The Big-Fish-Little-Pond Effect
The BFLPE model was specifically designed to understand the development of academic self-concept in school settings (Marsh et al., 1985). The BFLPE describes that equally able students have a lower academic self-concept when attending a school where the average ability level of other students is relatively high than when attending a school where the average ability level is relatively low. It is this negative effect of school-average achievement on individuals' self-concept that characterizes the BFLPE. According to Marsh and colleagues (2008), the minimal conditions needed to test the BFLPE are (1) a multilevel design with many schools and a substantial sample of students from each school; (2) an objective measure of achievement for each individual student that is directly comparable across different schools and an appropriate measure of academic self-concept; (3) tests of the effects of school-average achievement on academic self-concept after controlling for the effects of individual student achievement.
Social comparison processes through external frames of reference are assumed to underlie the BFLPE. The presupposition is that the same objective academic performance can lead to different academic self-concepts for different individuals depending on the external frame of reference used to judge their academic performance with (Marsh et al., 2008). When an average student in a high-ability school performs academically below the average achievement level of that school and he compares his academic ability with other students in that school (i.e., his frame of reference), this comparison will result in a lower academic self-concept. However, when the same student uses a low-ability school as a frame of reference, the student may well perform above the average of that school, resulting in a higher academic self-concept. According to Marsh and Craven (2001), in making these social comparisons, students will use the frame of reference most available to them to evaluate their ability, namely other students in their school. Mussweiler, Rüter, and Epstude (2004) assume that social comparisons influence self-evaluations in multiple ways. For example, Marsh and Craven (2001) found that a high-ability school has a counter-balancing positive effect on an individual students' academic self-concept in addition to the negative BFLPE. This positive effect is also called a reflected glory effect: One feels proud to belong to a high-achieving school and this glory of the school is then reflected on the individual's academic self-concept. In short, sometimes self-evaluations are assimilated toward a given standard (positive association). At other times, they are contrasted away from the standard (negative association). Assuming those two effects to operate in real life, Marsh and Craven (2001) postulate that the BFLPE comprises the net effect of a stronger negative contrast effect and a weaker positive assimilation effect. This implicates that the negative effect of school-average achievement on academic self-concept, often found in BFLPE research, may be the result of two underlying social comparison processes (i.e., contrast and assimilation) operating together, but with a different magnitude.
Research support for the BFLPE
The BFLPE model has a solid empirical underpinning as shown in numerous studies with different analytical approaches bearing empirical evidence over the years (Marsh & Craven, 2001; Marsh et al., 2008; Trautwein, Ludtke, Marsh, Koller, & Baumert, 2006). Additionally, the BFLPE seems a robust and general effect, being present for students of different ages, with various ability levels, from different types of education, etcetera (e.g., Marsh & Craven, 2001; Preckel, Zeidner, Goetz, & Schleyer, 2008).
Marsh and Hau (2003) analyzed a large multinational database of 103,558 15-year old students from 26 countries and found strong support for the BFLPE and its cross-cultural generalizability. Their study was based on the Program of Student Assessment (PISA) database compiled by the Organisation for Economic Cooperation and Development. Seaton, Marsh, and Craven (2009) also tested the cross-cultural generalizability of the BFLPE in an even larger sample of 41 culturally and economically diverse countries. In support of the BFLPE, the effect of school-average achievement on academic self-concept was negative for the total sample (mean effect size = -.49), negative for each of the 41 countries separately and statistically significant from zero in 38 countries. This large, culturally diverse sample of countries has the added advantage to differentiate between collectivistic and individualistic cultures and economically developing and developed nations. The BFLPE was found to be evident in all four. These findings demonstrate that cultural differences such as collectivism versus individualism do not influence the size nor the direction of the BFLPE. The BFLPE also appears to be persistent. Extended longitudinal studies show that the BFLPE becomes stronger the longer students attend selective schools and is maintained even two and four years after graduation from high school (Marsh et al., 2007; Marsh et al., 2008). Thus, the BFLPE findings appear to be robust, generalizable over a wide variety of cultures and samples, different individual student and contextual level characteristics, settings, countries and long-term follow-ups.
In addition to the abovementioned importance of promoting a positive academic self-concept and given the empirical evidence regarding the BFLPE, numerous practical educational policy implications can be derived from it, such as the way in which schools are organized (Wouters et al., 2009). Specifically, it has implications for grouping practices in secondary education and with regard to academically gifted students and students with special educational needs, but also in regular primary and higher education (Marsh et al., 2008). Thus, the BFLPE is not only theoretically important, its practical implications are worth mentioning as well.
Understudied topics
The BFLPE model states that in making comparisons about their academic abilities, students will use the frame of reference most available to them to compare their ability with, namely other students in their school (Marsh, 1991). As a result, past BFLPE research has mainly focused on the school as being an important frame of reference and has operationalized the school-average ability level as the standard of comparison. Nevertheless, students may use other frames as a reference, such as classmates and friends. Marsh and colleagues (2008) state that it is easy to generalize the BFLPE to classes and friends instead of schools, but this generalization remains understudied. Moreover, the BFLPE model makes no predictions about the relative importance of these different frames and how they will jointly affect the academic self-concept.
Because friendships become more complex, psychologically based and important for development during middle childhood (Markovitz, Benenson, & Dolensky, 2001; Rose, & Asher, 1999), it is to be expected that friendships form an important frame for social comparison and that they are likely to influence a students' academic self-concept. Moreover, there has been abundant evidence in the field of social psychology that friends are an important frame for social comparison and that friends influence a students' self-esteem (Acitelli & Steinberg, 2010; Altermatt, Broady, & Bellgard, 2007; Payne, 2005). For example, Lubbers, Kuyper, and Van Der Werf (2009) found that students in Grade 7 choose friends far more often to compare their grades with than would be expected based on chance. Mussweiler and Rüter (2003) also demonstrated that during self-evaluation, students activate knowledge about their best friend to compare themselves with.
Because school-aged children in middle childhood spend most of their time in class, it is to be expected that their classmates also form an important frame for social comparison and that they are likely to influence a students' academic self-concept.
Despite the significance of friends and classes as potential reference groups, in addition to schools, the relative importance of these multiple and simultaneously occurring reference groups has remained understudied in BFLP research. As an exception, Wu, Chang, and Huang (2006) did examine the importance of friends as a frame of reference in addition to class and school. They concluded that the academic achievement of friends has the strongest negative effect on students' self-esteem in comparison with class- and school-average achievement: When controlled for the effects of friends' average academic achievement on students' self-esteem, class-average and school-average academic achievement were no longer associated with self-esteem. However, Wu and his colleagues studied global self-esteem as the dependent variable instead of academic self-concept. Consequently, these findings may not hold for academic self-concept (Marsh et al., 2008).
Concerning the relative importance of the three different frames of reference (school, class, friends), we would hypothesize that classes are more plausible frames of reference for students to compare themselves with than schools, given the daily physical and social proximity of classmates, the vividness and the accessibility of information regarding academic achievement (e.g., tests, exams, …) (Dai & Rinn, 2008). This assumption is in line with what Zell and Alicke (2010) call "the local dominance effect": this effect suggests that social comparison with a few local vivid individuals such as classmates have a greater influence on one's self-assessments than comparisons with larger general groups such as schools. We would also hypothesize that friends are a more plausible frame of reference for students to compare themselves with than schools, given the abovementioned proximity, vividness and local dominance effect. Thus, when controlled for the effects of friends-average academic achievement and class-average achievement on students' academic-self-concept, we would expect school-average academic achievement to be no longer associated with academic self-concept.
Concerning the comparison between friends and classes when examining the relative importance of the three different frames of reference, we did not find any existing theoretical background to make a reasoned hypothesis about their relative importance. Thus, the investigation of the simultaneous effect of these two frames of reference would be explorative. However, we could say that friend groups consist of fewer and more local individuals than class groups and therefore we might expect friends to be a more important frame for social comparison. On the other hand, since students can deliberately choose their friends, the assumed positive assimilation-effect underlying the BFLPE might be stronger than the negative contrast-effect, resulting in a relatively small net BFLPE effect compared to the BFLPE with classmates as a frame of reference, with which students are forced to compare themselves.
The present study
The aim of the present study is to analyze the effect on academic self-concept of different frames of reference, including the effect of school-average achievement (as done in traditional BFLPE studies) as well as the effect of class-average and friends-average achievement. Also, we will consider the relative importance of those reference groups (school, class and friends) for students' academic self-concept. Given the theoretical background, we hypothesized that (H1) the effect of individual student achievement on academic self-concept would be positive, that (H2) the effect of school-average achievement on academic self-concept would be negative, that (H3) the effect of class-average achievement on academic self-concept would be negative and that (H4) the effect of friends-average achievement on academic self-concept would be negative as well. Additionally, we hypothesized that the negative effect of school-average achievement on academic self-concept would become non significant when including class-average achievement and friends-average achievement and we explored the relative importance of friends and classes as frames of reference (H5).
METHOD
Participants and procedure
Our study uses data derived from the longitudinal SiBO-project (Schoolloopbanen in het Basis Onderwijs; School trajectories in elementary education; Maes, Ghesquière, Onghena, & Van Damme, 2002). This is a large-scale study in Flanders - the Dutch-speaking part of Belgium - that followed about 6000 five to six year-old pupils in roughly 200 Flemish primary schools throughout primary education (Buyse, 2009). These longitudinal data were initially collected to gain insight into the factors influencing individual differences in educational development and were collected by researchers under the authority of the Flemish Government. In the current study we only used a subsample of the full longitudinal sample. More specifically we analyzed the data collected from children in 6th grade (N = 2987; Mage = 11,99 ; 1499 boys and 1486 girls) recruited from a representative sample of 174 classes in 112 schools. In Grade 6, additional variables and related instruments (e.g., academic self-concept measure, sociometric assessment) were included with the specific aim to test and refine the BFLPE model (in the context of the doctoral study of Sofie Wouters).
Instruments and measures
Information was obtained regarding children's sex, their age, their class, and their family SES (Reynders, Nicaise, & Van Damme, 2005).
In order to measure children's academic self-concept we used a subscale from the adjusted Dutch translation of Marsh's (1992) Self Description Questionnaire I (SDQ-I; Simons, & Fisette, 2001). The SDQ-I is considered to be one of the best self-concept instruments available and captures multiple dimensions of self-concept (Byrne, 1996). Research supports the factorial structure, reliability, and criterion-related validity of the SDQ-I (Byrne, 1996). The academic self-concept subscale used in the present study contains 5 items and was scored a 5-point Likert-type rating scale ranging from 1 ("strongly disagree") to 5 ("strongly agree"). A sample item is "I am good at all school subjects". The Cronbach's ï¡ of the academic self-concept subscale in Grade 6 was .81.
To measure students' individual academic achievement, standardized achievement tests in math and languages were used consisting of multiple-choice items. To measure language achievement, two subtests were used. The first subtest was an adapted version of a reading comprehension test from a standardized test system, which is often used to monitor student progress in Flanders and the Netherlands, developed by the Dutch Central Institute for Test Development (CITO). In particular, a shortened test developed specifically for Flanders (Staphorsius & Krom, 1998) was administered. The second subtest was an adjusted standardized reading test from the CITO system that measures students' technical reading abilities (Rymenans, 2002). Additionally, mathematics achievement was measured through a standardized math test developed by Paul Dudal. The Cronbach's ï¡ of this test was .88 (Cortois, Van Droogenbroeck, Verachtert, & Van Damme, (2011). The individual student achievement scores were calculated by averaging the students' scores on these three tests. There were two versions of the academic achievement tests. An easier and a more difficult version. To make the results comparable across pupils, an equation strategy based on the Item Response Theory (IRT) was used (Verachtert, Van Damme, Onghena, & Ghesquiere, 2009). In addition, we also created class-average and school-average academic achievement variables by averaging these individual academic achievement scores for each class/school.
Sociometric nominations were used to identify children's friends (Orobio de Castro, Goossens, & Olthof, 2007). More specifically, all students were given a list with their classmates' names and they were asked to circle the names of their friends. Based on children's answers on these questions, we identified friends groups using MAKEDYAD (Thissen, & Bendermacher, 2002). These groups comprised all friends a respondent had nominated, and thus included reciprocal friends (both students nominate each other as friends) as well as non-mutual friends (referentie naar Gest?). We then created a friends-average academic achievement variable by averaging the individual academic achievement scores for each friends group.
Data analyses
In the present study, individual student academic self-concept is the outcome variable, whereas predictor variables are sex, individual student achievement, friends-average achievement, class-average achievement and school-average achievement. All measures of academic self-concept and individual student achievement were standardized (M = 0, SD = 1). The intercorrelations of these variables are listed in Table 1. We used sex as a control variable because there is evidence that boys have a higher academic self-concept than girls (Hergovich, Sirsch, & Felinger, 2004).
We treated our data as having a hierarchically nested structure with students nested in classes and schools and we used multilevel analyses using the program Statistical Analyses Software (SAS). More specifically, we used a three level model to define the fixed and random effects regarding the variables listed above (level 1: student, level 2: class, level 3: school).
First, we split up the variance in academic self-concept per level in order to calculate the intraclass correlation (ICC). Although more than 98% of variance in academic self-concept was explained on the student level, we found that there were small but significant differences between schools and classes regarding academic self-concept (ICCschool < .05, p < .01; ICCclass < .05, p < .01). Therefore, we used multilevel methods to analyze our data.
The fixed effects included the main effects of sex, individual student achievement, friends-average achievement, class-average achievement, school-average achievement and the interaction effects of sex with friends-, class- and school-average achievement and of individual student achievement with friends-, class- and school-average achievement (see Table 2).
The random effects (see Table 2) included intercept terms which represent how much variance at each of the three levels (school, class and student) remains unexplained. To determine the variation of the effect of individual student achievement on academic self-concept from school to school, we randomized the main effect of individual student achievement at the school level. To determine the variation of the effect of friends-average and class-average achievement on academic self-concept from school to school, we also let the effects of friends-average and class-average achievement to be random at the school level. To determine the variation of the effect of individual student achievement on academic self-concept from class to class, we randomized the main effect of individual student achievement at the class level. To determine the variation of the effect of friends-average achievement on academic self-concept from class to class, we also let the effects of friends-average achievement to be random at the class level.
Results
In our preliminary analyses we found class-average achievement and school-average achievement to correlate strongly (r = .88) (see Table 1). This is understandable, considering 49.1 percent of the schools in our data only consisted of one class.
Consistent with the BFLPE model and our prior predictions, we found that (1) boys had a significantly higher academic self-concept than girls, (2) the effect of individual student achievement on academic self-concept was significantly positive, (3) the separate effects of school-average achievement, class-average achievement, and friends-average achievement on academic self-concept were significantly negative when controlling for the effect of individual achievement and sex (see Table 2). We then analyzed the relative effect of friends-average achievement, class-average achievement and school-average achievement by including all three predictors together in the equation (see Table 2). Consistent with our prior predictions we found that school-average achievement on academic self-concept became non significant. Surprisingly however, we also found that the negative effects of class-average achievement and friends-average achievement on academic self-concept became non significant. We did find a significant cross level interaction effect between individual student academic achievement and class-average achievement (see Appendix A). More specifically, we found that for higher achieving students who score one standard deviation (sd) higher than the group mean of individual academic achievement there was a stronger BFLPE: comparison with higher/lower achieving classmates lead to a decrease/increase of 0.50 sd in academic self-concept. For lower achieving students (scoring 1 sd below the group mean of individual achievement), however, there was a weaker BFLPE: comparison with higher/lower achieving classmates lead to a decrease/increase of only 0.14 sd in academic self-concept.
The random effects representing school-to-school and class-to-class variation in the effects of individual student achievement, friends-average achievement and class-average achievement in Model 2, 3 and 4 were small and not statistically significant. In Model 1 however, we found significant school-to-school and class-to-class variation in the effect of friends-average achievement on academic self-concept.
Because of the high correlation between class-average achievement and school-average achievement we repeated our analyses without school-average achievement as a predictor. Table 3 represents our final and less restricted two-level model in which school-average achievement was dropped. In Model 3 we analyzed the relative effect of friends-average achievement and class-average achievement and this time we did find a significant negative effect of class-average achievement on academic self-concept. The negative effect of friends-average achievement on academic self-concept became non significant, when including class-average achievement. We found one marginally significant interaction effect between friends-average achievement and individual student achievement (see Appendix B). More specifically, we found that for higher achieving students who score one standard deviation (sd) higher than the group mean of individual academic achievement there was a weaker BFLPE: comparison with higher/lower achieving friends lead to a decrease/increase of 0.01 sd in academic self-concept. For lower achieving students (scoring 1 sd below the group mean of individual achievement), however, there was a stronger BFLPE: comparison with higher/lower achieving friends lead to a decrease/increase of 0.13 sd in academic self-concept. We can conclude that when we do not control for school-average achievement, lower achieving students are not only influenced by their class-average achievement, but also to some extent by their friends' average achievement. For higher achieving students, friends do not seem an important frame of reference to compare their academic achievement.
The random effects representing class-to-class variation in the effects of individual student achievement and friends-average achievement were small and not statistically significant.
Table 1
Intercorrelations of all Variables
Variable
1
2
3
4
5
6
1. Sex
--
2. Individual achievement
.03a
--
3. Friends-average achievement
.04**a
.23**
--
4. Class-average achievement
.02a
.40**
.70**
--
5. School-average achievement
.03a
.35**
.62**
.88**
--
6. Academic self-concept
.09**a
.50**
.03
.07**
.05**
--
N
2985
2987
2968
2987
2987
2978
Note. aPoint-biserial correlation coefficients for the discrete dichotomous variable sex (Field, 2005).
* p < .05. ** p < .01. *** p < .001.
Tabel 2
The Effects of Sex, Individual Student Achievement, Friends-Average Achievement, Class-Average Achievement, and School-Average Achievement on Academic Self-concept (N=2957)
Model 1
Model 2
Model 3
Model 4
B
SE
B
SE
B
SE
B
SE
Fixed Effects
Main Effects
Constant
-.27***
.05
-.25***
.05
-.26***
.05
-.26***
.05
Sex
.18***
.03
.17***
.03
.17***
.03
.17***
.03
Individual student achievement
.52***
.02
.55***
.02
.54***
.02
.55***
.02
Friends-average achievement
-.20*
.09
--
--
--
--
-.04
.12
Class-average achievement
--
--
-.54***
.13
--
--
-.32
.28
School-average achievement
--
--
--
--
-.58***
.15
-.25
.29
Interaction Effects
Friends-average achievement x sex
.07
.06
--
--
--
--
.06
.08
Friends-average achievement x individual student achievement
.03
.03
--
--
--
--
.07
.04
Class-average achievement x sex
--
--
.11
.08
--
--
-.05
.18
Class-average achievement x individual student achievement
--
--
.01
.04
--
--
-.18*
.08
School-average achievement x sex
--
--
--
--
.12
.09
.11
.19
School-average achievement x individual student achievement
--
--
--
--
.06
.05
.17
.09
Random Effects
Level 2 school intercept
.03***
.01
.02**
.01
.03***
.01
.02**
.01
Level 2 class-average achievement
--
--
.04
.03
--
--
.01
.04
Level 2 friends-average achievement
.03*
.02
--
--
--
--
.01
.02
Level 2 individual student achievement
.01
.00
.00
.00
.01
.00
.01
.00
Level 2 class intercept
.03***
.01
.03***
.01
.03***
.01
.02
.01
Level 2 friends-average achievement
.05**
.02
--
--
--
--
.03
.02
Level 2 individual average achievement
.01
.01
.01
.01
.01
.01
.01
.01
Level 1 student intercept
.70***
.02
.69***
.02
.70***
.02
.70***
.02
Deviance
7449.1
7464.1
7475.1
7403.1
Note. * p < .05. ** p < .01. *** p < .001. All measures of academic self-concept and individual student achievement were standardized (M = 0, SD = 1). Sex: Female=1, Male=2
Tabel 3
The Effects of Sex, Individual Student Achievement, Friends-Average Achievement, and Class-Average Achievement on Academic Self-concept (N=2957)
Model 1
Model 2
Model 3
B
SE
B
SE
B
SE
Fixed Effects
Main Effects
Constant
-.28***
.05
-.26***
.05
-.26***
.05
Sex
.18***
.03
.17***
.03
.17***
.03
Individual student achievement
.52***
.02
.55***
.02
.55***
.02
Friends-average achievement
-.21*
.09
--
--
-.06
.12
Class-average achievement
--
--
-.56***
.13
-.49**
.18
Interaction Effects
Friends-average achievement x sex
.03
.03
--
--
.07
.08
Friends-average achievement x individual student achievement
.08
.06
--
--
.07â€
.04
Class-average achievement x sex
--
--
.11
.08
.03
.11
Class-average achievement x individual student achievement
--
--
.02
.04
-.06
.05
Random Effects
Level 2 class intercept
.03***
.01
.02**
.01
.02**
.01
Level 2 friends-average achievement
.03
.02
--
--
.03
.02
Level 2 individual average achievement
.01
.01
.00
.01
.01
.00
Level 1 student intercept
.69***
.02
.69***
.02
.68***
.02
Deviance
7449.1
7464.1
7405.2
Note. * p < .05. ** p < .01. *** p < .001. †p < .10. All measures of academic self-concept and individual student achievement were standardized (M = 0, SD = 1). Sex: Female=1, Male=2
Discussion
The present study provides a further test of the BFLPE model by examining different frames of reference (i.e., schools, classes, and friends) and their relative importance. Previous BFLPE research has mainly focused on the average achievement level of the school and its impact on students' academic self-concept, whereas the everyday school context of students consists of multiple potential frames of reference (e.g., schoolmates, classmates, and friends). Thus, the first aim of this present study was to investigate the effects of school, class and friends average achievement on the individual students' academic self-concept controlling for students' sex and individual academic achievement. Our findings were consistent with our hypotheses based on BFLPE research. More specifically, we found a positive effect of individual student achievement on academic self-concept and negative effects of each reference frame separately (i.e., school-average achievement, class-average achievement, and friends-average achievement) when controlling for individual achievement and sex. Also, our results showed that girls had a lower academic self-concept than boys.
Research on the relative importance of these reference groups for students' academic self-concept formation remains rare in BFLPE studies and thus the second aim of our study was to fill this gap. We hypothesized that classes are more plausible frames of reference for students to compare themselves with than schools, given the daily physical and social proximity of classmates, the local vividness, the accessibility of information regarding academic achievement (e.g., tests, exams, …) and the fact that, because of this daily physical and social proximity, students are forced to compare themselves with their classmates (Dai & Rinn, 2008; Zell & Alicke, 2010). We also hypothesized that friends are a more plausible frame of reference to compare themselves with than schools, given the fact that during middle childhood, peers and friendships become a more important context for development and frame for social comparison (Acitelli, & Steinberg, 2010; Altermatt, Broady, & Bellgard, 2007; Markovitz, Benenson, & Dolensky, 2001; Payne, 2005; Rose, & Asher, 1999). Concerning the comparison between friends and classes as frames of reference, we did not find any existing theoretical background to make a reasoned hypothesis about their relative importance. Thus, the investigation of the simultaneous effect of these two frames of reference in the three-level model was merely explorative. Our findings were mostly inconsistent with our theoretical hypotheses. More specifically, we found that when examining the simultaneous effects of the three frames of reference, none of the effects of friends-average achievement, class-average achievement nor school-average achievement remained significant. We did find an effect of class-average achievement when we took into account the individual student academic achievement. There seemed to be a strong BFLPE of class-average achievement on academic self-concept for high achieving students and a weak BFLPE of class-average achievement on academic self-concept for low achieving students. One possible explanation for this interaction effect might be that higher achieving students are more concerned with the grades of their classmates or with social comparison regarding academic achievement in general. A possible explanation for not finding a main effect of class-average achievement in the three-level model, might be that class-average achievement and school-average achievement correlate strongly (r = .88). 49.1 percent of the schools in our data only consisted of one class. Further investigation of de relative effect of different frames of reference in a sample of schools with multiple classes, more often found in secondary schools than in elementary schools, may lead to different conclusions.
We decided to repeat our analyses without school-average achievement as predictor. This time, we did find a significant main effect of class-average achievement when examining the simultaneous effects of the two frames of reference. Friends-average achievement became non-significant in the equation. One possible explanation for not finding a main effect of friends-average achievement in the two-level and three-level model might be found in our sample of friends. We identified the friend groups by giving all students a list with their classmates' names and asking them to circle the names of their friends. Possibly, if we had not limited the possible friends to classmates, but opened it up to the whole school, we might have gotten more realistic friend-groups, thus resulting in different conclusions regarding the relative importance of the three reference groups. Further study regarding this topic is needed to fully understand the impact of friends on academic self-concept. Another possible explanation for the non-significance of friends-average achievement when including class-average achievement might be that since students can deliberately choose their friends, the assumed positive assimilation-effect underlying the BFLPE might be stronger than the negative contrast-effect, resulting in a relatively small net BFLPE effect. In contrast, students do not freely choose their classmates and forced to compare themselves with them. Consequently, the assumed negative contrast-effect underlying the BFLPE might be stronger than the positive assimilation-effect, explaining the relatively strong net BFLPE effect of class-average achievement. However, we did find one marginally significant interaction effect between friends-average achievement and individual student achievement. It seemed that lower achieving students were not only influenced by their class-average achievement, but also to some extent by their friends' average achievement. For higher achieving students, friends did not seem an important frame of reference to compare their academic achievement.
Our findings provide strong empirical support for the importance of classmates in the formation of an individual student's academic self-concept and are of high practical concern, given the predictive value of academic self-concept for a variety of educational outcomes (e.g., ref). High achieving students seem to use classmates more often as a frame of reference than low achieving students. The latter are also influenced by the grades of their friends. Further study to possible buffers for the effect of class-average and friends-average achievement is necessary. Those results can then be used as an impulse for interventions. Perhaps if our education would focus less on achievement, performance and competition and would instead focus more on effort, dedication, cooperative learning and peertutoring, students would be less likely to be negatively influenced by the grades of their peers. Also, further investigation regarding possible mediating factors on the individual, friends, class and school level is needed.
For example, concerning grouping practices, it raises questions about the way in which schools treat children with special learning needs. Children with learning disabilities and mild mental retardation are more frequently being placed in regular classrooms full time increasing heterogeneity in classes. Given the BFLPE model, it is to be expected that these students might develop a lower academic self-concept, which may have negative consequences for the academic achievement in the long run.
