Journal of Applied Developmental Psychology 35 (2014) 273–283

Contents lists available at ScienceDirect

Journal of Applied Developmental Psychology

Collective student characteristics alter the effects of teaching practices onacademic outcomes

Eve Kikas a,⁎, Kätlin Peets a,b,d, Ernest V.E. Hodges b,c

a Tallinn University, Institute of Psychology, Narva mnt 25, 10120 Tallinn, Estoniab Utrecht University, The Netherlandsc St. John's University, Department of Psychology, 8000 Utopia Parkway, Jamaica, NY 11439, USAd University of Turku, Finland

⁎ Corresponding author.E-mail addresses: eve.kikas@tlu.ee (E. Kikas), katlin.pe

hodgese@stjohns.edu (E.V.E. Hodges).

http://dx.doi.org/10.1016/j.appdev.2014.04.0040193-3973/© 2014 Elsevier Inc. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 4 June 2013Received in revised form 28 April 2014Accepted 29 April 2014Available online xxxx

Keywords:Classroom contextCollective student characteristicsTeaching practicesAcademic skillsTask persistenceElementary school

The goal of this study was to examine the influence of collective student characteristics (academic skills and taskpersistence at the beginning offirst grade) and different teaching practices (child-centered, teacher-directed, andchild-dominated) on the development of academic skills and task persistence during thefirst two years in school.Wehypothesized that teaching practiceswould differentially impact the development of academic skills and taskpersistence depending on the collective needs of the classroom. Participants were 523 students (273 boys) from32 classrooms across Estonia. By using multilevel modeling, we found several interactions indicating that bothcontextual influences are important in determining subsequent academic functioning and task persistence butthat some teaching practices are more beneficial depending on the collective starting point of students. Thesefindings highlight the importance of studying different contextual influences hand in handwhen trying to under-stand what enhances young children's academic development.

© 2014 Elsevier Inc. All rights reserved.

Success during the first years of school has implications for subse-quent adaptive development (e.g., Jimerson, Egeland, & Teo, 1999).Thus, identifying factors that support (or undermine) academic devel-opment and learning during the early school years is of great impor-tance. Prior knowledge and skills clearly influence the developmentof children's academic skills (math, reading, spelling; Lerkkanen,Rasku-Puttonen, Aunola, & Nurmi, 2004; Passolunghi, Mammarella, &Altoè, 2008). In addition, children who show high task persistenceachieve better academic outcomes (McClelland, Acock, Piccinin, Rhea,& Stallings, 2013; Onatsu-Arvilommi & Nurmi, 2000; Schaefer &McDermott, 1999). Yet, less attention has been devoted to understand-ing how different characteristics of the classroom context shape earlyacademic development.

Classrooms provide both an academic as well as a social context forlearning. Classrooms differ with regard to a variety of characteristics(abilities, beliefs, interests, and behaviors) that children collectivelybring to the classroom. They also vary in the type of instruction,management and socio-emotional support teachers provide theirstudents. In fact, other students in the classroom and teachers consti-tute the most proximal environmental context (outside of home) foryoung children's academic and social development (Bronfenbrenner

ets@utu.fi (K. Peets),

& Morris, 1998; Hamre & Pianta, 2010; Harris, 1995; Wentzel &Watkins, 2002).

Importantly, other classmates and teachers in the early elementaryschool years have the potential to have a greater impact on studentsthan in later years because children remain with the same classmatesthroughout the day and one teacher is responsible for delivering allthe lessons. Although teachers' influence has long been acknowledged,there is a lacuna of studies separating the effects of teachers fromother classroom context effects (see Byrne et al., 2010). Moreover, rath-er than assuming independentmain effects, teaching practices are likelyto bemore (or less) effective in promoting academic skills and task per-sistence depending on the degree to which they match the collectiveneeds of the students (initial skills and task persistence of all the stu-dents in the classroom). Thus, guided by an ecological framework, thislongitudinal study was designed to examine interactive effects amongmultiple contextual-level influences on academic skills (math, textcomprehension, and spelling) and task persistence during the firsttwo years of elementary school.

Development of academic skills and task persistence

There is substantial evidence showing that technical reading skills(e.g., word decoding and reading) in the beginning of school predictfuture reading comprehension (e.g., Bianco, Pellenq, & Lambert,2012; Fuchs et al., 2012; Ortiz, Folsom, & Al Otaiba, 2012), spelling

274 E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

(Lerkkanen et al., 2004), and math (Lerkkanen, Rasku-Puttonen,Aunola, & Nurmi, 2005). Similarly, early mastery of calculation andword problems facilitates later success in math and reading compre-hension (Lerkkanen et al., 2005).

Motivational and affective mechanisms are also important determi-nants of children's learning and acquisition of academic skills (DiPerna,Volpe, & Elliott, 2005; Pintrich & Schunk, 2002). Although differentconceptualizations of motivation and related constructs exist, wefocused on learning behavior (i.e., how children approach differentlearning tasks), and more specifically on task persistence (see Schaefer& McDermott, 1999; Yen, Konold, & McDermott, 2004). Whereaschildren high in task persistence are characterized by showing effortand not giving up easily in the face of challenges and difficult tasks,children low in taskpersistence tend to quitwhen facedwith complicat-ed tasks. There is evidence that young children's task persistence con-tributes to academic achievement over and beyond their cognitiveabilities (Schaefer & McDermott, 1999; Yen et al., 2004). Children whodo not give up easily in the face of obstacles have better reading(Dally, 2006; Onatsu-Arvilommi & Nurmi, 2000) and math skills lateron (Aunola, Nurmi, Lerkkanen, & Rasku-Puttonen, 2003; Dally,2006; Onatsu-Arvilommi & Nurmi, 2000). The reverse is also true—better academic skills increase the likelihood that children stay ontask when faced with obstacles (Aunola et al., 2003; Dally, 2006;Onatsu-Arvilommi & Nurmi, 2000). Thus, when children can easilyfocus on tasks that are challenging, they are more likely to acquirenew skills that further facilitate persistence in tackling, and mastery ofchallenging tasks. Also, children with greater initial abilities are lesslikely to get frustrated and are consequently less likely to give upwhen faced with challenging tasks which, in turn, increases the proba-bility for further developing their academic skills.

Collective student characteristics

Upon school entry, interactions with peers increase dramatically(Rubin, Bukowski, Parker, & Bowker, 2008). Classmates provide one ofthe most important contexts for children's development and learning(Bronfenbrenner & Morris, 1998; Wentzel & Watkins, 2002; Wentzel& Wigfield, 1998). On a daily basis, children see how well their class-mates can read or perform math assignments and how persistentlythey work on different learning tasks. Children also receive consistentfeedback for their own skills and behaviors. Children engage in socialcomparison processes—they observe the persistence and performanceof others with academic tasks to gauge their own capabilities in relationto the group (Guay, Boivin, & Hodges, 1999; Ruble, Feldman, &Boggiano, 1976). For instance, when children see their classmatesperforming well, they are likely to believe that they too can succeed.Such increases in self-efficacy are, in turn, related to increased per-sistence in tackling challenging tasks and greater achievement(e.g., Schunk, 2003). Thus, children are likely to alter their goals, beliefs,and task persistence tomore closely align themselveswith group accept-ed goals, beliefs, and behaviors (Kindermann, 2003;Mercer, McMillen, &DeRosier, 2009; see also Ryan, 2001). Moreover, this is more likely whenchildren have a sense of belongingness and relatedness with other class-mates (Wentzel & Watkins, 2002; Wentzel & Wigfield, 1998).

Classrooms that are characterized by studentswho are eager to learnand exhibit high academic skills are likely to create a very different so-cializing context for children compared to classrooms that include stu-dents who struggle with difficult tasks and have poorer academicskills. Several studies show that children's development is promotedin classrooms with higher levels of academic skills (e.g., Foorman,York, & Santi, 2008; Mashburn, Justice, Downer, & Pianta, 2009) andlearning behavior (Barth, Dunlap, Dane, Lochman, & Wells, 2004;Bulotsky-Shearer, Bell, & Dominguez, 2012). The influence of collectivestudent characteristics is not limited to the development of academicskills but also extends to children's motivation, attitudes toward school,

self-perceptions, and psychologicalwell-being (Marsh,Martin, & Cheng,2008; Rutter & Maughan, 2002; Ryan, 2001).

Teaching practices

Teachers provide another important context for children's learning(Bronfenbrenner & Morris, 1998; Wentzel & Wigfield, 1998). Teachers,and the practices they use, also differ between classrooms. Weused the Early Childhood Classroom Observation Measure (ECCOM;Lerkkanen, Kikas, Pakarinen, Trossmann, et al., 2012; Stipek & Byler,2005) that assesses three dimensions of teaching practices—child-cen-tered, teacher-directed, and child-dominated (see Daniels & Shumow,2003; Lerkkanen, Kikas, Pakarinen, Poikonen, et al., 2012; Lerkkanen,Kikas, Pakarinen, Trossmann, et al., 2012; Stipek& Byler, 2005). These ap-proaches have their theoretical roots in constructivism, behaviorism, andmaturationism, respectively (Daniels & Shumow, 2003). The three prac-tices differ both in the amount and type of instruction, managementpractices, and the level of socio-emotional support teachers provide(Hamre & Pianta, 2010; Stipek & Byler, 2005). Child-centered andteacher-directed practices differ in the type of teacher activities, orin the degree to which teachers allow children to actively constructtheir knowledge (vs. teach basic skills), include children in variousdiscipline-related decision processes (vs. set the rules), and engage increating a positive social climate via individual support and encourage-ment of peer interactions. The primary difference between child-centered/teacher-directed and child-dominated practices lies in theamount of instruction, management, and socio-emotional supportteachers provide.Whereas the first two practices presume active teach-er participation, child-dominated practices are characterized by anoveremphasis on students' “natural development”, and teachers remainrelatively passive observers (Lerkkanen, Kikas, Pakarinen, Poikonen,et al., 2012; Lerkkanen, Kikas, Pakarinen, Trossmann, et al., 2012;Woolfolk Hoy & Weinstein, 2011).

Teachers who engage in child-centered practices are supporters ofchildren's academic and social development and view children as activecontributors to their own learning (McCombs, 2010; Stipek & Byler,2004; Woolfolk Hoy & Weinstein, 2011). In such classrooms, children'sinterests and personal experiences are valued (McCombs, 2010).Teachers give children opportunities tomake their own choices, provideemotional support, personal feedback and encouragement, and pro-mote mastery goal orientations that foster task-persistent behavior(Wentzel, 2010). In contrast, teacher-directed (or didactic) practicesare characterized by teachers' dominance and control. Teachers whoprefer such methods regard students as passive “receivers” and theyview themselves as fully responsible for students' success (Gettinger &Kohler, 2011; Woolfolk Hoy &Weinstein, 2011). Instructional practicesare based on the premise that more complicated learning tasksshould not be introduced to students before they have masteredthe basic academic skills. Teachers who use this approach focus onlecturing, demonstrations, and practice. Children are praised for giv-ing a correct answer rather than for their effort. Teachers rarely ad-just their teaching to the individual needs of children (Gettinger &Kohler, 2011; Stipek & Byler, 2004; Woolfolk Hoy & Weinstein,2011). Teachers who employ child-dominated practices provide littleguidance, control, or support. However, they are responsive to students'questions and demands (Lerkkanen, Kikas, Pakarinen, Poikonen, et al.,2012; Lerkkanen, Kikas, Pakarinen, Trossmann, et al., 2012; WoolfolkHoy & Weinstein, 2011).

Moreover, teachers vary in whether they predominantly use onepractice or implement a mix of approaches depending on what the sit-uation calls for (Stipek & Byler, 2004). Whereas child-centered andteacher-directed practices have been found to be highly negatively cor-related, associations of each with child-dominated practices are consid-erably weaker (Lerkkanen, Kikas, Pakarinen, Poikonen, et al., 2012; seealso Hauser-Cram, Sirin, & Stipek, 2003), suggesting that teachers whouse child-centered or teacher-directed methods can vary in the amount

275E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

of free time and choices they provide for students. Yet, what effect sucha combination of practices has on children's academic skills and taskpersistence has yet to be tested.

Collective student characteristics, teaching practices, andchildren's development

Previous studies have shown that whereas the benefits of child-centered practices appear to extend to different domains of functioning(academic, motivational, social; Hughes, 2002; Perry, Donohue, &Weinstein, 2007; Turner et al., 2002), the benefits of teacher-directedpractices are limited to academic functioning (sometimes to certainareas of academic functioning) and may hinder motivation conduciveto long-term achievement (e.g., Lerkkanen, Kiuru, et al., 2012; Stipek,Feiler, Daniels, & Milburn, 1995). The few studies that have evaluatedthe effects of child-dominated practices indicate that in classroomswhere teachers offer very little support and put too much emphasison free choice, children have difficulties in academic (e.g., Walker,2008) as well as social domains (e.g., Valeski & Stipek, 2001). However,rather than expecting that a particular teaching method is universallybeneficial or detrimental, we reasoned that the effects of teachingpractices are likely to depend on the collective needs of the classroom(e.g., whether children need more direction or autonomy). This is alsoin line with many theoretical accounts. For instance, motivational theo-ries (see Pintrich & Schunk, 2002) emphasize the importance of provid-ing children with adequate challenges in order for motivated andeffortful learning to occur. The idea that the most effective learningtakes place in the zone of proximal development (Vygotsky, 1978)also suggests that optimal developmental outcomes occur when schooltasks correspond to children's skill levels. Although these theories havebeen primarily applied to describing a match (or mismatch) betweenindividual characteristics and the environment, we used the same ideato test whether a match (or mismatch) between the two contextual in-fluences affects academic outcomes.

Although some research shows that the effects of teaching methodsare dependent on children's characteristics, such as age, cognitive abili-ties, and prior academic skills (Alfieri, Brooks, Aldrich, & Tenenbaum,2011; Connor, Morrison, & Katch, 2004; Kiuru et al., 2012), there is adearth of studies examining how the influence of different teachingpractices varies depending on collective student skills and learningbehavior (for an exception, see Mashburn et al., 2009). For instance,in classroomswheremany children have poor academic skills, teach-ing basic skills through drill and practice might be the most benefi-cial for the development of academic skills. When many childrenare struggling with task persistency, children might benefit themost when teachers support children's learning efforts and whenteachers are sensitive to children's needs (e.g., Crosnoe et al., 2010;Hamre & Pianta, 2005).

Although child-dominated practices should, in theory, be detrimen-tal to children's academic learning and persistence, the extant empiricalliterature on teaching methods has rarely addressed this issue. Thus,conclusions about whether adverse effects of child-dominated practicesare universal, or dependent on other factors (e.g., collective studentcharacteristics), would be premature. We reasoned that these teachingpractices are likely to bemost detrimental in classroomswhere studentshave poor academic skills or show low levels of task persistence. In con-trast, positive benefits might accrue in classrooms characterized bymany highly skilled ormotivated students who do not needmuch addi-tional skills-training during the first years of school (Connor et al.,2009). Such students are relatively (compared to unskilled and unmoti-vated students) more likely to have already established a routine ofworking persistently on challenging tasks, and may perceive that thelack of teachers' guidance and involvement reflects their autonomysupport rather than their disengagement. Autonomy support, in turn,is associated with positive academic and motivational outcomes (Deci& Ryan, 2000). Hence, in such classrooms, children might acquire new

knowledge andwork on tasksmore persistently when teachers providelower structural support and greater freedom (this is because thesechildren know how to use their time effectively and support eachother's learning efforts).

Aims and hypotheses

Thus far, the joint contributions of collective student characteristicsand teaching practices to the development of young children's achieve-ment and task persistence have not been simultaneously evaluated.However, understanding which teaching practices satisfy the collectiveneeds of the classroom can provide teachers with valuable informationabout when and how to modify their teaching methods. We includedthree types of academic outcomes to cover the important areas of thecurriculum—math, reading (i.e., text comprehension), and spelling.

We expected the effects of teaching practices to critically hinge on(i.e., interact with) collective student skills and task persistence. Morespecifically, we hypothesized that child-dominated practices wouldundermine the development of academic skills and task persistenceprimarily in classrooms characterized by poor initial academic skills orlow task-persistent behavior. In contrast, the negative effect of child-dominated practices might be attenuated (or even reversed) for class-rooms where students' initial skills or task persistence were high. Ourexpectationswere less clearwith regard to how collective student char-acteristics might alter the effects of teacher-directed and child-centeredpractices. One possibility is that teacher-directed practices might bebeneficial in classrooms where poor academic skills or low task persis-tence are the norm. In such classrooms, most students need strongdirection and guidance, and thus teacher-directed practices may facili-tate the development of academic skills and task-persistent behavior(for individual-level results, see Connor et al., 2009). Alternatively, inthese classrooms, students might need encouragement and teachers'sensitivity to students' efforts, and they would benefit the most whenteachers engage in child-centered practices (Wentzel &Watkins, 2002).

Finally, we tested interactions between different teaching practices.We expected that the positive effect of child-centered practices mightbe undermined, and negative effect of teacher-directed practices poten-tiated, when teachers also engage in high levels of child-dominatedteaching practices. This is because most first-grade students are notyet familiarwith the specifics and demands of a school (classroom) con-text, and without receiving any guidance, direction or support, thesechildren might give up easily and not know how to best develop theiracademic skills. Young childrenmight also perceive the implementationof a mix of approaches as confusing, especially if teachers are not sys-tematic in the way they use different methods.

Method

Subjects

Participants were 523 students (273 boys) from 32 classrooms (7 to25 students per classroom) across Estonia (M age at the beginning of thefirst grade = 7.49 years, SD = 0.52). All participating classrooms camefrom 23 mainstream schools. This study is part of a larger project “Kin-dergarten-School Study” that follows 352 children from kindergarten(at the end of their last kindergarten year; 6- to 7-year-olds) until theend of basic school education (Grade 9; 15- to 16-year-olds). In thisstudy we included only the classrooms that had observational data(from a larger sample of 870 children). In Estonia, children usuallyattend schools on the basis of proximity. In participating schools (aswell as inmost schools in Estonia), ability groupingwas not usedwithinor across the classrooms. The average classroom size in our sample(≈16 students per classroom) was somewhat lower than usual (how-ever, in reality, the size of the classrooms varies across schools withthe official upper limit being 24 students; Riigikogu, 2010/2013). Theclassroom structure does not usually change during the elementary

276 E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

school years. Children typically stay with the same classmates for atleast four years. In all the schools, the language of teaching (as well aschildren's home language) was Estonian. Our sample was representa-tive of the Estonian general population (European Social Survey,2008). A total of 13% of the mothers had a basic education (9 years offormal education, grades 1–9), 62% had a secondary education (highschool, grades 10–12), and 25% had a college or university education.

All teachers were female (M age at the beginning of firstgrade = 41.25 years). The majority of them (75%) had a master'sdegree and had worked, on average, for 17.47 years (SD = 9.47) at thestart of the school year. Students were taught by the same teacher dur-ing the first two grades.

Procedure

School principals were first informed about the project in whichthey were invited to participate. When principals had agreed, they in-formed teachers who, in turn, invited parents to participate. All partici-pating children received parental permission to take part in this study.

At the beginning of the first grade (in September–October of 2008),research assistants administered math and reading tests in each class-room. At the end of the second grade (April–May of 2009), childrencompleted math, text comprehension, and spelling tests (teachers ad-ministered these tests over two days but were not involved in scoring).In addition, teachers rated each student's task persistence at both timepoints. Assessments took place during regular school lessons that lasted45 min.

At the middle of the first-grade academic year (February–March), apair of observers conducted classroom observations following theprocedures described in the ECCOM manual (Stipek & Byler, 2005).Altogether, six observers with a master's degree in education orpsychology were carefully trained (including 6 h training and 3 h oflive observation practice). After practice coding, observers participatedin a second training session. If therewere discrepancies between coders,these were thoroughly discussed and resolved. In addition, when need-ed, we clarified the criteria once again and stressed the importance ofstrictly following the guidelines to maximize the consensus among theobservers.

All observation sessions started at the beginning of a school day andlasted three lessons (i.e., 135 min). As the number of lessons per dayvaries from three to five (the maximum number of lessons per weekfor Estonian first grades is 20; Riigikogu, 2010/2013), observationscovered either all the lessons or at least the majority of the lessonsgiven on one school day. Observers made written notes about the ob-served teaching practices. After the session (i.e., after three lessons),both observers provided independent ratings. All observations includedat least one Estonian lesson and one math lesson. The third lesson thatwas observed varied between classrooms (e.g., Estonian, math, music,arts and crafts, or science).

Measures

Academic skillsStandardized cognitive and achievement tests have yet to be devel-

oped in Estonia (see Kikas, 2006). Thus, all tests measuring first-gradeacademic skills were developed according to the academic demandslaid out in the National Curriculum for Preschool Institutions(Vabariigi Valitsus, 2008/2011) and all the tests assessing second-grade skills were developed on the basis of the National Curriculumfor Basic Schools and Upper Secondary Schools (Vabariigi Valitsus,2002/2010). In addition, because the structure of Estonian and Finnisheducational systems is quite similar (see Kikas & Lerkkanen, 2010) andthe two languages have similar phonologies andhighly regular orthogra-phies (Jürimäe, Kalmus, Koshelev, & Pandis, 2003 for Estonian; Seymour,Aro, & Erskine, 2003 for Finnish), we also followed the examples ofFinnish measures when developing specific tasks (e.g., Lerkkanen,

Poikkeus, & Ketonen, 2006; Poskiparta, Niemi, & Lepola, 1994). As wealso collected teacher ratings for academic tasks (e.g., how well thechild was doing in math), correlations between the teacher ratings andthe measures used in this study are given below to provide informationabout the validity of our measures.

First-grademath skills. Themath test consisted of 12 tasks, including fourcalculation tasks (e.g., children saw a picture of six squares and wereasked to draw five fewer squares), four number sequencing tasks(e.g., “Complete the line 2 4 … … 10 12”), and four word problems(e.g., “Mati had 10 apples. He gave one apple to Pille and one appleto Ain. How many apples does he have now?”). Children could receiveone point for each correct solution. For each child, we calculated thesum of correct solutions (maximum score = 12). Internal consistencywas good (Cronbach's α = .89). Actual scores ranged from 1 to 12.This measure correlated with teacher-rated math skills at .48, which issimilar in strength to other findings on associations between teacher re-ports and students' skills (average correlation is .53, see meta-analysisby Südkamp, Kaiser, & Möller, 2012).

Second-grade math skills. We developed a test that assesses children'sknowledge of facts and procedures (knowledge of geometrical objectsand calculation), as well as their skills at solving word problems. It in-cluded four simple arithmetic calculations (e.g., 64 + 7 = …), threecalculations with units of measurement (e.g., 1 kg − 400 g = … g),four word problems (e.g., “The first deck of cards has 20 cards and thesecond deck of cards has 34 cards. How many more cards are there inthe second deck of cards?”), four geometrical tasks (e.g., children sawnine geometrical objects and their task was to find triangles, quadran-gles, and pentagons), and three tasks with combinations of numbers(e.g., “By using all four numbers–0, 1, 6, 7–only once, write the smallestpossible number”). One point was given for each correct answer. Scoreswere summed (maximum score = 18). Internal consistency was good(Cronbach's α = .92). Actual scores ranged from 2 to 18. The correla-tion with teacher-rated math skills was .57.

First-grade reading skills. Initial reading skills were assessed with twotypes of tasks. First, children were shown five pictures of different ob-jects. Underneath the object (e.g., mesilane [bee]), there was a corre-sponding number of spaces representing the quantity of letters of theobject's name (e.g., _ _ _ _ _ _ _ _; mesilane). In addition, childrenwere shown one letter in the word (e.g., E) and were asked to markwhere, among the spaces, the letter should be placed. Second, childrenwere given a list of seven words that described the objects that werenecessary at school (e.g., vihik [copybook]), objects thatwere not neces-sary at school (e.g., suvi [summer]), and a non-word (raamatop; thisword looks very similar to the word “raamat” that means a “book”).Children read eachword andwrote downwhether the respective objectwas needed at school and received a point for each correct answer.Scores were summed (maximum score = 12). Internal consistencywas .82. Actual scores varied from 3 to 12. Our measure correlatedwith teacher-rated reading skills at .34.

Second-grade text comprehension. Children read a story (“Giraff”) byMiloš Macourek, formatted as a one-sentence poem. Four types oftasks were then used to assess comprehension skills. First, childrenwere given six sentences (including four correct sentences) and theywere asked if the sentence was correct according to the story they justread. Second, children were given half of the original sentence andfour choices for the second half of the sentence. They had to choosethe one that best matched the first half. Third, children were askedto define three lesser-known words that had been used in the text(arithmetic, clumsy, abundantly). Finally, children were asked toidentify the moral of the story. They had to choose the most suitableanswer from among four sentences (one correct answer). One pointwas given for each correct answer and scoreswere summed (maximum

277E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

score = 12). Internal consistency was .79. Actual scores ranged from 1to 12. This measure correlated with teacher-rated text comprehensionat .36.

Second-grade spelling skills. To assess second-grade spelling skills, chil-dren had to correct mistakes in a text. Two types of tasks were used.First, they were given 10mistakes that did not follow the rules of gram-mar (e.g., in consonant clusters, each consonant can only be representedby one letter). Second, children had to correct 10 mistakes made onvarious grades of sound length (e.g., pikkad instead of the correctword—pikad [long]). We summed the number of times each child ac-curately corrected the mistakes (maximum score = 20). Internalconsistency was good (Cronbach's α = .85). The range of actualscores was from 0 to 20. The correlation with teacher-reported spellingskills was .53.

First- and second-grade task-persistent behavior. At both time points,teachers assessed children's task persistence using the Behavioral Strat-egy Rating Scale (see Zhang, Nurmi, Kiuru, Lerkkanen, & Aunola, 2011for the original scale and its validation). Teachers could be consideredthe best source of information (when children are as young as ourparticipants) because on a daily basis, they see, and are thus the mostfamiliar with, how each child usually approaches different tasks. Also,teacher ratings are considered more reliable and valid compared toyoung children's own reports (Fulmer & Frijters, 2009). The scaleconsisted of five statements including two positively and threenegatively worded items. Teachers were first asked to think about andrecall how a certain child typically behaved in a classroom learning sit-uation, and then to rate the child's behavior on a five-point rating scale(1 = not at all, 5 = to a great extent). Some statements were positivelyworded and measured the degree to which children engaged in activeand persistent behaviors (e.g., “shows activity and persistence in activ-ities and solving tasks”). Other items were negatively worded and dealtwith children's lack of persistence, helplessness, and task-irrelevant be-haviors when engaging in school tasks (e.g., “easily loses his/her focuswhen the activity or task is not going well”). Before creating an averageacross the five statements, the three negatively worded items were re-verse scored (maximum score = 5). Internal consistency of the scalewas good at both time points (Time 1: Cronbach's α = .91; Time 2:Cronbach's α = .88). Scores ranged from 1 to 5 at both time points.

Collective student characteristicsTo index collective student characteristics, classroom means

were computed for first-grade math (range = 6.21 to 9.77), reading(range = 8.90 to 11.53), and task persistence (range = 2.64 to 4.82)by averaging respective scores across all the children in each classroom.

Teaching practicesObservations were conducted using the Early Childhood Observa-

tion Measure (ECCOM; Stipek & Byler, 2005). Teachers were ratedwith regard to three dimensions: Child-centered practices, Teacher-directed practices, and Child-dominated practices. We used 15 items perdimension. Items tapped the child's responsibility, classroom manage-ment, choice of activities, discipline strategies, support for communica-tion skills, support for interpersonal skills, student engagement,individualization of learning activities, learning standards, coherenceof instructional activities, teaching concepts, instructional conversation,literacy instruction, math instruction, and math evaluation. The exactwording of each item varied depending on the dimension. For example,the Choice of Activities item was described as “Mixture of teacher andchildren making choices” (reflecting child-centered practices), “Teachermakes most of the choices” (reflecting teacher-directed practices), and“Children make most of the choices” (reflecting child-dominatedpractices). Ratings were provided on a five-point scale based on the per-centage of time the described practices were observed (1 = practicesare rarely seen; 0–20% of time to 5 = practices predominate; 80–100%

of time). Each teacher received three scores—for child-centered practices,teacher-directed practices, and child-dominated practices. These scoresare interrelated because the sum of the percentages should be around100 (although the sum does not necessarily have to add up to 100).

Inter-rater reliability between pairs of observers was determined bycalculating intra-class correlations (ICC-s). To do so, we used a two-waymixed effect model (measure is fixed, observers are random), withthe absolute agreement definition and the average measure ICC.McGraw and Wong (1996) recommend this approach when ob-servers vary between classrooms and provide ratings on continuousscales. ICC-s were significant (p b .001) for all dimensions (child-centered: ICC = .80, teacher-directed: ICC = .84, child-dominated:ICC = .97).

For each of the 3 × 15 items, an average was computed across thetwo raters. These scores were then averaged across the respective 15items for each dimension. Internal reliabilities for the scales weregood (α-s N .94). The range of scores was from 1.17 to 4.30, 1.26to 4.77, and 1 to 4.67 for child-centered, teacher-directed, andchild-dominated practices, respectively. Child-dominated practiceshad a moderate negative correlation with child-centered practices(r = − .44, p b .001) but was unrelated to teacher-directed prac-tices (r = .08, p = .500). Similar to previous studies (e.g., Hauser-Cram et al., 2003), child-centered and teacher-directed teachingpractices were highly negatively correlated (r = − .82, p b .001).For this reason, we first used child-centered and child-dominatedscores in our analyses. However, in order to examine whether anyeffects of child-centered practices would be inversely replicatedwhen using teacher-directed practices, we constructed another setof models where we replaced the child-centered approach withthe teacher-directed approach.

Results

Data analysis strategy

We used multilevel modeling (Mplus 7.0; Muthén & Muthén,1998–2012) to test our focal hypotheses. We first constructed two-level models (one for each outcome) to evaluate the degree to whichsecond-grade outcomes were influenced by collective student charac-teristics (i.e., classroom means of academic skills and task persistence)and teaching practices (child-centered and -dominated practices),after adjusting for the child's initial academic skills (math, reading),task persistence, and sex (for a similar analytic strategy and interpreta-tion, see Ryan, 2001). Controlling for initial levels of outcomes (children'sinitial academic skills and task persistence) is a widely used methodolo-gy in longitudinal research that allows one to make stronger inferencesabout the direction of the effects. Because prior studies have shownthat girls, for instance, have better reading skills than boys (for an over-view, see Logan & Johnston, 2010), we included sex in the models to en-sure that any reported effects are not due to sex differences. Becausechild-centered and teacher-directed methods were highly negativelycorrelated, we also re-ran the same main-effect models where we re-placed child-centered practices with teacher-directed practices. Wethen added interaction terms (one by one) to our previous models totest whether collective student characteristics moderated the effects ofchild-dominated and child-centered practices. We also ran similar inter-action models involving teacher-directed practices. Finally, we testedwhether child-dominated practices moderated the effects of child-centered or teacher-directed practices (these were again tested in sepa-rate models).

In the multilevel models with covariates, all continuous individual-level predictors were grand-mean centered. By doing this, the interceptsrepresent second-grade classroom means, adjusted for individual-levelcovariates (these interceptsweremodeled as a functionof class-level var-iables). In addition, because we controlled for students' initial skills andtask persistence, associations between class averages and second-grade

Table 2Means, standard deviations, and correlations of classroom-level predictors.

M SD 1 2 3 4

1. Initial class-average math 8.36 0.882. Initial class-averagetask persistence

3.46 0.46 −.09

3. Child-centered 2.98 0.85 .16 −.59⁎⁎⁎

4. Child-dominated 1.53 0.86 .06 .20 −.44⁎⁎⁎

5. Teacher-directed 2.66 0.89 −.28 .50⁎⁎ −.82⁎⁎⁎ .08

Note. Nclassroom level = 32.⁎⁎ p b .01. ⁎⁎⁎p b .001.

278 E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

outcomes represent the degree to which classroom-level effects differfrom individual-level associations, or in other words, whether attendingone classroom vs. another has an additional or incremental effect on out-comes (Raudenbush & Bryk, 2002). Also, classroom covariates weregrand-mean centered (i.e., computed across the 32 classrooms).Classroom-level interaction terms were tested, due to limited power,one by one. In addition, when evaluating these interactions, all otherindividual- and classroom-level covariates were included in the models.We also computed effect sizes for the interaction effects using the formu-la provided by Tymms (2004; see also Trautwein, Gerlach, & Lüdtke,2008):

Δ ¼ 2 � B � SDpredictor=σe

where B is the unstandardized regression coefficient of the predictor(interaction term), SDpredictor is the standard deviation of the predictor,and σe is the residual standard deviation at the individual level. The sizeof the effects can be interpreted similar to Cohen's d values.

Significant interactions were followed up by testing simple slopesusing an online tool provided by Preacher, Curran, and Bauer (2006).Simple slopes were estimated at low (one SD below themean), average(at themean), and high (one SD above themean) levels of eachmoder-ator. When the simple slopes were not significantly different from zero,we examined whether any of the more extreme values in our data setwould fall into the regions of significance. All effects were consideredsignificant at p b .05.

There were some missing values on the individual-level predictors(first-grade academic skills and task persistence) and second-grade out-comes (6 to 20% of the values were missing) that were handled usingfull information maximum likelihood estimation (with robust standarderrors) that provides unbiased standard errors compared to other tradi-tional methods. Although by default, Mplus throws out cases that havemissing values on predictor variables, we were able to include themby employing the procedure described by Enders (2010, pp. 116–117).More specifically, we treated individual-level incomplete predictors(initial math, reading, and persistence) as latent factors (such a proce-dure “tricks the software program into treating the explanatory variableas an outcome,while still maintaining the variable's exogenous status inthe model”; Enders, 2010, p. 117). Each of the three factors were repre-sented by the single indicator (e.g., observedmath score)with the factorloading of one. Also, residual variances of the three indicatorswere fixedto zero so that the variances of latent factors represented the variancesof observed variables.

Intraclass correlations (ICC-s)

In an unconditional model (without any covariates), we found thatall second-grade outcomes had significant variance at the classroomlevel. More specifically, between-classroom variability accounted for16% of the variance inmath, 20% of the variance in text comprehension,12% of the variance in spelling, and 9% of the variance in task persis-tence. With regard to initial skills and task persistence, 11 (initial

Table 1Means, standard deviations, and correlations of individual-level variables.

M SD 1

1. Sex –

2. Initial math 8.35 2.22 −.13⁎⁎

3. Initial reading 10.78 1.51 .084. Initial task persistence 3.46 1.01 .075. Second-grade math 13.11 2.93 −.066. Second-grade text comprehension 7.87 1.91 .067. Second-grade spelling 13.06 5.04 .13⁎⁎

8. Second-grade task persistence 3.40 1.07 .08

Note. Nindividual level = 523. Second-grade outcomes were allowed to vary (and correlate) at th⁎p b .05. ⁎⁎p b .01. ⁎⁎⁎p b .001.

math skills) to 15% (initial task persistence) of the variance was attrib-utable to between-classroom differences. Because the classroom-levelvariance for initial reading skills was not significant (only 4% of the var-iance was due to between-classroom differences), we dropped class-average reading skills from our models. Means, standard deviations,and correlations (for both individual and classroom levels) are presentedin Tables 1 and 2.

We also evaluated whether classrooms with higher class-averagemath skills and task persistence had a greater proportion of studentswho scored high on math and task persistence. As expected, the per-centages of students who were above the 75th percentile on initialmath (score N 10) and task persistence (score N 4.20) in each class-room correlated highly with classroom averages (for math: r = .71,p b .001; for task persistence: r = .83, p b .001).

Two-level models with covariates

Individual-level effectsInterestingly, whereas initial math skills predicted all the outcomes,

reading skills failed to predict any (see Table 3 for results). In addition,first-grade task persistence had a positive association with all second-grade academic skills, over and above children's initial academic skills.Also, task persistence showed considerable stability across the twotime points. Finally, girls had better spelling skills in second gradethan boys. Altogether, individual-level predictors explained 12 to 45%of the variance in the second-grade outcomes.

Classroom-level main effectsNone of the contextual effects (class-average math and task persis-

tence) or teaching practices (child-centered and -dominated) were sig-nificant, except for one—children's second-grade comprehension skillswere, unexpectedly, lower in classrooms with higher aggregate levelsof initial math skills after controlling for students' first-grade mathscores and other covariates. Whenwe replaced child-centeredmethodswith teacher-directed practices in the models, we did not find a signifi-cant main effect. Altogether, classroom-level predictors explained 8 to23% of the variance in second-grade outcomes (i.e., intercepts).

2 3 4 5 6 7

.34⁎⁎⁎

.30⁎⁎⁎ .30⁎⁎⁎

.56⁎⁎⁎ .31⁎⁎⁎ .40⁎⁎⁎

.25⁎⁎⁎ .17⁎⁎ .30⁎⁎⁎ .32⁎⁎⁎

.35⁎⁎⁎ .26⁎⁎⁎ .37⁎⁎⁎ .40⁎⁎⁎ .26⁎⁎⁎

.44⁎⁎⁎ .28⁎⁎⁎ .60⁎⁎⁎ .56⁎⁎⁎ .38⁎⁎⁎ .46⁎⁎⁎

e classroom level. Sex was coded as 0 (boy) and 1 (girl).

Table 3Individual- and classroom-level associations.

Second-grade math Second-grade textcomprehension

Second-grade spelling Second-grade taskpersistence

b SE b SE b SE b SE

Intercept 13.18⁎⁎⁎ 0.30 7.77⁎⁎⁎ 0.21 12.49⁎⁎⁎ 0.49 3.32⁎⁎⁎ 0.09Student-level predictorsSex −0.12 0.29 0.24 0.18 1.24⁎⁎ 0.47 0.17 0.10Initial math 0.57⁎⁎⁎ 0.09 0.18⁎⁎ 0.06 0.52⁎⁎⁎ 0.15 0.14⁎⁎⁎ 0.02Initial reading 0.17 0.10 0.04 0.07 0.26 0.20 0.02 0.02Initial task persistence 0.74⁎⁎⁎ 0.16 0.38⁎⁎⁎ 0.10 1.43⁎⁎⁎ 0.32 0.54⁎⁎⁎ 0.05

Classroom-level predictorsInitial class-average math 0.13 0.31 −0.41⁎ 0.18 0.52 0.44 −0.09 0.09Initial class-average task persistence −0.36 0.53 0.26 0.36 −0.40 1.24 0.04 0.20Child-centered −0.09 0.40 −0.13 0.25 0.44 0.47 0.08 0.10Child-dominated −0.38 0.25 −0.23 0.16 0.21 0.44 0.001 0.11

R2 (individual level) .37 .12 .22 .45R2 (classroom level) .09 .23 .14 .09Intercept residual variance 1.28⁎⁎⁎ 0.29 0.72⁎ 0.28 3.10⁎ 1.21 0.08⁎⁎ 0.03

Note. Nindividual level = 523; Nclassroom level = 32. Unstandardized estimates are presented. All individual- and classroom-level continuous variables were grand-mean centered. Sex wascoded as 0 (boy) and 1 (girl).⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.

279E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

Interactions between teaching practices and collectivestudent characteristics

Interactions involving child-dominated practices. Class-average mathskills moderated the effect of child-dominated practices on later mathskills, spelling skills, and task persistence (see Table 4). We comparedsimple slopes at low (one SD below the mean), average (at the mean),and high (one SD above the mean) levels of collective math skills.When initial class-average math skills were low, child-dominatedpractices had a negative association with second-grade math andspelling skills, and task persistence (see Table 5 for follow-up re-sults). In contrast, in classrooms with high class-average math skillsat the beginning of first grade, child-dominated teaching practicespredicted higher spelling scores in second grade and increases intask persistence (when predicting second-grade math, the simpleslope was not significant even when using more extreme scores ob-served in the data set).

Interactions involving child-centered practices. Child-centered practicesinteracted with class-average task persistence when predictingsecond-grade task persistence (see Table 4). Our findings showed thatsimple slopes at low (b = −0.10, SE = 0.09, p = .280) and average(b = 0.07, SE = 0.08, p = .377) levels of the moderator were notsignificant. However, when collective task persistence was lower thanone SD below the mean (score b −0.77; observed scores ranged from−0.82 to 1.36), child-centered teaching practices became significantlyassociated with decreases in task persistence. In contrast, when first-grade classrooms were characterized by high task persistence, child-centered practices predicted relative increases in task persistence overtime (b = 0.24, SE = 0.11, p = .023).

Table 4Significant interactions between teaching methods and collective student characteristics, and a

Second-grade math Second-grade textcomprehension

Classroom-level interaction b SE Δ b SE

CC × CPersistCD × CMath 0.41⁎ 0.20 0.23TD × CMathTD × CPersistCC × CD −0.86⁎ 0.42

Note. Nindividual level = 523; Nclassroom level = 32. Unstandardized estimates are presented. CPermethod; CD—child-dominated teaching method; TD—teacher-directed teaching method. Δ =⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.

Interactions involving teacher-directed practices. The effects of teacher-directed practices varied as a function of collective math skills andtask persistence when predicting between-classroom differences inchanges in task persistence. More specifically, teacher-directed prac-tices predicted, albeit marginally, decreases in task persistence in class-rooms where initial mean levels of math skills were high (b = −0.21,SE = 0.12, p = 0.081). Simple slopes at other levels failed to reach sig-nificance and did not become significant when using the moderatorvalues lower than one SD below the mean (classrooms math scoresranged from −2.15 to 1.41). Follow-up analyses of the second interac-tion indicated that the teacher-directed approach was associated withincreases in task persistence in classroomswhere collective task persis-tence was initially relatively low (the slope became significant whenclass-average task persistence was below the value of−0.62; observedscores ranged from −0.82 to 1.36). In contrast, when initial class-average task persistence was relatively high, teacher directed practiceswere associated with decreases in task persistence over time (theslope became significant when class-average task persistence wasabove the value of 0.83).

Interactions among teaching practicesThe effects of child-centered teaching practices on second-grade text

comprehension, spelling, and task persistence were dependent on thedegree to which teachers also used child-dominated practices (seeTable 4). Follow-up results (see Table 6) demonstrated that child-centered practices predicted higher spelling skills and increases in taskpersistence when teachers rarely allowed children to dominate (thesimple slope at low levels of child-dominated practices was not signifi-cant when predicting second-grade text comprehension). In contrast,the child-centered teaching approach was associated with lowersecond-grade text comprehension and spelling skills (the slope became

mong teaching methods.

Second-grade spelling Second-grade task persistence

Δ b SE Δ b SE Δ

0.36⁎⁎ 0.11 0.450.97⁎⁎⁎ 0.23 0.28 0.25⁎⁎⁎ 0.06 0.41

−0.16⁎ 0.08 −0.30−0.30⁎⁎ 0.11 −0.44

−0.78 −2.15⁎ 0.98 −0.78 −0.67⁎⁎⁎ 0.14 −1.36

sist—class-average persistence; CMath—class-average math; CC—child-centered teachingeffect size.

Table 5Simple slopes: The effects of child-dominated teaching practices on Time-2 outcomes as afunction of class-average math skills.

Initial class-average math

Low level Average level High level

Outcome b SE b SE b SE

Second-grade math −0.83⁎ 0.32 −0.46⁎ 0.21 −0.10 0.22Second-grade spelling −0.85⁎ 0.35 0.003 0.29 0.86⁎ 0.36Second-grade task persistence −0.28⁎⁎ 0.09 −0.06 0.06 0.17⁎ 0.07

Note. Nindividual level = 523; Nclassroom level = 32. Unstandardized estimates are presented.Coefficients (b-s) refer to the relations between child-dominated teaching practices andclassroom averages of Time-2 outcomes (adjusted for covariates) under low (one SDbelow the mean), average, and high (one SD above the mean) levels of class-averagemath skills.⁎ p b .05. ⁎⁎ p b .01.

280 E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

significant when the score for child-dominated practices was above thevalue of 1.79; observed scores ranged from −0.53 to 3.13), as well aswith decreases in task persistence when teachers also used highlevels of child-dominated methods. When we replaced child-centeredwith teacher directed practices, no significant interactions with child-dominated practices were found.

Because teachers who had less teaching experience were morelikely to use child-dominated practices (r = − .37, p = .001), were-evaluated our interaction terms after controlling for teachers' experi-ence. All nine interactions remained significant, although twomarginallyso.

Effect sizes for interactionsWhen interactions were part of the models, classroom-level R2-s

ranged from .15 to .45. In addition, the average absolute value of Δwas 0.56, suggesting that the interaction effects were, on average, mod-erate in size (three out of nine effect sizes were around |0.80| or above).

Discussion

The major focus of this study was to evaluate the impact of twosources of contextual influences in the classroom on young children'sdevelopment of academic skills and task persistence. Importantly, wealso moved beyond tests of main effect contributions of collective stu-dent characteristics and teacher influences to evaluate the degree towhich varying teaching practices would differentially impact thedevelopment of academic skills and task persistence depending onaverage first-grade classroom skills and task persistence. Our results in-dicate that a sole focus onmain effects would have led to the erroneousconclusion that it matters little whether children are in a classroomcharacterized by either low or high initial levels of academic skills,whether task persistence is low or high, or whether teachers utilize

Table 6Simple slopes: The effects of child-centered teaching practices on Time-2 outcomes as afunction of child-dominated teaching practices.

Child-dominated practices

Low level Average level High level

Outcome b SE b SE b SE

Second-grade textcomprehension

0.28 0.36 −0.47 0.27 −1.2⁎ 0.53

Second-grade spelling 1.45⁎ 0.68 −0.40 0.59 −2.26 1.29Second-grade task persistence 0.40⁎⁎ 0.13 −0.18⁎ 0.08 −0.76⁎⁎⁎ 0.17

Note. Nindividual level = 523; Nclassroom level = 32. Unstandardized estimates are presented.Coefficients (b-s) refer to the relations between child-centered teaching practices andclassroom averages of Time-2 outcomes (adjusted for covariates) under low (one SDbelow the mean), average, and high (one SD above the mean) levels of child-dominatedpractices.⁎ p b .05. ⁎⁎ p b .01. ⁎⁎⁎ p b .001.

one teaching method over others. Rather, the several interactions wefound indicate that both contextual influences are important in deter-mining subsequent academic functioning and persistence but that cer-tain teaching practices are more beneficial (or detrimental) dependingon the collective starting point of students. Thus, the results enhanceour understanding of the better- and worse-fitting combinations ofteaching practices with collective student characteristics that should,in turn, assist teachers in supporting optimal, and thwarting suboptimal,achievement and motivation.

We found that the contributions of child-centered and teacher-directed practices to changes in task persistence depended on initialclass-average persistence or math skills. Our results suggest that child-centered practices should be the method of choice (and teacher-directed practices should be avoided) in classrooms characterized byhighly persistent children. But, for classrooms characterized by manychildren who have difficulties with working persistently on new andchallenging tasks, teacher-directed methods should be opted for.Although these two teaching practices failed to predict achievementoutcomes (alone or in combination with collective student characteris-tics), the differential courses of change in task persistence depending onteaching practices usedmay subsequently affect differential trajectoriesin achievement (Dally, 2006).

We should note that we did not find support for the theoretical as-sumption that children in classroomswith low skills or task persistencewould benefit mostly from child-centered practices that place highimportance on encouragement and sensitivity to children's needs(e.g., Hamre & Pianta, 2005). Such methods are theorized to satisfyneeds for relatedness and competence (Deci & Ryan, 2000) and, conse-quently, be beneficial for children's motivational development, includ-ing task persistence. Instead, our findings suggest that young childrenwho have relatively poor academic skills or difficulties with stayingfocused on learning new tasks, may actually need drill and practice,and strict classroom rules, in order to establish a habit of working (seealso Connor et al., 2004; Stipek et al., 1995). These children might beeasily distracted if they are given the opportunity to make their ownchoices and may not yet be capable of benefiting from teachers' guid-ance (rather than direction) and cooperative activities with others.This might be especially true if classrooms with low task persistencealso include many children who are aggressive–disruptive (cf. Määttä,Stattin, & Nurmi, 2006). However, it should be stressed that child-centered rather than teacher-directed practices enhance autonomywhich is important for the development of self-regulated learning andenhanced mastery orientation (Deci & Ryan, 2000). It is possible thatteacher-directed methods might make children who have difficultieswith task persistence work harder but, in the long run, these childrenwill work on tasks not because they want to master the new materialbut because they are motivated by external rewards (e.g., gainingteacher's praise) or avoiding punishment (e.g., teacher's criticism).Thus, a continued reliance on such a didactic approachmight underminestudents' intrinsic interest in learning and thereby have an adverse effecton later academic development (Stipek et al., 1995).

In addition, we found evidence that partially supports earlier studiesshowing that child-dominated practices tend to inhibit academic devel-opment andworsenmotivational and social problems (Valeski & Stipek,2001;Walker, 2008), but this was limited to classroomswith low initialclass-average math skills. Our results suggest that in classrooms withlow levels of skills, teachers should be especially careful to avoid givingchildren too much freedom and autonomy because this (even whenused with good intentions) can inhibit learning outcomes and task per-sistence. In classrooms characterized by high initial math skills, a lessstructured environment might allow children to effectively make useof their learning context (e.g., they might support each other's motiva-tion for learning). In addition, such children may acknowledge, andabide by, the rules in their classroom and no longer need discussions,or enforcement, of rules. In contrast, drilling and mechanical practicingmay seem boring to these children (Connor et al., 2004).

281E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

Moreover, our results suggest that granting excessive autonomy tochildren (relying on child-dominated practices)may transform theben-eficial effects of child-centered practices into deleterious ones, at least inthe beginning of elementary school. Teachers are likely to implement amix of different teaching practices for a variety of reasons. Someteachers may believe that engaging in different methods might bebeneficial for children's learning and academic development and theypurposefully use different teaching practices in different situations(e.g., depending on the subject domain or how easy or difficult thematerial is for children; whether the task is carried out in a larger vs.smaller group). Other teachersmight employ amix of methods becausethey are not experienced enough to be systematic in their approach.Although future research will need to understand the reasons for whyteachers use a mix of approaches, as well as the putative mechanismsthat account for their effects, our results do suggest that the combineduse of child-centered and child-dominated practices are not beneficialfor children's academic development and task persistence, at least inthe beginning of their school career. It might be confusing for youngchildren, especially if different approaches are not implemented in apredictable way. Or, teachers engaging in child-centered practicesmight reward their students with free, unstructured, time. Such exter-nal contingencies are likely to lead to a reduction in intrinsic motivation(Deci, Koestner, & Ryan, 1999), which in turn, reduces task persistenceand performance (Urdan, Ryan, Anderman, & Gheen, 2002).

In this study, we focused on effects of teaching practices, in conjunc-tion with collective student characteristics, on academic outcomes andlearning strategies. However, it is equally important to recognize thatcollective student characteristics are likely to influence the teachingmethods that teachers utilize. Indeed, transactional processes are ex-pected to occur within and between multiple levels of children's ecolo-gies (Bronfenbrenner & Morris, 1998). For example, in classroomswhere children are interested in learning and show high effort in com-pleting tasks, teachers may respond with greater use of child-centeredindividualized guidance, allowmore time for peer discussions, and pro-vide substantial feedback. This, in turn, is likely to enhance children'sself-efficacy andmotivation and thus, also support their task persistence.In other classrooms, teachersmight recognize that their students need toacquire basic skills and turn to teacher-directedmethods. In both cases, itis necessary for teachers to be attuned to the strengths and weaknessesof the children in their classrooms in order to make the adaptations totheir strategies for improving achievement and motivation.

Although not a main focus of this study, all individual level effects,except for initial reading skills, were consistent with earlier studies.Namely, initial math skills predicted higher second-grade reading com-prehension, spelling, and math (Lerkkanen et al., 2005), and highertask-persistent behavior (see Onatsu-Arvilommi & Nurmi, 2000). Also,task persistence showed considerable stability over time (Onatsu-Arvilommi & Nurmi, 2000) and children who showed less task persis-tence in first grade had lower reading, spelling (Dally, 2006), andmath scores (Aunola et al., 2003) in second grade. These results are inline with earlier findings (Hyson, 2008; McClelland et al., 2013;Schaefer & McDermott, 1999; Yen et al., 2004) and show that task per-sistence predicts achievement over and beyond initial skill levels.

Limitations and directions for future research

Some limitations of our study deserve mention. Similar to manyother studies, we examined classroom averages of students' skills andtask persistence. However, future studies could test whether the effectsof teaching practices are further dependent on the homogeneity vs.heterogeneity of students' strengths andweaknesses. Onemight expect,for instance, that child-dominated methods are more likely to facilitateacademic development when the majority of students exhibit highlevels of cognitive skills (i.e., variability is low). In addition, we limitedour examination of interactions to the classroom level. However, teach-ing practices may also have differential effects on academic outcomes

within the same classroomdepending on each child's individual charac-teristics. For instance, a child's weaknesses might be exacerbated inclassrooms where teachers use child-dominated practices. In addition,although child-dominated practices could, on average, bring benefitsto classrooms where students have high levels of academic skills, thebenefits might not apply to a child who is less advanced than his orher classmates. Thus, one important direction for future research is tosimultaneously test cross- and classroom-level interactions.

In addition, future research should include children from a widerrange of grade levels and follow them for multiple years to test the rep-licability and generalizability of our findings. For instance, it is possiblethat even when the child-dominated approach has a positive influenceon early academic development in classrooms where children havehigh initial skills, such practices might be detrimental in the long run(e.g., students might lose interest in learning over time if teachers donot provide sufficient direction, guidance, or emotional support). More-over, children's academic andmotivational development is also likely tobe influenced by the broader cultural context in which their classroomsare embedded (Bronfenbrenner &Morris, 1998; Hamre & Pianta, 2010).Although we would expect that the effects of teaching practices on aca-demic outcomes that are dependent on collective student characteris-tics to be relatively universal, future cross-cultural longitudinal workwill be needed to evaluate whether the best-fitting combinations ofteaching methods and collective student characteristics in other cul-tures are found to be similar to, or different from, those found in ourstudy.

Another set of limitations concerns assessment issues. Teachersassessed students' task persistence at the beginning of the first grade.However, teachers are likely to become more knowledgeable of eachstudent's learning habits over time. Although studies have confirmedthat teacher-rated task persistence is associated with parental ratings(Zhang et al., 2011), care should be taken in interpreting our results.Because teachers' reports might be influenced by additional factors(e.g., what they consider to be normative behavior, the level of teachers'tolerance, other characteristics of a child), it would be fruitful to com-plement teachers' reports with other methods (e.g., observations).Also, classroom observationswere conducted later than the assessmentof initial academic skills and task persistence. It is possible, for instance,that giving relatively greater freedom to students in highly skilled class-rooms could actually reflect teachers' adjustment to the classroom con-text. Hence, it will be necessary to examine teaching practices andclassroom characteristics over time to understand how they mutuallyinfluence each other. Similarly, in addition to examining how teacherschange their teaching methods according to the strengths and weak-nesses of children over time, it will be equally important to understandwhether teachers adjust their teaching methods from one classroom toanother (see also Marsh et al., 2008) or across different subjects for thesame classroom.

We evaluated the effects of child-centered and teacher-directedpractices in separate models because they were highly negatively corre-lated (see also Hauser-Cram et al., 2003). Future studies should testwhether these two approaches represent the same underlying construct(thus, they could be treated as opposite ends of the same continuum) orwhether they are separate constructs with distinct correlates.

Finally, this study focused on factors that influence academic devel-opment at school. Because parents can influence young children's aca-demic development, future studies would benefit from examininghow family- (e.g., parenting, motivation strategies used by parents)and school/classroom-related factors operate together when influenc-ing children's academic and motivational development. For instance,we would expect that the most optimal developmental outcomeswould occur if both teachers and parents provide a context that bestsuits children's needs. For example, children in academically competentclassrooms are likely to benefit most when teachers support the devel-opment of their autonomywhen parents do so aswell, whereas benefitsmay be minimized when parents utilize coercive disciplinary practices.

282 E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

Whether family context has an independent effect on children'soutcomes or whether it can enhance (or inhibit) the influence ofclassroom/school-related factors (e.g., teaching methods) will be impor-tant for future researchers from multiple cultures to evaluate to morefully understand the contributing factors and conditions that lead tochildren's optimal academic development.

Conclusions

Despite these limitations, this study underscores the importance ofunderstanding how different contextual factors interactively influenceyoung children's academic skills and task persistence. The current find-ings suggest that we should move beyond a search for universally ben-eficial teaching practices because the same practices can be enhancingor inhibiting depending on the collective student characteristics. Thisinformation is of high practical value, especially in elementary school.The initial school years represent a period when teachers can have amajor influence on children's development because young childrenare still receptive to teachers' behavior and guidance (Spinath &Spinath, 2005). In addition, in Estonia, one teacher is responsible forteaching almost all the subjects for the first three years. Thus, the impactthe teacher can have on students' development may be greater than incountries where children are taught by a different teacher every year(e.g., the U.S.A.). Of course, classrooms vary in terms of students' initialskills, knowledge, and task persistence that are beyond teachers' con-trol. However, by being aware of which teaching practices can facilitateor inhibit academic development in different contexts, teachers may beable tomaximize their positive influence, andminimize any detrimentalinfluence, on the lives of many students.

Acknowledgments

This studywas supported by institutional research funding IUT (3-3)of the EstonianMinistry of Education and Research and European SocialFund Program Eduko (via Archimedes Foundation, grant 30.2-4/549).We would like to thankMairi Männamaa for developing tests assessinginitial academic skills, Anu Palu for developing the second-grade mathtest, and Krista Uibu for developing the second-grade reading test.

References

Alfieri, L., Brooks, P. J., Aldrich, N. J., & Tenenbaum, H. R. (2011). Does discovery-based in-struction enhance learning? Journal of Educational Psychology, 103, 1–8. http://dx.doi.org/10.1037/a0021017.

Aunola, K., Nurmi, J. -E., Lerkkanen, M. -K., & Rasku-Puttonen, H. (2003). The roles ofachievement-related behaviours and parental beliefs in children's mathematical per-formance. Educational Psychology, 23, 403–421.

Barth, J., Dunlap, S., Dane, H., Lochman, E., & Wells, K. (2004). Classroom environment in-fluences on aggression, peer relations, and academic focus. Journal of SchoolPsychology, 42, 115–133. http://dx.doi.org/10.1016/j.jsp.2003.11.004.

Bianco, M., Pellenq, C., & Lambert, E. (2012). Impact of early code-skill and oral-comprehension training on reading achievement in first grade. Journal of Researchin Reading, 35, 427–455. http://dx.doi.org/10.1111/j.1467-9817.2010.01479.x.

Bronfenbrenner, U., & Morris, P. A. (1998). The ecology of developmental processes. InW.Damon (Series Ed.) & R.M. Lerner (Vol. Ed.), Handbook of child psychology: Vol. 1.Theoretical models of human development (5th ed., pp. 993–1028). New York, NY:Wiley.

Bulotsky-Shearer, R. J., Bell, E. R., & Dominguez, X. (2012). Preschool classroom behav-ioral context and school readiness outcomes for low-income children: A multilevelexamination of child- and classroom-level influences. Journal of EducationalPsychology, 104, 421–438. http://dx.doi.org/10.1037/a0026301.

Byrne, B., Coventry, W. L., Olson, R. K., Wadsworth, S. J., Samuelsson, S., Petrill, S. A.,et al. (2010). “Teacher effects” in early literacy development: Evidence from astudy of twins. Journal of Educational Psychology, 102, 32–42. http://dx.doi.org/10.1037/a0017288.

Connor, C., Piasta, S., Fishman, B., Glasney, S., Schatschneider, C., Crowe, E., et al. (2009).Individualizing student instruction precisely: Effects of child × instruction interac-tions on first graders' literacy development. Child Development, 80, 77–100. http://dx.doi.org/10.1111/j.1467-8624.2008.01247.x.

Connor, C. M., Morrison, F. J., & Katch, L. E. (2004). Beyond the reading wars: Exploringthe effects of child –instruction interactions on growth in early reading. ScientificStudies of Reading, 8, 305–336.

Crosnoe, R., Morrison, F., Burchinal, M., Pianta, R., Keating, D., Friedman, S., et al.(2010). Instruction, teacher–student relations, and math achievement trajectoriesin elementary school. Journal of Educational Psychology, 102, 407–417. http://dx.doi.org/10.1037/a0017762.

Dally, K. (2006). The influence of phonological processing and inattentive behaviour onreading acquisition. Journal of Educational Psychology, 98, 420–437. http://dx.doi.org/10.1037/00220663.98.2.420.

Daniels, D. H., & Shumow, L. (2003). Child development and classroom teaching: A re-view of the literature and implications for educating teachers. AppliedDevelopmental Psychology, 23, 495–526. http://dx.doi.org/10.1016/S0193-3973(02)00139-9.

Deci, E., & Ryan, R. (2000). The “what” and “why” of goals pursuits: Human needs andthe self-determination of behavior. Psychological Inquiry, 11, 227–268.

Deci, E. L., Koestner, R., & Ryan, R. M. (1999). A meta-analytic review of experiments ex-amining the effects of extrinsic rewards on intrinsic motivation. Psychological Bulletin,125, 627–668. http://dx.doi.org/10.1037/0033-2909.125.6.627.

DiPerna, J. C., Volpe, R. J., & Elliott, S. N. (2005). A model of academic enablers andmath-ematics achievement in the elementary grades. Journal of School Psychology, 43,379–392.

Enders, C. K. (2010). Applied missing data analysis. New York, NY: Guilford Press.Foorman, B., York, M., & Santi, K. (2008). Contextual effects on predicting risk for reading

difficulties in first and second grade. Reading and Writing: An Interdisciplinary Journal,21, 371–394. http://dx.doi.org/10.1007/s11145-007-9079-5.

Fuchs, D., Compton, D., Fuchs, L., Bryant, J., Hamlett, C., & Lambert, W. (2012). First-grade cognitive abilities as long-term predictors of reading comprehension and dis-ability status. Journal of Learning Disabilities, 45, 217–231. http://dx.doi.org/10.1177/0022219412442154.

Fulmer, S., & Frijters, J. (2009). A review of self-report and alternative approaches in themeasurement of student motivation. Educational Psychology Review, 21, 219–246.http://dx.doi.org/10.1007/s10648-009-9107-x.

Gettinger, M., & Kohler, K. (2011). Process-outcome approaches to classroom manage-ment and effective teaching. In C. Evertson, & C. Weinstein (Eds.), Handbook ofclassroom management. Research, practice, and contemporary issues. (pp. 73–95).New York, NY: Routledge.

Guay, F., Boivin, M., & Hodges, E. V. E. (1999). Social comparison processes and academicachievement: The dependence of the development of self-evaluations on friends'performance. Journal of Educational Psychology, 91, 564–568.

Hamre, B., & Pianta, R. (2010). Classroomenvironments anddevelopmental processes: Con-ceptualization andmeasurement. In J. Meece, & J. Eccles (Eds.),Handbook of research onschools, schooling, and human development (pp. 25–41). New York, NY: Routledge.

Hamre, B., & Pianta, R. C. (2005). Can instructional and emotional support in the firstgrade classroom make a difference for children at risk of school failure? ChildDevelopment, 76, 949–967. http://dx.doi.org/10.1111/j.1467-8624.2005.00889.x.

Harris, J. R. (1995). Where is the child's environment? A group socialization theory of de-velopment. Psychological Review, 102, 458–489. http://dx.doi.org/10.1037/0033-295X.102.3.458.

Hauser-Cram, P., Sirin, S. R., & Stipek, D. (2003).When teachers' and parents' values differ:Teachers' ratings of academic competence in children from low-income families. Journalof Educational Psychology, 95, 813–820. http://dx.doi.org/10.1037/0022-0663.95.4.813.

Hughes, J. N. (2002). Authoritative teaching: Tipping the balance in favour of school ver-sus peer effects. Journal of School Psychology, 40, 485–492.

Hyson, M. (2008). Enthusiastic and engaged learners: Approaches to learning in the earlychildhood classroom. New York, NY: Teachers College Press.

Jimerson, S., Egeland, B., & Teo, A. (1999). A longitudinal study of achievement trajecto-ries: Factors associated with change. Journal of Educational Psychology, 91, 116–126.

Jürimäe, M., Kalmus, V., Koshelev, J., & Pandis, M. (2003). Literacy in Estonia. [Supple-ment to international reports on literacy research: Estonia, Hungary, andBulgaria]. Reading Research Quarterly, 38, 288–290 (Retrieved May 30, 2013from http://www.reading.org).

Kikas, E. (2006). School Psychology in Estonia. In S. Jimerson, T. Oakland, & P. Farrell (Eds.),The handbook of international school psychology (pp. 91–102). Thousand Oaks, CA: SAGEPublications.

Kikas, E., & Lerkkanen, M. -K. (2010). Education in Estonia and Finland. In M. Veisson, E.Hujala, M. Waniganayake, P. Smith, & E. Kikas (Eds.), Perspectives in early childhoodeducation: Diversity, challenges and possibilities (pp. 33–46). Frankfurt, Germany:Peter Lang.

Kindermann, T. (2003). Development of children's social relationships. In J. Valsiner, & K.Connolly (Eds.), Handbook of developmental psychology (pp. 407–430). ThousandOaks, CA: Sage.

Kiuru, N., Aunola, K., Torppa, M., Lerkkanen, M. -K., Poikkeus, A. -M., Niemi, P., et al.(2012). The role of parenting styles and teacher interactional styles in children'sreading and spelling development. Journal of School Psychology, 50, 799–823. http://dx.doi.org/10.1016/j.jsp.2012.07.001.

Lerkkanen, M., Rasku-Puttonen, H., Aunola, K., & Nurmi, J. (2005). Mathematical perfor-mance predicts progress in reading comprehension among7-year olds. European Journalof Psychology of Education, 20, 121–137. http://dx.doi.org/10.1007/BF03173503.

Lerkkanen, M. -K., Kikas, E., Pakarinen, E., Poikonen, P. -L., & Nurmi, J. -E. (2012).Mothers' trust toward teachers in relation to teaching practices. Early ChildhoodResearch Quarterly. http://dx.doi.org/10.1016/j.ecresq.2012.04.005.

Lerkkanen, M. -K., Kikas, E., Pakarinen, E., Trossmann, K., Poikkeus, A. -M., Rasku-Puttonen, H., et al. (2012). A validation of the Early Childhood Classroom ObservationMeasure in Finnish and Estonian kindergartens. Early Education and Development, 23,323–350. http://dx.doi.org/10.1080/10409289.2010.527222.

Lerkkanen, M. -K., Kiuru, N., Pakarinen, E., Viljanranta, J., Poikkeus, A. -M., Rasku-Puttonen, H., et al. (2012). The role of teaching practices in the development ofchildren's interest in reading and mathematics in kindergarten. Contemporary

283E. Kikas et al. / Journal of Applied Developmental Psychology 35 (2014) 273–283

Educational Psychology, 37, 266–279. http://dx.doi.org/10.1016/j.cedpsych.2011.03.004.

Lerkkanen, M. -K., Poikkeus, A. -M., & Ketonen, R. (2006). ARMI - Luku- ja kirjoitustaidonarviointimateriaali 1. luokalle (ARMI−A tool for assessing reading and writing skills inGrade 1). Helsinki, Finland: WSOY.

Lerkkanen, M. -K., Rasku-Puttonen, H., Aunola, K., & Nurmi, J. -E. (2004). The develop-mental dynamics of literacy skills during the first grade. Educational Psychology, 24,793–810. http://dx.doi.org/10.1080/0144341042000271782.

Logan, S., & Johnston, R. (2010). Investigating gender differences in reading. EducationalReview, 62, 175–187. http://dx.doi.org/10.1080/00131911003637006.

Määttä, S., Stattin, J., & Nurmi, J. (2006). Achievement strategies in peer groups and ad-olescents' school adjustment and norm-breaking behavior. Scandinavian Journal ofPsychology, 47, 273–280. http://dx.doi.org/10.1111/j.1467-9450.2006.00517.x.

Marsh, H. W., Martin, A. J., & Cheng, J. S. (2008). A multilevel perspective on gender inclassroom motivation and climate: Potential benefits of male teachers for boys?Journal of Educational Psychology, 100, 78–95. http://dx.doi.org/10.1037/0022-0663.100.1.78.

Mashburn, A. J., Justice, L. M., Downer, J. T., & Pianta, R. C. (2009). Peer effects onchildren's language achievement during pre-kindergarten. Child Development, 80,686–702. http://dx.doi.org/10.1111/j.1467-8624.2009.01291.x.

McClelland, M., Acock, A., Piccinin, A., Rhea, S., & Stallings, M. (2013). Relations be-tween preschool attention span-persistence and age 25 educational outcomes. EarlyChildhood Research Quarterly, 28, 314–324.

McCombs, B. (2010). Learner-centered practices: Providing the context for positivelearner development, motivation, and achievement. In J. Meece, & J. Eccles (Eds.),Handbook of research on schools, schooling, and human development (pp. 60–74).New York, NY: Routledge.

McGraw, K., & Wong, S. P. (1996). Forming inferences about some intraclass correlationcoefficients. Psychological Methods, 1, 30–46. http://dx.doi.org/10.1037/1082-989X.1.1.30.

Mercer, S., McMillen, J., & DeRosier, M. (2009). Predicting change in children's aggres-sion and victimization using classroom-level descriptive norms of aggression andpro-social behavior. Journal of School Psychology, 47, 267–289. http://dx.doi.org/10.1016/j.jsp.2009.04.001.

Muthén, L. K., & Muthén, B. O. (1998–2012). Mplus user's guide (Sixth ed.). Los Angeles,CA: Muthén & Muthén.

Onatsu-Arvilommi, T., & Nurmi, J. -E. (2000). The role of task-avoidant and task-focusedbehaviors in the development of reading and mathematical skills during the firstschool year: A cross-lagged longitudinal study. Journal of Educational Psychology, 92,478–491. http://dx.doi.org/10.1037//0022-0663.92.3.478.

Ortiz, M., Folsom, J., & Al Otaiba, S. (2012). The componential model of reading:Predicting first grade reading performance of culturally diverse students from ecolog-ical, psychological, and cognitive factors assessed at kindergarten entry. Journal ofLearning Disabilities, 45, 406–417. http://dx.doi.org/10.1177/0022219411431242.

Passolunghi, M. C., Mammarella, I. C., & Altoè, G. (2008). Cognitive abilities as pre-cursors of the early acquisition of mathematical skills during first through secondgrades. Developmental Neuropsychology, 33, 229–250. http://dx.doi.org/10.1080/87565640801982320.

Perry, K., Donohue, K., & Weinstein, R. (2007). Teaching practices and the promotion ofachievement and adjustment in first grade. Journal of School Psychology, 45, 269–292.http://dx.doi.org/10.1016/j.jsp.2007.02.005.

Pintrich, P. R., & Schunk, D. H. (2002). Motivation in education: Theory, research, and ap-plications (2nd ed.). Columbus, OH: Merrill-Prentice Hall.

Poskiparta, E., Niemi, P., & Lepola, J. (1994).Diagnostiset testit 1. Lukeminen ja kirjoittaminen[Diagnostic tests 1. Reading and writing]. Turku, Finland: Centre of Learning Research.

Preacher, K. J., Curran, P. J., & Bauer, D. J. (2006). Computational tools for probing inter-action effects in multiple linear regression, multilevel modeling, and latent curveanalysis. Journal of Educational and Behavioral Statistics, 31, 437–448.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and dataanalysis methods (2nd ed.). Newbury Park, CA: Sage.

Riigikogu (2010/2013). Põhikooli- ja Gümnaasiumiseadus. [Law of basic schools andupper secondary schools]. Riigi Teataja I, 41, 240 (Retrieved May 30, 2013, fromhttps://www.riigiteataja.ee/akt/110072012020).

Rubin, K. H., Bukowski, W., Parker, J., & Bowker, J. (2008). Peer interactions, relation-ships, and groups. In W. Damon, & R. Lerner (Eds.), Child and adolescent development:An advanced course (pp. 141–180). New York, NY: Wiley.

Ruble, D. N., Feldman, N. S., & Boggiano, A. K. (1976). Social comparison between youngchildren in achievement situations. Developmental Psychology, 12, 192–197. http://dx.doi.org/10.1037/0012 ̶ 1649.12.3.192.

Rutter, M., & Maughan, B. (2002). School effectiveness findings 1979–2002. Journal ofSchool Psychology, 40, 451–475.

Ryan, A. (2001). The peer group as a context for the development of young adolescentmotivation and achievement. Child Development, 72, 1135–1150.

Schaefer, B., & McDermott, P. (1999). Learning behavior and intelligence as explanationsfor children's scholastic achievement. Journal of School Psychology, 37, 299–313.

Schunk, D. H. (2003). Self-efficacy for reading and writing: Influence of modeling, goalsetting, and self-evaluation. Reading and Writing Quarterly, 19, 159–172.

Seymour, P., Aro, M., & Erskine, J. M. (2003). Foundation literacy acquisition in Europeanorthographies. British Journal of Psychology, 94, 143–174.

Social Survey, European (2008). European social survey (ESS) data. Norwegian socialscience data services (Retrieved May 30, 2013 from http://ess.nsd.uib.no/ess/).

Spinath, B., & Spinath, F. M. (2005). Development of self-perceived ability in elementaryschool: The role of parents' perceptions, teacher evaluations, and intelligence.Cognitive Development, 20, 190–204. http://dx.doi.org/10.1016/j.cogdev.2005.01.001.

Stipek, D., & Byler, P. (2004). The early childhood classroom observation measure. EarlyChildhood Research Quarterly, 19, 375–397. http://dx.doi.org/10.1016/j.ecresq.2004.07.007.

Stipek, D., & Byler, P. (2005). Early childhood classroom observation measure: Codingmanual (Electronically from authors, 2005).

Stipek, D. J., Feiler, R., Daniels, D., & Milburn, S. (1995). Effects of different instructionalapproaches on young children's achievement and motivation. Child Development, 66,209–223. http://dx.doi.org/10.1111/1467-8624.ep9503233348.

Südkamp, A., Kaiser, J., & Möller, J. (2012). Accuracy of teachers' judgments of students'academic achievement: A meta-analysis. Journal of Educational Psychology, 104,743–762. http://dx.doi.org/10.1037/a0027627.

Trautwein, U., Gerlach, E., & Lüdtke, O. (2008). Athletic classmates, physical self-concept,and free-time physical activity: A longitudinal study of frame of reference effects.Journal of Educational Psychology, 100, 988–1001. http://dx.doi.org/10.1037/0022-0663.100.4.988.

Turner, J. C., Midgley, C., Meyer, D. K., Gheen, M., Anderman, E. M., Yongjin, K., et al.(2002). The classroom environment and students' reports of avoidance strategies inmathematics: A multimethod study. Journal of Educational Psychology, 94, 88–106.http://dx.doi.org/10.1037//0022-0663.94.1.88.

Tymms, P. (2004). Effect sizes in multilevel models. In I. Schagen, & K. Elliot (Eds.), Butwhat does it mean? The use of effect sizes in educational research (pp. 55–66).London, England: National Foundation for Educational Research.

Urdan, T., Ryan, A., Anderman, E., & Gheen, M. (2002). Goals, goal structures, and avoid-ance behaviors. In C. Midgley (Ed.), Goals, goals structures, and patterns of adaptivelearning (pp. 55–83). Mahwah, NJ: LEA.

Vabariigi Valitsus (2002/2010). Põhikooli ja gümnaasiumi riiklik õppekava. [Nationalcurriculum for basic schools and upper secondary schools]. Riigi Teataja I, 6, 21(Retrieved May 30, 2013, from https://www.riigiteataja.ee/akt/13276017).

Vabariigi Valitsus (2008/2011). Koolieelse lasteasutuse riiklik õppekava. [National curric-ulum for preschool institutions]. Riigi Teataja I, 23, 152 (RetrievedMay 30, 2013, fromhttps://www.riigiteataja.ee/akt/13351772).

Valeski, T. N., & Stipek, D. J. (2001). Young children's feelings about school. ChildDevelopment, 72, 1198–1213.

Vygotsky, L. (1978). Mind in society. Development of higher psychological processes.Cambridge, MA: Harvard University Press.

Walker, J. M. T. (2008). Looking at teacher practices through the lens of parenting style.The Journal of Experimental Education, 76, 218–240.

Wentzel, K. (2010). Students' relationships with teachers. In J. Meece, & J. Eccles (Eds.),Handbook of research on schools, schooling, and human development (pp. 75–91).New York, NY: Routledge.

Wentzel, K., & Watkins, D. (2002). Peer relationships and collaborative learning as con-texts for academic enablers. School Psychology Review, 31, 366–377.

Wentzel, K. R., & Wigfield, A. (1998). Academic and social motivational influences onstudent's academic performance. Educational Psychology Review, 10, 155–175.

Woolfolk Hoy, A., & Weinstein, C. (2011). Student and teacher perspectives on classroommanagement. In C. Evertson, & C.Weinstein (Eds.),Handbook of classroommanagement.Research, practice, and contemporary issues. (pp. 181–219). New York, NY: Routledge.

Yen, K., Konold, T., & McDermott, P. (2004). Does learning behavior augment cognitiveability as an indicator of academic achievement? Journal of School Psychology, 42,157–169. http://dx.doi.org/10.1016/j.jsp.2003.12.001.

Zhang, X., Nurmi, J. -E., Kiuru, N., Lerkkanen, M. -K., & Aunola, K. (2011). A teacher-report measure of children's task-avoidant behavior: A validation study of the Behav-ioral Strategy Rating Scale. Learning and Individual Differences, 21, 690–698. http://dx.doi.org/10.1016/j.lindif.2011.09.007.