central role that quality instruction plays in students’ achievement, particularly in cases where students attend “at risk” schools.

References

Auwater

Arguete

(2008) Effects of student gender and socioeconomic status on teacher perceptions. The Journal of Educational Research 101(4): 243–246.

Barnett

Bauer

Ehrhardt

(1996) Keystone targets for change: Planning for widespread positive consequences. School Psychology Quarterly 11: 95–117.

Baskett

Bryant

White

(2005) Half-day to full-day kindergarten: An analysis of educational change scores and demonstration of an educational research collaboration. Early Child Development and Care 175: 419–430.

Bennett

Elliott

Peters

(2005) Classroom and family effects on children’s social and behavioral problems. The Elementary School Journal 105(5): 461–480.

Biggs

(1987) Student Approaches to Learning and Studying, Melbourne: Australian Council for Educational Research.

Bulotsky-Shearer

Fantuzzo

(2011) Preschool behavior problems in classroom learning situations and literacy outcomes in kindergarten and first grade. Early Childhood Research Quarterly 2: 61–73.

Chen

McNamee

(2011) Positive approaches to learning in the context of preschool classroom activities. Early Childhood Education Journal 39: 71–78.

Claessens

Duncan

Engel

(2008) Kindergarten skills and fifth grade achievement: Evidence from the ECLS-K. Economics of Education Review 28(4): 415–427.

Curley

(2005) Theories of urban poverty and implications for public housing policy. Journal of Sociology and Social Welfare 32(2): 97–119.

10.

Dart

Burnett

Purdie

(2000) Students’ conceptions of learning, the classroom environment, and approaches to learning. Journal of Educational Research 93(4): 262–270.

11.

Denton

West

(2002) Children’s reading and mathematics achievement in kindergarten and first grade. U.S. Department of Education, NCES 2002–125, Washington, DC: U.S. Government Printing Office.

12.

DiPerna

Lei

Reid

(2007) Kindergarten predictors of mathematical growth in the primary grades: An investigation using the Early Childhood Longitudinal Study—Kindergarten Cohort. Journal of Educational Psychology 99(2): 369–379.

13.

DiPerna

Volpe

Elliott

(2002) A model of academic enablers and elementary reading/language arts achievement. School Psychology Review 31(3): 298–312.

14.

Enders

Tofigi

(2007) Centering predictor variables in cross-sectional multilevel models: A new look at an old issue. Psychological Methods 12(2): 121–138.

15.

Fantuzzo

Perry

McDermott

(2004) Preschool approaches to learning and their relationship to other relevant classroom competencies for low-income children. School Psychology Quarterly 19(3): 212–230.

16.

Finn

Pannozzo

(2004) Classroom organization and student behavior in kindergarten. The Journal of Educational Research 98(2): 79–91.

17.

Fusaro

Royce

(1995) A reanalysis of research data. Perceptual and Motor Skills 81: 358–358.

18.

George

Greenfield

(2005) Examination of a structured problem-solving flexibility task for assessing approaches to learning in young children: Relation to teacher ratings and children’s achievement. Applied Developmental Psychology 26: 69–84.

19.

Koth

Bradshaw

Leaf

(2008) A multilevel study of predictors of student perceptions of school climate: The effect of classroom-level factors. Journal of Educational Psychology 100(1): 96–104.

20.

Lee

Burkam

Ready

(2006) Full-day vs. half-day kindergarten: In which program do children learn more? American Journal of Education 112(1): 163–208.

21.

Li-Grining

Votruba-Drzal

Maldonado-Carreno

(2010) Children’s early approaches to learning and academic trajectories through fifth grade. Developmental Psychology 46(5): 1062–1077.

22.

Lynch

(2010) Kindergarten teachers’ beliefs about students’ knowledge of print literacy and parental involvement in children’s print literacy development. The Alberta Journal of Educational Research 56(2): 157–171.

23.

Magnuson

Ruhm

Waldfogel

(2007) The persistence of preschool effects: Do subsequent classroom experiences matter? Early Childhood Research Quarterly 22(1): 18–38.

24.

McCoach

O’Connell

Reis

(2006) Growing readers: A hierarchical linear model of children’s reading growth during the first 2 years of school. Journal of Educational Psychology 98(1): 14–28.

25.

McDermott

Beitman

(1984) Standardization of a scale for the study of children’s learning styles: Structure, stability, and criterion validity. Psychology in the Schools 21(1): 5–14.

26.

McDermott

(1984) Comparative functions of preschool learning style and IQ in predicting future academic performance. Contemporary Educational Psychology 9: 38–47.

27.

National Governors Association Center for Best Practices, Council of Chief State School Officers (2010) Common Core State Standards in English Language Arts. Washington, DC: National Governors Association Center for Best Practices, Council of Chief State School Officers.

28.

National Institute of Child Health and Human Development (2000) Teaching children to read: An evidence-based assessment of the scientific research literature on reading and its implications for reading instruction. Report of the National Reading Panel, NIH Publication No. 00-4769. Washington, DC: U.S. Government Printing Office.

29.

National Task Force on School Readiness (1991) Caring Communities: Supporting Young Children and Families, Alexandria, VA: National Association of State Boards of Education.

30.

Pett

Lackey

Sullivan

(2003) Making Sense of Factor Analysis, Thousand Oaks, CA: Sage Publications.

31.

Ponitz

Rimm-Kaufman

Grimm

(2009) Kindergarten classroom quality, behavioral engagement, and reading achievement. School Psychology Review 38: 102–120.

32.

Rabiner

Coie

(2000) Early attention problems and children’s reading achievement: A longitudinal investigation. Journal of the American Academy of Child and Adolescent Psychiatry 39(7): 859–867.

33.

Raudenbush

Bryk

(2002) Hierarchical Linear Models: Application and Data Analysis Methods, Thousand Oaks, CA: Sage Publications.

34.

Ready

LoGerfo

Burkham

(2005) Explaining girls’ advantage in kindergarten literacy learning: Do classroom behaviours make a difference? The Elementary School Journal 106(1): 21–38.

35.

Schaefer

McDermott

(1999) Learning behavior and intelligence as explanations for children’s scholastic achievement. Journal of School Psychology 37(3): 299–313.

36.

Spira

Bracken

Fischel

(2005) Predicting improvement after first-grade reading difficulties: The effects of oral language, emergent literacy, and behavior skills. Developmental Psychology 41(1): 225–234.

37.

US Department of Education, National Center for Education Statistics (2002) Early Childhood Longitudinal Study–Kindergarten Class of 1998–99 (ECLS–K). Psychometric Report for Kindergarten Through First Grade, NCES 2002–05, by Donald A. Rock and Judith M. Pollack, Educational Testing Service, Elvira Germino Hausken, Project Officer. Washington, DC: US Department of Education, National Center for Education Statistics.

38.

US Department of Education, National Center for Education Statistics (2010a) Combined user’s manual for the ECLS-K eighth-grade and K-8 full sample data files and electronic codebooks. Available at http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2009004 (accessed 7 November 2013).

39.

US Department of Education, National Center for Education Statistics (2010b) State Education Reforms (SER): Table 5.3. Kindergarten requirements, by state: 2010. Available at: http://nces.ed.gov/programs/statereform/tab5_3.asp (accessed 7 November 2013).

40.

US Department of Education, National Center for Education Statistics (2012) Early Childhood Longitudinal Programs (ECLS): Kindergarten Class of 1998–99. Available at http://nces.ed.gov/ecls/kindergarten.asp (accessed 7 November 2013).

41.

US Department of Health and Human Services (2010) The Head Start Child Development and Early Learning Framework: Promoting Positive Outcomes in Early Childhood Programs Serving Children 3-5 Years Old. Washington, DC: US Department of Health and Human Services.

42.

Welsh

Nix

Blair

(2010) The development of cognitive skills and gains in academic school readiness for children from low-income families. Journal of Educational Psychology 102(1): 43–53.

43.

Wirt J, Choy S, Rooney P, et al. (2004) The condition of education 2004. U.S. Department of Education, National Center for Education Statistics, NCES 2004-077. Washington, DC: U.S. Government Printing Office. Available at: http://nces.ed.gov/programs/coe/2004/pdf/03_2004.pdf_ (accessed 7 November 2013).

44.

Xue

Meisels

(2004) Early literacy instruction and learning in kindergarten: Evidence from the Early Childhood Longitudinal Study—Kindergarten class of 1998–1999. American Educational Research Journal 41(1): 191–229.

45.

Yan

Lin

(2005) Effects of class size and length of day on kindergartener’s academic achievement: Findings from Early Childhood Longitudinal Study. Early Education and Development 16(1): 49–68.