Negative effects of school-average achievement on academic self-concept: a comparison of the big-fish-little-pond effect across Australian states and territories.

Marsh, Herbert W.

Attending academically selective schools is intended to have positive effects, but a growing body of theoretical and empirical research demonstrates that the effects are negative for academic self-concept. The big-fish-little-pond effect (BFLPE), based on social comparison theory, posits that equally able students will have lower academic self-concepts in academically selective schools than in nonselective schools. Here we test the validity of these predictions for representative samples of 15-year-olds from eight Australian states and territories by using multilevel modelling. Consistent with the BFLPE, the effects of individual student achievement were positive but the effects of school-average achievement were negative. Although there were small differences between states/territories in academic achievement, there were no significant differences between states/territories in the negative effects of school-average ability.

Keywords

ability grouping

academic achievement

school effectiveness

selective admission

self concept

social theories

**********

The importance of high self-concept as a desirable outcome variable is evident in diverse settings, including education, child development, mental and physical health, social services, organisations, industry, and sport (Branden, 1994). For example, educational policy statements throughout the world list self-concept enhancement as a central goal of education and an important vehicle for dealing with social inequities experienced by disadvantaged groups. This study evaluates the generalisability of theoretical predictions about the effects of school-average achievement based on the big-fish-little-pond effect (BFLPE), by using large representative samples of 15-year-olds from eight Australian states/territories.

Frame of reference effects in the formation of academic self-concept: The BFLPE

Self-concept cannot be adequately understood if the role of frames of reference is ignored. The same objective characteristics and accomplishments can lead to disparate self-concepts, depending on the frame of reference or standards of comparison that individuals use to evaluate themselves. In an educational context, Marsh (1984; Marsh & Parker, 1984) proposed a frame of reference model called the big-fish-little-pond effect (BFLPE) to encapsulate frame of reference effects posited in social comparison theory. In the BFLPE, academic self-concept is influenced substantially by the ability levels of other students in the immediate context, in addition to one's own ability and academic accomplishments.

The historical and theoretical background for this research program (see Marsh, 1974, 1984, 1991, 1993; Marsh & Parker, 1984) derives from research in psychophysical judgement (e.g. Helson, 1964; Marsh, 1974; Parducci, 1995), social judgement (e.g. Morse & Gergen, 1970; Sherif & Sherif, 1969; Upshaw, 1969), sociology (Alwin & Otto, 1977; Hyman, 1942; Meyer, 1970), social comparison theory (e.g. Festinger, 1954; Suls, 1977), and the theory of relative deprivation (Davis, 1966; Stouffer, Suchman, DeVinney, Star, & Williams, 1949). Marsh (1984) first proposed the theoretical model underlying the BFLPE, where he hypothesised that students compare their own academic ability with the academic abilities of their peers and use this social comparison impression as one basis for forming their own academic self-concept. A negative BFLPE (a contrast effect) occurs where equally able students have lower academic self-concepts when they compare themselves with more able students, and higher academic self-concepts when they compare themselves with less able students. For example, if average ability students attend a school where the average ability level of other students is high (hereafter referred to as a 'high-ability school') so that their academic abilities are below the average of other students in the school, it is predicted that this educational context will foster social comparison processes that will lead to academic self-concepts that are lower than if the same students attended an average-ability school. Conversely, if these students attend a low ability school, then their abilities will be above average in relation to other students in the school and social comparison processes will result in higher academic self-concepts. Hence academic self-concepts depend not only on a student's academic accomplishments but also the accomplishments of those in the school that the student attends. According to this model, academic self-concept will be affected positively by individual achievement (higher achieving children will have higher academic self-concepts). However, academic self-concept should be affected negatively by school-average achievement (equally able students will have lower academic self-concepts in a school where the average ability is high and higher academic self-concepts in a school where the average ability is low).The BFLPE is typically operationalised (see Marsh & Craven, 2002) as a path model (see Figure 1) in which the effects of individual student achievement on academic self-concept are predicted to be positive whereas the effects of school-average achievement are predicted to be negative. Empirical support for this negative effect of school-average achievement on academic self-concept (the BFLPE) comes from numerous studies based on a variety of experimental/analytical approaches (see recent review by Marsh & Craven, 2002).

[FIGURE 1 OMITTED]

Diener and Fujita (1997, p.350) reviewed BFLPE research in relation to the broader social comparison literature. In relation to the broad spectrum of social comparison research, they emphasised that Marsh's BFLPE provided the clearest support for predictions based on social comparison theory in an imposed social comparison paradigm. The reason for this, they surmised, was that the frame of reference, based on classmates within the same school, is more clearly defined in BFLPE research than in most other research settings. Clearly the importance of the school setting is that the relevance of the social comparisons in school settings is much more ecologically valid than manipulations in typical social psychology experiments involving introductory psychology students in contrived settings. Indeed, except for opting out altogether, it is difficult for students to avoid the relevance of achievement as a reference point within a school setting or the social comparisons provided by the academic accomplishments of their classmates. A major focus of BFLPE research has been on the substantively important and surprising implications of this research, which undermines the assumed advantages of attending academically selective schools at one end of the ability spectrum and mainstreaming academically disadvantaged students at the other end (see also Robinson, Zigler, & Gallagher, 2000).

Two theoretically important issues are the extent to which the BFLPEs are specific to academic self-concept and the extent to which the BFLPE varies across different individual student ability levels. Consistent with theoretical predictions and a growing emphasis on the multidimensionality of self-concept, the BFLPE is very specific to academic components of self-concept. Marsh and Parker (1984; Marsh, 1987) showed that there were large negative BFLPEs for academic self-concept, but little or no BFLPEs on general self-concept or self-esteem. Marsh, Chessor, Craven, and Roche (1995; see also Craven, Marsh, & Print, 2000) reported two studies of the effects of participation in gifted-and-talented programs on different components of self-concept over time and in relation to a matched comparison group. There was clear evidence for negative BFLPEs, in that academic self-concept of students in the gifted and talented programs declined over time and in relation to the comparison group. These BFLPEs were consistently large for academic components of self-concept, but were small and largely nonsignificant for four nonacademic self-concepts and for general self-esteem.

Whereas the BFLPE predicts declines for academic self-concept for students in academically selective settings, there is some theoretical and empirical disagreement about whether these effects vary according to the initial achievement levels of individual students within these settings. Coleman and Fults (1985), for example, predicted and found that students in the top half of academically selective classes experienced little or no decline in self-concepts. In contrast, Marsh (1984, 1987, 1991, 1993; Marsh et al., 1995; Marsh & Rowe, 1996) argued that attending selective schools should lead to reduced academic self-concepts for students of all achievement levels based on several different theoretical perspectives. For a large, nationally representative (US) database, Marsh and Rowe (1996) found that the BFLPE was clearly evident for students of all achievement levels and that the size of the BFLPE varied only slightly with individual student achievement. In two studies demonstrating BFLPEs in students attending gifted-and-talented programs, Marsh et al. (1995) found no significant interaction between the size of the BFLPE and the achievement level of individual students. However, Marsh, Koller, and Baumert (2001) found small interaction effects in their large German study based on three waves of data collected on three occasions in the first year after reunification of the East and West German school systems. Whereas the size of the BFLPE--the negative effect of class-average achievement--did not vary with individual student achievement at any of the three times considered separately or for changes between Time 1 and Time 2 self-concept, the size of this negative effect diminished slightly for the most able students over the three occasions.

An Australian perspective

Australia continues to experience a substantial growth in the numbers of both gifted-and-talented (GAT) classes and secondary selective schools. This growth reflects strong parental interest in, and political support for, special educational settings for academically able students. Several early studies were undertaken by the New South Wales (NSW) Government to evaluate special educational provisions for GAT students. Sampson (1969) matched students in regular classes (who had declined offers to participate in specialised GAT classes) with students who participated in GAT classes in Years 5 and 6 (the last two years of primary school), on the basis of information collected in Year 4. He found that the two groups did not differ significantly in subsequent Commonwealth Secondary Scholarship Examination scores, Higher School Certificate scores, or school persistence. However, regular-class students performed significantly better than GAT class students on the aggregate School Certificate in Year 10, although this difference occurred primarily with boys. In subsequent research, Sampson (1977) compared a random sample of 240 students from eight selective high schools and a comparison group of comprehensive high school students matched on the basis of gender, age, socioeconomic status, IQ, and prior achievement. He found that there were no statistically significant differences between selective and comprehensive high school students on subsequent School Certificate scores, Higher School Certificate scores, or school retention rates. This result was consistent across initial ability levels and for boys and girls. This research contributed to a ministerial report recommending that selective schools should be phased out, but this recommendation was not enacted due in part to parental pressure to maintain these schools. In both studies, Sampson emphasised that he was unable to consider affective variables (e.g. self-concept) that he suggested might be enhanced by attending selective schools.

Much of the BFLPE research, not surprisingly, is based on Australian settings (e.g. Craven, Marsh, & Print, 2000; Marsh, 1984; Marsh & Parker, 1984; Marsh et al., 1995). In a critique of this research, Gross (1997) argued against the validity of the BFLPE in Australian high school settings, which prompted a rejoinder by Marsh and Craven (1998). Gross (1992; see also Gross, 1993) argued:

 It might be anticipated that exceptionally gifted children who have
 been radically accelerated would score high on the index of academic
 self-esteem. By contrast, they display positive but modest scores,
 between the mean for their age groups and .7 of a standard deviation
 above ... Interestingly, it is the children who have not been
 radically accelerated whose academic self-esteem is unusually
 inflated. (p. 97)

However, Marsh and Craven countered that, although Gross argued that students in non-accelerated settings have 'inflated' academic self-concepts, her results support the BFLPE. Despite being four standard deviations above the mean IQ, radically accelerated children have 'radically deflated' academic self-concepts that are only slightly above average because they compare their academic skills with those of their older, more able classmates. In contrast, the non-accelerated gifted students have realistically high academic self-concepts because they compare their abilities with those students from a normal range of abilities. Thus radical acceleration is likely to produce substantial declines in academic self-concept that are consistent with the BFLPE. Marsh and Craven suggested that the implication of Gross's argument was that it is somehow bad for gifted children to have academic self-concepts commensurate with their high levels of academic achievement and good for them to experience a substantial decline in academic self-concept, but that Gross provided no evidence in support of this implication.

Gross (1997) subsequently evaluated shifts in academic and nonacademic self-concepts for students in their first year of selective high schools and comprehensive high schools. She implied that her results did not support the big-fish-little-pond effect (BFLPE). However, Marsh and Craven (1998) noted that a careful evaluation of her published data and statistical analyses shows that her results were clearly consistent with BFLPE predictions. In particular, Marsh and Craven emphasised that, consistent with BFLPE predictions, Gross's published results showed that: (a) students from academically selective schools experienced significant declines in academic self-concept over time; (b) these shifts were more negative for academic than for nonacademic self-concepts; and (c) the declines in selective schools were more negative than in comprehensive schools, particularly for academic selfconcept (which increased slightly in comprehensive schools). Hence Marsh and Craven (1998) concluded that a critical reanalysis of data and findings published by Gross provided clear support for the BFLPE.

This controversy about selective schools was reignited by the Vinson Report (Inquiry into the Provision of Public Education, 2000), the most far-reaching public inquiry into education in NSW in the last 50 years. Based in part on BFLPE research summarised in this article, the Vinson Report recommended a moratorium on new selective high schools and selectively cutting back on existing ones--recommendations that were not endorsed by the NSW Government. In an editorial comment published in The Sydney Morning Herald responding to this report and the ongoing controversy about selective schools, Marsh (2002) noted:

 It has been more than a quarter of century since the NSW Government
 conducted research into this critical educational issue. Why has
 there not been a systematic evaluation of the effects of selective
 schools on a whole range of outcomes, including achievement,
 academic self-concept, educational aspirations, university
 attendance, etc.? Where is the solid research upon which to base
 decisions about maintaining, increasing, or decreasing the numbers
 of students attending selective schools? Given the controversy
 surrounding the Vinson report, there is a clear challenge to the NSW
 Government to initiate such a research program. Or maybe we will
 just wait until the next time this issue erupts and again ask why
 educational decisions are based on political expediency instead of
 good research. (p. 11)

An international perspective

There exists a growing body of international support for the BFLPE (see Marsh & Craven, 2002) that substantiates the generalisability of the theoretical predications and empirical results. Jerusalem (1984) examined the self-concepts of West German students who moved from nonselective, heterogeneous primary schools to secondary schools where selection was on the basis of academic achievement. At the transition point, students selected to enter the high ability schools had substantially higher academic self-concepts than did those entering the low ability schools. However, by the end of the first year in the new schools, no differences in academic self-concepts for the two groups were present. Path analyses indicated that the direct influence of school type on academic self-concept was negative. The most able students in the low ability schools were less able, but had much higher academic self-concepts, than the least able children in the high ability schools.

In 1991, former East and West German students experienced a remarkable social experiment--the reunification of very different school systems after the fall of the Berlin Wall. Self-concepts were collected at the start, middle, and end of the first school year after reunification (Marsh, Koller & Baumart, 2001). East German students had not previously been grouped according to ability. For them, the BFLPE was initially small, then moderate, and then substantial by the end of the year. West German students had attended schools based on ability grouping for the two years prior to the reunification. For them, the BFLPE was substantial at all three times. A large East-West difference in the size of the BFLPE at the start of the year disappeared completely by the end of the year. The evolvement of the BFLPE in the East and West German settings supported the social comparison process posited to underlie the BFLPE and its cross-cultural generalisability.

In Hong Kong, schools are more highly segregated in relation to ability than anywhere else in the world. However, collectivist cultural values prevail that are posited to counter social comparison processes (compared with more individualistic values in most western countries). Marsh, Kong, and Hau (2000) followed a large, nationally representative sample of Grade 7 students through high school (7997 students, 44 high schools, 4 years) based on a Chinese translation of the Self Description Questionnaire (SDQII). Psychometric properties of responses to the instrument were similar to--or even stronger than--those based on the Australian research that was the basis of the SDQII instrument. Consistent with the BFLPE, school-average ability (based on measures collected in Grade 6, prior to the start of high school) had negative effects on academic self-concept during high school.

Zeidner and Schleyer (1999) tested the BFLPE in a large-scale study based on a nationally representative sample (N = 1020) of Israeli gifted students participating in either special homogenous classes for the gifted or mixed ability classes. Path analyses indicated that gifted students in mixed ability classes evidenced markedly higher academic self-concepts, lower anxiety and higher school grades than gifted students in specialised classes.

The focus of BFLPE research has been on the negative effects of ability segregation on academically gifted students, but the theoretical basis of the BFLPE also has important policy implications for the placement of learning disadvantaged students as well (see also Robinson, Zigler, & Gallagher, 2000). Tracey, Marsh, and Craven (2003) extended this research to focus on BFLPE predictions for special classes for academically disadvantaged students (see also Chapman, 1988; Marsh & Johnston, 1993). They studied 211 special education students in Grades 2-6 who had an intellectual disability (IQ of 56 to 75). These students were either in regular classroom settings or in full-time support units. Consistent with BFLPE predictions, students in special classes had significantly higher self-concepts for all three academic scales (Reading, Math, School) but did not differ significantly from other special education students in regular classes for Parents, Physical Ability, and Physical Appearance self-concepts. Somewhat unexpectedly, special education students in special classes had significantly higher Peer and General self-concepts.

Broader policy implications of the BFLPE

The results of the BFLPE are very important for understanding the formation of academic self-concept and testing frame-of-reference models. However, classroom teachers, policy makers and parents might ask 'So what?'. What are the consequences of attending high-ability schools on other academic outcomes and how are these related to academic self-concept? Educators and particularly parents often assume that there are academic benefits associated with attending higher-ability schools. After all, academic achievement, aspirations and subsequent attainment are typically higher in these schools. This naive analysis, however, fails to take into account the initially higher abilities and other pre-existing differences of students who attend academically selective high schools. A better test would be to compare academic outcomes after controlling for the pre-existing differences.

Marsh (1991) considered the influence of school-average ability on a much wider array of outcomes in a very large-scale longitudinal study--the High School and Beyond Study, in which 36 students each from 1000 high schools were surveyed in Year 10 (T1),Year 12 (T2), and again two years after graduation from high school (T3).The outcomes in this study included many of the most important outcomes of education. After controlling background and initial ability, the effects of school-average ability were negative for almost all of the Year 10,Year 12, and postsecondary outcomes: 15 of the 17 effects were significantly negative and 2 were non-significant. School-average ability most negatively affected academic self-concept (the BFLPE) and educational aspirations, but school-average ability also negatively affected general self-concept, selection of advanced coursework, school grades, standardised test scores, occupational aspirations and college attendance. Even after controlling for all Year 10 outcomes, school-average ability negatively affected many subsequent outcomes. This implies that school-average ability continues to affect negatively Year 12 and post-secondary outcomes beyond the negative effects experienced at Year 10. Consistent with the proposal that these negative effects were substantially mediated by academic self-concept, controling for the negative effects of school-average ability on academic self-concept substantially reduced the size of negative effects on other outcomes.

The present investigation

According to the BFLPE (see Figure 1), the predicted effect of a student's own individual achievement is positive (+ in Figure 1), whereas the effect of school-average achievement is negative (- in Figure 1). in the present investigation we test these predictions in a nationally representative sample of 15-year-old Australians and evaluate the generalisability of the effects across eight Australian states and territories. Based on theory and previous research, the following a priori hypotheses were developed:

1 (a) The effect of individual student achievement on academic self-concept is predicted to be positive. (b) Whereas there are likely to be some differences between states/territories, the direction of the effect is predicted to be consistently positive and substantial.

2 (a) The effect of school-average achievement on academic self-concept (the BFLPE) is predicted to be negative. (b) Whereas there are likely to be some differences between states/territories, the direction of the effect is predicted to be consistently negative.

3 The size of the BFLPE (the negative effect of school-average achievement) is predicted to be reasonably consistent across different levels of individual student achievement (i.e. the interaction between individual student achievement and school-average achievement is predicted to be small or non-significant). In particular, higher-achieving students are predicted to suffer BFLPEs as well as lower-achieving students.

Although not a major focus of the study, it is also an interesting research question to evaluate the size of differences in academic self-concept and achievement as a function of the state/territory and the school. In pursuing these hypotheses and research questions, multilevel modelling was used--a state-of-the-art statistical methodology specifically designed to evaluate multilevel data. This provided a partition of variance associated with different effects into components associated with the individual student (level 1) and the school (level 2). In particular, it provided appropriate tests of the extent to which the effects of individual student achievement and school-average achievement differ according to state/territory.

Methods

Data source and sample

The study was based on the Program of Student Assessment (PISA) database, compiled by the Organisation for Economic Cooperation and Development (OECD), that consisted of nationally representative responses by 15-year-olds collected in Australia and other countries in 2000 (see OECD, 2001a, 2001b, for a description of the database and variables). The PISA database was collected in response to the need for internationally comparable evidence of student performance and related competencies within a common framework that is internationally agreed upon. Selection of the measures was made on the basis of advice from substantive and statistical expert panels and results from extensive pilot studies. Substantial efforts and resources were devoted to achieving cultural and linguistic breadth in the assessment materials, stringent quality-assurance mechanisms were applied in the translation of materials into different languages, and data were collected under independently supervised test conditions. Paper-and-pencil assessments consisted of a combination of multiple-choice items and written responses. Whereas all students completed some reading assessment items (which were the focus of the 2000 data collection), only random samples of students completed mathematics or science assessments. In addition, countries were given the option of collecting materials on a Cross Curriculum Competencies questionnaire which included the academic self-concept items that are the focus of the present investigation; 26 of 32 countries in the PISA project--including Australia--administered this additional survey.

The present investigation is based on students who completed the academic self-concept items on the Cross Curriculum Competencies questionnaire and standardised academic achievement tests that were developed specifically for the PISA. The self-concept items were from the highly regarded Self Description Questionnaire II (Byrne, 1996; see also Marsh, 1991, 1993; Marsh, Plucker, & Stocking, 2001). However, because of limitations on the length of the total instrument, only three items were used to represent academic self-concept. Nevertheless the reliability for this scale was consistently reasonable across all 26 countries (mean coefficient alpha = .76; SD = .04) and did not differ significantly from the reliability of responses in Australia (.74).

Because of the nature of the PISA project, nine versions of the achievement tests were administered, which contained different combinations of verbal, mathematics, and science test items. Because of the focus of the PISA study, all students completed at least some verbal test items whereas only about half of the students completed mathematics and science tests. Using test-equating procedures based on item response theory, scores based on each version of the various achievement tests were put onto a comparable scale. For present purposes, a total achievement test score was obtained by taking an average score for all the achievement test scores available for each student. As recommended in the database documentation (OECD, 2001a, 2001b), analyses were conducted using sample weights to obtain unbiased estimates of population parameters. For purposes of the present investigation, the effective sample size for each Australian state/territory was set equal to the number of cases for that state/territory prior to weighting so that the weighted sample size was the same as the unweighted sample size (i.e. the average weight across all cases was 1.0).Analyses were based on responses from 4916 students from 223 schools (from a total sample of 5106 students from 231 schools) who had no missing data on variables considered here (4921 students completed the self-concept items) and attended high schools in which there were at least 10 students who had complete data. The number of students in each of the 223 schools varied from 11 to 40 (M = 22, SD = 3.83).

Statistical analysis

In the present investigation, consistent with the design of the PISA data, we consider a two-level multilevel model in which students (level 1) are nested within schools (level 2). In general, it is inappropriate to pool responses of individual students without regard to schools or countries unless it can be shown that schools and countries do not differ significantly from each other. If, for example, there were systematic differences between schools, then the typical single-level analyses that ignored this clustering of students into schools would probably be invalid (violating statistical assumptions in a way that increases the likelihood of finding a significant effect when there is none). Furthermore characteristics associated with individual students are likely to be confounded with those based on schools. From a practical perspective, a systematic multilevel approach allows researchers to pursue new questions about how effects vary from school to school, state/territory differences in the pattern of results, and the characteristics of schools that are associated with this variation. This is particularly important in studies such as the present investigation in which critical variables are associated with both the individual student level (academic self-concept and achievement) and the school level (school-average achievement). In the present investigation, for example, the use of a multilevel approach provides appropriate tests to determine the extent to which the observed pattern of relations between academic self-concept and achievement generalises across schools and across the different Australian states and territories. Hence the multilevel approach provides a much richer and more appropriate approach to testing support for the BFLPE than would be possible with traditional single-level approaches that ignore the fact that students are clustered within schools (for further discussion, see Goldstein, 1995; Goldstein et al., 1998; Raudenbush & Bryk, 2002). Of particular relevance in the present investigation is the effect of school-average ability. According to the BFLPE, the effect of school average ability on individual student academic self-concept should be negative. The present investigation is unique in being able to evaluate the extent to which this effect varies across eight Australian states and territories.

Based on a large, nationally representative sample of United States high school students, Marsh and Rowe (1996) found a nonlinear relation between academic self-concept and achievement. Whereas the relation was strictly monotonic, increments in ability had less effect on academic self-concept near the lower end of the ability range than near the top. They speculated, however, that this reflected idiosyncratic scaling issues in the normally distributed standardised test scores and negatively skewed self-concept ratings. In support of this suggestion, they found no nonlinearity in relations between school grades (which were also negatively skewed) and self-concept responses. Based on these results, we also included a nonlinear (quadratic) component of individual student achievement as a predictor of academic self-concept in the present investigation.

In testing the effect of states/territories, the set of eight states/territories was treated as a categorical predictor variable represented by a set of eight dummy variables (for further discussion, see Goldstein et al., 1998; see also Aiken & West, 1991). The dummy variables were scored 1 (from a particular state/territory) or 0 (not from the particular state/territory).Thus each student had a value of 0 for seven of the dummy variables and a value of 1 for one of the eight dummy variables. For purposes of statistical analyses one of the states/territories was 'left out' (i.e. the eight levels of the categorical variable were represented by seven degrees of freedom). New South Wales was selected as the left-out category because it had the largest sample size of the eight states/territories. Whereas the selection of the state/territory to be left out had no effect on the overall significance of the effect of states/territories (i.e. the set of dummy variables), effects associated with individual dummy variables must be interpreted in relation to the left-out state. Thus, for example, a statistically significant effect for a given state means that the effect in that state was significantly different from the effect in New South Wales. In different models, the main effects of states/territories and the interaction of this variable with individual student achievement and school-average student achievement was considered. Finally, in supplemental analyses, analyses were conducted separately for each of the eight states and territories

For purposes of the present investigation, individual student self-concept was the main outcome (dependent) variable, whereas predictor variables in different models included individual student achievement (linear and quadratic terms), school-average achievement, the interaction between school-average and individual achievement (to test whether the BFLPE varies with individual student ability), and the eight Australian states/territories and their interactions with other predictor variables. We began by standardising (z-scoring) all variables to have M = 0, SD = 1 across the entire sample (see Marsh & Rowe, 1996; see also Aiken & West, 1991; Raudenbush & Bryk, 2002). Product terms were used to test interaction and quadratic effects. In constructing these product variables, we used the product of individual (z-score) standardised variables (and the product terms were not re-standardised). Similarly, school-average achievement for each of the 223 schools was computed as the average score of the (standardised) achievement test scores for students attending those schools without restandarising these scores (i.e. individual and school-average achievement scores are in the same metric).

Results and discussion

I begin with an overview of the main results of the study as an advanced organiser, and then provide more detailed summaries of the statistical analyses. Across the eight Australian states/territories, there were no significant differences in academic self-concept, but small differences in academic achievement (Table 1). Achievement scores were significantly higher than the Australian average in the Australian Capital Territory and New South Wales and significantly lower in Victoria and Northern Territory, but no differences were greater than .3 SD. Results in Figure 2 indicated that there were substantial differences between individual schools in terms of school-average levels of achievement, but not for school-average levels of academic self-concept. Across all Australian states and territories, there was good support for the BFLPE (Table 1). In particular, the effect of individual achievement on academic self-concept was positive in each of the eight states/territories and the total sample (path coefficient = .40), whereas the effect of school-average achievement (the BFLPE) was negative across all states/territories and the total sample (path coefficient = -.35).

[FIGURE 2 OMITTED]

In order to evaluate the BFLPE, a nested series of a priori models was posited in which new predictor variables were added to each subsequent model (Table 2). Results summarised in Table 2 indicate the statistical significance of new predictor variables added to each model, and the effects of the predictor variables are summarised in Tables 3 and 4.

Effects of individual student achievement

In Model 1 (Table 2), individual student achievement scores were used to predict individual student academic self-concept. The effects of both linear and quadratic components of achievement were significantly positive. In support of Hypothesis la, the effect of academic achievement on academic self-concept was positive (linear component), but the quadratic component (see Figure 2) indicated that the effect of achievement on academic self-concept was stronger for higher achieving students and weaker for lower achieving students. Whereas there was substantial consistency in the effect of achievement across the 223 schools, the significant variance component associated with school intercepts (Table 3) indicated that there were school-to-school differences in academic self-concept after controlling for academic achievement (also see Figure 2).

Effects of school-average achievement

In Model 2 (Table 2), school-average achievement was added to the predictor variables. In support of Hypothesis 2a and the BFLPE, the effect of school-average achievement was negative and highly significant (Table 2). The residual variance component associated with schools decreased substantially (from .025 to .010) due to the introduction of school-average achievement, which indicates that many of the differences associated with schools in Model 2 were explained by the negative effect of school-average achievement in Model 3 (See Figure 3).

[FIGURE 3 OMITTED]

In Model 3 (Table 2), the interaction between school-average achievement and individual student achievement (linear and quadratic terms) was added to the predictor variables. Of particular importance, in support of Hypothesis 3, the individual achievement (linear) x school-average achievement interaction term was not statistically significant. Hence the size of the BFLPE did not vary linearly with levels of individual student achievement. There was, however, a significant individual achievement (quadratic) x school-average achievement effect. The nature of this interaction indicated that the negative effects of school-average ability were slightly smaller for students of intermediate ability levels. Given the large sample size, these results indicated that the negative effect of school-average ability was consistent across the range of student achievement levels. In particular, there was no evidence that the BFLPE had systematically larger effects on lower-achieving students than on higher-achieving students.

Effects of eight Australian states and territories

In Models 4-7 (Table 2), I explored the generality of the effects across the eight Australian states/territories. In each of these models, the set of eight states/ territories was represented by a set of seven dummy variables in which one of the set of predictor variables was 'left out' (i.e. the eight levels of the categorical variable states/territories are represented by seven degrees of freedom). New South Wales was selected as the left-out category because it had the largest sample size of the eight states/territories. Whereas the selection of the state/territory to be left out had no effect on the overall significance of the effect of states/territories (i.e. the set of dummy variables, Table 2), effects associated with individual dummy variables must be interpreted in relation to the left-out state (i.e. a statistically significant effect for a given state/territory means that the effect in that state was significantly different from the effect in New South Wales). Finally, in supplemental analyses, Model 3 was applied separately to each of the eight states and territories (and key effects from these separate analyses were summarised in Table 1). In Model 4, the main effect of the states/territories was not statistically significant, which indicates that there were no differences between states/territories in terms of academic self-concept.

In Model 5, the interaction between individual achievement (linear component) and states/territories (i.e. the set of seven dummy variables) was added to the predictor variables considered in Model 4. The overall effect of this interaction was statistically significant (Table 2), and the effects associated with three individual states/territories were statistically significant (Table 4). Relative to New South Wales (the left-out category), the positive effects of individual student achievement on academic self-concept were smaller (less positive) in the Australian Capital Territory, Western Australia, and Tasmania. Importantly, however, the results based on separate analyses for each of the eight states/territories (Table 1) indicated that the effects of individual students' achievement on academic self-concept were highly significant and positive in each of the states/territories. These results provide partial support for Hypothesis 1b.

In Model 6, the interaction between school-average achievement and states/territories (i.e. the set of seven dummy variables) was added to the predictor variables considered in Model 5. The overall test of significance for this interaction effect was not statistically significant (Table 2). Hence, in support of Hypothesis 2b, the size of the negative effect of school-average achievement (the BFLPE) did not differ significantly in the eight Australian states/territories.

Summary, conclusions and implications

International interest in the BFLPE and its relevance to educational settings throughout the world provides exciting new opportunities to evaluate the generalisability of theoretical predictions and empirical findings. The availability of the Australian PISA data provided a unique opportunity to evaluate rigorously the generalisability of the BFLPEs across eight Austrahan states/territories. Of central importance to the present investigation, the effects of school-average achievement were significantly negative and the sizes of these negative effects did not vary significantly across the eight states/territories. Given the size and representativeness of the PISA data and the rigorous statistical analyses, the results apparently provided strongest support for the generalisability of the BFLPE in an Australian setting.

School-average achievement effects represent complicated combinations of the social comparison processes emphasised here, the quality of the education (e.g. resources, curriculum, expertise of the teachers) and perhaps family background factors. Furthermore these effects are likely to be confounded (see Gamoran, Nystrand, Berends, & LePore, 1995). In BFLPE studies based on true longitudinal data in which the first wave was collected before the start of high school (e.g. Marsh, 1994; Marsh et al., 2000), it is possible to unconfound characteristics of the student cohort from subsequent school effects. These results show that it is the initial ability levels of students attending a school that drive the BFLPE rather than subsequent school effects (although Marsh and Craven, 2002, argue that school policies in academically selective schools can exacerbate the BFLPE). However, because features other than the achievement grouping per se are likely to have a positive effect on subsequent outcomes, potential biases are likely to be conservative in relation to the negative BFLPE. Therefore, because of the direction of this potential bias, interpretations of the negative effects of school-average achievement on academic self-concept are likely to underestimate the true BFLPE.

Another feature of the present investigation is that the achievement tests were specifically designed to reflect generic skills rather than the curriculum in any particular school or country. Furthermore students had no opportunity to study for any of these examinations, they were given no feedback on the examination, and they knew that it would have no consequences for them. Marsh (1987, 1990, 1993; Marsh & Yeung, 1998) argued that relations between academic self-concept and academic achievement were likely to be stronger when achievement was based on: classroom performance measures such as school grades, which are more easily influenced by characteristics such as quality and amount of student effort; high-stakes achievement tests that have important implications for students' future education and career path; and tests that closely match the school curriculum in a particular school. Following from this reasoning, the BFLPE is likely to be smaller when based on 'low stakes' tests that do not specifically match the school curriculum like those used in the present investigation. This consideration is, however, a double-edged sword in that achievement tests that did match the curriculum of each school and state/territory would not have provided a common basis of comparison.

BFLPE research provides an alternative, contradictory perspective to educational policy on the placement of students in special education settings that is being enacted throughout the world. Remarkably, despite the very different issues, this clash between our research and much existing policy exists at both ends of the achievement continuum (also see Robinson, Zigler, & Gallagher, 2000). In gifted education research and policy, there is an increasing trend towards the provision of highly segregated educational settings--special gifted-and-talented classes and academically selective schools for very bright students. This policy direction is based in part on a labelling theory perspective, which suggests that bright students will have higher self-concepts and experience other psychological benefits from being educated in the company of other academically gifted students. Yet our BFLPE and empirical evaluation of the effects of academically selective settings (e.g. Marsh, Chessor, Craven, & Roche, 1995) showed exactly the opposite effects. Placement of gifted students in academically selective settings results in lower academic self-concepts, not higher academic self-concepts. In recent research and policy for academically disadvantaged students, there is a worldwide inclusion movement to integrate these students into mainstream, regular classroom settings. Although economic rationalist perspectives appear to be the underlying motive for such decisions, the espoused rhetoric is based on a direct application of labelling theory. According to labelling theory, academically disadvantaged children are likely to be stigmatised and suffer lower self-concepts as a consequence of being placed in special classes with other academically disadvantaged students. Yet theory underpinning the BFLPE and empirical evaluation of the effects of including academically disadvantaged students in regular mainstream classrooms showed exactly the opposite effects (Tracey, Marsh, & Craven, 2003). Placement of academically disadvantaged children into regular classrooms results in lower academic self-concepts, not higher academic self-concepts. Furthermore the negative effects of inclusion on Peer self-concept reported by Tracey et al. suggested that academically disadvantaged children in regular classrooms actually feel socially excluded, not included. Whereas not all gifted and talented students will suffer lower academic self-concepts when attending academically selective high schools, many will. Similarly not all academically disadvantaged students will suffer lower academic self-concepts when attending regular, mixed-ability classes, but many will. BFLPE research, however, provides an important alternative perspective to existing policy directions that have not been adequately evaluated in relation to current educational and psychological research. The present investigation is particularly important in demonstrating the generalisability across Australian states/ territories of the theoretical and empirical basis of our claims.

Table 1 Summary of multilevel models applied separately to each
state/territory

 Mean Mean
 Number Number academic academic
State/ of of achieve- self-
territory students schools ment concept

1 Australian Capital
 Territory 500 23 .20 * .07
2 New South Wales 806 38 .15 * .00
3 Victoria 782 34 -.15 * .00
4 Queensland 761 34 -.04 -.04
5 South Australia 606 28 .04 -.02
6 Western Australia 579 27 .07 .00
7 Tasmania 535 24 -.13 .03
8 Northern Territory 287 15 -.29 * -.04

Total 4916 223 .00 .00

 Effects (in prediction
 of academic self-
 concept) for individual
 states/territories

 School
 Individual Individual average
 student student achieve-
State/ achieve- achieve- ment
territory ment ment (BFLPE)

1 Australian Capital
 Territory .27 * .14 * -.40 *
2 New South Wales .50 * .05 -.35 *
3 Victoria .51 * .13 * -.36 *
4 Queensland .38 * .07 * -.47 *
5 South Australia .45 * .07 -.44 *
6 Western Australia .30 * .06 -.29 *
7 Tasmania .31 * .04 -.32 *
8 Northern Territory .44 * .08 -.14

Total .40 * .07 -.35 *

Note: Individual student achievement = Individual student achievement
(linear component); Individual student [achievement.sup.2] = Squared
Individual student achievement (quadratic component). All outcome
and predictor variables were standardised (M = 0, SD = 1) across the
entire sample of 4916 individual students (product terms and
school-average variables were not restandardised). A multilevel
analysis of achievement test scores indicated that the set of eight
states/territories was statistically significant and when each
state/ territory was considered separately, four were statistically
significant. A similar analysis on academic self-concept responses
resulted in no significant differences among the states/territories.

* p<.05

Table 2 Summary of different multilevel models considered

(see Tables 3 and 4)

Model LLH DLLH Ddf Description of model

0 13948 Variance component (baseline) model
 with no predictor variables
1 13299 649 * 4 Model 0 + linear individual
 achievement (LIA) (a) and Quadratic
 IA (QIA)
2 13232 67 * 1 Model 1 + school-average
 achievement (SAA)
3 13222 12 * 2 Model 2 + SAAxLIA and SSAxQIA
4 13218 4 7 Model 3 + main effects of states/
 territories (State) (b)
5 13200 18 * 7 Model 4 + interaction of states/
 territories and LIA (StatexLIA)
6 13194 6 7 Model 5 + interaction of states/
 territories and SAA (StatexSSA)

Note: LLH = -2*log likelihood ratio. DLLH = difference in LHR between
the model being tested and the model with which it is compared (as
indicated in the Model description). Ddf = difference in degrees of
freedom (number of new parameters estimated in the new model).

(a) The linear component of individual student achievement (LIA) was
random at the school level (level 2) so that there were three
additional parameters associated with the introduction of this effect
(its effect, a variance component at level 2, and a covariance
component between this variable and variance in school intercept
terms). All other variables were fixed effects so that the
introduction of one additional variable resulted in one additional
parameter estimate and a change of 1 degree of freedom.

(b) The set of 8 states and territories was represented by a set of 7
dummy variables (scored 0 or 1) in which one state was arbitrarily left
out to serve as a basis of comparison (see Table 4).

* p < .05

Table 3 Academic self-concept: The effects of individual student
achievement and school-average achievement

 Model 1 Model 2
Model
variables Effect SE Effect SE

Fixed effects
Ind Ach .375 .015 .418 .016
Ind [Ach.sup.2] .061 .010 .063 .010
School MAch (BFLPE) -.302 .035
SMAch x IAch
SMAch x I[Ach.sup.2]
Constant -.059 .019 -.066 .018
Residual variance components
Lev 2 school .025 .006 .010 .005
Ind Ach .007 .005 .007 .004
Lev 1 students .851 .018 .849 .018

 Model 3
Model
variables Effect SE

Fixed effects
Ind Ach .396 .017
Ind [Ach.sup.2] .072 .012
School MAch (BFLPE) -.353 .038
SMAch x IAch .004 .034
SMAch x I[Ach.sup.2] .058 .017
Constant -.072 .018
Residual variance components
Lev 2 school .010 .005
Ind Ach .006 .004
Lev 1 students .848 .018

Note: Ind ach = Individual student achievement (linear component);
Ind [ach.sup.2] = Squared Individual student achievement (quadratic
component) ; School MAch = School-average achievement (the
big-fish-little-pond effect--BFLPE; indicated by shading) ; SMAch
x IAch = Interaction between School-average achievement and Individual
student achievement; SMAch x I[ach.sup.2] = Interaction between
School-average achievement and Squared Individual student achievement.
All outcome and predictor variables were standardised (M = 0, SD = 1)
so that effects correspond to standardised beta weights. All parameter
estimates are statistically significant when they differ from zero
by more than two standard errors (SEs). Analyses are based on responses
by 4916 students from 223 schools.

Table 4 Differences between states and territories (Model 6)

 Interactions
 with:

 Main effects Individ Ach

Dummy variables Effect SE Effect SE

ACT .074 .061 -.141 .060
Victoria .047 .053 .025 .054
Queensland -.021 .052 -.064 .054
South Australia -.008 .056 .011 .059
Western Australia .027 .057 -.154 .057
Tasmania .054 .059 -.132 .059
Northern Territory .092 .088 -.058 .068 .366

 Interactions
 with:

 School-avg
 Ach

Dummy variables Effect SE

ACT .078 .141
Victoria .040 .127
Queensland .014 .142
South Australia .015 .142
Western Australia .154 .152
Tasmania .125 .136
Northern Territory .092 .188

Note: Ach = achievement. Effect = effect in the path model. SE =
standard error (effects greater than 2 SE are statistically significant
at p < .05). In Model 6 (see also Tables 1 and 2), the set of 8 states
and territories was represented by a set of 7 dummy variables (scored 0
or 1) in which New South Wales was arbitrarily selected as the 'left
out' category to serve as a basis of comparison (see earlier
discussion). Whereas the selection of which state was to be left out
had no effect on the overall test of significance, tests of the
individual components had to be made relative to the left-out category.

Acknowledgements

This research was funded in part by grants from the Australian Research Council. I would like to thank Kit-Tai Hau, Wolfram Schulz, Ken Rowe, and Upali Jayasinghe for comments on an earlier version of the article. Requests for further information about this investigation should be sent to the author.

References

Aiken, L. S. & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage.

Alwin, D. F. & Otto, L. B. (1977). High school context effects on aspirations. Sociology of Education, 50, 259-273.

Branden, N. (1994). Six pillars of self-esteem. New York: Bantam.

Byrne, B. M. (1996). Measuring self-concept across the lifespan: Issues and instrumentation. Washington, DC: American Psychological Association.

Chapman, J. W. (1988). Learning disabled children's self-concepts. Review of Educational Research, 58, 347-371.

Coleman, J. M. & Fults, B. A. (1985). Special class placement, level of intelligence, and the self-concept of gifted children: A social comparison perspective. Remedial and Special Education, 6, 7-11.

Craven, R G., Marsh, H. W., & Print, M. (2000). Selective, streamed and mixed-ability programs for gifted students: Impact on self-concept, motivation, and achievement. Australian Journal of Education, 44, 51-75.

Davis, J. A. (1966). The campus as a frog pond: An application of theory of relative deprivation to career decisions for college men. American Journal of Sociology, 72, 17-31.

Diener, E. & Fujita, F. (1997). Social comparison and subjective well-being. In B. P. Buunk & F. X. Gibbons (Eds.), Health, coping, and well-being: Perspectives from social comparison theory (pp. 329-358). Mahwah, NJ: Erlbaum.

Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117-140.

Gamoran, A., Nystrand, M., Berends, M., & LePore, P. C. (1995). An organizational analysis of the effects of ability grouping. American Educational Research Journal, 32, 687-715.

Goldstein, H. (1995). Multilevel statistical models. London: Arnold.

Goldstein, H., Rasbash, J., Plewis, I., Draper, D., Browne, W., Yang, M., Woodhouse, G., & Healy, M. (1998). A user's guide to MLwiN. London: University of London, Institute of Education.

Gross, M. (1992). The use of radical acceleration in cases of extreme intellectual precocity. Gifted Child Quarterly, 36, 91-99.

Gross, M. (1993). Exceptionally gifted children. London: Routledge.

Gross, M. (1997). How ability groups turns big fish into little--or does it? Of optical illusions and optimal environments. Australasian Journal of Gifted Education, 6(2), 18-30.

Helson, H. (1964). Adaptation-level theory. New York: Harper & Row.

Hyman, H. (1942). The psychology of subjective status. Psychological Bulletin, 39, 473-474.

Inquiry into the Provision of Public Education in New South Wales (Tony Vinson, Chair). (2002). Report of the Vinson Inquiry: Sponsored by the New South Wales Teachers Federation and the Federation of P&C Associations of New South Wales (3 vols.). Retrieved from http://www.pub-ed-inquiry.org/reports/final_reports/

Jerusalem, M. (1984). Reference group, learning environment and self-evaluations: A dynamic multi-level analysis with latent variables. In R. Schwarzer (Ed.), The self in anxiety, stress and depression (pp. 61-73). Amsterdam: Elsevier Science Publishers.

Marsh, H. W. (1974). Judgmental anchoring: Stimulus and response variables. Unpublished doctoral dissertation, University of California, Los Angeles.

Marsh, H. W. (1984). Self-concept: The application of a frame of reference model to explain paradoxical results. Australian Journal of Education, 28, 165-181.

Marsh, H. W. (1987). The big-fish-little-pond effect on academic self-concept. Journal of Educational Psychology, 79, 280-295.

Marsh, H. W. (1990). A multidimensional, hierarchical self-concept: Theoretical and empirical justification. Educational Psychology Review, 2, 77-172.

Marsh, H. W. (1991). The failure of high ability high schools to deliver academic benefits: The importance of academic self-concept and educational aspirations. American Educational Research Journal, 28, 445-480.

Marsh, H. W. (1993). Academic self-concept: Theory, measurement and research. In J. Suls (Ed.), Psychological perspectives on the self (Vol. 4, pp. 59-98). Hillsdale, NJ: Erlbaum.

Marsh, H. W. (1994). Using the National Educational Longitudinal Study of 1988 to evaluate theoretical models of self-concept: The Self-Description Questionnaire. Journal of Educational Psychology, 86, 439-456.

Marsh, H. W. (2002, July 26). Comment: Pupils can lose their shine in elite schools. Sydney Morning Herald, Editorial Section, p. 11. Retrieved from http://www.smh.com.au/articles/2002/07/25/1027497382341.html

Marsh, H. W., Chessor, D., Craven, R. G., & Roche, L. (1995). The effects of gifted and talented programs on academic self-concept: The big fish strikes again. American Educational Research Journal, 32, 285-319.

Marsh, H. W. & Craven, R. G. (1998). The big fish little pond effect, optical illusions, and misinterpretations: A response to Gross [1997]. Australasian Journal of Gifted Education, 7, 6-15.

Marsh, H. W. & Craven, R. (2002). The pivotal role of frames of reference in academic self-concept formation: The big fish little pond effect In F. Pajares & T. Urdan (Eds.), Adolescence and education (Vol. 2, pp. 83-123). Greenwich, CT: Information Age.

Marsh, H. W. & Johnston, C. F. (1993). Multidimensional self-concepts and frames of reference: Relevance to the exceptional learner. In F. E. Obiakor & S. Stile (Eds.), Self-concept of exceptional learners: Current perspectives for educators. Dubuque, IA: Kendall/Hunt.

Marsh, H. W., Koller, O., & Baumart, J. (2001). Reunification of East and West German school systems: Longitudinal multilevel modeling study of the big fish little pond effect on academic self-concept. American Educational Research Journal, 38, 321-350.

Marsh, H. W., Kong, C-K, Hau, K-T. (2000). Longitudinal multilevel modeling of the big fish little pond effect on academic self-concept: Counterbalancing social comparison and reflected glory effects in Hong Kong high schools. Journal of Personality and Social Psychology, 78, 337-349.

Marsh, H. W. & Parker, J. (1984). Determinants of student self-concept: Is it better to be a relatively large fish in a small pond even if you don't learn to swim as well? Journal of Personality and Social Psychology, 47, 213-231.

Marsh, H. W., Plucker, J. A., & Stocking, V. B. (2001). The Self-Description Questionnaire II and gifted students: Another look at Plucker, Taylor, Callahan & Tomchin's (1997) 'Mirror, mirror on the wall'. Educational and Psychological Measurement, 61, 976-996.

Marsh, H. W. & Rowe, K.J. (1996). The negative effects of school-average ability on academic self-concept--an application of multilevel modelling. Australian Journal of Education, 40, 65-87.

Marsh, H. W. & Yeung, A. S. (1998). Longitudinal structural equation models of academic self-concept and achievement: Gender differences in the development of math and English constructs. American Educational Research Journal, 35(4), 705-738.

Meyer, J. W. (1970). High school effects on college intentions. American Journal of Sociology, 76, 59-70.

Morse, S. & Gergen, K.J. (1970). Social comparison, self-consistency, and the concept of self. Journal of Personality & Social Psychology, 16, 148-156.

Organisation for Economic Cooperation and Development. (2001a). Knowledge and skills for life: Results from the first OECD programme for international student assessment (PISA) 2000. Paris: Author.

Organisation for Economic Cooperation and Development. (2001b). PISA international data base (Technical Report). Paris: Author.

Parducci, A. (1995). Happiness, pleasure, and judgment: The contextual theory and its applications. Mahwah, NJ: Erlbaum.

Raudenbush, S. W. & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: Sage.

Robinson, N. M., Zigler, E., & Gallagher, J. J. (2000). Two tails of the normal curve: Similarities and differences in the study of mental retardation and giftedness. American Psychologist, 55, 1413-1424.

Sampson, J. F. (1969). The high school examination achievements of opportunity "C" class pupils. Sydney: NSW Department of Education.

Sampson, J. F. (1977). The secondary school performance of pupils in selective and comprehensive schools. In New South Wales Department of Education, The education of the talented child (pp. 111-131). Sydney: Author.

Sherif, M. & Sherif, C. W. (1969). Social psychology. New York: Harper & Row.

Stouffer, S. A., Suchman, E.A., DeVinney, L.C., Star, S.A., & Williams, R.M. (1949). The American soldier: Adjustments during army life (Vol. 1). Princeton: Princeton University Press.

Suls, J. M. (1977). Social comparison theory and research: An overview from 1954. In J. M. Suls & R. L. Miller (Eds.), Social comparison processes: Theoretical and empirical perspectives (pp. 1-20). Washington, DC: Hemisphere.

Tracey, D. K., Marsh, H.W., & Craven, R. G. (2003). Self-concepts of preadolescent students with mild intellectual disabilities: Issues of measurement and educational placement. In H. W. Marsh, R. G. Craven, & D. M. McInerney (Eds.), International advances in self research (Vol. 1, pp. 203-230). Greenwich, CT: Information Age.

Upshaw, H.S. (1969). The personal reference scale: An approach to social judgment. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 4, pp. 315-370). New York: Academic Press.

Wills, T. A. (1981). Downward comparison principles in social psychology. Psychological Bulletin, 90, 245-271.

Zeidner M. & Schleyer, E.J. (1999). The big-fish-little-pond effect for academic self-concept, test anxiety and school grades in gifted children. Contemporary Educational Psychology, 24, 305-329.

Professor Herbert W. Marsh is Director of the SELF Research Centre, University of Western Sydney, Bankstown Campus, Penrith, New South Wales, 1797. E-mail: h.marsh@uws.edu.au.