Learning and Individual Differences 22 (2012) 490–497 Contents lists available at SciVerse ScienceDirect Learning and Individual Differences journal homepage: www.elsevier.com/locate/lindif Associations of temperament traits and mathematics grades in adolescents are dependent on the rater but independent of motivation and cognitive ability Mirka Hintsanen a, b,⁎, Saija Alatupa b, Markus Jokela b, Jari Lipsanen b, Taina Hintsa b, Mare Leino c a b c Helsinki Collegium for Advanced Studies, P.O. Box 4 (Fabianinkatu 24), 00014 University of Helsinki, Helsinki, Finland IBS, Unit of Personality Work and Health Psychology, P.O. Box 9 (Siltavuorenpenger 1 A), 00014 University of Helsinki, Helsinki, Finland Institute of Social Work, Tallinn University, Narva mnt 25, 10120 Tallinn, Estonia a r t i c l e i n f o Article history: Received 6 July 2011 Received in revised form 24 January 2012 Accepted 3 March 2012 Keywords: Temperament Mathematics Academic achievement School performance Cognitive ability a b s t r a c t The current study examines associations between self- and teacher-rated temperament traits (activity, inhibition, negative emotionality, persistence, distractibility, and mood) and mathematics grades. The sample includes 310 ninth grade students (mean age 15.0) from several schools in Finland. Analyses were conducted with multilevel modeling. Except for mood, all temperament traits were associated with math grades independent of motivation and standardized cognitive ability test performance (intelligence). However, some relations were found only for teacher-reports and some only for student-reports. The rater dependent differences are of interest and emphasize a need for data from multiple raters before drawing conclusions on how temperament relates to school grades. The results suggest that temperament should be taken into account in schooling and in teacher education. © 2012 Elsevier Inc. All rights reserved. 1. Introduction Academic achievement is an early milestone in developmental pathways leading to adult psychological and social outcomes. For example, it predicts factors related to social adjustment, such as social competence and peer acceptance (Chen, Rubin, & Li, 1997). Moreover, school performance is associated with adjustment problems, health, and well-being later in life. Lower academic achievement has been shown to predict somatic health risks (Alatupa et al., 2010; Bryant, Schulenberg, Bachman, O'Malley, & Johnston, 2000; Huurre et al., 2010) and depression (Chen, Rubin, & Li, 1995). Academic achievement is also predictive of future educational attainment (Marjoribanks, 2005), socioeconomic position (Guglielmi, 2008; Judge & Hurst, 2007), and the risk of unemployment in adulthood (Caspi, Wright, Moffitt, & Silva, 1998). In the future, especially the importance of education related to science, technology, engineering and math is increasing rapidly as the occupations related to these fields demand more and more employees (Carnevale, Smith, & Melton, Abbreviations: DOTS-R, Revised Dimensions of Temperament Survey; TABC-R, Temperament Assessment Battery for Children Revised. ⁎ Corresponding author at: Helsinki Collegium for Advanced Studies, University of Helsinki, P.O. Box 4, FIN-00014 University of Helsinki, Finland. Tel.: + 358 9 191 29517; fax: + 358 9 191 29521. E-mail addresses: mirka.hintsanen@helsinki.fi (M. Hintsanen), saija.alatupa@helsinki.fi (S. Alatupa), markus.jokela@helsinki.fi (M. Jokela), jari.lipsanen@helsinki.fi (J. Lipsanen), taina.hintsa@helsinki.fi (T. Hintsa), eram@tlu.ee (M. Leino). 1041-6080/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.lindif.2012.03.006 2011). For this reason, skills in mathematics and related subjects have special importance for the individual as well as for the society. It is important to identify factors that predict mathematics achievement in order to be able to intervene and help students to perform better. Adjustment to school and school performance can be seen to be dependent on the goodness of fit between an individual's cognitive and socio-emotional capacity and an individual's environment (Thomas & Chess, 1977, 1980). A poor fit, resulting from disparity between individual characteristics and environmental demands, may lead to negative outcomes (Thomas & Chess, 1980). In the school setting, a poor fit may result in reduced motivation, distractions in the learning process and diminished learning capacity. In this context, temperament traits play an especially important role. Temperament describes the how of behavior in contrast to the what and why of behavior that are related to ability and motivation, respectively (Thomas & Chess, 1977). Temperament traits are comparatively stable biologically based behavioral tendencies that appear early in life (Goldsmith et al., 1987) and have heritable underpinnings (Keller, Coventry, Heath, & Martin, 2005). Furthermore, temperament is thought to form a basis for personality development. According to an extensive review by Strelau (1998), temperament is more closely related to biology and it is postulated to be present already in early childhood, whereas personality is more strongly affected by social factors, and emerges later as a result of socialization and learning. These kinds of differences separate temperament traits from most personality traits (Strelau, 1998). Although most temperament research has focused on children, temperament remains relevant also in adolescence and adulthood (Thomas & Chess, 1977). M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 Temperament traits predict attention (Curtindale, Laurie-Rose, Bennett-Murphy, & Hull, 2007), interests, motivation (Elliot & Thrash, 2002), cognitive strategies (Davis & Carr, 2002), and therefore also working styles (Kristal, 2005). Some studies have shown that associations of temperament and factors related to working styles are moderate or large in magnitude (Davis & Carr, 2002). Temperament traits may contribute to adjustment to the requirements of school environment or create extra challenges to it. In the school context, a combination of traits forming a factor called task orientation, has been shown to be of particular relevance (Oliver, Guerin, & Gottfried, 2007). Task orientation is one of the three temperament factors describing an ideal student who is, in a teacher's view, easy to teach (Keogh, 1994). Students with high task orientation can engage in and finish tasks despite disruptions, and they are able to moderate and direct their activity according to the task at hand (Keogh, 1983). Therefore, task orientation can be seen as having close connections with working styles. High task orientation is formed by a temperament trait combination of high persistence, low activity and low distractibility (Keogh, 1983). Activity refers to a tendency to show high motor activity, whereas persistence refers to the ability to concentrate and engage in tasks for long periods of time (Martin & Bridger, 1999). Characteristic to distractibility is the tendency to easily shift attention to surrounding stimuli instead of being able to concentrate on the task at hand (Windle, 1992; Windle & Lerner, 1986). Another two temperament constellations related to a teachers' view of an ideal, teachable student are personal–social flexibility and reactivity, the first of which includes temperament traits high adaptability, tendency to approach rather than withdraw and positive mood, whereas the latter constellation includes negative mood and tendency to respond to slightest cues with intensity (Keogh, 1983). Traits related to personal–social flexibility and reactivity may be associated to working styles but they are more likely to be related to social interaction and emotional adjustment, which may then be reflected in motivation and teacher's evaluations. We assess personal–social flexibility with two traits: low inhibition and high mood. Inhibition is a tendency to be careful and to hesitate in unknown and social situations (Martin & Bridger, 1999), whereas mood refers to a tendency to experience and show positive affect (Windle, 1992; Windle & Lerner, 1986). We assess reactivity with one trait: negative emotionality refers to a tendency to easily experience negative feelings such as anger and irritation (Martin & Bridger, 1999). Empirical evidence of the association between temperament and academic achievement is limited (Martin & Holbrook, 1985; Zhou, Main, & Wang, 2010). Previous research on mathematics achievement, has shown most support (in general, with moderate to strong associations) for better math achievement being associated with traits forming high task orientation, i.e. high persistence, low distractibility and low activity (Bramlett, Scott, & Rowell, 2000; Guerin, Gottfried, Oliver, & Thomas, 1994; Martin, 1989; Martin & Holbrook, 1985; Mullola et al., 2010). Few studies and findings on negative emotionality, mood, and inhibition exist. Higher negative emotionality, lower mood, and higher inhibition have been found to correlate with lower math grades although the correlations have been only small in size (Mullola et al., 2010). While previous studies suggest that temperament traits may be related to mathematics achievement, there are some important limitations in existing research. The majority of studies on temperament and math achievement have included small samples of approximately 100 participants or less (Bramlett et al., 2000; Checa, RodríguezBailón, & Rueda, 2008; Guerin et al., 1994; Martin & Holbrook, 1985; Martin, Nagle, & Paget, 1983; Mevarech, 1985). Data have often been collected from one (Checa et al., 2008; Martin & Holbrook, 1985) or only few schools (Martin et al., 1983; Mevarech, 1985) and few classes, which may limit the generalizability of the findings. 491 Furthermore, most previous publications on temperament and mathematics achievement have focused on age groups between the ages of four and 10 (Bramlett et al., 2000; Martin & Holbrook, 1985; Martin et al., 1983; Mevarech, 1985; Rudasill, Gallagher, & White, 2010). However, the associations of temperament with mathematics achievement might be expected to become stronger with age. In mathematics, one topic is practiced for a long time, after which, it is expected to be mastered. Furthermore, learning new things is based on skills learned before. If one lags behind, it might be difficult to catch up, and these difficulties are likely to accumulate over time. Consequently, the strongest associations with temperament may be seen most clearly in the final school years. Although only few studies have examined this issue, it has been reported that the association between temperament traits and math performance seem to get somewhat stronger with increasing age (Guerin et al., 1994; Maziade, Cote, Boutin, Boudreault, & Thivierge, 1986). A recent large-scale study reported associations between teacher-rated temperament traits and mathematics grades (Mullola et al., 2010) but this study did not take into account student's selfrated temperament and cognitive ability assessed with standardized test. In the current study, we examine associations of self- and teacher-rated temperament with math grades in a sample of 310 ninth graders from several schools and classes in Finland. Self- and teacher-ratings of temperament are included in order to improve the validity and to examine whether the findings are dependent on the rater. Very few previous studies have included self- or parent-ratings alongside with teacher-ratings (Bramlett et al., 2000). Temperament measures that are used are based on the temperament theory by Thomas and Chess (1977) that describes nine temperament traits. Our study included fewer traits for instance leaving out rhythmicity that was considered somewhat difficult to assess in school setting where curricular activities largely define the daily rhythms. We use Martin's temperament inventory that has been especially developed for the school setting (Martin & Bridger, 1999). Measures were complemented with two relevant scales from another inventory (Windle, 1992) as these traits were not included in Martin's measure. We include student motivation and standardized math-related cognitive test performance as they may be explaining factors in the association between temperament and math grades. We aim to examine which temperament traits are associated with math grades, and whether these associations are independent of motivation and cognitive ability test performance. Based on previous research we hypothesized that higher persistence and mood and lower distractibility, activity, negative emotionality and inhibition would be associated with higher math grade. The Finnish school system is very homogenous which has been demonstrated for example as the lowest between-school variance in science performance among 54 countries in PISA study in 2006 (OECD, 2009). Finnish schools show little quality differences and little selection related to age, gender, motivation or ability of the students. Residential areas are not highly segregated and the vast majority of students are enrolled to the nearest school. Private schools are virtually non-existent. In all schools, teaching is organized according to the same national curriculum. Furthermore, all teachers receive university education reducing the differences between teachers. This comparatively homogenous environment reduces school and teacher related variance and therefore offers a good setting for examining individual variance in academic achievement. 2. Material and methods 2.1. Participants The subjects were 354 ninth graders who participated in a sub-study of the research project “The Finnish Study on Temperament and School 492 M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 Achievement” (Alatupa, Karppinen, Keltikangas-Järvinen, & Savioja, 2007). Finnish school system consists of 9 years of compulsory schooling starting when the children are seven-year-olds. Lower-comprehensive school (grades 1–6) encompasses the first 6 years and uppercomprehensive school (grades 7–9) the last 3 years. Therefore, ninth grade is the final year of compulsory schooling during which the students are aged 15 or 16 years. In upper-comprehensive schools, teachers are subject teachers and every subject is, therefore, taught by a different teacher. The data were collected in 2005–2006 in upper-comprehensive schools in the areas of Lahti and Kirkkonummi in southern Finland. All schools in these areas were first listed and then randomly selected for the current study. In case a school refused to participate, the next school on the list was selected. As a result, the sample consisted of 354 students in 11 schools. Swedish-speaking and special schools were not sampled. All students voluntarily completed a test battery during regular class sessions. The participants with information on all study variables formed the final sample (N=309). The mean age of the participants was 15.0 years (S.D.=0.34). The sample included 144 girls (46.6%) and 165 boys (53.4%). In the final sample, the students were from 11 different schools and from 18 classes. In addition to the students, their teachers were asked to participate as additional raters. Teacher-ratings were obtained in 5 months time period between October and February. There were 18 teachers in the final data of the current study. The same teacher is responsible for teaching a certain group of students throughout the whole upper-comprehensive school, i.e. through grades 7–9. Therefore, the teachers had been teaching these students more than two school years. A majority of the teachers were women (n=16). The mean age of the teachers was 43.2 years. The teachers were not paid for their contribution. 2.2. Measures 2.2.1. Mathematics grades In a written sheet, the students were asked to report their math grade in their latest school report (report from the autumn term or from the preceding spring term). The final grades from ninth school year had not been assigned during the data collection (lasting from October to end of February). The grades range between 4 and 10 (4 = fail, 5–6 = poor, 7–8 = good, and 9–10 = excellent). 2.2.2. Temperament Self- and teacher-reports were used to assess students' temperament. Scale meanings and the numbers of items are presented in Table 1. Information on temperament was collected using two different measures. First, four scales of the Temperament Assessment Battery for Children Revised (TABC-R) (Martin & Bridger, 1999) were used: inhibition (see Table 1 for the items), persistence, negative emotionality, and activity. Some of the items of TABC-R were slightly modified to be more age appropriate for the current participants. Secondly, two scales from the Revised Dimensions of Temperament Survey (DOTS-R) (Windle, 1992) were applied: mood and distractibility. All items were answered on a five-point scale, ranging from 1 (strongly disagree) to 5 (strongly agree). Temperament scores were calculated only for those participants who had answered to at least 50% of the items of a scale. Others were excluded from the analyses. In The Finnish Study on Temperament and School Achievement data (i.e. the data of the current study), the following internal reliabilities (Cronbach's α) have been previously reported for inhibition, persistence, negative emotionality, activity, mood and distractibility, respectively: 0.83, 0.60, 0.65, 0.51, 0.91 and 0.72, for self-reported temperament (Hintsanen, Alatupa, Pullmann, Hirstiö-Snellman, & Keltikangas-Järvinen, 2010) and 0.90, 0.93, 0.81, 0.86, 0.96, and 0.91 for teacher-rated temperament (Mullola et al., 2010). 2.2.3. Motivation General school motivation was self-reported with six-items (α = 0.69), e.g. “My school work is very important to me” and “If I want, I can improve my results”. Answers were rated on a five-point scale, ranging from 1 (strongly disagree) to 5 (strongly agree). The motivation scale was developed for this particular study. 2.2.4. Cognitive ability test performance Cognitive ability was assessed with three standardized tests (R1, R3, and V3) developed for assessing components of cognitive performance in vocational guidance in the age group of the current study (Pulliainen, 1994). Tests R1 and R3 assess inference skills. R1 is a task of code language and R3 is a task of analogies. Test V3 includes tasks where complicated instructions have to be followed. There are various kinds of V3 tasks including inference tasks and math problems (Pulliainen, 1994). The test scores were standardized so that a mean of 0 and standard deviation of 1 were achieved. After that, the mean of these three test scores was calculated and the formed variable (α = 0.74) was used in the analyses. 2.2.5. Non-attendance The amount of days absent from school during the last semester was self-reported by the students on a following scale: 1—none (n = 51), 2—a few days (n = 201), 3—a few weeks (n = 53), 4—several months (n = 5). 2.3. Statistical analyses The associations between independent variables (teacher-rated and self-rated temperament) and dependent variable (math grade) were assessed with random-intercept multilevel modeling using restricted maximum-likelihood estimation (REML). Separate models were fitted for teacher-rated and self-rated temperament with students being nested within teachers and school classes (the sample included only one teacher for each class). Multilevel modeling calculates the standard errors of regression coefficients by taking into account the clustered nature of the data, i.e., the non-independence of the first-level observations (students) nested within second-level groups (teachers/classes). Intra-class correlation was 0.09 (i.e. the second-level, i.e. teacher or class, explained 9% in math grade). All student-level continuous independent variables were standardized before conducting the analyses. First, a random intercept model was calculated. Second, a random coefficient model was calculated. Model and covariance structure (diagonal, unstructured) was chosen by considering changes in Akaike's information criteria along with maximizing the amount of significant covariance parameters (p b 0.1 as a criterion). Chosen models and covariance structures are reported in Tables 4 and 5. Each teacher-rated temperament trait was examined separately with the following variables included as first-level variables in addition to the temperament trait in question: model 1) student's age and gender, model 2) student's age, gender, motivation, ability test score and amount of non-attendance by the student. The same procedure was followed in examining self-rated temperament. Estimates of effect sizes for the whole model (Adjusted R 2) and for the temperament traits (R 2 change) were calculated with linear regression analyses as there are still not well-established methods for calculating effect sizes in multilevel modeling (Hox, 2010). Gender interactions predicting math achievement were in general not significant except for interaction with self-rated inhibition (p = 0.043) which we considered to be a chance finding due to the number of interaction analyses performed. Therefore, all analyses were performed for a combined sample of girls and boys. All analyses were conducted with a mixed procedure in SPSS 15.0 software. M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 493 Table 1 Temperament scale meanings in self- and teacher-rated scales. Temperament Assessment Battery for Children Revised (TABC-R) Self-rated inhibition Being shy with strange adults Being uncomfortable with strange adults Socializing with new people in different situations (r) At home approaching new visiting people (r) Not feeling comfortable with new situations and people Being off-balance with new people and situations Being relaxed with new people (r) Being shy with new people Teacher-rated inhibition Being shy with strange adults Talking loudly about own experiences (r) Avoiding new people and situations Time taken before comfortable in a new situation Not hesitating to begin new activities (r) Difficulty of knowing students feelings Not hesitating to perform in class (r) Being shy with new students Movements being slow Self-rated persistence Staying a long time with same activity Practicing new activities for a long time Changing activities when an activity is difficult (r) Giving up easily in front of difficult tasks (r) Enthusiasm to learn even difficult tasks Teacher-rated persistence Returning to interrupted activity Work being disturbed by classroom noises (r) During breaks spends short time in each activity (r) When teacher explains, attention span short (r) Ability to continue an activity for long time Ability to concentrate on teaching despite noise Beginning but not finishing (r) Being distracted easily (r) Self-rated negative emotionality Getting irritated when tired Not being easily relented Showing anger when corrected Showing anger when someone limits activities Difficulty to focus on things selected by someone else Difficult to comfort when upset Protesting when having to leave an enjoyable activity Being carefree (r) Teacher-rated negative emotionality Accepts a replacement instead of demanding other student's tools (r) Getting upset easier than other students Showing anger when not liking something Does not mind if looses (r) Student gets upset when teacher comments on behavior Overreacts to stressors Gets upset with other students Being moody Self-rated activity Liking active tasks as compared to sitting still Moving fast Moving when talking Fidgeting while sitting Entering room fast and loudly Liking to run in stairs Teacher-rated activity Difficult to sit still Can sit still for a long time (r) During breaks prefers quiet activities (r) Sits still while teacher is explaining (r) Revised Dimensions of Temperament Survey (DOTS-R) Self-rated mood Laughing and smiling easily Not laughing and smiling easily (r) Smiling Not laughing (r) Being on positive mood Laughing and smiling during a day Being happy Teacher-rated mood Laughing and smiling easily Not laughing and smiling easily (r) Smiling Not laughing (r) Being on positive mood Laughing and smiling during a day Being happy Self-rated distractibility Not turning attention away (r) Turning attention away Not interrupting task at hand (r) Things happening around not turning attention away from task (r) Ability to concentrate (r) Teacher-rated distractibility Not turning attention away (r) Turning attention away Not interrupting task at hand (r) Things happening around not turning attention away from task (r) Ability to concentrate (r) (r)—Reversed item. 3. Results 3.1. Bivariate correlations Table 2 presents study characteristics and Table 3 presents bivariate correlations between study variables. Higher teacher-rated persistence and mood and lower activity, inhibition, negative emotionality, and distractibility were correlated with higher math grade. Of the student-rated temperament traits only higher persistence and inhibition and lower distractibility were correlated with higher math grade. Higher motivation and especially higher ability test score were also correlated with higher math grade. Correlations between self- and teacher-ratings of a given temperament trait were the following: activity r = .120, p b .05; inhibition r = .371, p b .01; negative emotionality r = .109, p = ns; persistence r = .219, p b .01; distractibility r = .231, p b .01; mood r = .317, p b .01. All teacherrated temperaments correlated with ability test score, whereas of the student-rated temperament traits, only persistence correlated with it. 3.2. Associations between temperament and math grade Table 4 presents the results of multilevel modeling analyses for teacher-rated temperament traits. In multilevel modeling, a random intercept model was chosen for all analyses on self-rated temperament and for a majority of analyses on teacher-rated temperament as can be seen in Tables 4 and 5. High teacher-rated persistence and mood and low teacher-rated activity, negative emotionality and 494 M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 Table 2 Characteristics of the study variables. Variable (scale range) Mean (SD) N (%) 144 166 (46.5) (53.5) Table 4 Multilevel modeling analyses of teacher-rated temperament traits predicting math grade (N = 310). Model 1 Gender Girl Boys Age (14–16) 15.0 Motivation (1–5) 3.5 Ability (− 1.9 to 3.0) 0.1 Teacher-rated temperament (1–5) Activity 2.4 Inhibition 2.8 Negative emotionality 2.3 Persistence 3.4 Distractibility 3.0 Mood 3.7 Student-rated temperament (1–5) Activity 2.9 Inhibition 2.9 Negative emotionality 2.7 Persistence 3.4 Distractibility 3.1 Mood 4.0 Math grade (4–10) 7.6 (0.3) (0.6) (0.8) (1.0) (0.8) (0.7) (0.9) (0.9) (0.9) Estimate Activity Inhibition Negative emotionality Persistence Distractibility Mood Model 2 p Estimate b.001⁎⁎⁎ −.626 −.144 −.611 .951a −.849a .161 .054 b.001⁎⁎⁎ b.001⁎⁎⁎ b.001⁎⁎⁎ .044⁎ −.353 .002 −.289 .518 −.490 .024 p b.001⁎⁎⁎ .977 b.001⁎⁎⁎ b.001⁎⁎⁎ b.001⁎⁎⁎ .681 Model 1—controlled for student's age and gender. Model 2—controlled for student's age, gender, motivation, ability test score, and nonattendance. Note: random intercept model was chosen for all analyses except for those marked with ª. a Random coefficient model with covariance type diagonal. ⁎⁎⁎ p b .001. ⁎ p b .05. (0.5) (0.7) (0.5) (0.6) (0.6) (0.7) (1.3) ranged from 0.005 for distractibility to 0.012 for inhibition. Variance explained by the fully adjusted model before entering temperament was 47.7% (Adjusted R 2 = 0.477). distractibility were associated with better math grade in age and gender adjusted models. Except for mood, associations remained significant in the fully adjusted model. In the age and gender adjusted model, effect sizes (R 2 change) for the significant temperament traits ranged from 0.020 for mood to 0.304 for distractibility and 0.329 for persistence. Variance explained by the model before entering temperament was 1.1% (Adjusted R 2 = 0.011). In the fully adjusted model, the effect sizes for the significant temperament associations ranged from 0.009 for negative emotionality to 0.078 for persistence and 0.083 for distractibility. Variance explained by the model before entering temperament was 47.6% (Adjusted R 2 = 0.476). Table 5 presents the results of multilevel modeling analyses for self-rated temperament traits. High self-rated persistence and inhibition and low self-rated activity and distractibility were associated with better math grade in age and gender adjusted analyses. Except for activity, temperament associations remained significant when other control variables were added. In the age and gender adjusted model, effect sizes (R 2 change) for the significant temperament traits ranged from 0.009 for activity to 0.094 for persistence. Variance explained by the model before entering temperament was 1.1% (Adjusted R 2 = 0.011). In the fully adjusted model, effect sizes for the significant temperament associations 4. Discussion Our findings demonstrate that temperament is associated with math grades even when motivation and cognitive ability are taken into account. Thus, the relation between temperament and math grade is independent of motivation and standardized cognitive ability test performance. This is an important observation because cognitive ability is a strong predictor of academic achievement (Deary, Strand, Smith, & Fernandes, 2007). Our results imply that in addition to cognitive ability, math grades reflect other factors. Self- and teacher-rated temperaments were associated with math-related cognitive ability test performance but associations to math grades were, in general, stronger. All temperament traits were related with math grades but associations for negative emotionality and mood were only found in teacher-reports, whereas associations for inhibition were only found in self-reports. There are at least four possible mechanisms that may explain the associations between temperament and math grades. First, temperament may be associated with math grades via cognitive working styles (e.g. tendency to use verbalization or visualization). There is evidence on associations between temperament traits and preferences to use certain strategies (Davis & Carr, 2002). Table 3 Bivariate correlations. 1. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. Age Gendera Absenteeism Motivation Ability Activity Inhibition Negative emotionality Persistence Distractibility Mood Math grade − 0.11 0.07 − 0.03 − 0.11 0.02 0.04 0.05 − 0.03 0.02 − 0.07 − 0.08 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. − 0.11 0.07 − 0.05 − 0.03 − 0.09 − 0.11 − 0.11 0.26⁎⁎ 0.03 0.25⁎⁎ 0.00 0.20⁎⁎ 0.17⁎⁎ − 0.18⁎⁎ 0.06 0.03 − 0.07 − 0.04 0.06 − 0.15⁎⁎ 0.11 0.11 0.13⁎ − 0.19⁎⁎ − 0.11 − 0.15⁎ − 0.14⁎ 0.33⁎⁎ 0.18⁎⁎ 0.06 0.17⁎⁎ 0.07 − 0.29⁎⁎ − 0.10 0.23⁎⁎ 0.00 0.10 0.04 0.18⁎⁎ − 0.28⁎⁎ − 0.24⁎⁎ 0.19⁎⁎ − 0.08 0.12⁎ − 0.07 0.46⁎⁎ 0.62⁎⁎ − 0.05 − 0.09 0.26⁎⁎ − 0.28⁎⁎ − 0.04 − 0.20⁎⁎ 0.28⁎⁎ − 0.30⁎⁎ 0.12⁎ 0.12⁎ − 0.10 0.04 0.19⁎⁎ − 0.10 0.15⁎⁎ − 0.19⁎⁎ 0.17⁎⁎ 0.01 − 0.07 0.25⁎⁎ − 0.28⁎⁎ − 0.12⁎ − 0.23⁎⁎ 0.39⁎⁎ − 0.38⁎⁎ 0.07 0.46⁎⁎ 0.05 − 0.28⁎⁎ − 0.20⁎⁎ − 0.29⁎⁎ 0.40⁎⁎ − 0.35⁎⁎ 0.23⁎⁎ 0.61⁎⁎ −0.40⁎⁎ 0.61⁎⁎ − 0.84⁎⁎ 0.83⁎⁎ 0.22⁎⁎ − 0.43⁎⁎ 0.08 0.05 − 0.10 − 0.65⁎⁎ − 0.14⁎ 0.00 0.35⁎⁎ 0.16⁎⁎ − 0.71⁎⁎ 0.65⁎⁎ − 0.19⁎⁎ − 0.33⁎⁎ 0.00 − 0.23⁎⁎ − 0.18⁎⁎ − 0.92⁎⁎ 0.06 0.58⁎⁎ − 0.09 0.08 0.07 0.04 − 0.42⁎⁎ 0.05 − 0.01 − 0.56⁎⁎ 0.16⁎⁎ − 0.07 0.13⁎ − 0.05 0.29⁎⁎ − 0.21⁎⁎ − 0.01 Values below the diagonal represent correlations with teacher-rated temperament traits of the student (n = 310). Values above the diagonal represent correlations with studentrated temperament traits (n = 309). a Higher values represent girls. ⁎⁎ Correlation is significant at the 0.01 level (2-tailed). ⁎ Correlation is significant at the 0.05 level (2-tailed). M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 Table 5 Multilevel modeling analyses of student-rated temperament traits predicting math grade (n = 309). Model 1 Estimate Activity Inhibition Negative emotionality Persistence Distractibility Mood −.168 .164 −.072 .413 −.322 −.083 Model 2 p .024⁎ .025⁎ .343 b.001⁎⁎⁎ b.001⁎⁎⁎ .268 Estimate p −.062 .134 .047 .135 −.118 −.109 .279 .014⁎ .418 .024⁎ .039⁎ .054 Model 1—controlled for student's age and gender. Model 2—controlled for student's age, gender, motivation, ability test score, and nonattendance. Note: random intercept model was chosen for all analyses. ⁎⁎⁎ p b .001. ⁎ p b .05. Second, temperament may be associated with math grades through the ability to adapt to the many demands the school environment sets to the students. For instance, one has to be able to sit still and concentrate despite of the noisy environment. Third, the teacher's perceptions of student temperament may bias their grading of a student's school performance (the so-called halo effect). Teachers may, for instance, anticipate better performance from students with certain temperament traits. Some of the differences between teacher-rated and student-rated temperament trait associations are in line with such an interpretation, although this remains speculative, as we did not directly assess how teacher's opinion on student temperament affects grading. High teacher-rated activity and negative emotionality were consistently associated with lower math grades, which is in line with some (Martin, 1989; Mullola et al., 2010) but not all (Rudasill et al., 2010) previous studies. By contrast, self-rated negative emotionality, or activity were not related to math grades, when adjusted for cognitive ability and motivation. In agreement with these findings, high activity and high negative emotionality have been related to more negative teacher-evaluations of student's educational competence, motivation and maturity (Mullola et al., in press). Furthermore, students' temperament has been shown to be reflected in the attitudes of the teachers, high activity and distractibility being related to more negative attitude (Martin et al., 1983). It should, however, be noted that in Finland, teachers are instructed to evaluate how the student works and regulates his/her working (which is highly dependent on temperament) and to take that into account in grade giving. If following these instructions leads to evaluating temperament instead of evaluating actual learning, the temperament-related bias in grade giving is built in the evaluation instructions. It may therefore not reflect teachers' own tendency to confuse evaluations of temperament to that of performance. It should be deliberated whether it is appropriate to include heritable temperament in criteria for school grades. Fourth, it is also possible that the direction between temperament assessment and school grades flows the other way around, so that high-ability students are evaluated more positively by the teachers. This interpretation is in line with our finding that when rated by the teachers, all temperament traits were correlated with cognitive ability test score, whereas when temperament traits were rated by student these associations were not found, with the exception of trait persistence. However, we controlled for cognitive ability, which makes it unlikely that this kind of halo effect would explain our results. Self-rated inhibition was positively associated with math grades, whereas the association was negative, albeit non-significant, for teacher-assessed inhibition. High inhibition could have a positive effect on school work because more inhibited children are probably not as likely to engage in task-irrelevant social activity during 495 lessons as less inhibited children. Moreover, a layman's view of a mathematically talented person is often characterized by high inhibition. It may be that those good in math form a self-concept reflecting this view. High persistence and low distractibility (components of task orientation constellation) were associated with higher grades (similarly as the third component of task orientation, namely low activity that is discussed above). As mentioned, these relations were independent of the rater of temperament. These results are in line with previous research that has found consistent associations for these traits with math performance (Guerin et al., 1994; Martin, 1989; Martin & Holbrook, 1985; Mullola et al., 2010) and between task orientation and math performance in Scholastic Achievement Test (Oliver et al., 2007). In teacher-ratings, a single temperament trait could explain as much as 7–8% of a math grade even after taking intelligence and motivation into account. In practice, temperament explains even a larger proportion because temperament is associated to motivation (Elliot & Thrash, 2002) which in turn may affect the math grade. It seems that especially teachers might mix cognitive ability in their evaluations of temperament, which is visible in the correlations presented in Table 3. Indeed, in teacher-ratings, when motivation and intelligence were not included in the model, temperament trait explained up to 33% of math grade. The possibility of temperament biasing grade giving should be paid attention to. Student characteristics may be related to an interaction between a student and a teacher in many ways: for instance, better learning opportunities may be offered to those with certain characteristics and assessment may be affected as well (Morgan, 2000). It has been suggested that if reasons for inequitable treatment would be better understood, it could be more easily avoided (Morgan, 2000). The possibility that grades are associated with a teacher's subjective opinion of student temperament unrelated to performance should be examined. Moreover, increasing goodness of fit between student temperament and the demands set by school environment is important and there are many ways for achieving this. For instance, giving students with high activity possibilities to move might make it easier for them to adjust. Furthermore, increasing emotional support may be useful, as classroom emotional support may buffer possible negative effects of lower temperamental attention on academic performance (Rudasill et al., 2010). The quality of a teacher–student relationship may be related to student coping (Ruus et al., 2007) and improvements in it may help all students. Possible ways to improve temperamental fit in schooling has been discussed in detail previously (Kristal, 2005). 4.1. Limitations and methodological considerations An important limitation of the current study is the cross-sectional design, which prevents examining temporal relations. It seems more likely that temperament would predict math grades than that math grades would predict temperament: math grades form only a small part of a person's experiences and temperament has been shown to be rather stable (Hintsanen et al., 2009; Josefsson et al., in press). However, temperament and math grades could be explained by an unmeasured third factor not included in the current study. For example, negative life events could be one such factor. Another limitation is that we could not control for the socioeconomic position of the students. However, taking motivation and ability test performance into account is likely to compensate for this omission to a large extent, as socioeconomic position is likely to influence school performance through these factors. TABS-R was originally designed for significantly younger children. Although the items were altered as more age appropriate for the current sample, some items may still not be as well suited for 496 M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 adolescents. In general, the reliabilities of the temperament trait measures that were used in the current study have been acceptable expect for self-reported activity. The reliabilities for self-rated persistence and negative emotionality are also in the lower range of acceptable. Therefore, results calculated with self-ratings of these traits should be viewed cautiously. It should also be noted that the mathematics grades were selfreported by the students. It has been shown that although selfreported grades reflect actual grades they also include some bias (Kuncel, Credé, & Thomas, 2005). Therefore, the current results should be replicated with grades taken from school records. We are not aware of any other study that would have examined both teacher-ratings and self-ratings of temperament in relation to math grades. Differences in results based on self-ratings and teacher-ratings were notable and correlations between selfand teacher-ratings of temperament were small to moderate in size. It is yet unclear, which ratings are the more accurate ones, or whether there are situational differences in self- and teacherratings. For example, an inhibited student might give a more inhibited impression at school than in more familiar surroundings. Furthermore, teacher-ratings may reflect school achievement whereas students' self-ratings may to a larger extent reflect views held by their peers. TABC-R includes a separate school relevant set of items for teacher-reports. Therefore, self-ratings and teacher-ratings were assessed with different items. This could be one explaining factor for the low correlations between self- and teacher-reported temperaments. However, traits assessed with DOTS-R (mood and distractibility) were assessed with the same items in self- and teacher-reports and the correlations were not higher than in TABC-R indicating that differences between self- and teacher-reports are not likely to be caused by item differences. Furthermore, student–teacher agreement seems to be of similar magnitude in several other areas of study, such as student social status (Hintsanen et al., 2010), somatic complaints, delinquent behavior (Grigorenko, Geiser, Slobodskaya, & Francis, 2010) and internalizing and externalizing problems (Tompkins, Hockett, Abraibesh, & Witt, 2011). A strength of the current study is the sample that includes older participants than those usually examined. The final school years are especially important as the student's future educational possibilities depend on them. Furthermore, the sample is larger than in most previous studies and includes students from several schools and classes. Furthermore, the analyses were performed with multilevel modeling, which gives a possibility to take a higher level variance properly into account unlike more traditional methods which may lead to biased findings in examining multilevel data. 4.2. Conclusions The current study shows that high persistence and inhibition as well as low activity, negative emotionality and distractibility are associated with better math grades independent of motivation and cognitive ability. This implies that temperament plays an important role in math grades. This is apparently the first study that examines both teacher-rated and self-rated temperament in association to math grades taking cognitive ability into account. The raterdependent differences we found are of interest and emphasize the need for data from multiple raters. Longitudinal research is called for to clarify the temporal directions of the associations found. Acknowledgements This work was supported by research funds of the University of Helsinki (490017). References Alatupa, S., Karppinen, K., Keltikangas-Järvinen, L., & Savioja, H. (2007). Koulu, syrjäytyminen ja sosiaalinen pääoma - Löytyykö huono-osaisuuden syy koulusta vai oppilaasta? Helsinki: Sitran raportteja 75. Alatupa, S., Pulkki-Råback, L., Hintsanen, M., Ravaja, N., Raitakari, O. T., Telama, R., et al. (2010). School performance as a predictor of adulthood obesity: A 21-year followup study. European Journal of Epidemiology, 25, 267–274. Bramlett, R. K., Scott, P., & Rowell, R. K. (2000). A comparison of temperament and social skills in predicting academic performance in first graders. Special Services in the Schools, 16, 147–158. Bryant, A. L., Schulenberg, J., Bachman, J. G., O'Malley, P. M., & Johnston, L. D. (2000). Understanding the links among school misbehavior, academic achievement, and cigarette use: A national panel study of adolescents. Prevention Science, 1, 71–87. Carnevale, A. P., Smith, N., & Melton, M. (2011). STEM: Science Technology Engineering Mathematics. Retrieved 5th January, 2012, from. http://cew.georgetown.edu/STEM/ Caspi, A., Wright, B. R. E., Moffitt, T. E., & Silva, P. A. (1998). Early failure in the labor market: Childhood and adolescent predictors of unemployment in the transition to adulthood. American Sociological Review, 63, 424–451. Checa, P., Rodríguez-Bailón, R., & Rueda, R. (2008). Neurocognitive and temperamental systems of self-regulation and early adolescents' social and academic outcomes. Mind, Brain, and Education, 2, 177–187. Chen, X., Rubin, K. H., & Li, B. -S. (1995). Depressed mood in Chinese children: Relations with school performance and family environment. Journal of Consulting and Clinical Psychology, 63, 938–947. Chen, X., Rubin, K., & Li, D. (1997). Relation between academic achievement and social adjustment: Evidence from Chinese children. Developmental Psychology, 33, 518–525. Curtindale, L., Laurie-Rose, C., Bennett-Murphy, L., & Hull, S. (2007). Sensory modality, temperament, and the development of sustained attention: A vigilance study in children and adults. Developmental Psychology, 43, 576–589. Davis, H., & Carr, M. (2002). Gender differences in mathematics strategy use: The influence of temperament. Learning and Individual Differences, 13, 83–95. Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and educational achievement. Intelligence, 35, 13–21. Elliot, A. J., & Thrash, T. M. (2002). Approach-avoidance motivation in personality: Approach and avoidance temperaments and goals. Journal of Personality and Social Psychology, 82, 804–818. Goldsmith, H. H., Buss, A. H., Plomin, R., Rothbart, M. K., Thomas, A., Chess, S., et al. (1987). Roundtable: What is temperament? Four approaches. Child Development, 58, 505–529. Grigorenko, E. L., Geiser, C., Slobodskaya, H. R., & Francis, D. J. (2010). Cross-informant symptoms from CBCL, TRF, and YSR: Trait and method variance in a normative sample of Russian youths. Psychological Assessment, 22, 893–911. Guerin, D. W., Gottfried, A. W., Oliver, P. H., & Thomas, C. W. (1994). Temperament and school functioning during early adolescence. Journal of Early Adolescence, 14, 200–225. Guglielmi, R. S. (2008). Native language proficiency, English literacy, academic achievement, and occupational attainment in limited-English-proficient students: A latent growth modeling perspective. Journal of Educational Psychology, 100, 322–342. Hintsanen, M., Alatupa, S., Pullmann, H., Hirstiö-Snellman, P., & Keltikangas-Järvinen, L. (2010). Associations of self-esteem and temperament traits to self- and teacherreported social status among classmates. Scandinavian Journal of Psychology, 51, 488–494. Hintsanen, M., Pulkki-Råback, L., Juonala, M., Viikari, J., Raitakari, M., & KeltikangasJärvinen, L. (2009). Cloninger's temperament traits and preclinical atherosclerosis: The Cardiovascular Risk in Young Finns Study. Journal of Psychosomatic Research, 67, 77–84. Hox, J. J. (2010). Multilevel analysis: Techniques and applications. New York: Routledge. Huurre, T., Lintonen, T., Kaprio, J., Pelkonen, M., Marttunen, M., & Aro, H. (2010). Adolescent risk factors for excessive alcohol use at age 32 years. A 16-year prospective follow-up study. Social Psychiatry and Psychiatric Epidemiology, 45, 125–134. Josefsson, K., Jokela, M., Cloninger, C. R., Hintsanen, M., Salo, J., Hintsa, T., et al. (in press) Temperament and character: Differences in adulthood developmental trends over a 10-year follow-up. Development and Psychopathology. Judge, T. A., & Hurst, C. (2007). Capitalizing on one's advantages: Role of core selfevaluations. Journal of Applied Psychology, 92, 1212–1227. Keller, M. C., Coventry, W. L., Heath, A. C., & Martin, N. G. (2005). Widespread evidence for non-additive genetic variation in Cloninger's and Eysenck's personality dimensions using a twin plus sibling design. Behavior Genetics, 35, 707–721. Keogh, B. K. (1983). Individual differences in temperament — A contribution to the personal, social, and educational competence of learning disabled children. In J. D. McKinney, & L. Feagans (Eds.), Current topics in learning disabilities (pp. 33–55). New Jersey: Ablex publishing. Keogh, B. K. (1994). Temperament and teachers' views of teachability. In W. B. Carey, & S. C. McDevitt (Eds.), Prevention and early intervention: Individual differences as risk factors for the mental health of children — A Festschrift for Stella Chess and Alexander Thomas. New York: Brunner/Mazel. Kristal, J. (2005). The temperament perspective: Working with children's behavioral styles. Baltimore: Paul H. Brookes Publishing. Kuncel, N. R., Credé, M., & Thomas, L. L. (2005). The validity of self-reported grade point averages, class ranks, and test scores: A meta-analysis and review of the literature. Review of Educational Research, 75, 63–82. Marjoribanks, K. (2005). Family background, academic achievement, and educational aspirations as predictors of Australian young adult's educational attainment. Psychological Reports, 96, 751–754. M. Hintsanen et al. / Learning and Individual Differences 22 (2012) 490–497 Martin, R. P. (1989). Activity level, distractibility, and persistence: Critical characteristics in early schooling. In G. A. Kohnstamm, J. E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 451–461). Chichester: John Wiley & Sons Ltd. Martin, R. P., & Bridger, R. C. (1999). Temperament Assessment Battery for Children— Revised: A tool for the assessment of temperamental traits and types of young children. Unpublished manual. Martin, R. P., & Holbrook, J. (1985). Relationship of temperament characteristics to the academic achievement of first-grade children. Journal of Psychoeducational Assessment, 3, 131–140. Martin, R. P., Nagle, R., & Paget, K. (1983). Relationships between temperament and classroom behavior, teacher attitudes, and academic achievement. Journal of Psychoeducational Assessment, 1, 377–386. Maziade, M., Cote, R., Boutin, P., Boudreault, M., & Thivierge, J. (1986). The effect of temperament on longitudinal academic achievement in primary school. Journal of the American Academy of Child Psychiatry, 25, 692–696. Mevarech, Z. R. (1985). The relationship between temperament characteristics, intelligence, task-engagement and mathematics achievement. British Journal of Educational Psychology, 55, 156–163. Morgan, C. (2000). Better assessment in mathematics education? A social perspective. In J. Boaler (Ed.), Multiple perspectives on mathematics teaching and learning (pp. 225–242). Westport: Ablex Publishing. Mullola, S., Ravaja, N., Lipsanen, J., Alatupa, S., Hintsanen, M., Jokela, M., et al. (in press). Gender differences in teachers' perceptions of students' temperament, educational competence and teachability. British Journal of Educational Psychology. Mullola, S., Ravaja, N., Lipsanen, J., Hirstiö-Snellman, P., Alatupa, S., & KeltikangasJärvinen, L. (2010). Teacher-perceived temperament and educational competence as predictors of school grades. Learning and Individual Differences, 20, 209–214. OECD (2009). Highlights from education at a glance 2008. 497 Oliver, P. H., Guerin, D. W., & Gottfried, A. W. (2007). Temperamental task orientation: Relation to high school and college educational accomplishments. Learning and Individual Differences, 17, 220–230. Pulliainen, H. (1994). Kykytestistö AVO-9. Helsinki: työministeriö ja Psykologien Kustannus Oy. Rudasill, K. M., Gallagher, K. C., & White, J. M. (2010). Temperamental attention and activity, classroom emotional support, and academic achievement in third grade. Journal of School Psychology, 48, 113–134. Ruus, V. -R., Veisson, M., Leino, M., Ots, L., Pallas, L., Sarv, E. -S., et al. (2007). Student's well-being, coping, academic success, and school climate. Social Behavior & Personality, 35, 919–936. Strelau, J. (1998). Temperament: A psychological perspective. New York: Plenum Press. Thomas, A., & Chess, S. (1977). Temperament and development. New York: Brunner/Mazel. Thomas, A., & Chess, S. (1980). The dynamics of psychological development. New York: Brunner/Mazel. Tompkins, T. L., Hockett, A. R., Abraibesh, N., & Witt, J. L. (2011). A closer look at corumination: Gender, coping, peer functioning and internalizing/externalizing problems. Journal of Adolescence, 34, 801–811. Windle, M. (1992). Revised Dimensions of Temperament Survey (DOTS-R): Simultaneous group confirmatory factor analysis for adolescent gender groups. Psychological Assessment, 4, 228–234. Windle, M., & Lerner, R. (1986). Reassessing the dimensions of temperamental individuality across the life span: The Revised Dimensions of Temperament Survey (DOTS-R). Journal of Adolescent Research, 1, 213–230. Zhou, Q., Main, A., & Wang, Y. (2010). The relations of temperamental effortful control and anger/frustration to Chinese children's academic achievement and social adjustment: A longitudinal study. Journal of Educational Psychology, 102, 180–196.