Quantcast

Impact of Attitudes of Peers on Language Achievement: Gender Differences

December 29, 2007

By Van De Gaer, Eva Pustjens, Heidi; Van Damme, Jan; De Munter, Agnes

ABSTRACT The authors examined whether gender differences in language achievement were related not only to gender differences in attitudes toward schooling but also to the attitudes toward schooling of peers (i.e., peers in classes and in schools). The authors used multilevel analysis on data compiled from a longitudinal research project in secondary education. The primary results revealed that all boys who had negative school-related attitudes were underachievers in language. Furthermore, boys were more influenced by the attitudes of their peers in classes than were girls. The attitudes of peers in schools did not affect the gender gap in language achievement. Keywords: achievement, gender differences, language, multilevel analysis, peer group

Over the last 2 decades, discourse on gender and schooling has shifted dramatically. Although in earlier literature girls’ achievement and subject choice were the main focus, concern has moved to the underachievement of boys (Whitelaw, Milosevic, & Daniels, 2000). The concern for boys’ underachievement originates from the observation of the closing gender gap in mathematics (and to a lesser extent in sciences) and the continuing excellence of girls in languages.

Cole (1997) used several national representative samples and found that girls have closed the mathematics and science gap over the last 30 years, but a fairly large gap still exists in writing skills. Sutherland (1999) reported a changing gender gap in countries such as Great Britain, Australia, New Zealand, the United States, Canada, the Netherlands, Japan, France, Germany, and Jamaica. In this study, we focused on gender differences in language achievement because the underachievement of boys is most prominent in languages.

Researchers have suggested several explanations for the gender gap in language achievement. Gender differences in verbal intelligence and learning styles (Halpern, 1997), interest for languages (Dwyer & Johnson, 1997; Head, 1999; Murphy & Elwood, 1998), motivation and attitude toward languages (Lamb, 1997; Walsh, Hickey, & Duffy, 1999), behavior (McIntyre & Tong, 1998), perception of the difficulty and usefulness of the subject (Eccles et al., 1983), and expectations (Beyer, 1998; Mittelberg & Lev-Ari, 1999) are associated with gender differences in language achievement. Researchers also have focused on masculine identities and the way in which they relate to the underachievement of boys. Salisbury and Rees (1999) described a culture of laddishness or a macho culture that can be characterized as a culture in which it is not “cool” for a male student to be perceived as too interested or involved in schoolwork (see also Jackson, 2002; Martino, 1999). Effortless achievement can be tolerated, but studying for school is for girls or “sissies” (Francis, 1999; Paetcher, 1998).

That boys have more negative attitudes toward schooling than do girls is well documented. In general, girls work harder, are better organized, spend more time on homework, are less distracted in the classroom, and break school rules less often than do boys (Arnot, David, & Weiner, 1999; Clark & Trafford, 1995; Davies & Brember, 2001). Although negative attitudes and a “lad culture” among boys are often associated with their underachievement, it has rarely been tested quantitatively (Renold, 2001; Whitelaw et al., 2000). We tested quantitatively whether boys’ negative attitudes toward schooling are associated with their underachievement in language.

However, not only attitudes toward schooling of individual boys and girls but also attitudes toward schooling of peers (in classes and schools) can influence achievement. In particular, researchers have related the antischool subculture among boys and peer pressure to boys’ underachievement. Qualitative and ethnographic researchers have revealed that it is harder to deviate from the group norm for boys than for girls and that boys are more concerned with their status in the male peer group than are girls in the female peer group (Warrington, Younger, & Williams, 2000; Younger & Warrington, 1996). Boys admitted in interviews that “having a laugh,”"playing up” the teacher, and other disruptive behavior played an important part in their social status in the male peer group (Francis, 1999). The possibility that an antischool culture may be prevalent among some girls cannot be ruled out but, in general, working hard seems more acceptable for girls than for boys.

It would be a generalization and a simplification to state that all boys (and not a single girl) underachieve and have antischool attitudes. Boys construct their masculinities at school in distinct ways (Connell, 2002). Jackson (2002), for instance, found that in Great Britain, mainly middleclass White boys construct their masculine identity in a different way than do the “failing” boys who are mainly White and working class, along with African Caribbean boys. According to Sewell (1998), that finding is an oversimplification (see Power, Edwards, Whitty, & Wigfall, 1998; Wright, Weekes, & McGlaughin, 2000).

The diversity and difference within and among groups of boys is seldom acknowledged or addressed (Gilbert & Gilbert, 1998). We considered diversity among boys by examining how boys with school- related attitudes ranging from very negative to very positive achieve in language compared with girls. We investigated whether boys are more influenced by the attitudes of their peers toward schooling than are girls. Therefore, we contributed to the understanding of boys’ underachievement by examining whether peer pressure is one of the main factors affecting boys’ achievement.

Purpose and Research Questions

Most research on peer groups and masculinities is on a small scale (e.g., one school) and qualitative (e.g., interviews with students and teachers). Although a few largescale studies on gender and achievement exist (e.g., Gorard, Rees, & Salisbury, 2001; Warrington et al., 2000), the relationship between boys’ underachievement and boys’ attitudes toward schooling is often assumed but rarely tested quantitatively (Renold, 2001; Whitelaw et al., 2000). We tested quantitatively whether there is an association between boys’ underachievement in language and their antischool- related attitudes. Furthermore, we argue that school-related attitudes not only at the individual level but also at the group level (i.e., peers in classes and in schools) influence the achievement of boys and girls. No data on the relationships of friends is available, but information is available on which students are (a) in the same class, (b) taught by the same teacher, and (c) in the same school.

Ryan (2001) defined a peer group as an individual’s small, relatively intimate group of peers who interact with each other on a regular basis. Although all small groups of students in classes and schools are not defined as peer groups, one cannot deny that these settings are social places in which students influence each other (Lee, 2000). Therefore, we examined the effect of mean school- related attitudes of classes and schools on the achievement of boys and girls, assuming that students in classes and schools influence each other as they do in peer groups.

In a first step, we tested Grade 8 students to determine whether boys are lower achievers in language than are girls and whether boys have less positive school-related attitudes than do girls. In a second step, we considered background characteristics of boys and girls (i.e., cognitive ability, socioeconomic background, language spoken at home, curriculum, age) and previous Dutch language and mathematics achievement. That allowed us to test whether language achievement of boys is lower than that of girls with similar background characteristics and previous achievement. In a third step, we examined whether the gender differences in language achievement are related to gender differences in attitudes toward schooling. If the language achievement of girls is higher only because girls have more positive attitudes toward schooling, then the gender gap would reduce to nonsignificance when controlling for the attitudes toward schooling of boys and girls. In a fourth step, we investigated whether boys are more influenced than are girls by the school-related attitudes of peers. We expected that boys are lower achievers in classes containing students who have negative attitudes toward schooling, whereas boys achieve higher in classes with students who have positive attitudes toward schooling.

We formed the following hypotheses:

1. Boys have less positive attitudes toward schooling and achieve less in language than do girls.

2. When gender differences in school-related attitudes are considered, the advantage of girls in language is reduced to nonsignificance.

3. The language achievement of boys is influenced more by the attitudes of classes and schools than by that of girls: Boys perform worse in classes with students displaying a low mean attitude toward schooling than in classes with students exhibiting a high mean attitude toward schooling. For girls, the mean attitudes of students have less impact on their language achievement than it does for boys. Method

We compiled data from the Longitudinaal Onderzoek Secundair Onderwijs [Longitudinal Research in Secondary Education; LOSO] project, in which a cohort of more than 6,000 students (aged 12-21 years) were studied during and after secondary education (Van Damme & Onghena, 2002). The main focus of the LOSO project was to describe and explain the achievement of a large sample of students in secondary education in Flanders, which constitutes the northern part of Belgium bordering the Netherlands to the north. In Flanders, inhabitants speak Dutch,1 whereas in the southern part of Belgium, inhabitants speak French, and in a very small eastern part of Belgium, they speak German. Each section of Belgium has its own education system. Because our study is limited to Flanders, we investigated gender differences in native language skills.

We began gathering data in the 1990-1991 school year. We selected the sample of schools according to school characteristics representative of Flanders, such as the size and type of schools, curriculum offered, and representation of Catholic and public schools. Within the schools, we selected all the students in Grade 7 (i.e., the 1st year or the start of secondary education) in September 1990 (i.e., a cohort of 6,411 students) to participate in the LOSO project. At the start of secondary education (i.e., beginning of Grade 7), we gathered information from parents in a questionnaire regarding (a) occupational status, (b) highest obtained education level of both parents, (c) parents’ cultural activities, and (d) language spoken at home. In addition, we tested students’ intelligence at the start of secondary education. We measured language and mathematics at the beginning and end of Grade 7 and at the end of Grade 8. Students in Grade 8 filled out a well- being questionnaire tapping school-related attitudes. We focused on explaining Dutch language achievement of boys and girls at the end of Grade 8; only students who did not repeat Grade 7 were considered. The data set consisted of 4,072 students (2,011 boys and 2,061 girls), 321 classes, 180 Dutch teachers, and 49 schools. In Grade 8, students could follow one of two types of curricula: A or B. The A curriculum was academically orientated, whereas the B curriculum was vocationally orientated and intended for students with learning difficulties. The A curriculum attracted not only higher achieving students but also more girls than did the B curriculum. In our data set, the majority (89%) of students followed the A curriculum (n = 3,641); in the B curriculum, 61.9% (n = 267) of the students were boys, whereas in the A curriculum, 47.9% (n = 1,744) were boys.

The dependent variable was the score on a Dutch achievement test (DUTCH2) administered at the end of Grade 8 (age 14). Because there are no national examinations in Belgium, the Dutch achievement test was constructed especially for the LOSO project. The test consisted of curriculum-relevant, multiple-choice items that included grammar, linguistic performance, reading comprehension, and spelling, which were approved by a board of inspectors and teachers assuring high content validity. We constructed different versions of the achievement test for the A and the B curriculula that we adapted to the curriculum. Both versions showed high internal consistency (A curriculum, Cronbach’s a = .90; B curriculum, Cronbach’s a = .82). Because of the large overlap in items between the two versions, we calibrated the scores on both versions and converted them into item response theory (IRT) scores using BIMAIN (Zimowski, Muraki, Mislevy, & Bock, 1994). We used the IRT scores as the dependent variable.

At the student level, our main variable of interest was gender, which we dummy coded with 0 for boys and 1 for girls. The other student-level variables could be divided into control variables and explanatory variables. The control variables were (a) initial cognitive ability, (b) socioeconomic status (SES) of the student’s family, (c) language spoken at home, (d) age at the start of secondary education, and (e) curriculum. Prior mathematics (MATH1) and language achievement (DUTCH1), which we measured at the end of Grade 7, were also controlled for previous achievement to adjust the language achievement at the end of Grade 8. The initial cognitive ability was a combination of scores on an intelligence test (i.e., the Getlov intelligence test; Lancksweerdt, 1990) and on a Dutch language test and a mathematics achievement test administered at the start of secondary education (beginning of Grade 7). The Getlov intelligence test has a high internal consistency (Cronbach’s a = .82) and a strong validity (see Lancksweerdt, 1990); the Dutch and mathematics achievement tests have high internal consistency (Cronbach’s a, .93 and .90, respectively). Because the three measures showed high intercorrelations, we combined them into one variable that indicated the initial cognitive ability of the students at the start of secondary education.

The SES of the family was a weighted composite score of indicators (education and the occupation level of both parents, monthly income, and cultural capital of the family). The language spoken at home was coded as a dummy variable (0 for families who spoke a language other than Dutch at home and 1 for families who spoke only Dutch at home); so was age at the start of secondary education (0 for started secondary education with delay, 1 for started secondary education without delay) and curriculum (0 for B curriculum and 1 for A curriculum). The explanatory variables were the following attitudes at the student level: (a) interest in learning tasks (e.g., “I enjoy doing most of the subjects in this school”), (b) relationship with teachers (e.g., “I think that most of the teachers are very helpful when I have problems with the school work”), (c) wellbeing at school (e.g., “I am glad to go to this school”), (d) attentiveness in the classroom (e.g., “I find it difficult to keep my mind on my work during a whole lesson”), (e) motivation toward learning tasks (e.g., “I work hard for all subjects to get good results”), (f) attitude toward homework (e.g., “When I have homework, I start as soon as possible”), and (g) social integration in the class (e.g., “I get along well with my classmates”). We based the seven variables on 52 items of a well- being questionnaire, which was an adaptation of the LOSO (Smith & Vorst, 1982), and supplemented it with new items (formulated with the help of a questionnaire by Janssen, 1982). There were betweeen 4 and 10 items per indicator. Cronbach’s as were between .82 and .89. (For more information on the indicators, see Van Damme, De Fraine, Van Landeghem, Opdenakker, & Onghena, 2002). Appendix A shows the item composition of each indicator.

At the class and school levels, we constructed aggregates of the seven student-level attitudes. We calculated the mean of the attitudes of all the students in a class or school (including students not belonging to our data set of 4,072 students) for each class and school. The prefixes CL- and SCH-indicate that a variable was aggregated at the class or school level, respectively. TH- indicates teacher level.

Aggregated measures or compositional variables are commonly used in multilevel analysis because they are important contextual variables and often have an effect on achievement (Snijders & Bosker, 1999; Teddlie & Reynolds, 2000). We used the group mean for a general idea of the mean attitudes of classes and schools, even though other aggregate measures could have been used. For instance, the standard deviation of attitudes of students within a class can be a significant influence on achievement. Another possibility is to use only the mean values of relevant peers in classes (such as friends) instead of the whole group (Campbell & Alexander, 1965). However, we did not have data on friendship relations or on the most relevant peers of the students.

Hauser (1970) suggested using direct measurements rather than aggregated measurements (see also Firebaugh, 1980; Glick, 1985; Manski, 1993). Hauser argued that contextual effects can be artificial consequences of group differences in composition on individual important variables not considered at the individual level. We included seven control variables at the individual level that were reliable and important predictors of achievement. In other words, we met Hauser’s objection by constructing a strong individual- level model. We also used aggregated variables because one of our primary aims was to discover whether attitudes at the group level had an effect on achievement. We preferred to have the same variables on the higher (group) levels as on the lowest (individual) level. If we had used direct measures of attitude toward schooling at the class and school levels, we would have had difficulty comparing the effects of the attitudes at student and group levels. Appendix B provides an overview of all variables considered in the study; for each of the seven schoolrelated attitude items, we provide one item that is a brief description of each indicator.

The LOSO data were hierarchically structured: Students were nested within classes (student level), classes were nested within teachers (class level), teachers were nested within schools (teacher level), and schools were at the highest level (school level). Because students were grouped within classes, teachers, and schools, the observations were not independent. To account for the dependency between observations, we used multilevel analysis (Hox, 2002; Kreft & de Leeuw, 1998). Moreover, multilevel analysis allowed us to study individual and group effects on gender differences simultaneously. We centered all continuous variables around the grand mean. We used MLwiN software to estimate the models (Rasbash et al., 2000). We analyzed a four-level hierarchical model but did not use any variables at the teacher level (except TH-Gender). We included teacher level in the model because previous research showed that ignoring an important intermediate level in a multilevel model influenced not only the variances of the adjacent (i.e., class and school) levels but also the standard errors of the variance parameters and the regression coefficients of the fixed variables at the adjacent (i.e., class- and school-level variables) levels (Opdenakker & Van Damme, 2000; Van den Noortgate, Opdenakker, & Onghena, 2006). Our investigation of the null model (i.e., a multilevel model without any predictors) revealed that we needed to consider teacher level. A comparison between two null models that did and did not include teacher level showed that including the teacher level improved the fit of the model significantly, ???(1, N = 4,072) = 6.89, p

To test the first hypothesis (i.e., gender differences in language achievement and attitudes), we did not use t tests because that would have ignored the nonindependence in our data. Instead, we constructed eight separate multilevel models with the language achievement at Grade 8 and each of the seven student-level attitudes as the dependent variable. For each dependent variable, we estimated a model with four levels (student, class, teacher, and school) with gender, CL-Gender, TH-Gender, and SCH-Gender as the independent variables. CL-Gender, TH-Gender, and SCH-Gender were the aggregated variables of the student variable gender at the class, teacher, and school level, respectively. CL-Gender represented the proportion of girls in a class, TH-Gender represented the proportion of girls in the group taught by a teacher, and SCH-Gender represented the proportion of girls at a school.

We included aggregated variables of student gender together with the student-level variable to make a distinction between within- and between-group regression. Snijders and Bosker (1999) stated that the coefficients for the within- and between-group regressions can be completely different. “For theoretically important variables in multilevel studies, it is the rule rather than the exception that within-group regression coefficients differ from betweengroup regression coefficients” (p. 56). Without considering the aggregated variables, the regression coefficient for gender would be an average of the within- and between-group relations; therefore, the Snijders and Bosker concluded that it is necessary to consider within- and between-group regressions jointly.

We tested the second hypothesis (i.e., whether gender gap in language achievement is related to gender differences in attitudes toward schooling) by the following model:

DUTCH2^sub ijkl^ = (dependent variable) beta^sub 0jkl^ + (intercept) beta^sub 1^cogn^sub ijkl^ + … + beta7DUTCH1^sub ijkl^ + (covariates) beta^sub 8l^gender^sub ijkl^ + beta^sub 9^attitude^sub ijkl^ + (main effects) beta^sub 9^(gender^sub ijkl^*attitude^sub ijkl^) + (interaction effect) r^sub ijkl^ (student-level residual) (1)

beta^sub 0^jkl = gamma^sub 00^kl + u^sub 0^jkl (random intercept at class level) (2)

gamma^sub 00l^kl = gamma^sub 0001^ + ?^sub 00kl^ (random intercept at teacher level) (3)

gamma^sub 0001^ = gamma^sub 0000^ + w^sub 000l^ (random intercept at school level) (4)

When (2-4) are replaced in (1) and rearranging the terms,

DUTCH2^sub lijkl^ = gamma^sub 0000^ + beta^sub 1^cogn^sub ijkl^ + … + beta^sub 7^DUTCH1^sub ijkl^ + beta^sub 8^gender^sub ijkl^ + beta^sub 9^attitude^sub ijkl^ + beta^sub 10^(gender^sub ijkl^*attitude^sub ijkl^) + r^sub ijkl^ + u^sub 0jkl^ + v^sub 00kl^ + w^sub 000l^ (5)

Dutch achievement at the end of the second year of secondary education (DUTCH2) is our dependent variable. The four indexes, ijkl, refer to student i in class j taught by teacher k in school j. Our independent variables are the seven control variables, gender, and an attitudinal variable. (The model is called a random intercept model because only the intercept varies at the different levels.) The indexes jkl of the intercept beta0jkl indicate that the intercept varies between classes, teachers, and schools. In Equation 2, we split beta0jkl into a class mean intercept, gamma00kl, and a deviation of each class from this class mean intercept, namely, u0jkl. The class mean intercept, gamma00kl, which varies between teachers, could be split into a teacher mean intercept, gamma000l, and a teacher-specific deviation score, v00kl (see Equation 3). Finally, the teacher mean intercept is divided into a school mean intercept, gamma0000, and a school-specific deviation score, w000l (see Equation 4). The other coefficients in the model that we estimated were the coefficients of the (a) covariates, (b) main effect of gender and an attitude, and (c) interaction between gender and the attitude. The student-level residual is represented by rijkl. Equation 5 shows that the only difference between our model and the classical multiple-regression model is that random effects are estimated at the class (u0jkl), teacher (v00kl), and school levels (w000l).

In a third hypothesis, we tested whether the gender gap varied between classes and schools and whether this variation could be explained by class- and school-level characteristics. We limited the statistical model formulation to the random slope at the class level and to the introduction of class-level variables. The statistical formulation of the random slope model at the school-level was analogous.

DUTCH2^sub ijkl^ = beta^sub 0jkl^ + beta^sub 1^cognijkl + … + beta^sub 7^DUTCH1ijkl+ beta^sub 8jkl^gender^sub ijkl^ + beta^sub 9^attitude^sub ijkl^ + beta^sub 10^(genderijkl*attitudeijkl) + r^sub ijkl^ (6)

beta^sub 0jkl^ = gamma^sub 00kl^ + gamma^sub 01kl^CL- attitude^sub jkl^ + u^sub 0^jkl (random intercept at class level) (7)

beta^sub 8jkl^ = gamma^sub 8000^ + gamma8^sub 100^CL- attitude^sub jkl^ + u^sub 8jkl^ (random slope at class level) (8)

gamma^sub 00kl^ = gamma^sub 000l^ + v^sub 00kl^ (random intercept at teacher level) (9)

gamma^sub 000l^ = gamma^sub 0000^ + w^sub 000l^ (random intercept at school level) (10)

Equation 6 is similar to Equation 1. Only the coefficient of gender (beta8jkl) has an extra index, jkl, indicating that the coefficient of gender varies between classes, which are nested within teachers and schools. In Equation 8, the coefficient is split into a class mean intercept, gamma8000, and a deviation score of the class mean, u8jkl. Also new is the class-level variable (i.e., the class mean attitude) that is a predictor of the random intercept and the random slope at the class level (cf. see Equations 7 and 8). When Equations 7-10 are submitted into Equation 6, the model becomes as follows:

DUTCH2^sub ijkl^ = (gamma^sub 0000^ + w^sub 000l^ + v^sub 00kl^ + gamma^sub 01kl^CL-attitude^sub jkl^ + u^sub 0jkl^) + beta^sub 1^cogn^sub ijkl^ + . . . +beta^sub 7^DUTCH1^sub ijkl^ + (gamma^sub 8000^ +gamma ^sub 800^CL-attitude^sub jkl^ + u^sub 8jkl^)gender^sub ijkl^ +beta attitude^sub ijkl^ + beta^sub 10^(gender^sub ijkl^*attitude^sub ijkl^) + r^sub ijkl^ (11)

= gamma^sub 0000^ + (intercept) beta^sub 1^cogn^sub ijkl^ + . . . + beta^sub 7^DUTCH1^sub ijkl^ + (covariates) gamma^sub 8000^gender^sub ijkl^ + beta^sub 9^attitude^sub ijkl^ + (main effects at student level) beta^sub 10^(gender^sub ijkl^*attitude^sub ijkl^) + (interaction effect at student level) gamma^sub 0jkl^CL- attitude^sub jkl^ + (main effect at class level) gamma^sub 8100^(gender^sub ijkl^*CL-attitude^sub jkl^) + (cross-level interaction) r^sub ijkl^ + u^sub 0jkl^ + v^sub 00kl^ + w^sub 000l^ +u8jklgender^sub ijkl^ (random effects) (12)

Equation 12 shows that the main effect of the class mean attitude and the interaction effect between gender and the class mean attitude is called a cross-level interaction because it constitutes variables of different levels on language achievement. To rule out the possibility that the cross-level interactions are reflections of the interactions between gender and student-level attitude, we included the interaction at the student level in the model. In the random part, the term u8jkl, genderijkl reflects a random interaction between classes and gender.

Results

Descriptive Statistics

Table 1 shows descriptive information on student, class, and school-level variables. For each continuous variable, Table 1 provides the number of observations, means, and standard deviations (prior to grand mean centering); for categorical variables, the number of observations and their percentages are shown. Regarding student-level variables, Table 1 shows the descriptive information for the total group as well as for boys and girls separately to obtain more insight into the differences between boys and girls. Table 1 indicates that (a) girls (n = 2,061, 50.6%) and boys (n = 2,011, 49.4%) were represented equally in the data set; (b) the majority of the students spoke Dutch at home (n = 3,735, 91.7%), compared with only 8.3% of the students who spoke another language at home; (c) the majority of students started secondary education on the first day of school (90.3%); (d) the majority of the students chose the academically orientated curriculum (89.4%); and (e) students had a mean score of 0.49 on the Dutch achievement test at the end of Grade 8 (DUTCH2).

Concerning gender differences, we concluded that there were no differences in language spoken at home between boys and girls, more boys than girls delayed secondary education, and fewer boys than girls chose the A curriculum. We found no gender differences in students’ SES and mathematics achievement at the end of Grade 7 (MATH1), but boys showed higher mean cognitive abilities than did girls. Conversely, girls showed higher mean scores on language achievement at the end of Grades 7 and 8 and on all school-related attitude items except interest in learning tasks. Although comparison of the mean scores between boys and girls may provide an idea about gender differences in language achievement and school- related attitudes, the significance of these gender differences needs to be tested. Therefore, we used multilevel analyses to investigate whether gender differences in language achievement and school-related attitudes are statistically significant. Hypothesis 1: Gender Differences in Language Achievement and Attitudes

As expected, we found that girls scored significantly higher in language than did boys at the end of Grade 8, ??(1, N = 4,072) = 78.78, p

We found no significant differences between boys and girls in the same class regarding their interest in learning tasks, ??(1, N = 4,072) = 0.48, p = .49, or in their attentiveness in the classroom, ??(1, N = 4,072) = 1.08, p = .30. In addition, the within- and between-group regression coefficients of gender were indistinguishable, indicating that the gender composition at the class, teacher, and school levels did not add an extra effect above the effect of gender on language achievement and on student attitudes. To conclude, we found evidence that girls had more positive school-related attitudes than did boys.

Hypothesis 2: Can Gender Differences in Attitudes Account for Gender Differences in Language Achievement?

In Hypothesis 2, we tested whether gender differences in language achievement were related to gender differences in attitudes toward schooling by comparing the gender coefficient between two models. In Model A, we tested the effect of the student-level variable gender on language achievement after considering background characteristics and previous achievement. In Model B, we also considered the five attitudinal variables in which boys and girls differed (see previous section). Our results showed that the coefficient for gender decreased from 0.309 (Model A) to 0.278 (Model B) but remained significant, ??(1, N = 4,072) = 112.51, p

Studying the specific unique mediating effects using the Sobel t test (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002) revealed that only social integration in the classroom, t(4,072) = 2.19, p

Interactions Between Gender and Attitudes Toward Schooling on Language Achievement

We subsequently examined interactions between gender and each of the seven student-level attitudes on language achievement. Results showed seven significant interactions, all displaying the same pattern: relationships between interest in learning tasks, ??(1, N = 4,072) = 14.04, p

Despite the narrowed gender gap, girls continued to outperform boys. The only exception was among boys and girls who were attentive in the classroom. Figure 1 shows the interaction effect between gender and attentiveness in the classroom on language achievement. Girls and boys with a mean level of attentiveness in the classroom (0 on the horizontal axis in Figure 1 represents mean level of attentiveness in the classroom) differed significantly in language achievement. That advantage for girls became larger the less boys and girls were attentive in the classroom. However, boys with the highest level of attention in the classroom did not perform significantly worse in language than did girls with the same high level of attention in the classroom, M gender difference = 0.049, SE = 0.048. In other words, girls did not significantly achieve higher in language than did boys when boys and girls were attentive in the classroom.

In sum, findings suggest that attentiveness in the classroom is a particularly important predictor of boys’ language achievement. Conversely, for girls, attentiveness in the classroom was not associated with language achievement. When boys were attentive in the classroom, they achieved equally well as the girls did in language. Only a small percentage of the boys were attentive in the class. These results offer important implications for educational practice (see Discussion).

Hypothesis 3: Are Boys More Influenced by the Attitudes of Classes and Schools Than Are Girls?

In a final hypothesis, we tested whether the gender gap varied between classes and schools and whether this variation can be explained by class- and school-level characteristics. Table 3 (Model 1) shows that the gender gap varied between classes: Slope for gender between classes was significant, ??(2, N = 4,072) = 68.28, p

Table 3 (Model 2) shows that the gender gap also varied significantly between schools, with a mean gender gap of .32 and a variance of .02, ??(2, N = 4,072) = 16.04, p

Next, we investigated whether variation of the gender gap at the class and school levels could be explained by class- and school- level characteristics. As expected, results showed that the mean attitudes of classes had a stronger impact on the language achievement of boys than of girls. We found two significant cross- level interactions: The difference between girls and boys was smaller in classes with a good mean relationship with teachers, ??(1, N = 4,072) = 4.32, p

Discussion and Conclusions

We tested whether gender differences in achievement were related to gender differences in attitudes toward schooling and to the attitudes toward schooling of peers in classes and schools. That association is often assumed in the boys’ debate, which is discussed frequently in the research literature. This term refers to the discourse on gender differences having moved from a concern for girls to a concern for underachievement of boys. It has rarely been tested quantitatively. We investigated the association in a sample of more than 4,000 students by using reliable measures of a wide range of school-related attitudes and intake characteristics. Our study, however, suffers from the limitation that the school-related attitudes are general instead of subject-specific. The latter type of attitudinal measures may constitute more powerful predictors. Previous research showed that boys viewed languages, and reading in particular, as girls’ activities and that these attitudes were related to language achievement (Head, 1999). Another limitation is our lack of data on friendship relations of boys and girls. Data on friendships might have resulted in a more detailed and differentiated picture of the influence of group-level characteristics on the language achievement of boys and girls. For instance, boys and even girls who belonged to peer groups with negative attitudes toward schooling were underachievers.

Overall, our hypotheses were confirmed. The gender gap in language achievement was associated with schoolrelated attitudes in three ways. First, fewer boys than girls had positive attitudes toward schooling. Once we considered differences in school-related attitudes, gender gap in language achievement became smaller but remained significant, indicating that the effect of gender on language achievement was mediated only partly by gender differences in attitudes. Although we controlled for differences in intelligence, we did not control for other factors that may have accounted for the remaining gender difference in language achievement, such as differences in expectations, perceived difficulty and the usefulness of the subject, lack of male role models, or the way that boys and girls respond to distinct teaching and learning styles.

Second, the gender gap in language achievement became smaller the more boys and girls had positive attitudes toward schooling. Specifically, boys who were attentive in the classroom did not achieve lower scores in language than did girls. Third, the results showed that the attitudes of peers in classes had a stronger effect on the language achievement of boys than of girls. We found no gender differences in language achievement in classes containing students who had a good relationship with teachers, whereas girls outperformed boys in classes containing students who had a poor relationship with teachers. Thus, our results demonstrate that a number of factors are important for understanding why boys and girls achieve differently in language tests. We found that primary factors influencing boys’ language achievement were not only less positive school-related attitudes of boys but also the stronger impact of school-related attitudes of peers in classes. The finding that boys did not achieve lower scores in language when they were attentive in the classroom has important implications for educational practice. Intervention programs that focus on remediating lower achievement of boys should emphasize their attitudes toward schooling.

Our results indicate that when boys can be motivated to be attentive during class, put more effort into schoolwork, and complete their homework, their achievements should improve. A possible strategy to enhance boys’ attitudes toward schooling could be to adapt teaching methods and adjust topics of reading materials during language classes to subjects such as sports, adventure, detectives, action, or humorous stories. In a traditional learning situation, learning proceeds primarily through verbal communication and passively by listening and observing the teacher. However, both of those characteristics are disadvantageous for boys, who prefer an active approach toward learning, such as experiential learning and the use of computers (Weaver-Hightower, 2003).

Although the fact that boys have less positive attitudes toward schooling than do girls has been documented in Great Britain, Australia, New Zealand, the United States, Canada, the Netherlands, Japan, France, Germany, and Jamaica (Sutherland, 1999), few researchers (e.g., Ryan, 2001) have concentrated on the effect of school-related attitudes of peers in classes and in schools on gender differences in achievement. Most researchers have concentrated on the effect of cognitive characteristics of peers such as intelligence (i.e., ability grouping) or the effect of SES of peers on achievement (e.g., Burns & Mason, 2002; Caldas & Bankston, 1997). However, the question of whether noncognitive characteristics of peers such as their school-related attitudes affect achievement has been less studied. Van Houtte (2004) did not investigate the effect of schoolrelated attitudes of peers, but she examined the school and study culture that make up school-related attitudes of peers, and she arrived at the same conclusions as we did. However, her study was also implemented in a Flemish context. Therefore, researchers need to investigate to what extent our findings can be generalized across countries.

Another result was that not only individual students’ attitudes but also attitudes of classmates had an impact on achievement. Enhancing school-related attitudes and creating a class and school culture in which students can achieve without fear of ridicule is important. Class groups are social environments in which students spend much time with one another and influence one another, as we observed in regular peer groups. Although we did not examine peer- group influences, our results regarding class-group influences strongly resemble those from studies that related boys’ poor achievements to peer-group attitudes (e.g., Younger & Warrington, 1996). Thus, teachers and school board members should monitor class- group attitudes and the social and learning climate that characterizes class groups, along with the attitudes of individual students. According to our results, establishing a positive relationship with teachers is important. Furrer and Skinner (2003) concluded that a good relationship with teachers improves identification with school and motivation and learning in the classroom. Therefore, teachers should be aware of the importance of maintaining good relationship with students.

We also found that the mean school-related attitudes of classes but not of schools influenced gender differences in mathematics achievement. The finding that interactions occurred with class- group characteristics but not with school characteristics can be explained by the fact that class groups are social groups that are closer to the individual student, as compared with entire school groups. Hence, characteristics of class groups can play a larger role than school characteristics in shaping achievement. Researchers have found that class-level variables have a greater impact on children’s development than do school-level variables (Teddlie & Reynolds, 2000). Stringfield (1994) argued that factors that are most proximal to the students, such as class and teacher, are likely to be more important than are the more distant factors, such as school characteristics.

A possible explanation for boys being more influenced by school- related attitudes of peers than are girls is that boys are more concerned with their status in the male peer group than are girls in the female peer group (Warrington et al., 2000). In classes with students having a good relationship with teachers, it may be acceptable for boys to show an interest in learning tasks, and it may not be considered a bad thing to achieve well. Likewise, in classes with students having a bad relationship with teachers, boys might be laughed at when they study and achieve well. Teacher gender may also play a role. For instance, boys might have better relationships with male than with female teachers. Researchers need to clarify the role of teacher gender to explain positive and negative effects of classes. An alternative explanation of the larger contextual effects for boys is the differential sensitivity hypothesis (Dar & Resh, 1994). This hypothesis states that low- achieving or underprivileged students, or both, are more school (context) dependent than are high-achieving and privileged students. “Effective or ineffective schools are especially effective or ineffective for these students” (Scheerens & Bosker, 1997, p. 96). Because boys are low achievers in language, compared with girls, this may explain why boys were more influenced by context than were girls. Another factor that researchers need to address is that more boys than girls are represented in lower ability courses that usually include students with less positive school-related attitudes (Boaler, Wiliam, & Brown, 2000). We found that 62% of all students in the lower (B) curriculum were boys and that students in classes from the lower curriculum had less positive school-related attitudes toward schooling than did students in the higher tracks. Because more boys than girls were in the lower curriculum and students in the lower curriculum might have been more influenced by peers who had less positive school-related attitudes, the negative influence of classes with negative school-related attitudes on the language achievement of boys may be a reflection of boys’ achieving less than did girls in lower curriculum classes. Although lower curriculum classes had less positive relationships with teachers and were less attentive in the classroom than were classes from the higher curriculum, a large overlap remained. For instance, not all classes consisting of students having a poor relationship with teachers and being inattentive were classes from the lower B curriculum classes; A curriculum classes also had students with negative school-related attitudes. We found the same pattern of interactions within the two curricula. We controlled for the curriculum as well as for cognitive ability of students and still found that boys were more influenced by their peers in classes than were girls.

The results of more elaborated analyses, which we did not report here, showed that interactions between gender and school-related attitudes on language achievement were similar in higher and lower curricula. Although we do not exclude the possibility that curriculum may play a role in the explanation of gender differences in language achievement, results suggest that boys were more influenced by the attitudes of peers in the higher as well as in the lower curriculum classes.

Our findings underscore the importance of considering various factors for understanding differences between boys and girls in language achievement. The findings clearly advise against general statements such as “Boys perform worse than girls” and illustrate that more nuance is needed when discussing gender differences in language achievement. Instead, our findings emphasize that not all boys achieve lower than do girls in language; we found no gender differences in language achievement when boys were attentive in class and when they were in classes containing students who had a good relationship with teachers.

NOTES

1. The official language in Flanders is Dutch. The Belgian variant of Dutch is called Flemish. The difference between Flemish and Dutch that is spoken in the Netherlands can be compared with the difference between American and British English.

2. We calculated the variance at the class level for girls by estimating a new multilevel model in which the dummy variable for gender was recoded (0 = girls, 1 = boys).

3. Although we have single-sex classes in our data set, the covariance at the class level between intercept and gender slope applies only to mixed classes.

REFERENCES

Arnot, M., David, M., & Weiner, G. (1999). Closing the gender gap: Postwar education and social change. Cambridge, England: Polity Press.

Beyer, S. (1998). Gender differences in self-perception and negative recall biases. Sex Roles, 38, 103-133.

Boaler, J., Wiliam, D., & Brown, M. (2000). Students’ experiences of ability grouping-disaffection, polarization and the construction of failure. British Educational Research Journal, 26, 631-648.

Burns, R. B., & Mason, D. A. (2002). Class composition and student achievement in elementary schools. American Educational Research Journal, 39, 207-233.

Caldas, S., & Bankston, C. (1997). Effect of school population socioeconomic status on individual academic achievement. The Journal of Educational Research, 90, 269-277.

Campbell, E. Q., & Alexander, C. (1965). Structural effects and interpersonal relationships. American Journal of Sociology, 71, 284- 289.

Clark, A., & Trafford, J. (1995). Boys into modern languages: An investigation of the discrepancy in attitudes and performance between boys and girls in modern languages. Gender & Education, 7, 315-326.

Cole, N. S. (1997). The ETS gender study: How females and males perform in educational settings. Princeton, NJ: Educational Testing Service.

Connell, R. W. (2002). Gender. Cambridge, England: Polity Press.

Dar, Y., & Resh, N. (1994). Separating and mixing students for learning: Concepts and research. Pedagogisch Tijdschrift, 19, 109- 126.

Davies, J., & Brember, I. (2001). The closing gender gap in attitudes between boys and girls: A 5-year longitudinal study. Educational Psychology, 21, 103-114.

Dwyer, C. A., & Johnson, L. M. (1997). Grades, accomplishments, and correlates. In W. Willingham & N. Cole (Eds.), Gender and fair assessment (pp. 127-156). Hillsdale, NJ: Erlbaum.

Eccles, J. E., Adler, T. F., Futterman, R., Goff, S. B., Kaczala, C. M., Meece, J. I., et al. (1983). Expectancies, values, and academic behaviors. In J. T. Spence (Ed.), Achievement and achievement motives. Psychological and sociological approaches (pp. 75-146). San Francisco: Freeman.

Firebaugh, G. (1980). Groups as contexts and frog ponds. New Directions of Methodology of Social and Behavioral Science, 6, 43- 52.

Francis, B. (1999). Lads, lasses and (new) labour: 14-16-year- old students’ responses to the “laddish behaviour and boys’ underachievement” debate. British Journal of Sociology of Education, 20, 355-373

Furrer, C., & Skinner, E. (2003). Sense of relatedness as a factor in children’s academic engagement and performance. Journal of Educational Psychology, 95, 148-162.

Gilbert, R., & Gilbert, P. (1998). Masculinity goes to school. London: Routledge.

Glick, W. H. (1985). Conceptualizing and measuring organizational and psychological climate: Pitfalls in multilevel research. Academy of Management Review, 10, 601-616.

Gorard, S., Rees, G., & Salisbury, J. (2001). Investigating the patterns of differential attainment of boys and girls at school. British Educational Research Journal, 27, 125-139.

Halpern, D. F. (1997). Sex differences in intelligence: Implications for education. American Psychologist, 52, 1091-1102.

Hauser, R. M. (1970). Context and consex: A cautionary tale. American Journal of Sociology, 75, 645-654.

Head, J. (1999). Understanding the boys: Issues of behaviour and achievement. London: Falmer Press.

Hox, J. J. (2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Erlbaum.

Jackson, C. (2002). Laddishness as a self-worth protection strategy. Gender & Education, 14, 37-51.

Janssen, P. (1982). Vragenlijst studiebeleving [Study experience questionnaire]. Leuven, Belgium: Leuven, Afdeling Psychodiagnostiek en Psychologische Begeleiding, Centrum voor Schoolpsychologie.

Kreft, I., & de Leeuw, J. (1998). Introducing multilevel modeling. London: Sage.

Lamb, S. (1997). Gender differences in mathematics participation: An Australian perspective. Educational Studies, 23, 105-126.

Lancksweerdt, P. (1990). GETLOV (Gemeenschappelijke Testbatterij Lager Onderwijs. Onderzoeksverslag). [Getlov (Common Test Battery Primary Education). Research report.] Deinze, Belgium: P.M.S. 1.

Lee, V. (2000). Using hierarchical linear modeling to study social contexts: The case for school effects. Educational Psychologist, 35, 125-141.

MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83- 104.

Manski, C. F. (1993). Identification problems in the social sciences. Sociological Methodology, 23, 1-56.

Martino, W. (1999). “Cool boys,” party animals,”"squids” and “poofers.” Interrogating the dynamics and politics of adolescent masculinities in school. British Journal of Sociology of Education, 20, 239-263.

McIntyre, T., & Tong, V. (1998). Where the boys are: Do cross- gender misunderstandings of language use and behavior patterns contribute to the overrepresentation of males in programs for students with emotional and behavioral disorders? Education & Treatment of Children, 21, 321-334.

Mittelberg, D., & Lev-Ari, L. (1999). Confidence in mathematics and its consequences: Gender differences among Israeli Jewish and Arab youth. Gender & Education, 11, 75-93.

Murphy, P., & Elwood, J. (1998). Gendered learning outside and inside the school: Influences on achievement. In D. Epstein, J. Elwood, V. Hey & J. Maw (Eds.), Failing boys: Issues in gender and achievement (pp. 162-181). Buckingham, Philadelphia: Open University Press.

Opdenakker, M.-C., & Van Damme, J. (2000). The importance of identifying levels in multilevel analysis: An illustration of the effects of ignoring the top or intermediate levels in school effectiveness research. School Effectiveness and School Improvement, 11, 103-130.

Paetcher, C. (1998). Educating the other. Gender, power and schooling. London: Falmer Press.

Power, S., Edwards, A., Whitty, G., & Wigfall, V. (1998). Schoolboys and schoolwork: Gender identification and academic achievement. International Journal of Inclusive Education, 2, 135- 153.

Rasbash, J., Browne, W., Goldstein, H., Yang, M., Plewis, I., Healy, M., et al. (2000). A user’s guide to MlwiN. London: Institute of Education.

Renold, E. (2001). Learning the “hard” way: Boys, hegemonic masculinity and the negotiation of learner identities in the primary school. British Journal of Sociology of Education, 22, 369-385. Ryan, A. M. (2001). The peer group as a context for the development of young adolescent motivation and achievement. Child Development, 72, 1135-1150.

Salisbury, J., & Rees, G. (1999). Accounting for the different attainment of boys and girls at school. School Leadership and Management, 19, 403-427.

Scheerens, J., & Bosker, R. (1997). The foundations of educational effectiveness. Oxford: Pergamon.

Sewell, T. (1998). Loose canons: Exploding the myth of the “black macho” lad. In D. Epstein, J. Elwood, V. Hey, & J. Maw (Eds.), Failing boys? Issues in gender and achievement (pp. 111-127). Buckingham, England: Open University Press..

Smith, J., & Vorst, H. (1982). Schoolvragenlijst voortgezet onderwijs [School questionnaire secondary education]. Nijmegen, The Netherlands: Berkhout Nijmegen.

Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and advanced modeling. London: Sage.

Stringfield, S. (1994). The analysis of large data bases in school effectiveness research. In D. Reynolds, B. P. M. Creemers, P. S., Nesselroldt, E. C. Schaffer, S. Stringfield, & C. Teddlie (Eds.), Advances in school effectiveness research and practice (pp. 55-72). London: Pergamon Press.

Sutherland, M. B. (1999). Gender equity in success at school. International Review of Education, 45, 431-443.

Teddlie, C., & Reynolds, D. (2000). International handbook of school effectiveness research. London: Falmer Press.

Van Damme, J., De Fraine, B., Van Landeghem, G., Opdenakker, M.- C., & Onghena, P. (2002). A new study on educational differences in secondary schools in Flanders: An introduction. School Effectiveness and School Improvement, 13, 383-397.

Van Damme, J., & Onghena, P. (Eds.). (2002). Educational effectiveness in secondary schools in Flanders [Special issue]. School Effectiveness and School Improvement, 13(4).

Van den Noortgate, W., Opdenakker, M.-C., & Onghena, P. (2006). The effects of ignoring a level in multilevel analysis. School Effectiveness and School Improvement, 16, 281-303.

Van Houtte, M. (2004). Why boys achieve less at school than girls: The difference between boys’ and girls’ academic culture. Educational Studies, 30, 159-173.

Walsh, M., Hickey, C., & Duffy, J. (1999). Influence of item content and stereotype situation on gender differences in mathematical problem solving. Sex Roles, 41, 219-240.

Warrington, M., Younger, M., & Williams, J. (2000). Students attitudes, image and the gender gap. British Educational Research Journal, 26, 393-407.

Weaver-Hightower, M. B. (2003). Crossing the divide: Bridging the disjunctures between theoretically oriented and practice-oriented literature about masculinity and boys at school. Gender & Education, 15, 408-423.

Whitelaw, S., Milosevic, L., & Daniels, S. (2000). Gender, behaviour and achievement: A preliminary study of student perceptions and attitudes. Gender & Education, 12, 87-113.

Wright, C., Weekes, D., & McGlaughin, A. (2000). “Race,” class and gender in exclusion from school. London: Falmer Press.

Younger, M., & Warrington, M. (1996). Differential achievement of girls and boys at GCSE: Some observations. Perspective of one school. British Journal of Sociology of Education, 17, 299-314.

Zimowski, M. F., Muraki, E., Mislevy, R. J., & Bock, R. D. (1994). BIMAINTM2: Multiple-group IRT analysis and test maintenance for binary outcomes. Chicago: Scientific Software International.

EVA VAN DE GAER

HEIDI PUSTJENS

JAN VAN DAMME

Centre for Educational Effectiveness and Evaluation, K.U.Leuven, Belgium

AGNES DE MUNTER

Research Centre Women and Education, K.U.Leuven, Belgium

Address correspondence to Eva Van de gaer, K.U.Leuven, CO&E, Dekenstraat 2, Postbus 3773, B-3000 Leuven, Belgium. (E-mail: Eva.Vandegaer@ped.kuleuven.be)

Copyright (c) 2007 Heldref Publications

EVA VAN DE GAER is a postdoctoral research fellow at the Centre for Educational Effectiveness and Evaluation, K.U.Leuven, Belgium. Her current work focuses on educational effectiveness, equal educational opportunities, and value-added models. HEIDI PUSTJENS is a doctoral researcher also at the Centre for Educational Effectiveness and Evaluation. Her interests include educational effectiveness research and multilevel analysis. JAN VAN DAMME is professor at the Department of Educational Sciences of the K.U. Leuven, Belgium, and head of the Centre for Educational Effectiveness and Evaluation. His major research interests are educational effectiveness and evaluation and the longitudinal study of the educational career of students. AGNES DE MUNTER is professor, Department of Educational Sciences of the K.U.Leuven, Centre for Methodology of Educational Research, and head of the Research Centre for Woman and Education. Her major research interests are women’s studies and methods of education research.

(ProQuest: Appendix omitted.)

Copyright Heldref Publications Nov/Dec 2007

(c) 2007 Journal of Educational Research, The. Provided by ProQuest Information and Learning. All rights Reserved.




comments powered by Disqus