Learning-Related Behaviors and Literacy Achievement in Elementary School-Aged Children Deborah Stipek and Stephen Newton, School of Education, Stanford University. Amita Chudgar College of Education, Michigan State University. In press, Early Childhood Research Quarterly http://www.elsevier.com/wps/find/journaldescription.cws_home/620184/description Stanford University makes this peer-reviewed final draft available under a Creative Commons AttributionNoncommercial License. The published version will be available from the publisher, subscribing libraries, and the author. Running head: LEARNING-RELATED BEHAVIORS Abstract This longitudinal study investigated associations between children’s learning-related behaviors and literacy achievement in an ethnically diverse sample of low-income children throughout elementary school. Children’s literacy and learning-related behavior (e.g., working independently, seeking challenges) were assessed when they were in kindergarten or first grade and again in the third and fifth grades. The results showed fair consistency over time in both learning-related behaviors and literacy skills. Learning-related behaviors in one grade predicted literacy achievement in the subsequent grade in which it was assessed, but literacy skills did not predict subsequent learning-related behaviors. 2 Running head: LEARNING-RELATED BEHAVIORS 3 Learning-Related Behaviors and Literacy Achievement in Elementary School-Aged Children In this era of accountability, educators are being asked to increase students’ academic achievement and to reduce the socioeconomic and racial achievement gap. The federal No Child Left Behind law does not require student testing until third grade, but teachers in the early elementary grades are being pressured to emphasize the literacy and math skills that will ultimately be tested (Miller & Almon, 2009). Because the achievement gap exists when children enter school (Stipek & Ryan, 1997), districts and states are also broadening access to preschool as a strategy for improving children’s academic success in school (Barnett, Epstein, Friedman, Stevenson Boyd, & Hustedt, 2008). Evidence for the benefits of preschool education is strong now (see reviews by Barnett, 1995; Farran, 2000), but controversy continues about which dimensions of children’s development should be emphasized. While policy makers focus on academic achievement, early childhood educators complain, along with some early elementary school grade teachers, that skills and behaviors on other dimensions are also important, both in their own right and because they contribute to academic success (Stipek, 2006a, 2006b). Empirical evidence on the behaviors in early childhood that contribute to learning can be used to guide decisions about which dimensions of children’s development deserve special attention in efforts to improve achievement. The current study examines learning-related behaviors, such as working independently, seeking challenges, and accepting responsibility— what teachers and parents might refer to as good work habits. The learning-related behaviors examined in this study are connected to a large array of behaviors and cognitive processing skills under the rubric of self-regulation. There has been a proliferation of constructs and measures related to self regulation in recent years, in part because of evidence connecting various aspects of self-regulation to academic performance. The constructs and measures can be categorized, admittedly somewhat arbitrarily, into three large Running head: LEARNING-RELATED BEHAVIORS 4 groups. The first is emotional self-regulation (e.g., Cole, Martin, & Dennis, 2004), sometimes referred to as “hot,” affectively mediated, self-regulation skills. The second is cognitive processing skills, sometimes referred to as “cold” cognitively mediated self-regulation skills (Ponitz et al., 2008), such as meta-cognition (Brown, 1987), executive functions (e.g., Posner & Rothbart, 2007), and effortful and inhibitory control (Blair, 2003; Happaney, Zelazo, & Stuss, 2004; McClelland et al., 2007). The third is classroom behavior, such as sitting still, working independently, and listening to the teacher (McClelland, Morrison, & Holmes, 2000). These behaviors are sometimes combined in studies with items assessing academic social skills or social responsibility, such as cooperation and conformity to rules(e.g., McClelland & Morrison, 2003; Wentzel, 1991). While conceptual distinctions can be made, these various aspects of self-regulation are no doubt interconnected. For example, presumably cognitive processing skills, which are typically assessed in laboratory settings, contribute to behaviors children manifest in the classroom, such as paying attention and completing tasks without becoming distracted. Few studies have examined the association between performance on laboratory tasks and classroom behavior directly. One exception is a recent study by Ponitz et al. (2008), which found significant associations between young children’s (age 3 to 6 years) performance on the “Touch your Toes!” task (presumed to tap attention and working memory) and teachers’ ratings of classroom behavior (e.g., complying with adult directions, observing rules, and completing tasks). Although the interconnections among the self-regulation constructs are not well known, evidence suggests that some of the various measures of self-regulation are associated with academic achievement (Blair, 2002; Bull & Scerif, 2001; McClelland et al., 2006; McClelland et al., 2007; Matthews, Ponitz, & Morrison, 2009; National Institute of Child Health & Human Development [NICHD] Early Child Care Research Network, 2003). We chose to study behaviors that can be directly observed in classrooms and that have ecological validity for teachers. We wanted to assess behaviors that teachers believe are important (Foulks & Morrow, 1989; Lewit & Baker, 1995) and often lacking (Rimm-Kaufman, Pianta, & Cox, 2000), and that Running head: LEARNING-RELATED BEHAVIORS 5 might be amenable to improvement by teacher interventions in the regular classroom context. Classroom behavior was also selected because the study spans the elementary school years, and most laboratory measures are only appropriate for children in preschool and the early elementary grades. We focused on literacy achievement as the measure of academic success because it is critical to learning in all subject areas and it is a primary focus in the early grades of school. Although a number of studies have shown that learning-related behavior predicts academic achievement, the direction of causality has not been tested directly. It is possible that the causal direction is in the opposite direction, with achievement influencing learning-related behavior, or more likely, the relationship is reciprocal. There is consistent evidence in the motivation literature that success and the ensuing feelings of pride and self-confidence promote greater motivation and effort to learn (Mac Iver, Stipek, & Daniels, 1991; see Stipek, 2002, for a review). Also, children who perform relatively well academically are presumably reinforced with high grades and social approval for learning-related behaviors, such as seeking challenge and working independently, and these behaviors may increase as a consequence. In brief, academic success may reinforce good behaviors, which in turn produce further academic success. Academic success and the accompanying reinforcing behaviors of teachers toward students who perform well might also foster better relationships between students and their teachers, and thus a desire to please their teacher by behaving in ways that conform to expectations. If high academic achievement reinforces and enhances productive learning-related behaviors, we would expect academic achievement in one grade to predict learning-related behavior at a later grade with previous learning-related behavior held constant. If the causal relationship is reciprocal, predictions should be significant in both directions, with learningrelated behavior predicting later achievement, and achievement predicting later learning-related behavior. The current study is longitudinal, with children entering the study in kindergarten or first grade, and then being reassessed in third- and again in fifth grade. Mostly minority children, all Running head: LEARNING-RELATED BEHAVIORS 6 from low-income families, are included in the study because these children are most at risk of school failure (Lee, 2002; Lee & Burkam, 2002) The three time points allowed us to assess the possibility that the direction of the relationship between learning-related behavior and literacy skills is different for early and later grades, another issue that has not been previously addressed. Consistent with previous findings we expected children who exhibited good learning-related behaviors at the beginning of elementary school to learn more and thus achieve literacy skills at a faster rate than children who began school with poor learning-related behaviors. By third grade children have had many opportunities for good academic behavior to be reinforced by teachers. We reasoned, therefore, that the direction of the relationship between learning-related behavior and literacy skills may change, or at least become more reciprocal in the later grades. The association was therefore examined separately for children as they moved from kindergarten or first grade to 3rd grade and from 3rd grade to 5th grade. Direct comparisons of the associations between learning-related behavior and achievement at different grades have not been made in previous studies. Next we discuss previous findings on associations between academic achievement and classroom behaviors that are similar to what we refer to as learning-related behavior. Associations between Learning-Related Behaviors and Academic Achievement Several studies have found significant contemporaneous associations between young children’s learning-related behavior and academic performance. Normandeau and Guay (1998) reported that first-graders’ “cognitive self-control” (the ability to plan, evaluate and regulate problem-solving activities, attend to tasks, persist, and resist distraction) was associated with their academic achievement, net of their intellectual skills assessed in kindergarten. Howse, Lange, Farran, and Boyles (2003) found that teachers’ ratings of kindergarteners’ (but not second graders’) motivation (e.g., “is a self-starter;” “likes to do challenging work”) predicted concurrent reading achievement with receptive vocabulary (but not previous reading achievement) held constant. And Ladd, Kochenderfer, and Coleman (1996) found that teachers’ Running head: LEARNING-RELATED BEHAVIORS 7 ratings of children’s engagement and independence (e.g., “seeks challenges,” “works independently”) predicted their academic progress and performance in kindergarten. In a longitudinal study of children from kindergarten through second grade conducted by McClelland et al. (2000), teachers’ ratings of kindergarten children’s work-related skills (compliance with work instructions, memory for instructions, and completion of games and activities) were significantly associated with children’s academic performance in kindergarten, controlling for IQ. Work-related skills in kindergarten also predicted academic performance at the end of second grade, with kindergarten academic scores controlled. In a more recent study, McClelland et al. (2006) found that learning-related behavior in kindergarten predicted reading and mathematics scores in sixth grade and growth in reading and math between kindergarten and second grade, but not between second and sixth grade. The measure they used was very broad, including social interaction and participation in play activities as well as task behavior (e.g., working independently and organizing work products). Matthews et al. (2009) report that classroom self-regulation (e.g., completes tasks successfully) in the fall of kindergarten predicted growth in math skills over the academic year. Green and Francis (1988) found that learning style (e.g., settles down well at an activity that needs concentration, willing to try on his/her own, copes with something new without getting nervous or upset) in 5- and 6-year-olds predicted reading scores four years later, when children were 9 and 10 years old. The study did not, however, hold constant previous reading scores. Social-class, Race, and Gender Difference in Learning-Related Behavior Only a few studies report group differences in learning-related behaviors. The McClelland et al. (2000) study assessed associations at the beginning of kindergarten between work-related skills and a variety of child and family variables. They found that children rated low on work-related behaviors were more likely than children with medium or high ratings to be Black and to have parents with relatively low education and occupational status. Connell and Prinz (2002) similarly reported that ethnic minority children had lower behavior regulation, but the effects disappeared when social class was controlled. McClelland et al. (2006) similarly Running head: LEARNING-RELATED BEHAVIORS 8 found that kindergarten children rated low on learning-related skills had relatively poorly educated mothers. Analyses of the nationally representative Early Childhood Longitudinal Study (ECLS) likewise revealed that kindergarten children from families with multiple risks (parents who have not completed high school, low-income or welfare-dependent, single-parents, speak a language other than English) were less likely to be rated as eager to learn and attentive than children from families with one or no risks (National Center for Education Statistics, 2001). In their study of children in kindergarten through second grade, Howse et al. (2003) did not find a significant difference in teachers’ ratings of at-risk (low-income) and middle-class children’s motivation (e.g., self-initiation, preference for challenge). But the at-risk and not-at-risk children came from different schools, and teachers may have used different standards to rate children. Given the limited range of family income (and concomitantly parent education) of children in the present study, we expected income effects to be modest. But previous studies suggest that children who are in the most impoverished circumstances have more negative outcomes than children who are also living below the federal poverty line but relatively better off (Dearing, Taylor, & McCartney, 2006; McLeod & Shanahan, 1996). If income effects were found, we expected the children from the lowest-income families in our low-income sample to be rated lower on learning-related behavior. Several studies have reported gender differences in children’s task-related behavior. Ponitz et al. (2008) report that 3-6-year-old girls performed better on their laboratory task assessing behavioral self-regulation than boys. Matthews et al.(2009) report a gender difference favoring girls for both teacher ratings and a direct measure of self-regulation. Keogh (1994) studied 360 children who were in four different school placements--general elementary school, general preschool, special education elementary school, and special education preschool. Teachers rated girls higher than boys on a measure of task orientation (“school-appropriate behaviors”), which included items such as: able to begin and complete classroom tasks, alert and attentive to classroom proceedings, completes work on time, eager, enthusiastic, enjoys school work, follows directions, and willingly participates in classroom activities. Girls were also rated Running head: LEARNING-RELATED BEHAVIORS 9 as more eager to learn, attentive in class, and persistent in completing tasks by kindergarten teachers in the ECLS study (National Center for Education Statistics, 2001), as having better work-related skills in the McClelland et al. (2000) study, and as more self-directed in a study by Birch and Ladd (1997). Fantuzzo, Bulotsky-Shearer, Fusco, and McWayne (2005) did not find gender differences in ratings of Head Start children’s willingness to take on and complete tasks, but this study was done at the preschool level, when expectations for task completion might have been less clear. If gender differences in learning-related behavior were found in the present study, we expected them to favor girls. Summary There is a fair amount of evidence demonstrating that learning-related behaviors are associated with contemporaneous academic achievement, and some evidence that these behaviors in the early grades of school predict later academic achievement. Although most researchers assume that children’s behaviors cause academic achievement, the opposite and reciprocal causal directions have not been assessed. To examine the directionality of the association between learning-related behavior and achievement in the current study we assessed the degree to which learning-related behavior predicted literacy achievement in a later grade, controlling for previous literacy achievement, and the degree to which literacy achievement predicted later learning-related behavior, controlling for previous learning-related behavior. Because we had assessed children on both variables at three points in time, we were able to assess the nature of the associations within two separate time spans. In the few studies that have reported gender or social class differences in learning-related behavior, boys and children from economically disadvantaged or high-risk homes were rated relatively low in work-related skills. We accordingly predicted relatively low learning-related behavior ratings for boys and the most disadvantaged children. Method Sample Running head: LEARNING-RELATED BEHAVIORS 10 Overview. The data used in this study are from a longitudinal, multi-state effort to track development in low-income children through elementary school. The children included in the present study had been enrolled in a previous study of low-income children; their selection in the original study was based only on having incomes (based on needs/income ratio) below the federally established poverty line. All of the children in the original study who could be found and whose parents consented were included in the present study, which followed children from kindergarten or first grade through the fifth grade. Some of the children had already completed kindergarten when the present study began in 1996, and thus could not begin their participation until the first grade. Kindergarteners and first graders were combined for analyses. The children who entered the study at either kindergarten or first grade were re-assessed in third-grade and again in the fifth grade. Those who began the study in kindergarten versus first grade did not differ in terms of gender, race, or family income and their learning-related behavior scores did not differ throughout the duration of the study. Children. Altogether there were 379 children in the study. Of the 267 children who entered the study in kindergarten, 136 were girls and 130 were boys and one did not have gender recorded. Of the 112 children who began the study in first grade, there were 48 girls and 63 boys, and one did not have gender recorded. The mean age of the children in kindergarten was 6 years and 1 month; in first grade it was 6 years and 9 months. The children lived in rural and urban communities in three states, two in the northeast and one on the west coast. The sample was ethnically diverse. The children included in these analyses were 36.5% Caucasian, 34.7% African-American, 23.6% Latino, 1.5% of mixed ethnic backgrounds, and 0.5% other ethnicity, and 3.2% missing this information. At the beginning of the study, 75% of the Latino children were given all assessments in Spanish; they were either rated by their teachers as having no or very little fluency or as having modest fluency (difficulty making themselves understood in English), or the child expressed a preference for being assessed in Spanish. By fifth grade all of the children preferred being assessed in English and their teachers Running head: LEARNING-RELATED BEHAVIORS 11 rated them as fluent or fairly fluent (has only occasional problems either with grammar, vocabulary, or pronunciation). All of the children came from families who met federal poverty guidelines, based on income and household size. At the beginning of this study 76% of the families had annual incomes below $15,000 and about half were receiving federal assistance (AFDC). The distribution of income did not change much, increasing only $3,604 per year, on average. Parent education was low; 50.6% of mothers had not completed high school; 33% had completed high school, 14.5% completed some college, and 1.2% had bachelors or graduate degrees. Children’s receptive vocabulary skills (which are strongly associated with other measures of cognitive skills) were relatively low on average; 60-month standardized PPVT (Peabody Picture Vocabulary Test) scores averaged 86.02 (SD = 14.21). Overall, attrition in the study was modest (less than 19% over a period of six years). Analyses comparing children who were lost to those who were still in the sample at 5th grade revealed no significant differences in family income, gender, or cognitive skills (measured at 60 months, before the present study began). A chi square analysis of ethnic differences in attrition rates was, however, significant, X2(4) = 18.62, p ≤ .001. A higher proportion of Latino children were lost than either African American or Caucasian children. Teachers and schools. Teachers and schools were invited to participate because they had a study child enrolled. In kindergarten and 1st grade the children were distributed across 58 schools and 48 school districts. By 5th grade children were distributed across 101 schools. The modal number of study children in a classroom was 1. The number of teachers involved was as follows: kindergarten, 71; 1st grade, 107; 3rd grade, 127; 5th grade, 135. Across grades, nearly all teachers held Bachelor’s Degrees and about half held Master’s Degrees. There was a huge range in teacher experience from novice teachers to teachers near retirement. Most children in the study attended public schools that served a relatively high proportion of low-income children. Averaging over the 5-year period of data collection between 1996 and Running head: LEARNING-RELATED BEHAVIORS 12 2001, 74% of schools had 50% or more students eligible for free or reduced lunch (higher than the national average, which was about one-third). Procedures Principals of schools that were attended by a child who had parent approval to participate in the study were sent a letter, explaining the purpose of the longitudinal study and our desire to include the child. Only two principals refused or were not possible to reach. District approval processes were followed when required. When district and principal approval was achieved, teachers were contacted by letter and a follow-up phone call. Informed consent was received by all participating teachers and child assent by all participating children. Data were collected each year in the spring, at the end of kindergarten or 1st grade and again in 3rd and 5th grades. Informed parental consent was obtained for all participating children. Of the children assessed in the 3rd - and 5th grade, 23 had been retained in kindergarten, 1st or 2nd grade, and 5 more were retained in 3rd or 4th grade. For these children, there was an additional year between the first and second or second and third set of assessments. Literacy achievement was assessed individually at the child’s school by a trained experimenter. The number of experimenters varied from 27 in kindergarten and 1st grade to 16 in 5th grade. Experimenters had one full day of training before a trainer observed them assessing a child. Before being certified by the lead trainer, experimenters had to satisfy high standards for interacting in a supportive way with children and following protocols. They were observed intermittently following certification to ensure consistent effectiveness and appropriately supportive interactions with young children. Teachers completed a questionnaire about each study child in their classroom, which included ratings of children’s learning-related behavior. Questionnaires were mailed or given to teachers and returned by mail. Information on family income was obtained in an interview with a parent, usually the mother, conducted in homes or over the telephone by a trained interviewer and in the native language of the parent (English or Spanish). Running head: LEARNING-RELATED BEHAVIORS 13 Measures Learning-related behavior. Four items (e.g., works independently, seeks challenges, accepts responsibility for a given task, tuned in to what’s going on in the classroom) from the Teacher Rating Scale of School Adjustment (TRSSA; Birch & Ladd, 1997; Ladd, Birch, & Buhs, 1999; Ladd, Kochenderfer, & Coleman, 1997) were used as a measure of learning-related behavior. All items have a 3-point response scale (doesn’t apply, applies sometimes, and certainly applies). Previous research using the TRSSA scales show high reliabilities for children the same age as in this study as well as associations with child outcomes, such as feelings about school and academic achievement (Ladd et al., 1997). In Ladd et al. (1999), the TRSSA scores mediated the relationship between cognitive maturity and academic achievement and between the quality of teacher-child relationships and achievement. Ladd et al., (1997) report that scores on the TRSSA were positively associated with children’s social skills and relationships, and their own reports of how much they liked school; they were negatively associated with children’s desire to avoid school. In this previous study, for boys but not girls, TRSSA scores were negatively correlated with children’s reports of having conflict with peers. Although the original scale includes additional items, the alphas in previous studies were so high that additional items were somewhat redundant. We limited the scale to four items to reduce the burden on teachers. Reliabilities (alpha) for the four items used in the present study were high, .87, .85, .87 and .85, for kindergarten, 1st-, 3rd- and 5th grades, respectively. The mean of the ratings were used in analyses. The significant correlations across years, shown in Table 1, provide some evidence from the current study of the validity of the measure, given that different teachers were rating children in entirely different classroom contexts. Literacy achievement. Two subtests of the Woodcock-Johnson psycho-educational battery-revised (Woodcock & Johnson, 1990) were used for the 3rd- and 5th graders: the letterword reading and the passage comprehension. The first test asks children to decode single words. In the second, children read a passage and fill in a missing word. The Woodcock-Johnson letterword identification subscale was used more as a model for the kindergarten and 1st-graders. Running head: LEARNING-RELATED BEHAVIORS 14 (Pilot testing revealed very little variation on the letters part of the test, and a floor effect on the words.) Instead of asking children to identify only the 9 letters in the Woodcock Johnson, they were shown all 26 letters of the alphabet. Kindergarten and 1st-grade children were also given a larger number of simple (2 - 4 letters) words to identify than are provided in the Woodcock Johnson test. All subscale scores were converted to standardized scores and then combined within each grade to create an overall mean literacy achievement score. Alphas for the literacy scores at each grade were as follows: kindergarten, .82; 1st-grade, .82; 3rd-grade, .81; 5th-grade, .74. Family income. The parent interviewed (usually the mother) was asked about the total family income for the previous year coming from any source (including from jobs, business, farm, rent, dividends, alimony, child support, welfare, social security payments and unemployment insurance). Respondents were given 10 ranges (e.g., less than $3,000, $3,0016,000, etc.) from less than $3,000 to over $50,000, and asked in which range their income fell. Because most parents answered the question multiple years, responses were averaged. The 10 ranges were then collapsed into three for analyses ($9,000 or less; $9,001-20,000; and more than $20,000). Note that in this sample, the highest income group is relatively low compared to the general population. Results The primary research question concerned the direction of effects between learning-related behavior and literacy skills. Does learning-related behavior promote better literacy learning or does relatively good mastery of literacy skills, and the presumed concomitant rewards for good behavior, promote more learning-related behavior, or is the relationship reciprocal? Data Imputation As is common in longitudinal datasets, subjects often were missing one or more values. We explored these cases with missing values to determine whether the data were missing at random or whether there were systematic differences between students missing data as compared with those who were not. We conducted independent sample t-tests comparing students missing Running head: LEARNING-RELATED BEHAVIORS 15 data with those who were not missing data for each grade level (K-1, 3, 5) and key measure (literacy, learning-related behavior) to determine whether the groups differed on the other key measures. For example, we compared students who had literacy scores in grade 3 with those who were missing them on the five remaining measures of learning-related behavior and literacy. We ran these comparisons for the sample overall and also, separately by ethnicity (African American, Latino, White), resulting in a total of 120 comparisons (30 tests x four groups). Of these 120 comparisons, only one was significant at the p < .05 level, which would be expected by chance. We concluded that the data were missing at random and therefore it was appropriate to replace the missing data using multiple imputation. We replaced missing values using the “ice” multiple imputation procedure in Stata to create five datasets. Multiple imputation, when used in appropriate contexts, is considered a superior method for dealing with missing values as compared with traditional approaches such as mean substitution, case-wise deletion, and even more recent approaches such as single imputation (Acock, 2005). The final sample with imputation was 379. Regression Analyses To assess evidence for the causal order of learning-related behavior and literacy skills we adopted a strategy similar to that developed by Marsh (1990; Marsh, Byrne & Yeung, 1999). Whereas they used structural equation modeling, we ran four regression analyses to predict literacy and learning-related behavior in grade 3 and in grade 5, with literacy and learningrelated behavior in prior years as predictors. This approach allowed us to examine evidence for causal ordering separately for the earlier and the later grades, as well as to replicate the findings at two different grade intervals. We did not use a nested design for analyses because the modal number of children in a classroom was one, and intraclass correlations across schools were all below .10 and averaged .04. In assessing the relationship between literacy and learning-related behavior, we held constant the effects of gender, ethnicity, and household income in kindergarten by including them in the model as predictors, giving us the following model: Running head: LEARNING-RELATED BEHAVIORS 16 Y i = B 0 + B1 X 1 i + B 2 X 2 i + B 3 X 3 i + B 4 X 4 i + B 5 X 5 i where Y i = Predicted outcome (literacy or learning-related behavior in Z-scores), B 0 = Intercept, B 1 = slope coefficient for prior literacy, X 1 i = prior score for literacy for student i, B 2 = slope coefficient for prior learning-related behavior, X 2 i = prior score for learning-related behavior for student i, B 3 = coefficient for gender (female=1), X 3 i = gender of student i, B 4 = coefficient for ethnicity (African American, Latino), X 4 i = ethnicity of student i, B 5 = coefficient for Kindergarten household income, = household income for student i. Using the three categories of income described earlier, we created dummy-coded variables for low income, medium income, and high income, and in the regression model used low income as the reference group. For ethnicity, African American and Latino were entered as predictors, with Caucasian as the reference group. The multiple imputation procedure created five data sets, and because of this, standard errors would be artificially low without modeling the effects of the imputation process. In order to aggregate the results from these different imputations in an appropriate way, we used the micombine procedure in Stata that implements the aggregation method recommended by Rubin (1987). This procedure averages the parameter estimates for the separate regressions done for each imputed data set and adjusts the standard errors by utilizing information on within-imputation and between-imputation variation. In this way, it accounts for the variability across imputations, increasing standard errors to the extent that results differ across the imputed data sets. The aggregated results for the four regression analyses using this method are shown below in Table 2, and graphically in Figure 1. The regression results showed a similar pattern of relationships in the lower and upper grades, K-1 to grade 3, and grade 3 to grade 5, although there were some differences in the magnitude of the relationships. Our primary concern was with the coefficients represented by the diagonal lines (between variables, over time), which is the extent to which literacy and learningrelated behavior predict each other in a later grade. Learning-related behavior at the prior grade Running head: LEARNING-RELATED BEHAVIORS 17 was a very strong predictor of later literacy skills with prior literacy skills held constant for both 3rd and 5th grade. Literacy skills at the prior grade predicted learning-related behavior with previous learning-related behavior held constant at 5th, but not as 3rd grade. The stability coefficients for both literacy and learning-related behavior were statistically significant and positive in all cases. The coefficients for learning-related behavior were very similar in size (0.32 vs. 0.31), whereas for literacy the coefficient predicting grade 5 based on grade 3 was much larger than the coefficient predicting grade 3 literacy from literacy in Kindergarten-grade 1 (0.26 vs. 0.69). We conducted analyses to determine whether stability patterns might be the result of collinearity between these variables and other predictors by removing the other predictors from the model, and then predicting literacy and learning-related each by the prior score on the same construct. We found essentially the same results and therefore we ruled out collinearity as an explanation for the pattern of results for stability coefficients. Gender, Race, and Grade Effects We also conducted a 2 x 3 x 3 repeated measures analysis of variance (ANOVA) on learning-related behavior, with gender, race, and income as between-subject variables and grade as a within-subject variable. (The 16 children who were not Latino, African American or Caucasian were eliminated from this analysis.) The gender, F(1, 338) = 4.09, p < .001, and the grade, F(16.82, 385), p < .01, main effects were significant. Girls (M = 2.45) were rated higher in learning-related behavior than boys (M = 2.28), and children in K-1 (M = 2.34) were rated higher than children in both 3rd (M = 2.15) and 5th (M = 2.13) grade. The gender by race interaction was also significant, F( 2, 322 ) = 3.95, p < .05. Latino boys (M = 2.36) were rated higher than African American (M = 2.15) and White (M = 2.11) boys (p < .05). The race differences were not significant for girls. The race and income main effects, all other two-way interactions and the three-way interaction were not significant. A second 2 x 3 x 3 repeated measures ANOVA on literacy using the same predictors did not find any statistically significant differences. Discussion Running head: LEARNING-RELATED BEHAVIORS 18 This study builds on prior research finding associations between self-regulation and children’s learning (e.g., Blair, 2002; Bull & Scerif, 2001; Green & Francis, 1988; McClelland et al., 2006; McClelland et al., 2007; National Institute of Child Health & Human Development [NICHD] Early Child Care Research Network, 2003). The present study is different from most previous studies because it focused exclusively on children living in poverty. It also goes beyond previous studies by assessing evidence related to alternative causal directions. We assessed evidence for the hypothesis that children who performed well in literacy, and were presumably reinforced for their good behavior, would increasingly exhibit positive learning-related behavior, as well as for the hypothesis that children who exhibited positive learning-related behaviors develop more literacy skills. By analyzing changes from K-1 to 3rd grade separately from changes from 3rd to 5th grade, we were able to examine age differences in the nature of the relationship, as well as embed a replication in this study. The evidence clearly and consistently supported the view that positive learning-related behavior promotes literacy achievement. Learning-related behaviors in kindergarten and 1st grade predicted literacy skills in 3rd grade, over and above the effects of kindergarten and 1st grade literacy. The same was found for children moving from 3rd to 5th grade. The results cannot be explained by the three possible confounds included as covariates: gender, ethnicity, and family income. There was also some modest support for the hypothesis that high achievement led to an increase in positive learning behavior in the later grades in elementary school. Presumably by 3rd grade, children with strong literacy skills have been reinforced with teacher praise, good grades, and the like for relatively high performance in school. The reinforcement, in turn, may explain the increase in the behaviors that preceded the good performance. This interpretation is speculative because we do not have data on grades or other forms of reinforcement. If it is valid, it suggests a kind of multiplicative effect of beginning school with strong learning-related behaviors. Not only may the produce better reading skills directly, but the higher reading skills lead to even stronger learning-related behaviors. Running head: LEARNING-RELATED BEHAVIORS 19 These findings, along with other studies finding associations between children’s learning behavior in the early grades and their academic success in school (e.g., Green & Francis, 1988; McClelland et al., 2000, 2006; Matthews et al., 2009), provide strong evidence for the importance of learning-related behavior, and they suggest that we may be able to promote literacy skills by helping children develop good work habits in preschool and elementary school. Although the study does not provide information on how learning-related behaviors contributed to the development of the wide range of literacy skills assessed, there are a number of possible explanations for the effect. For example, children who pay attention, are responsible, and who work independently, should be more likely to complete their work and generally to take good advantage of opportunities to learn. Children who seek challenge are also more likely to have opportunities to expand their skill levels. The children in this study all came from very low income families, so it is not surprising that learning-related behavior did not vary systematically among the income levels involved and that income could not explain the effects of learning-related behavior on literacy skill development. The findings do, however, suggest that learning-related behaviors are important for low-income children. Boys were rated somewhat lower than girls on our measure of learning-related behavior. Few studies have assessed gender differences, but our findings are in the same direction as those, reviewed above, which reported differences. Gender difference in such behaviors may partially explain previous studies that have found girls to have closer and less conflictual relationships with their teachers (e.g., Birch & Ladd, 1997). It is likely that teachers prefer students who exhibit the learning-related behaviors assessed in the present study because they make students easier and more enjoyable to teach (Keogh, 1994). Beginning school with these behaviors better developed could give girls an advantage that increases with time. In addition to helping them master the material they are taught, these behaviors may also lead to favored treatment and positive relationships with teachers, making school generally a more pleasant and comfortable place for girls. Running head: LEARNING-RELATED BEHAVIORS 20 We do not attempt to explain the higher learning-related behavior ratings given to the kindergarten and 1st graders because it is likely that teachers’ interpretation of the items and their expectations for age appropriate behavior varied by grade. The significant grade effect does not necessarily mean that children’s learning-related behaviors declined after the first two grades of school, although the gap between children’s behavior and teachers’ expectations may have increased. With regard to gender and ethnicity differences, we found one potentially important significant difference. Latino boys were rated higher by teachers in learning-related behavior than White and African-American boys. The differences may reflect cultural expectations and norms. There is some evidence that responsibility and respect of teachers is stressed in Latino homes (Goldenberg & Gallimore, 1995) and in a related study Latino children were rated higher than Euro-American classmates on classroom social skills (e.g., participates well in a group; O’Neil, Welsh, Parke, Wang, & Strand, 1997). It is possible that the Latino boys had been socialized at home to take responsibility and to respect the demands the teacher made on them as students. The methods used have several limitations. First, the measure of learning-related behavior was modest and focused on behaviors that can be observed in the classroom. Although the four items seemed to be sufficient to capture meaningful variance in children’s behavior, a more robust measure is advisable. We caution researchers, however, to maintain clarity in their measures and avoid mixing constructs (e.g., attention, motivation) which may be correlated with learning-related behaviors but are conceptually distinct. Although it is one of the few standardized achievement measures available in both English and Spanish, the Woodcock Johnson measure is also very limited, and an older version of the measure was used in this study than the measure that is currently available. The findings are also limited to children from lowincome families who attend schools with a relatively high concentration of low-income children. The findings are, therefore, not necessarily generalizable to more middle-class samples or school contexts. Running head: LEARNING-RELATED BEHAVIORS 21 Research on self-regulation is burgeoning, and would be helped by future studies that include independent measures of the different components of self-regulated learning, analyze associations among them, and assess differential predictions of achievement and other important child outcomes. Studies using multiple measures of self-regulation would help identify which of the cognitive, affective, and motivational variables considered part of self-regulation are most directly connected to learning gains. Practical Implications Despite the limitations, the findings of this study suggest the importance of efforts to help young children develop positive classroom behaviors in preschool and the early grades of elementary school. We do not have systematic evidence on how to promote behaviors such as independence, responsibility, and challenge-seeking in young children. But programs designed to promote self-regulation skills are being developed. Tools of the Mind, for example, trains teachers to implement exercises that are designed to help children develop strategies for controlling their own behavior (e.g., telling themselves out loud what they should do), and enhancing their memory and attention. A study involving random assignment of both teachers and children demonstrated the value of the program on a measure of executive function (Diamond, Barnett, Thomas, & Munro, 2007), which most likely underlies the kinds of learningrelated behavior examined in the present study. Bedrova and Leong (2006) summarize strategies teachers have been encouraged to use to support the development of self-regulation, including using props (e.g., pictures that remind children what they are supposed to be doing, such as an ear to remind them to listen), and analyzing and redesigning activity settings that might undermine self-regulation, such as circle time that is too long or choice times that provide too few or too many options. They also suggest that teachers introduce interventions slowly, withdrawing their own involvement to give students increasing amounts of autonomy. Our own observations of classrooms suggest additional classroom qualities that might promote learning-related behavior. First, children are more likely to develop independence and a Running head: LEARNING-RELATED BEHAVIORS 22 sense of responsibility in classrooms in which they are given some discretion but held accountable for their work than in classrooms in which children are told what to do and how and when to do it all day long, or who are given discretion but not held accountable. Likewise, we expect children to seek challenges in classrooms in which they have options, are encouraged to attempt more difficult tasks, and where there are no negative consequences for failure. In classroom contexts in which children risk peer ridicule or teacher irritation when they have difficulty with a task, or in which only successful outcomes are rewarded (e.g., by teacher praise, good grades or having papers displayed on the bulletin board), we would expect children to select tasks that they are confident they can complete successfully and to be concerned and dependent on the teacher to make sure they are doing what the teacher has in mind. We have observed preschool and kindergarten classrooms that are organized to promote independence and responsibility. Materials are directly accessible to children, and they are expected to return materials to their places. Tasks (e.g., at centers) are clearly defined and can be done with minimal teacher assistance. Children are taught directly how to use classroom resources (e.g., letters on the walls, information on the board, computers) and each other to solve problems and complete tasks. Teachers hold children accountable for their learning by regularly reviewing work and giving feedback and by interacting with children while they are working on tasks to assess understanding and make suggestions. Preschool is not too soon to invest in children’s learning related behavior. This study, like many others, found that literacy skills at the beginning of school predicted literacy skills at 5th grade. Other studies have shown predictions from school entry to high school (see Stipek, 2001, for a review). Indeed, the authors of one meta-analysis estimated that about half of the total black-white math and reading gap at the end of high school is explained by the gap between blacks and whites at school entry (Phillips, Crouse, & Ralph, 1998). There are many reasons why early behavior and academic skills might affect later achievement. Children who enter with relatively productive behavior and good skills may be able to take better advantage of the academic curriculum. Their behavior and skills also engender higher expectations in teachers, Running head: LEARNING-RELATED BEHAVIORS 23 which can lead to placement in higher ability groups and more challenging instruction (see Stipek, 2001). There are other reasons to promote independence, responsibility and challenge-seeking in children. These behaviors will serve children well in and out of school and throughout their lives. But even if academic achievement is the only concern, the results of this study suggest that learning-related behaviors merit direct and proactive attention. Running head: LEARNING-RELATED BEHAVIORS 24 References Acock, A. C. (2005). Working with missing values. Journal of Marriage and the Family, 67, 1012-1028. Barnett, S. (1995). Long-term effects of early childhood programs on cognitive and school outcomes. Future of Children, 5, 25-35. Barnett, S., Epstein, D., Friedman, A., Stevenson Boyd, J., & Hustedt, J. (2008). The state of preschool. New Brunswick, NJ: National Institute for Early Education Research. Bodrova, E., & Leong,D. (2006). Self-regulation as a key to school readiness: How early childhood teachers can promote this critical competency. In M. Zaslow & I. MartinezBeck (Eds.), Critical issues in early childhood development. (pp. 203-224). Baltimore, MD: Brookes Publishing Company. Birch, S., & Ladd, G. (1997). The teacher-child relationship and children’s early school adjustment. Journal of School Psychology, 35, 61-79. Blair, C. (2002). School readiness: Integrating cognition and emotion in a neurobiological conceptualization of children’s functioning at school entry. American Psychologist, 57, 111-127. Blair, C. (2003). Behavioral inhibition and behavioral activation in young children: Relations with self-regulation and adaptation to preschool in children attending Head Start. Developmental Psychobiology, 42(3), 301-311. Brown, A. (1987). Metacognition, executive control, self-regulation, and other more mysterious mechanisms. In F. Reiner & R. Kluwe (Eds.), Metacognition, motivation, and understanding (pp. 65-116). Hillsdale, NJ: Erlbaum. Bull, R., & Scerif, G. (2001). Executive functioning as a predictor of children's mathematics ability: Inhibition, switching, and working memory. Developmental Neuropsychology, 19, 273–293. Running head: LEARNING-RELATED BEHAVIORS 25 Cole, P. M., Martin, S. E., & Dennis, T. A. (2004). Emotion regulation as a scientific construct: Methodological challenges and directions for child development research. Child Development, 75, 317–333. Connell, C., & Prinz, R. (2002). The impact of childcare and parent-child interactions on school readiness and social skills development for low-income African American Children. Journal of School Psychology, 40(2), 177-193. Dearing, E., Taylor, B.A., & McCartney, K. (2006). Within-child associations between family income and externalizing and internalizing problems. Developmental Psychology, 42(2), 237-252. Diamond, A., Barnett, S., Thomas, J., & Munro, S. (2007). Preschool program improves cognitive control. Science, 318, 1387-1388. Fantuzzo, J., Bulotsky-Shearer, R., Fusco, R., & McWayne, C. (2005). An investigation of preschool classroom behavioral adjustment problems and social-emotional school readiness competencies. Early Childhood Research Quarterly, 20, 259-275. Farran, D. (2000). Another decade of intervention for children who are low-income or disabled: What do we know now? In J. Shonkoff & S. Meisels (Eds.), Handbook of early childhood intervention (2nd ed., pp. 510-548). New York: Cambridge University Press. Foulks, B., & Morrow, R. (1989). Academic survival skills for the young child at risk for school failure. Journal of Educational Research, 82, 158-165. Goldenberg, C., & Gallimore, R. (1995). Immigrant Latino parents' values and beliefs about their children's education: Continuities and discontinuities across cultures and generations. In P. R. Pintrich, & M. Maehr (Eds.), Advances in motivation and achievement: Culture, ethnicity, and motivation, Vol 9 (pp. 183-228). Greenwich, CT: JAI Press. Green, L., & Francis, J. (1988). Children’s learning skills at the infant and junior stages: A follow-on study. British Journal of Educational Psychology, 58, 120-126. Happaney, K., Zelazo, P. , & Stuss, D. (2004). Development of orbitofrontal function: Current themes and future directions. Brain and Cognition, 55, 1–10. Running head: LEARNING-RELATED BEHAVIORS 26 Howse, R., Lange, G., Farran, D., & Boyles, C. (2003). Motivation and self-regulation as predictors of achievement in economically disadvantaged young children. The Journal of Experimental Education, 71, 151-174. Keogh, B. (1994). Temperament and teachers' views of teachability. In W. Carey & S. McDevitt (Eds.), Prevention and early intervention: Individual differences as risk factors for the mental health of children (pp. 246-256). New York: Brunner/Mazel. Ladd, G., Birch, S., & Buhs, E. (1999). Children’s social lives in kindergarten: Related spheres of influence. Child Development, 70, 1373-1400. Ladd, G., Kochenderfer, B., & Coleman, C. (1996). Friendship quality as a predictor of school early adjustment. Child Development, 67, 1103-1108. Ladd, G., Kochenderfer, B., & Coleman, C. (1997). Classroom peer acceptance, friendship, and victimization: Distinct relational systems that contribute uniquely to children’s school adjustment? Child Development, 68, 1181-1197. Lee, J. (2002). Racial and ethnic achievement gap trends: Reversing the progress toward equity? Educational Researcher, 31, 3-12. Lee, V., & Burkam, D. (2002). Inequality at the starting gate: Social background differences in achievement as children begin school, Washington, D.C.: Economic Policy Institute Lewit, E. M., & Baker, L. S. (1995). School readiness. The Future of Children, 5, 128-139. Mac Iver, D., Stipek, D., & Daniels, D. (1991). Explaining within semester changes in student effort in junior high school and senior high school courses. Journal of Educational Psychology, 83, 201-211. Marsh, H. (1990). Causal ordering of academic self-concept and academic achievement: A multiwave, longitudinal panel analysis. Journal of Educational Psychology, 82, 646-656. Marsh, H., Byrne, B., & Yeung, A. (1999). Causal ordering of academic self-concept and achievement: Reanalysis of a pioneering study and revised recommendations. Educational Psychology, 34, 154-157. Running head: LEARNING-RELATED BEHAVIORS 27 Matthews , J., Ponitz, C. Morrison, F. (2009). Early Gender Differences in Self-Regulation and Academic Achievement, Journal of Educational Psychology, 101, 689-704. McClelland, M., Acock, A., & Morrison, F. (2006). The impact of kindergarten learning-related skills on academic trajectories at the end of elementary school. Early Childhood Research Quarterly, 21, 471-490. McClelland, M., Cameron, C., Connor, C., Farris, C., Jewkes, A., & Morrison, F. (2007). Links between behavioral regulation and preschoolers' literacy, vocabulary, and math skills. Developmental Psychology, 43, 947-959. McClelland, M. M., & Morrison, F. J. (2003). The emergence of learning-related social skills in preschool children. Early Childhood Research Quarterly, 18(2), 206–224. McClelland, M., Morrison, F., & Holmes, D. (2000). Children at risk for early academic problems: The role of learning-related social skills. Early Childhood Research Quarterly, 15, 307-329. McLeod, J. D., & Shanahan, M.J. (1996). Trajectories of poverty and children’s mental health. Journal of Health and Social Behavior, 37 (September), 207-220. Miller, E., & Almon, J. (2009). Crisis in the kindergarten. College Park, MD: Alliance for Childhood. National Center for Education Statistics (2001). Entering Kindergarten: Findings from the Condition of Education. Washington DC: U.S. Department of Education. National Institute of Child Health and Human Development Early Child Care Research Network (2003). Do children's attention processes mediate the link between family predictors and school readiness? Developmental Psychology, 39, 581–593. Normandeau, S., & Guay, F. (1998). Preschool behavior and first-grade school achievement: The mediational role of cognitive self-control. Journal of Educational Psychology, 90, 111121. Running head: LEARNING-RELATED BEHAVIORS 28 O’Neil, R., Welsh, M., Parke, R., Wang, S., & Strand, C. (1997). A longitudinal assessment of the academic correlates of early peer acceptance and rejection. Journal of Clinical Child Psychology, 26, 290-303. Phillips, M., Crouse, J., & Ralph, J. (1998). Does the black-white test score gap widen after children enter school? In C. Jencks & M. Phillips (Eds.). The black-white test score gap (pp. 229-272). Washington DC: Brookings Institution Press. Ponitz, C., McClelland, M., Jewkes, A., Connor, C., Farris, C., & Morrison, F. (2008). Touch your toes! Developing a direct measure of behavioral regulation in early childhood. Early Childhood Research Quarterly, 23(2), 141-158. Posner, M., & Rothbart, M. (2007). Educating the human brain. Washington DC: American Psychological Association. Rimm-Kaufman, S., Pianta, R., & Cox, M. (2000). Teachers’ judgments of problems in the transition to kindergarten. Early Childhood Research Quarterly, 15, 147-166. Rubin, D. (1987). Multiple Imputation for Nonresponse in Surveys. New York: Wiley. Stipek, D. (2001). Pathways to constructive behavior: Importance of academic achievement in the early elementary grades. In A. Bohart & D. Stipek (Eds.), Constructive and destructive behavior: Implications for family, school, and society (pp. 291-315). Washington DC: American Psychological Association. Stipek, D. (2002). Motivation to learn: Integrating theory and practice (4th edition). Needham Heights, MA: Allyn & Bacon. Stipek, D. (2006a). Accountability Comes to Preschool: Can We Make it Work for Young Children? Phi Delta Kappan, 87 (10), 740-744, 747. Stipek, D. (2006b). No child left behind comes to preschool. Elementary School Journal, 106(5), 455-465. Stipek, D., & Ryan R. (1997). Economically disadvantaged preschoolers: Ready to learn but further to go. Developmental Psychology, 33, 711-723. Running head: LEARNING-RELATED BEHAVIORS Wentzel, K. (1991). Social competence at school: Relation between social responsibility and academic achievement. Review of Educational Research, 61, 1-24. Woodcock, R. W., & Johnson, M. B. (1990). Woodcock-Johnson psycho-educational batteryrevised. Allen, TX: DLM Teaching Resources. 29 Running head: LEARNING-RELATED BEHAVIORS 30 Table 1 Correlation Matrix, Means, and Standard Deviations for Learning-Related Behavior and Literacy Achievement K-1st 3rd 5th L-R behavior Literacy ach. L-R behavior Literacy ach. L-R behavior Literacy ach. (272) (347) (310) (323) (320) (285) (N)1 K-1 L-R behavior Literacy ach. 0.34*** 3rd L-R behavior 0.36*** 0.15* Literacy ach. 0.39*** 0.51** 0.27*** L-R behavior 0.20*** 0.19** 0.34*** 0.19*** Literacy ach. 0.39*** 0.46** 0.34*** 0.71*** 0.18** Mean 2.34 .08 2.15 -.05 2.14 -.06 SD .57 2.66 .53 1.68 .53 1.67 Range 1-3 -4.7-8.9 1-3 -4.9/6.1 1-3 -5.5-3.6 5th *p < .05; ** p < .01; *** p < .001 Note: Means for literacy achievement are close to 0 because they are based on standardized scores. 1 N is before imputation Running head: LEARNING-RELATED BEHAVIORS 31 Table 2 Regression Coefficients (standard errors in parentheses) Dependent Variable Literacy Learning-Related Behavior _____________________________________________________ Predictor Grade 3 Grade 5 Grade 3 Grade 5 _____________________________________________________________________________ Prior Literacy .26*** (.04) .69*** (.02) .01 (.05) Prior L-R Behavior .80*** (.19) .41*** (.06) .32*** (.02) .31*** (.02) Female .05 (.16) -.01 (.06) .03 (.01) .10*** (.02) .13*** (.02) African American -.20 (.19) -.50*** (.07) Latino -.61** (.23) -.03 (.07) Parent Income (med) -.03 (.19) -.01 (.06) -.03 (.03) .00 (.03) Parent Income (high) -.01 (.25) -.05 (.08) .01 (.04) .04 (.04) -1.74*** (.51) -.69*** (.13) Constant ** p < .01; *** p < .001 -.05 (.03) .04 (.03) .13*** (.03) .02 (.03) 1.35*** (.06) 1.38*** (.06) Running head: LEARNING-RELATED BEHAVIORS 32 Figure Caption Figure 1. Parameter estimates from regression analyses of Teacher Ratings of Learning-Related Behavior, and Child’s Literacy Achievement in 1st, 3rd, and 5th Grades, holding constant ethnicity, gender, and parent income Grade K/1 Literacy Grade3 .26*** Literacy .01 .32*** .69*** Literacy .03** .80*** L-R Behavior Grade 5 .41*** L-R Behavior L-R Behavior .31***