The Big-Fish-Little-Pond Effect: What Do We Know and Where Do We Go from Here? PDF

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/225145441 The Big-Fish-Little-Pond Effect: What Do We Know and Where Do We Go from Here? Article in Educational Psychology Review · September 2008 DOI: 10.1007/s10648-008-9071-x CITATIONS READS 153 9,718 2 authors: David Dai Anne N. Rinn University at Albany, The State University of New York University of North Texas 86 PUBLICATIONS 3,039 CITATIONS 89 PUBLICATIONS 2,109 CITATIONS SEE PROFILE SEE PROFILE All content following this page was uploaded by Anne N. Rinn on 06 November 2023. The user has requested enhancement of the downloaded file. Educ Psychol Rev (2008) 20:283–317 DOI 10.1007/s10648-008-9071-x REVIEW ARTICLE The Big-Fish-Little-Pond Effect: What Do We Know and Where Do We Go from Here? David Yun Dai & Anne N. Rinn Published online: 19 February 2008 # Springer Science + Business Media, LLC 2008 Abstract The big-fish-little-pond effect (BFLPE) refers to the theoretical prediction that equally able students will have lower academic self-concepts in higher-achieving or selective schools or programs than in lower-achieving or less selective schools or programs, largely due to social comparison based on local norms. While negative consequences of being in a more competitive educational setting are highlighted by the BFLPE, the exact nature of the BFLPE has not been closely scrutinized. This article provides a critique of the BFLPE in terms of its conceptualization, methodology, and practical implications. Our main argument is that the BFLPE, while having added to our understanding of the origins of self-concepts, disproportionately emphasizes one aspect of social comparison to the exclusion of many other intervening factors. In light of our critique, we suggest a broader conception of social comparison effects on academic self-concept that emphasizes a more active role of individuals in regulating their social cognition and motivation, as well as a more distinct effect of social-contextual influences. We also suggest alternative research designs that would incorporate contextual, developmental, and individual differences as potential moderators or mediators of the BFLPE. Keywords Social comparison (imposed vs. self-engaged). Academic self-concept. Coping. Academic motivation and achievement. School selectivity. Ability grouping D. Y. Dai (*) Department of Educational and Counseling Psychology, University at Albany, State University of New York, ED233, Albany, NY 12222, USA e-mail: [email protected] A. N. Rinn Department of Social Sciences, University of Houston-Downtown, Suite South 625, 1 Main Street, Houston, TX 77002, USA e-mail: [email protected] 284 Educ Psychol Rev (2008) 20:283–317 What is the BFLPE? The big-fish-little-pond effect (BFLPE) was proposed by Marsh and Parker (1984; Marsh 1984, 1987) based on the assumption that students use a local “frame of reference” to form their self-concepts on a relevant dimension of the self. Marsh and Parker operationalized the frame of reference or standard of comparison to be school-average ability level1. Thus, the general theoretical prediction based on the BFLPE is that equally able students will have lower academic self-concepts in higher average-ability schools but higher self-concepts in comparatively lower average-ability schools. In plain language, whether one perceives oneself as a big fish depends on the size of the other fish (more accurately, the average size of the other fish) in the pond. The stronger the peer group as a frame of reference, the lower one’s academic self-concept. By the same token, weaker peer groups lead to higher academic self-concepts. Since Marsh and his colleagues proposed the BFLPE model, numerous studies have been conducted to date (see Marsh and Hau 2003; Marsh et al. 2005; Marsh et al. 2007, for most recent studies), and the areas where the BFLPE is investigated have also been extended from academic to artistic and athletic, among others (e.g., Chanal et al. 2005). The research interest generated by the BFLPE model is largely due to its practical implications. Marsh (1987; Marsh et al. 2004) has repeatedly warned that parents who are eager to send their children to selective schools should be aware of possible negative consequences associated with the BFLPE. Moreover, based on the fact that people are constantly working in social settings that potentially provide social comparative information and salient cues for social comparison, the BFLPE should be ubiquitous. In any place where upward comparison is either forced or voluntarily engaged, such as attending selective schools, self- contained gifted programs or classes, academic acceleration, or other forms of advanced placement, mainstreaming or full inclusion of students with disabilities or other special needs, the BFLPE should be relevant. Indeed, Marsh and colleagues (e.g., Marsh et al. 2007) also alluded to the negative consequences of mainstreaming or full inclusion on academic self-concepts for students with special education needs (see Marsh et al. 2006 for an empirical study of students with mild intellectual disabilities). With this background in mind, we will first provide an overview of the research literature pertaining to the BFLPE. This overview is not purported to be exhaustive but will point out discrepancies and raise questions about the BFLPE. We will then provide a critique of the BFLPE model in terms of its conceptualization (theory), methodology, and practical implications. Based on our critique, we will suggest a possible broadening of the research scope that would permit a more refined understanding of where, when, and why (and for whom) the BFLPE will or will not occur. We will discuss specific methodological ramifications and practical significance of these suggested changes. What Do We Know About the BFLPE? Evidence supporting the BFLPE In Marsh and his colleagues’ research, the BFLPE is typically operationalized as a negative effect of school-average ability (or achievement) on academic self-concept, controlling for 1 In the BFLPE studies, Marsh used the terms ability and achievement interchangeably to indicate levels of academic performance students have attained or capable of attaining. Accordingly, they operationalized the BFLPE as caused sometimes by school-average ability and sometimes by school-average achievement. In this article, we follow this practice but will point out where the distinction becomes important. Educ Psychol Rev (2008) 20:283–317 285 the positive effect of individual student ability (or achievement) in a path model (see Marsh and Hau 2003; Marsh and Craven 2002). In earlier studies, path analysis was used for this analysis (e.g., Marsh and Parker 1984; Marsh 1987). With the advent of more sophisticated analytic techniques, multi-level modeling (i.e., hierarchical linear modeling, HLM) has become a standard procedure (Marsh and Hau 2003; Marsh et al. 2007). Evidence accumulated since 1984 based on this path model (from hereon the BFLPE paradigm) unequivocally supports the BFLPE hypothesis (see Marsh and Craven 2002 for review). Using the multi-level modeling of the opposite effects of individual student ability and school-average ability on academic self-concept, Marsh and Hau (2003) tested the cross-cultural generality of the BFLPE with data from 26 countries. They found consistent evidence of the BFLPE across nations examined. In a recent longitudinal modeling of the BFLPE, Marsh et al. (2007) not only found persistent negative effects of being in more selective high schools 2 years beyond graduation, these effects also seemed robust over gender and individual student achievement levels. In addition to examining the BFLPE in the realm of academic ability or achievement and self-concept, other researchers have examined the BFLPE in such areas as physical abilities and artistic ability. These findings would seem to lend further support to the BFLPE model. For example, Chanal et al. (2005) examined the gymnastics self-concepts of 405 participants in 20 gymnastics classes. They found gymnastics self-concept was negatively predicted by class-average gymnastics skills and this negative relationship grew larger during the 10-week gymnastics program. Marsh and colleagues argue that academic self-concept is important not only because it has consequences for one’s general psychological well-being, but it also carries motivational properties. Academic self-concept has been shown to affect one’s educational aspirations (Davis 1966; Marsh 1987), academic interest (Marsh et al. 2005; Marsh et al. 2007), course selection (Marsh and Yeung 1997a), and achievement over time (Marsh and Yeung 1997b). Although the causal predominance of academic achievement over academic self-concept or vice versa is a controversial issue (see Caslyn and Kenny 1977), it is likely that the relationship is bi-directional and reciprocal, rather than unidirectional (Hamachek 1995; House 2000), and mediated by motivational processes such as self-appraisal and self-motivation (Pajares 1996). Bandura and Jourden (1991) found, for example, different social comparative feedback (consistently surpassing others or progressively declining in performance) significantly impacted individuals’ self-efficacy and performance aspirations (i.e., aspired-for levels of performance) in organizational decision making. Evidence constraining the BFLPE Marsh and colleagues also recognized the possibility that being with more competitive peers in school does not exclusively lead to declined self-concepts on relevant dimensions of comparison. Therefore, they labeled a positive “reflected glory” effect in contrast to the negative BFLPE. These two counterbalancing effects parallel what are typically called in the social comparison literature “assimilation effects” and “contrast effects.” A contrast effect occurs when one sees oneself as different from a comparison target, and an assimilation effect occurs when one sees oneself as similar to a comparison target. Marsh et al. (2000) conducted a large-scale study of about 8,000 Hong Kong high school students, juxtaposing both the BFLPE and reflected glory effects. Using both a measure of school average ability and that of perceived school status, they were able to identify a “reflected 286 Educ Psychol Rev (2008) 20:283–317 glory” effect along with the BFLPE. However, their analysis supports a stronger BFLPE compared to a reflected glory effect. Other research supports a more social-contextual view of contrast versus assimilation effects, rather than treating these effects as invariant. McFarland and Buehler (1995), for example, found the BFLPE was attenuated among people who value a social group to which they belong but was the strongest when there is a weak emotional bond toward the social group. Stapel and Koomen (2001) found, in a series of experiments, that whether social comparison has contrast or assimilation effects depends on activation of the personal or social self. Activation of the social self (“we”) more likely induces assimilation effects, and activation of the personal self (“I”) more likely induces contrast effects. Likewise, independent construal of the self leads to lower self-evaluation under upward comparison, but interdependent construal of the self reduces the contrast effect, and no difference in self- evaluative consequences between upward and downward comparison (Cheng and Lam 2007). The BFLPE model assumes that except for a possible “reflected glory” effect, which is largely based on how one perceives the status of the school in which they find themselves, inevitable upward comparison in higher average-ability schools is always self-deflating; that is, leading to declined self-concepts. This assumption is challenged by several studies. Huguet et al. (2001) studied the social comparison behavior of students from two French middle school during transition to new schools. They found many students engaged in social comparison with academically “better-off” friends for the sake of self-evaluation and self-improvement. This upward comparison is not self-deflating as long as the comparer identifies with the comparison target and has an internal locus of control (i.e., one is able to catch up with the better-off targets through efforts). Blanton et al. (1999) argued that social comparison could occur at multiple levels. They asked a large sample of ninth grade students to report (a) how they compare with most of other students in specific academic subjects (this is the frame of reference assumed by the BFLPE hypothesis), and (b) who their most and least favorable comparison targets are (this is the self-selected comparison process and target beyond the radarscope of the BFLPE paradigm). They found that these two frames of reference are not conflicting, as both independently predicted improved academic performance. Of interest here is the finding that selection of peers who were doing well predicts improved achievement. This finding led the authors to argue that shifting comparison targets downward for self-enhancement is not productive academically. Even if it might temporarily protect one’s well-being or self-worth, it does not facilitate self-improvement and, in the long run, will negatively affect academic achievement. In the same vein, Altermatt and Pomerantz (2005) studied a large sample of fifth, sixth, and seventh grade students and found that low achievers who established and maintained friendships with high-achieving friends evaluated themselves less positively, a variant of the BFLPE; but they performed better academically than low achievers with similarly low achieving friends, a tradeoff not captured by the BFLPE model. (See also later discussion of Burleson et al. 2005; for descriptive details of the BFLPE-related studies, see Appendices A and B). Evidence that reveals the limitations of the BFLPE paradigm In addition to the literature that shows a more complex pattern of social comparison effects than the BFLPE predicts, there is a bulk of social comparison research that does not provide direct evidence regarding the BFLPE per se, but nevertheless reveals important limitations of the BFLPE model. These limitations have to do with the lack of detailed mechanisms Educ Psychol Rev (2008) 20:283–317 287 between whether and how social comparison is made and whether and how it impacts self- concept formation or changes. Butler (1992) conducted an experimental study of how mastery versus ability goal condition affects students’ interest in different types of social comparison information. She was able to demonstrate that under mastery goal condition, students are more interested in social comparison information that is task-diagnostic; that is, what helps one improve their performance. In contrast, under ability goal condition, students are more interested in ability-diagnostic information, that is, how well one performs compared to others. Regner et al. (2007) observed similar patterns of relationships in natural settings. Their findings suggest that social information seeking is fundamentally selective, depending on the current needs, purposes, and concerns. The issue of self-concepts and self-perceptions becomes salient only when ability goal is activated; in other words, there is an ability-diagnostic focus (Bandura 1993). In a later study, Butler (1995) also found students’ social comparative information seeking has both informational and motivational foci, and the two may co-exist, which echoes the findings of Blanton et al. (1999). Ruble and Flett (1988) studied developmental trends of using social comparison for self- evaluation of academic competence with a cross-sectional sample of second, fourth, and sixth grade students representing high, medium, and low ability levels in arithmetic. They found that high ability children engaged in the most self-evaluative information seeking, and with increasing age, high ability children were more likely to engage in autonomous comparison (i.e., comparing products of one’s own performance on several occasions) than social comparison. In contrast, low- and medium-ability children across ages showed a consistent interest in social comparison in terms of looking at others’ works displayed in the room, labeled as representing different levels of ability. Ruble and Flett speculated that children of low- and medium ability are more likely than high-ability children to continually experience uncertainty of their levels of ability and therefore maintain a strong interest in social comparison. Social comparison information is more relevant than autonomous comparison to assessing relative ability as a stable, dispositional characteristic. As children become cognitively more mature through the elementary school years, there is an increasing recognition of ability as a stable trait (Rholes et al. 1990), a recognition crucial for forming enduring academic self-concepts. These findings do not contradict the BFLPE hypothesis. As a matter of fact, they provide a mechanism for the BFLPE in that when one moves to a high-ability school or program, one is more likely to experience uncertainty about one’s ability, and thus more likely to engage in social comparison with available peers. However, these findings do reveal the theoretical gap left by the BFLPE paradigm. Supporting Ruble and colleagues’ finding on developmental trends in social comparison, Nicholls and Miller (1984) were able to show with a clinical interview technique that inferring ability as capacity is constrained by cognitive development. For example, only those who have developed the most differentiated conceptions of ability are able to infer that, with equal achievement outcomes, high effort implicates low ability and vice versa. Inferring a personal quality such as ability or making causal self-attributions is a process underlying academic self-concept formation (Marsh 1984). Although Marsh, in the early years of research on the BFLPE, also mentioned the possible developmental constraint on how self-concept is shaped (Marsh and Parker 1984; Marsh 1987), developmental considerations have not been a main concern in his research. Marsh (1987) speculated that the BFLPE is more likely to occur in elementary school years than later years, as children at that age are still in the formative years regarding self-concepts. This view is at variance with findings of research by Nicholls and Miller and by Ruble and her colleagues 288 Educ Psychol Rev (2008) 20:283–317 on developmental underpinnings of social comparison and self-concept formation. Although the BFLPE and developmental research on social comparison have differing foci, and thus pitting one against the other may not be justified, it is fair to argue that without a clear understanding of developmental underpinnings of social comparison processes and consequences, the BFLPE hypothesis would be weak in its theoretical foundation. The BFLPE hypothesis is based on the assumption that social comparison is situationally imposed, and the standard of comparison is the average ability or achievement of the peer group. However, in a study of spontaneous social comparison behaviors of first- year college students, Foddy and Crundall (1993) found most students compared themselves with either similar or better-off students. There was a distinct lack of interest in downward comparison, regardless of ability levels, presumably because downward comparison is not informative for performance evaluation, albeit its self-enhancing effect, a finding consistent with Ruble and Flett’s (1988) suggestion that students’ social information seeking is an active process, rather than a passive one. In a study of K-2 elementary school children, Pomerantz et al. (1995) found consistent evidence that with cognitive maturity, children become increasingly aware of positive and negative effects of social comparison; they adjust their social comparison behavior accordingly to suit their self-evaluation goals. Although coming from different research traditions and drawing different conclusions, the above studies are similar in that they all involve detailed observations of social comparison behaviors in situ, where students were engaged in some academic tasks and demonstrated an emergent need for social comparison. Most of these studies also attempted to detail the developmental and cognitive underpinnings of how social comparison information is attended to and processed for self-evaluative, self-improvement, and self- protection purposes. Even though they do not necessarily directly support or refute the BFLPE hypothesis, they portrayed a more complex picture of how social comparison is made than the BFLPE hypothesis assumes. Applications of the BFLPE theory to attending gifted programs There have been several studies investigating the BFLPE in gifted programs. One can argue that gifted education provides an ideal test bed for the BFLPE theory, as participating in a self-contained or short-term gifted program, or getting accelerated into a more advanced class and grade level, gets close to the essence of the metaphor of a big fish in a little pond suddenly turned median or small when thrown into a big pond with many big or bigger fish. Marsh et al. (1995) conducted a quasi-experiment of 29 students participating in a gifted and talented (GT) class, and found a measurable decline in academic self-concept but not in non-academic self-concepts. However, to what extent the decline reflects an enduring change is still open to question. Gibbons et al. (1994) found that participating in a gifted program initially led to lowered self-concept, particularly for male students; however, the self-concept bounced back to the baseline 6 months after the program. Zeidner and Schleyer (1998) found, with a large sample of fourth through sixth grade Israeli students, that students in the gifted programs had lower academic self-concepts than their counterparts in mixed ability classes, a finding in line with the BFLPE hypothesis. However, there is also evidence suggesting that there is no prevailing BFLPE of participating in gifted programs or schools as predicted by the theory. Cunningham and Rinn (2007) studied a group of adolescents who participated in a 3-week summer program. They measured their academic self-concepts at entry point and at the end of the program, and did not find significant changes in academic self-concept over time. They suggested that more than one reference group can be involved when students in gifted programs make Educ Psychol Rev (2008) 20:283–317 289 social comparison, which can balance out the BFLPE. Related to this issue are the effects of ability (homogeneous) grouping on self-concept as compared with that of heterogeneous grouping. Findings seem mixed in that regard (see Kulik and Kulik 1991, 1997). Since early research typically used measures of general self-concept or self-esteem rather than academic self-concepts as outcome measures, the findings in either way do not directly support or refute the BFLPE hypothesis, which emphasizes academic self-concepts as outcomes. In comparing gifted college students who were enrolled in an honors program and gifted college students who were not enrolled in an honors program, Rinn (2007) found gifted college students who are enrolled in an honors program have significantly higher academic self-concepts than gifted college students who are not enrolled in an honors program, even after controlling for SAT scores. Although not a direct examination of the BFLPE, these findings lend some support to the notion that the BFLPE might not exist at the university level. Marsh (1987) noted that the BFLPE might be smaller for older students, as they “typically have some basis for the assessment of their own academic skills that is independent of the performances of their classmates, and they often know how the average ability level of their classmates compares with some broader frame of reference” (p. 282). The mixed findings in gifted programs suggest, along with the research reviewed earlier, that the social comparison processes and consequences are more complex than the BFLPE theory predicts. In line with the social comparison literature, when the artistic ability self- concepts of 141 artistically talented adolescents during a 6-week program were examined, Burleson et al. (2005) found both support for the BFLPE and evidence that it is at variance with what is predicted by the BFLPE. Their findings suggest that mere exposure to highly talented peers does not predict a decrease in related self-concepts, but how individuals construe the social comparison information (the upward comparison target as inspiration versus as a source of the feeling of inferiority) affects positive and negative changes in self- concept, which is consistent with what is typically found in the social comparison literature, further implicating individual differences in how a given social situation is construed. One enduring question regarding the BFLPE is whether the observed decline in self- concept is a temporary state or has an enduring impact on the individual’s motivation and self-esteem (Dai et al. 1998). While Marsh and colleagues (e.g., Marsh et al. 2005; Marsh et al. 2007) argued that the negative effects tend to get worse over time, the long-term fallout of participating in gifted program does not seem in keeping with this prediction. Moon et al. (1994) noted that the short-term effects of a gifted program on the self-concepts of gifted students may be negative or non-existent, but the long-term effects may actually be positive. In their study of the long-term effects of an enrichment program for gifted elementary school students, Moon et al. found the long-term benefits of participating in a gifted program to be mostly positive, including an increased self-concept, increased motivation, an increase in basic thinking skills, and an increase in autonomous learning. Although Marsh et al. (1995) did identify a group of 9-year-old students suffering a decline in academic self-concept after being accelerated into a gifted class (indeed a decline in non- academic self-concepts as well), most studies of accelerated students do not find an consistent pattern of lowered self-concepts and prolonged negative motivational con- sequences (see Brody and Benbow 1987). In fact, long-term prospects based on research are quite positive (Lubinski 2004). Most of the accelerated students in the SMPY study, when reaching their adulthood, were satisfied with the decision of acceleration (Lubinski 2004). To be sure, it can be argued that research on the effects of gifted programs typically does not test the BFLPE hypothesis in a stringent manner. However, no apparent long-term negative self-related or motivation-related consequences of participating in gifted programs do challenge the external or ecological validity of the BFLPE model. 290 Educ Psychol Rev (2008) 20:283–317 Discrepancies between the BFLPE paradigm and the social comparison literature When research shows discrepancies in findings in a systematic way, it is natural to wonder whether it is merely a methodological artifact (e.g., different sample characteristics, different measurement tools) or reflects much deeper conceptual and methodological chasms or differing assumptions and premises. Dai (2004) pointed out the BFLPE research program has had minimal contact with the social comparison literature. This assessment is based on the extent to which the BFLPE researchers have integrated or incorporated empirical evidence as well as conceptual and methodological considerations in modifying and refining the BFLPE model or paradigm over time. Despite Marsh’s (Marsh et al. 2004) argument to the contrary, discrepancies between the BFLPE paradigm and the social comparison literature are evident with respect to at least the following aspects of social comparison: a) Imposed vs. self-engendered social comparison. The social comparison literature typically considers social comparison as an active process, though encountered or imposed social comparison that makes social comparison unavoidable is also considered as legitimate (Wood 1996). Festinger (1954), for example, argued that people seek social comparison information because they experience uncertainty and ambiguity in evaluating their own performance and verifying their own competence. In other words, if people are certain about their abilities, there is no need to engage in social comparative information seeking. Marsh et al. (2004) cited Diener and Fujita (1997) as supporting what they called “situationally imposed or forced comparison” (p. 270) underlying the BFLPE. However, although acknowledging the existence of imposed social comparison, Diener and Fujita argued that “to the extent people employ social comparison as a flexible coping strategy in the health domain as well as other areas, the effects of imposed social comparison diminish or disappear” (p. 330). Supporting this argument, Gibbons et al. (1994) found that students who did not compare favorably with others reduced the amount of social comparison they made or shifted their comparison targets, suggesting social comparison processes involve active coping, for good or ill. Although self-engendered social comparison was alluded to on various occasions in Marsh and colleagues’ studies (e.g., Marsh et al. 2000), it has never been formally incorporated into their research design as an integral component. b) Directionality of social comparison. In an imposed social comparison, one’s academic standing among peers is uniformly gauged based on the local norms (presumed to be school-based according to Marsh 1987). The frame of reference is a given entity, imposed by the impinging environment, an assumption that justifies using the path coefficient from school-average ability or achievement to academic self-concepts as a proxy measure of the BFLPE (provided that average school ability varies from school to school). In contrast, in self-engendered models of social comparison, social comparison can be engaged in with similar others, upward with better-off peers, or downward with worse-off peers at any given point in time, depending on a host of contextual and individual difference variables, and comparison targets are selected based on internal standards, guides, and current concerns and motives (Banaji and Prentice 1994; Brown and Dutton 1995; Collins 1996; Wood 1996). Furthermore, effects of the directionality of social comparison also vary, with downward comparison more likely serving self-protection and upward comparison serving information and self-improvement needs. Thus, directionality of social comparison does not automat- ically produce positive or negative consequences on self-concept in a linear fashion, as the BFLPE hypothesis proposes. Educ Psychol Rev (2008) 20:283–317 291 c) Mitigating conditions for the BFLPE and situational and individual moderators and mediators. Except for pointing out a counterbalancing “reflected glory effect,” Marsh and colleagues do not specify any other possible situational or personal variables that might moderate the BFLPE. What mediates the BFLPE is also implicitly assumed rather than explicitly tested (e.g., the social comparison processes). In general, the research strategy of the BFLPE program is to show generality and ubiquity of the BFLPE over gender, ability levels, and cultures (Marsh and Hau 2003; Marsh et al. 2007), rather than finding out details of how it works psychologically (i.e., addressing the issue of internal validity), as evidenced by their preference for large-scale data sets and statistical manipulation to tease out the effects (e.g., Marsh 1987, 1994; Marsh and Hau 2003; Marsh et al. 2007), rather than conducting carefully designed experimental or observational studies especially aimed at elucidating the nature of the BFLPE. This strategy differs from that of the mainstream social comparison research programs (typically conducted by social and personality psychologists), which tend to seek detailed understanding of contexts, motivations, mechanisms, and constraints of social comparison and related social cognition through experimentation (see Wood 1996). d) The role of self-motives and self-regulation. The social comparison literature usually considers selective attention, differential perception, construal, and interpretation of social situations and social information as essential for understanding why and how people engage in social comparison (Taylor et al. 1996). Active regulation of social comparison behaviors based on motives and current concerns constitute an important aspect of social comparison, and also determines social information uptake and processing. This aspect of social comparison is ignored in the BFLPE paradigm. Instead, self-concept formation works like an impression imposed upon the individual’s mind due to some assumed contrast effects of unspecified social encounters with information about one’s academic standing. The construct of academic self-concept in the BFLPE paradigm is so remote from the person who is supposed to form and maintain this self-concept or self-perception that self-motives, purposes of self-evaluation, and self-regulation of social comparison become obscured. In short, the social comparison literature generally maintains a perspective of social comparison as an agent-based process, whereas the BFLPE paradigm conveys a mechanistic flavor. Why is the BFLPE Paradigm Flawed? So far, we have described the research on the BFLPE and identified the discrepancies and their sources. In this section, we provide a critique of the BFLPE. The purpose of this critique is not to discredit this valuable line of research efforts by Marsh and colleagues for almost a quarter of a century, but to point out areas where improvements and modifications can be made to further our understanding of the BFLPE in a broader context of research on academic achievement, self-concept, academic motivation, and general psychological well- being. Our critique starts with a conceptual analysis of three general models of social comparison, followed by a close look at the standard methodology adopted to investigate the BFLPE, including the path model for detecting the BFLPE (Fig. 1). Since Marsh and colleagues also make a strong argument about the disadvantages of being in a “big pond,” so to speak, based on the BFLPE research, we also identify what is omitted or neglected in this argument. 292 Educ Psychol Rev (2008) 20:283–317 (Positive Effect) ++ Individual student Individual student ability Academic self-concept ++ _ (Negative Effect, BFLPE) School-average ability Fig. 1 Theoretical prediction of the big-fish-little-pond effect (BFLPE); the BFLPE is predicted to be smaller in effect size than the effect of individual student ability. Adapted from Marsh and Hau (2003) Conceptual problems with the BFLPE theory To simplify our conceptual analysis, we can roughly identify three models of social comparison processes in academic contexts: the situationally imposed model underlying the BFLPE hypothesis; the coping model, typically used in social comparison research; and the motivational model, which puts self-concept squarely in the context of many other self- related constructs such as self-efficacy, self-attribution, and self-theory in academic and other settings where personal performance or competence is at stake. Situationally imposed models Situationally imposed models of social comparison are not without some good foundation. Sometimes social information becomes so compelling that one cannot avoid but relate it to oneself in a way significantly impacting how one views her- or himself (Wood 1996). However, there is a set of conditions that constrain when situationally imposed models become viable. These may include contextual factors such as salience of comparison information, a performance-driven atmosphere, an ability-diagnostic focus, as well as cognitive sophistication that permits discerning enduring dispositions and ability as capacity (Nicholls 1984). This situationally imposed social comparison can get to the point that social comparison standards and cultural norms become the same (Diener and Fujita 1997). In general, a tightly structured social environment renders situationally imposed social comparison more likely. Conversely, in a loosely structured social environment, where no distinct centralized information control and dissemination exists, there is likely an array of information from which people can pick and choose, and individual differences are more likely to come into play. However, Marsh’s BFLPE paradigm treats situationally imposed social comparison as applicable to all academic situations as long as they meet one condition: school-average ability or achievement differs across schools. This criterion does not consider how salient and predominant the social comparison cues are and how accessible social comparison information is. Consequently, defining frame of reference or standard of comparison as fixated at the school level (i.e., school-average ability) becomes a matter of research convenience rather than psychological necessity. The question of how school-average ability or achievement as a framework of Educ Psychol Rev (2008) 20:283–317 293 reference is psychologically mediated is assumed rather than explicated through evidence. One might argue, using the physical and social proximity and information accessibility as criteria, that the class level constitutes a more viable frame of reference for social comparison. Indeed, Marsh and colleagues’ own research shows that it works just as well (e.g., Lűdtke et al. 2005). While situationally imposed or forced comparisons may exist and impact self-concept to some extent, social psychology research also shows that individuals perceive and interpret their worlds and themselves differently, depending on different guides, standards, and goals, “normative guides” being only one of them (Higgins et al. 1986). Thus, a student’s reference group can be viewed in several different ways. For example, Williams and Montgomery (1995) define one’s frame of reference as his or her “particular learning environment” (p. 403). Kuh et al. (2000) developed a student typology based on the activities in which students engage while in college. Based on this typology, Kuh et al. argue that one’s activities expose a person to different groups of people with whom to compare oneself. Thus, a student’s typology is their reference group. The BFLPE is based on the assumption that people compare themselves externally only with a local norm in their immediate environment (e.g., school average). External comparison, however, could actually involve any number of people or groups or combinations of these. In evaluating one’s abilities, it would be helpful to know where one stands relative to all ability levels, instead of just where one stands relative to others of similar ability. In other words, it is possible to have multiple reference groups for social comparison. Bassis (1977) argues that students within a particular environment are likely to realize where their environment falls on the selectivity continuum, at least in a broad sense. “Thus the argument is that, to some extent at least…students extend the boundaries of their comparison reference groups beyond their local frog ponds” (p. 1322). Bassis found that college students, in forming their reference groups, are likely to incorporate across- institution comparisons when evaluating themselves. This effect was later labeled by Marsh (Marsh et al. 2000; Marsh et al. 2001) as the “reflected glory effect.” Likewise, students in a gifted program might not experience the BFLPE if they extend their reference groups outside of their immediate peer group within the gifted program and assimilate with the members of the gifted program. However, the implication of flexibility using comparison standards, local or global, goes beyond “reflected glory.” It suggests that unless the social or cultural norms are extremely compelling, overwhelming individuals’ flexible choice, people will selectively use varied comparison criteria under different circumstances that better serve self-evaluation purposes (Diener and Fujita 1997). A case in point is that people even create imaginary comparison targets to facilitate self-evaluative and motivational purposes (Levine et al. 1993; Wood 1996)! The use of imaginary targets convincingly indicates an active role of the person. Individual choices of comparison standards naturally lead to the second model of social comparison: the coping model. Coping models Coping models of social comparison are mainly used in the domain of health and psychological well-being (Buunk and Gibbons 1997). However, coping with academic challenges and unfavorable comparison and self-evaluation is also a main focus in the realm of educational psychology (e.g., Boekaerts 1993). Broadly defined, coping refers to situations where the challenges one faces seem to exceed the resources available to the person or threaten one’s psychological well-being. Thus, Boekaerts suggested that students 294 Educ Psychol Rev (2008) 20:283–317 always have dual concerns in academic settings, concerns about learning and self-improvement and concerns about subjective well-being. Her theory coincides with self-theory developed in social psychology, which highlights three self-motives: self-improvement, self-enhancement, and self-verification (Banaji and Prentice 1994). Of the three, social psychological researchers used to focus on self-verification and self-enhancement, and not until a decade ago did self- improvement resurface as a major focus (see Collins 1996), since the original, classic theory of social comparison proposed by Festinger (1954) stressed a “unidirectional drive upward” (p. 124) as underlying social comparative information seeking and self- verification. Therefore, in a coping model of social comparison, one would face a dilemma: upward social comparison is an optimal comparison strategy if self-improvement (along with self-verification) is the goal, but it potentially carries hedonistic costs (Brown and Dutton 1995). In contrast, downward comparison is an optimal strategy when self- enhancement is the goal (but see Taylor et al. 1996), but does not carry much informational value other than a comforting feeling that there are people worse off than oneself. Alternatively, people can also diminish and disengage from social comparison to protect themselves, and even downplay the importance of the dimension on which self-evaluation is made (Gibbons et al. 1994). Thus coping models produce a variety of scenarios regarding the fallout of social comparison. Consider successful coping in big-fish-little-pond situations: the person gets a “reality check” and repositions herself or himself in a productive manner (“there are always people better or stronger out there”), maintaining a focus on learning and self-improvement, or even better, uses better-off persons as an inspiration or as models to emulate (Burleson et al. 2005). In this case, the consequences of being in a “big pond” are benign, or even beneficial. There are, of course, many cases of failed coping. For example, for some individuals the “reality check” is truly a “reality shock”: the contrast (self-other differences) is so sharp and the target is so out of reach that assimilation (i.e., identifying better-off peers as similar to oneself) is difficult to achieve. In other words, self-improvement or catching-up does not seem realistic and a sense of inferiority takes hold (Burleson et al.). This is an acute form of the BFLPE, which may lead to lowered self-efficacy, negative affective self-reactions, and lowered educational aspirations (Bandura and Jourden 1991). Of course, effects of upward social comparison can be moderated by self-esteem. Cancer patients with high self-esteem felt better about their chance of recovery when seeing better-off patients, while their counterparts with low self-esteem felt worse after seeing other patients recovering (Buunk et al. 1990). The same upward social comparison produces diametrically different outcomes. When one takes a close look at self-motives and conflicting goals of making social comparison, and positive and negative consequences associated with them, one inevitably comes to a realization that the BFLPE paradigm vastly overlooked the coping aspect of social comparison. Diener and Fujita (1997) compared imposed social comparison models and coping models and concluded that the coping model appears to be more accurate primarily because people are so flexible in natural settings in the information they use to determine their satisfaction. In addition, the coping model recognizes the substantial individual differences that exist in the choice of comparison targets. (p. 330) Motivational models The term “motivational model” is used here to refer to a body of the literature that intersects social comparison and academic motivation. Defined as such, these models overlap with the Educ Psychol Rev (2008) 20:283–317 295 coping models discussed above (e.g., self-improvement vs. self-enhancement motives) but carry different theoretical backgrounds than the coping research and are developed predominantly in academic contexts. Weiner (1990) identified self and self-related motivational processes as the then dominant trend in motivational research in education. Therefore, it should be particularly relevant to the BFLPE hypothesis, which is mainly concerned with academic self-concept. The self in the motivational research involves many self-related constructs, such as self-efficacy, self-expectancy, self-attribution, self-theory, and possible selves. Following the suggestion by William James, the motivation literature distinguishes between an enactive aspect of the self (“I want to do it” or “I can do it”) and a reflective aspect of the self (“What does it say about me?”); constructs of self-efficacy and self-expectancy lean toward the former, and are typically considered situation- or task-specific and measured before performance, and self-attribution and self-concept belong to the latter, typically gauged after performance, and reflect a more enduring aspect of the self (Dai et al. 1998). Rather than studying self-concept as a stand-alone construct, the motivational researchers attempt to situate it in specific performance contexts, in conjunction with other self-related constructs to reveal both the enactive and reflective aspect of the self and their affective and motivational consequences (e.g., Pajares and Miller 1996; see Pajares 1996 for a review). These two aspects of the self are likely to have a reciprocal relationship. Enactive self-appraisal and expectations can influence or bias post-performance self- attributions and self-concepts (Feather and Simon 1971; Pajares and Miller 1996); and the latter, once becoming internalized, off-line tendencies, can also influence and bias situational, on-line self-appraisal and expectations, leading to direct behavioral conse- quences, such as course selection (Marsh and Yeung 1997a). Indeed, in early years, Marsh (1984) envisioned a dynamic equilibrium model in which academic achievement, academic self-concept, and self-attributions for the causes of success or failure form reciprocal relations, such that a change in one aspect leads to changes in the others. Marsh suggested that effects of moving from a lower ability school to a higher-ability one could be differential, some suffering declined self-concept, some developing an urgent need for protecting or maintaining their positive self-concept, and others may seek self-improvement. Thus, social comparison information is only one of many possible factors—new levels of academic challenges, direct and vicarious experiences with the task at hand, and social- evaluative feedback from others—that produce “disturbance” on this dynamic equilibrium. However, this vision has never been materialized in Marsh’s research, surely not in the BFLPE research paradigm. Social-cognitive and motivational underpinnings of self-concept become clearer when one realizes that self-concept formation is fundamentally a causal self-attribution process (i.e., inferring stable personal attributes); therefore, whether a personal attribute is perceived as stable and controllable influences future motivation and behavior (Weiner 1979). For example, Eccles (Parsons) et al. (1984) found adolescents to be more likely to perceive math as taking more ability (i.e., a stable, enduring personal attribute) than English. This finding suggests that students perceive difficulty of domains differentially, and therefore infer ability differentially, depending on domains of performance involved. Although Marsh also emphasizes domain-specificity of academic self-concepts and has engaged in several studies of the BFLPE with a focus on mathematics achievement and self-concept (e.g., Marsh et al. 2007), his domain-specificity argument is mainly based on empirical evidence from factor analysis of self-concepts. The BFLPE, indeed Marsh’s general internal and external frames of reference, theory of self-concept has little to say about why people make differential domain-specific self-attributions. 296 Educ Psychol Rev (2008) 20:283–317 Dweck’s (1999) incremental versus entity theory of ability and self is one of the most influential motivational theories in education. According to this theory, self-attributions people make about themselves and others are partly determined by their implicit theories of whether things are static or dynamic, fixed or incremental. Thus, an incremental view of ability more likely induces mastery efforts and self-improvement motivation, and a fixed view of ability more likely induces a performance goal and, if the expectation is not favorable, an ego-protection motivation. From this point of view, the same social comparison information can lead to completely different construal. Construal of social comparison and self-evaluative information can be geared either toward a task focus and self-improvement, or toward an ability-diagnostic focus on relative ability. In the same vein, Markus and Nurius (1986) argued that people not only form self-representations of who they are, but also envisage who they can be (i.e., possible selves;), a distinct dynamic, future-oriented view of self. Thus, upward social comparison can truly be inspiring (Burleson et al. 2005). Related to this point are individual and cultural differences with regard to inferring invariant attributes from social comparison. To the extent people are predisposed to see personal attributes as either fixed and stable or fluid and malleable, the measure of self-concept may carry different meanings and valences for different individuals and across different cultures (Dai 2002; Dweck et al. 1995; Guimond et al. 2007). Marsh et al. (2004) acknowledged that cultural differences in conceptions of ability may affect the size of the BFLPE, but the BFLPE is robust across cultures. Marsh et al. (1995) also suggest that students with mastery versus performance or ego orientation may more likely use self- improvement as a basis for their academic self-concepts. However, the BFLPE paradigm does not grant a place for a dynamic construal of self to reveal itself. How one construes social comparative information is an issue of self-regulation of cognition and motivation. Part of social and intellectual maturity is to develop internal standards for one’s personal strivings, and to become less dependent on external standards such as social norms. This means that comparison standards will developmentally become more autonomous (Zimmerman and Schunk 2004), more “criterion-based” (i.e., what the person tries to accomplish, regardless of what standards others hold), rather than “norm-referenced,” which is more vulnerable to external disturbance. Indeed, this is the only way a dynamic but stable equilibrium of self can be maintained, not easily perturbed by external stimuli and changes. We have discussed so far why the BFLPE reflects only part of much larger picture of how individuals make social comparison in academic and other settings, how cognitive and social development changes the way social comparison is engaged, how different motivations are involved in making social comparison. In the following section, we will look at the design and methodology Marsh and colleagues have consistently used for decades in studying the BFLPE. Design and data analysis problems with the BFLPE paradigm Since the first study of the BFLPE of which we know (Marsh and Parker 1984) up to the most recent one (Marsh et al. 2007), we have seen a consistent pattern of using large-scale data sets and a standard path model of estimating the BFLPE. Using large-scale national and international data sets has many advantages, such as generalizability of findings and statistical power, among others. Using a standard path model also has many advantages, such as ensuring comparability across studies, and accumulation of evidence over time. However, when it comes to the particular issue of how the BFLPE is estimated, the design and data analysis typically adopted in the BFLPE research reveals several weaknesses and flaws. Educ Psychol Rev (2008) 20:283–317 297 Lack of specification of contexts There is no specification of contexts where the BFLPE is more likely or less likely to occur. The ability of Marsh and his colleagues to identify crucial contextual information is limited because the large-scale national or international data they have used typically do not provide detailed information about contextual information relevant to the BFLPE or self-concept in general, such as salience of social comparison information, degrees of academic competition, or critical transition points, developmentally or contextually. To be sure, using East and West Germany data during the transition period, and the related German educational system (e.g., Gymnasium schools; e.g., Marsh et al. 2001), or identifying a group of students in a gifted program (Marsh et al. 1995) represents good efforts to highlight the importance of contexts or contextual changes. However, most of the BFLPE studies are indiscriminative of contextual features other than school-average ability or achievement, which is typically the only basis for estimating the BFLPE. The levels of “school selectivity” are surely an important index, just like socioeconomic status (SES), different ethnic or cultural groups, or class size. But it is at best what Bronfenbrenner (1989) called a “social address (p. 193).” Relying on school selectivity as a basis for inferring the BFLPE is particularly problematic when we know that what reference group(s) students will use for self-evaluation is necessarily complex and depends on social-contextual, developmental, and personal factors, as we discussed earlier in our conceptual analysis. The dilemma of whether to use standardized test scores or school grades as a basis of student ability and achievement, discussed by Marsh et al. (2007) further reveals the ambiguity as to which index is more accessible and meaningful as a basis for the development of academic self-concept, and a dilemma for selecting proper achievement indices in research. Social comparison is implicitly assumed rather than observed The most problematic aspect of the BFLPE paradigm is that social comparison is inferred, not observed or measured; and explanation and interpretation of data is based on blanket assumptions rather than direct evidence. For example, it is difficult to know whether the BFLPE is due to more downward comparison in less selective schools or more upward comparison in more selective schools. The BFLPE hypothesis makes three default assumptions, namely, that (a) students engage in social comparison using the same frame of reference based on school norms and that (b) levels of engagement in social comparison are comparable across individuals, and (c) the directionality of social comparison is a function of individuals’ ability or achievement in an academic subject vis-à-vis a related school norm. Only by making these assumptions can one interpret the relationship between school-average ability and academic self-concepts to indicate the BFLPE. However, to the extent the social comparison process is implicitly assumed rather articulated and measured, the interpretation of the BFLPE as caused by social comparison is largely a speculation rather than based on solid evidence. The measurement of academic self-concept also exacerbates the problem. When self- concepts in specific academic subjects are measured, one might infer that it is based on their achievement. However, when broad academic self-concept is concerned (e.g., one item in the Self-Description Questionnaire reads “I learn things quickly in most school subjects”; see Marsh 1990), one is less certain that it is a result of social comparison, as people are more likely to give a biased opinion without any social comparison information as a basis (Diener and Fujita 1997). In general, self-enhancement biases are more likely to build into measures of traits that are vague and ambiguous. Statements implicating social comparison can be made without actual social comparison behavior (Wood 1996). Regardless, there is no sure 298 Educ Psychol Rev (2008) 20:283–317 evidence in the BFLPE paradigm that individuals engage in social comparison. An observed correlation between achievement and related self-concept does not make it any clearer. Problems with the statistical procedure used to tease out the BFLPE Marsh and his colleagues have been able to produce a consistent pattern of findings with more than 20 years of research, pointing to the alleged BFLPE (Marsh 1987, 1991; Marsh et al. 2000; Marsh et al. 2005, Marsh et al. 2007; see Marsh and Craven 2002 for a review). In such a case, one could, of course, conclude that the robustness of the findings provides a solid empirical grounding for the BFLPE. Alternatively, however, one could also surmise that homogeneity of the methodology could also produce consistent data patterns, not because they reflect an important aspect of the reality, but because they are the product of its methodology. Of course, truth might as well stand somewhere in-between. However, a close look at the path model (Fig. 1) set up to test the BFLPE reveals a statistical procedure that is disconcerting, to say the least. In most of the studies of the BFLPE conducted by Marsh and his colleagues, the BFLPE is operationalized as the effect of school-average ability (achievement data were often used as proxy measure of ability) on a relevant academic self-concept, along with the effect of individual-level student ability, which is assumed to be the comparative basis for individual self-concept, with individual-level ability (or achievement) also having a causal path onto school-average ability, as in most situations, the school-average ability (or achievement) is an aggregated measure derived from individual-level student ability or achievement; and thus the two are by nature positively correlated. A problem with the path model is that one cannot have any certainty that partialing out, or controlling for, the effect of the alleged “individual-level ability” would not simultaneously partial out positive effects of attending a more selective school, not just in terms of the “reflected glory” or assimilation effect, but in terms of the educational benefits of having a more rigorous curriculum and presumably better instruction on students’ achievement, which is, as we point out above, typically used as a measure of individual student ability, and subsequently their academic self-concept. Indeed one should expect more selective schools to do better in this regard (otherwise why do so many parents pay a fortune to send their children to selective schools?). By partialing out individual student ability or achievement, potential credit due to attending more selective schools is taken away. Not only so, at a technical level, multicollinearity occurs when two predictor variables are positively correlated, leading to unstable parameter estimates or even anomalies. Consider an early study of the BFLPE by Marsh (1987). Individual student ability had a correlation of 0.56 with school-average ability, and a correlation of 0.51 with academic self- concept. School-average ability had a correlation of 0.14 with academic self-concept. However, in the standard path analysis modeling the BFLPE, after controlling for the effect of individual-level student ability, the regression coefficient of school average ability on academic self-concept became negative (β=−0.23). This is a clear case of a suppression effect in multiple regression models because a simple (zero-order) positive correlation turned negative as a beta weight in multiple regression. Similar suppression effects can be seen when Marsh et al. (2000) found that when perceived school status (i.e., a measure of school selectivity) was entered into the standard BFLPE path model, school status had a positive effect on academic self-concept, but there was a measurable increase in the negative effects of school-average ability on academic self-concept. Marsh et al. argued that “[b]y including a separate measure of perceived school status, we partialed out some of the reflected glory effects associated with school-average achievement so that it became a better Educ Psychol Rev (2008) 20:283–317 299 (less confounded) basis for inferring social comparison contrast effects leading to a more negative BFLPE” (pp. 346–347). A problem with such an interpretation is again the suppression effect whereby the already suppressed effect of school-average ability got further suppressed (i.e., more negative) due its positive correlation with perceived school status. Although suppression in multiple regression is a complex statistical phenomenon that takes many forms and is subject to different interpretations (see Cohen et al. 2003), suffice it to say that when it happens, extra care needs to be taken to make sure that a statistical artifact is not taken as self-explanatory evidence for substantive arguments. Marsh (1987) argued that controlling for individual-level student ability is a necessary correction for individual ability in order to detect the BFLPE. Putting aside the issue of whether educational benefits of attending more selective schools get partialed out along the way, it begs the question of what this “negative effects” means. It is incumbent on the researchers to provide a detailed diagnostic analysis and data breakdowns to show that indeed a negative effect of school selectivity (a between-school effect) can be teased out from the positive effects of individual student ability (a within-school effect), rather than relying on a suppression effect to make the case. Effect sizes do not warrant a strong argument for the BFLPE Even if for the sake of discussion we disregard the aforementioned problem, an effect size of −0.20 or even −0.25 (one can visualize a scatterplot) does not make a compelling case. One can imagine that some data points on the scatterplot strongly indicates the BFLPE, but there are many individuals who deviate from the prediction, as the prediction is far from perfect. These cases of deviation should not be just taken as “statistical error” but as cases of not conforming to the BFLPE prediction. Another reason that the effect size does not make a compelling case is that a predictive relationship cannot be equated with a causal relationship based on correlational data. Thus, the predictive efficacy of school-average ability on academic self-concept should not be seen as automatically supporting a causal interpretation of the BFLPE nature. The relatively low effect sizes of the BFLPE typically found using the BFLPE methodology, even taken as a valid measure, suggest that between being in a higher ability school (or other settings with academically competitive peers) and academic self-concepts lie a host of intervening factors moderating and mitigating the alleged negative effects of school selectivity on academic self-concepts. The coping and motivational models discussed earlier have much bearing on this issue. Marsh seems to be satisfied with the effect sizes he has consistently obtained and argues for the robustness of the BFLPE. While the robustness of an effect is clearly desirable and comforting in science, we suggest that specification of where (e.g., under what circumstances), when (e.g., at what juncture of one’s life), and who (e.g., what kinds of individuals are more susceptible to the BFLPE) in the methodology would increase the predictive efficacy of the model, permitting a stronger theoretical argument based on stronger quantitative evidence. Summary To summarize this critique of the methodology of the BFLPE paradigm, the contexts that produce the BFLPE are underspecified in most BFLPE research (with a few exceptions; e.g., Marsh et al. 1995); the processes of social comparison (e.g., making self-attributions in light of social comparative information) mediating self-concept formation are assumed rather than observed; and the statistical model that operationalizes and estimates the BFLPE 300 Educ Psychol Rev (2008) 20:283–317 is open to serious challenges. The issues of causality are not adequately addressed in the research design. Marsh and his colleagues acknowledge on many occasions that due to the correlational nature of the data, causality and directionality of the causal relationships cannot be determined. Therefore, using a stringent criterion, the findings based on the path model of the BFLPE are suggestive rather than definitive (let alone conclusive). To enhance the internal validity of the BFLPE claim, more rigorous procedures need to be used. Although Marsh and his colleagues occasionally used alternative procedures (Marsh et al. 1995, 2000) that provided some convergent operations and innovative procedures and measurements, more needs to be done to improve the design and diversify the methodology. Indeed one might argue that except for Marsh’s laudable efforts to generalize the BFLPE across gender, ability, and culture, the research program has provided fewer additional insights regarding the BFLPE than should be expected since its inception (e.g., compare Marsh and Parker 1984 and Marsh et al. 2007). Problems with the policy argument based on the BFLPE research Marsh et al. (2004) argued that the robustness of the BFLPE across the board is sufficient for a strong policy message regarding the advantages and disadvantages of attending a selective school or advanced programs. Dai (2004) argued that in order to make the practical implications of the BFLPE clear, the question of when and for whom it is likely to occur should be answered. These competing arguments represent very different assessments of the practical implications of the BFLPE. While acknowledging that the BFLPE is probably real and may take a toll on one’s self-concept and motivation, we argue that it needs to be put in perspective rather than blown out of proportion. A one-sided argument based the BFLPE can be misleading in the following ways: (a) It creates an impression that going to more selective schools or programs carries more disadvantages than advantages in terms of academic self-concept and motivation. The issue really depends on two factors based on the coping model discussed earlier: possible educational gains (including enhanced achievement and enhanced academic self-concept, which has strong empirical support; see Caslyn and Kenny 1977; Marsh and Yeung 1997b) and possible concerns about one’s capabilities in light of social comparison or a situationally induced sense of inferiority. There are individual differences in terms of what kinds of educational and social environments constitute a sufficient, but not overwhelming, challenge, and how individuals cope in social situations while trying to maximize their educational gains. For some, it is a trade-off, and for others, coping will solve the problem (after all, there will always be bigger fish out there). This is why we argue that specification of where, when, and for whom provides crucial information for determining whether an individual is fit to take on the challenge of attending a selective school or program. (b) The BFLPE model also sends out a message that besides “reflected glory,” being with equally or more competent peers than oneself always hurts one’s academic self- concept if one does not compare favorably with peers. The theory neglects the benefits of being with more competent peers, such as observational learning or using them as models or inspiration for self-improvement. Even when the BFLPE does occur, it may not be all negative, as it helps individuals get a “reality check” that is informative and helpful in the long run. The reality check after moving to a bigger pond could be painful to a fish, sometimes with a deflated self-image, but one would argue that it is better than to stay in a little pond and enjoy one’s ignorance and complacence. Educ Psychol Rev (2008) 20:283–317 301 (c) An argument purely based on the BFLPE without qualification is in effect advocating a zero-sum game, leading to a path of no real practical solution. As one reviewer points out, if, to avoid the BFLPE, higher ability students were to decide to choose less competent or average ability ones as their peers (e.g., go to less selective schools or colleges), what would that do to their less competent peers? Some of the big fish there now suddenly turn small or median! Should they move to schools with still less competent peers? Where is an equilibrium that would satisfy everyone? For general education, Marsh et al. (2007) seemed to suggest that heterogeneous grouping would solve the problem. But would the contrast effect not even get greater in that arrangement? For special education, Marsh and his colleagues cautioned about putting students with special education needs in mainstreaming or full inclusion situations, yet do these students not know why they ended up in a self-contained classrooms? For good or ill, social comparison is integral part of distinguishing oneself from others, and its effects on self-concept, academic or otherwise, will sometimes be negative. While we agree that parents should be advised about a possible “reality shock” if they decide to send their children to selective schools, at the policy level, the practical utility of the BFLPE theory is limited, for decisions about appropriate placement should be made on an individual-by-individual basis. Where Do We Go from Here? A critical review would serve no good if it cannot build upon and extend the existing research, provide rich heuristics for future research, and point out possible avenues one might take. In the following concluding section, we attempt to provide a picture of how the BFLPE research can be broadened and refined in light of the relevant extant traditions of research and the inter-theoretical dialogue this review has attempted to promote. General issues and principles guiding future research A major issue revolving the assumption underlying the BFLPE is whether social comparison is imposed by an impinging environment, which carries information too compelling to ignore, as is presumably the case in many academic contexts; or whether social comparison is self-engendered by individuals who are motivated to make social comparison for specific self-evaluative or other adaptive purposes, and as a result wittingly or unwittingly pick specific comparison targets or particular comparison information. Realistically, two types of social comparison (e.g., how they compare with most others, and with whom they prefer to compare) can co-exist (Blanton et al. 1999). But we can posit a general rule for which one would be predominant, namely situational forces or individual dispositions. To the extent social or cultural norms are salient and dominating individuals’ strivings and agendas, imposed social comparison is more likely to be predominant. To the extent multiple, competing values and norms exist, there will be more individual latitude of selectively attending to and interpreting social-evaluative or comparative information, hence a more salient role of individual differences. On the end of social context, it is important to identify cultural differences that regulate social and self-evaluative cognition; for example, collectivist versus individualistic cultures, interdependent versus independent construal of self (Burleson et al. 2005) or the correspondence bias in making causal attribution (Choi et al. 1999). On the end of individual differences, it will be informative to 302 Educ Psychol Rev (2008) 20:283–317 examine for whom engagement in social comparison information seeking would be of particular informational and motivational value. Studying this way, we might be in a better position to tell when the BFLPE prevails regardless of individual dispositions and when coping and motivational models of social comparison better explain the data. More specifically, the future research would do well to define and measure the social comparison processes mediating academic self-concept. For example, a new model could incorporate social comparison processes both as a mediator and as a moderator. As a mediator, social comparison explains how the BFLPE is cognitively mediated. As a moderator, depending on what kinds of social comparison one is engaged, the effects may be completely different. For those who report using a local norm, for example, achievement and academic self-concept will be more closely related than for those who do not use a local norm, but instead highlight the reflected glory as a self-enhancement strategy, or those whose purposes of social comparison is geared more toward self-improvement. More generally, we need to better delineate how self-concept is formed and changed, and what tends to maintain its equilibrium (e.g., consistency and coherence) and what tends to change its equilibrium (e.g., conflicts and discrepancies; see Pervin 2000). Thus the former contributes to assimilation effects and the latter to contrast effects on self-concept. This implies the researchers take both macro-developmental and micro-developmental views of how one perceives himself or herself, both in terms of “who I am,” and “what I can be” (i.e., possible selves; Markus and Nurius 1986) given new opportunities and challenges in development. Methodological considerations and strategies Although we have mentioned earlier that the default assumption of social comparison without evidence is the most problematic aspect of the BFLPE paradigm, in a larger scheme of things, we also view an exclusive use of a nomothetic approach to the self-concept research as a limitation of the BFLPE paradigm. Nomothetic approaches to psychological phenomena start with universal assumptions and use deductive logic to deduce specific consequences of some generality. Thus, the BFLPE can be naturally deduced from the assumption that people use their local frame of reference as a basis for their academic self-concept. In contrast, idiographic approaches start with particulars and emphasize unique individuality in how they construe social information (Allport 1962). The nomothetic versus idiographic orientation in many ways determines our methodology. We suggest that, when used properly, the two approaches can be complementary. Specific to the BFLPE research, we propose three levels of description or analysis. At a nomothetic level, we suggest using individual growth modeling (IGM) as an alternative to multi-level modeling, because individual growth modeling captures individual differences in self-concept stability and changes, and directionality of the changes, if any (Willett and Sayer 1994; see also Singer and Willett 2003). Technically, it requires three waves of self-concept measures obtained at some critical junctures of one’s life (e.g., the entry to a selective school). The IGM provides a measure of individual differences in self- concept at a baseline (intercept differences), and a measure of individual differences in self- concept trajectories or slopes (change upward, downward, or no change). Various antecedents such as gender, self-esteem, or the degree or kind of social comparison made, can all serve as predictors. When multiple schools with distinctly different cultures are studied, multi-group modeling can be run to see whether parameter estimates remain unchanged or reflect significant moderation effects. At an idiographic level, we suggest conducting intensive case studies of individuals who are in various critical developmental transitions or facing life tasks such as Educ Psychol Rev (2008) 20:283–317 303 dealing with more competitive peers than they used to. One can argue that the more complex the statistical maneuvering based on some nomothetic assumptions, the more difficult the interpretation that ties statistics back to the immediate phenomenology under investigation. These idiographic accounts of individuals’ construal and con- structions of self will provide a more intimate picture of success and failure of coping and motivation in the face of challenges, the BFLPE being one of them. The BFLPE research mainly used Marsh’s (1990) self-report measures of academic self-concepts (different versions of the SDQ), with a forced choice format (i.e., Likert-scale) based on the nomothetic assumption of academic self-concept as a trait or trait-like quality. In contrast, the research on “possible selves” uses a more idiographic, open-ended format (Cross and Markus 1991), which is more likely to generate authentic accounts of how individuals think about themselves and their own identity. Rather than isolating academic self-concept as a stand-alone entity, the idiographic approach puts it back as part of the selfhood or personhood, rendering academic self-concept meaningful only in that personal context. A middle ground between the nomothetic and idiographic approaches can be found in latent class analysis (Muthén and Muthén 2000) that combine person-centered and variable- centered approaches. This non-parametric method is similar to cluster analysis in that it can identify homogeneous subgroups; thus, who suffered a decline in academic self-concept and who did not can be identified and characteristics of each group further investigated. However, different from traditional cluster analysis, the latent class analysis also permits detection of latent trajectory classes. Thus, for example, those who have a longitudinal pattern of instability can be differentiated from those who have a longitudinal pattern of stability in self-concept. This approach avoids the nomothetic tendency to isolating variables as if these variables function independently as entities; on the other hand, it also avoids the idiographic tendency to get lost in the details of numerous individual accounts and fails to identify regularities and principles. Conclusion To sum up, the research that modifies the BFLPE paradigm should allow us to identify situational circumstances and distal and proximal predictors of the BFLPE, and related mediators and moderators. In other words, it should address the issues of where, when, and for whom, as well as how and why it happens. It should facilitate identification of vulnerable groups of individuals who are susceptible to the BFLPE, and facilitate placement decisions and guide educational interventions. Only then does the practical utility of BFLPE become clear. But more importantly, research that goes beyond the BFLPE paradigm should allow us to see, given new life challenges and tasks, academically and socially, how one negotiates (or fails to negotiate) a path of optimal educational development that is maintaining a positive yet realistic self-concept and outlook for one’s future, but also capitalizing on one’s opportunity to learn and grow, and how one’s coping and self-regulatory skills develop through cognitive and social development. The BFLPE is only part of this much larger story about personal and academic development. Acknowledgments We would like to thank Dr. Herbert Marsh and four anonymous reviewers for their constructive feedback on the previous version of this article. We also thank Drs. Robert Pruzek and John Willett for their expert opinions on technical aspects of methodology and statistical analysis related to the current topic. 304 Appendix A Table 1 A Summary of Research that Supports the Big-Fish-Little-Pond Effect (BFLPE) Publication Research focus Data sources/ setting/sample Design and data analysis Results Marsh et al. Long-term BFLPE on Over 2,000 students from 147 Individual- and school-level BFLPE present 2 years (2007) math self-concept and German high schools (study 1) multi-level modeling of effects beyond high school graduation negative related academic followed 2 years beyond of achievement on academic (study 1) outcomes graduation and 4 years beyond self-concept, measured by SDQ (study 2) Use of both standardized test Gender somewhat moderates the scores and school grades effects, with stronger effects for girls Modeling long-term effects of BFLPE as predictive efficacy of school-average achievement school-average achievement or school type on math self-concept ranged from −0.02 to −0.28, with a tendency of lesser effects over time (study 2) Margas et al. Are the BFLPE and 46 physically gifted students Questionnaires, correlational, Different variations of self-evaluations (2006) Social Identity Theory participating in a high ability 1 year longitudinal but same final self-evaluation (perceived social status, class and in regular classes Social identification measured High-ability class—strong positive self-categorization) twice, prior to entry into the social identification effect at first complementary? program and at the end of the study but decreases over time. Self- categorization mainly explains the variation in self-evaluation Altermatt and Achievement-related 929 children from 5th, 6th and Report card grades of children Low achievers who established and Pomerantz implications of establishing 7th grade, in the midwest and their friends were obtained and maintained friendships with high- (2005) children completed achieving friendships with high- Questionnaire measures of their friends evaluated themselves less achieving versus low- self-evaluative beliefs and preference positively, but performed better achieving classmates for challenge academically, Educ Psychol Rev (2008) 20:283–317 Table 1 (continued) Publication Research focus Data sources/ setting/sample Design and data analysis Results than low achievers with similarly low- achieving friends. This tradeoff is less obvious for high achievers Lűdtke et al. BFLPE based on class-average A German sample of 2,150 HLM, multi-level analysis Attending academically selective (2005) mathematics of lower secondary students educational programs has a negative achievement from 112 classes who effect on academic self-concept. participated in the TIMSS. Effect grew over time. Negative Mathachievement and math effect of school-average ability self-concept data; video data on was similar in size for relatively teachers’ frame of reference more and less able students Teachers’ role in self- (TFR) Mathematics self-concept was The extent to which teachers adapt Educ Psychol Rev (2008) 20:283–317 concept development measured using a 4-item scale individualized frame of reference has that is written in German and positive effects on self-concept commonly used to measure self-concept with German-speaking samples Marsh et al. Juxtapose self-concept with German 7th-grade students (study Structural Equation Modeling of two Prior self-concept significantly affected (2005) academic interest 1: N=5,649 and study 2: waves of data (Time 1 and Time 2) subsequent math interest, school N=2,264) grades, and standardized test score Particularly on self-concept, Self-concept was measured using a Prior math interest had only a small interest and achievement in German adaptation of the SDQ-III effect on subsequent math self-concept math. Integrates self- concept with the developmental motivation literature Marsh and Hau Test of cross-national Approx 4,000 Three-level multilevel model in The effects of school-average (2003) generalizability which students (Level 1) were nested achievement on academic within schools (Level 2) and schools self-concept were negative were nested within countries (Level 3) 15-year-olds from each of the 26 Academic self-concept was measured (M=−0.20) in 26 countries countries (N=103,558) using the three best items from the academic self-concept subscale of the SDQ-II 305 306 Table 1 (continued) Publication Research focus Data sources/ setting/sample Design and data analysis Results Catsambis et al. Compared the social NELS 1988 data; 1,052 Correlations, and regression There was significant differences in (2001) attributes of similar male eight-grade schools and 24,500 analyses students’ attitudes towards and female students who students Mathematics self-concept was mathematics, educational aspiration, attend schools with tracked measured by asking students if self-esteem, locus of control, and versus untracked they “looked forward to mathematics engagement in school. When males mathematics classes class, whether they were afraid to were grouped with peers of similar ask questions in mathematics class, and high ability in a subject area whether mathematics would that defines their competence, they be useful to their future” (p. 91) seem to lose their competitive edge Marsh et al. To determine the size and 2,778 seventh graders in 161 Longitudinal data, multilevel modeling Negative “big-fish-little-pond effect” (2001) direction of the BFLPE classrooms were tested at three (BFLPE) measurement points: at the beginning, in the middle, and at the end of the 1991/1992 school year Reunification and differences in Mathematics self-concept was measured Attending classes where class-average East and West German Schools using a 4-item scale that is written in math achievement was higher led to German that was also used by Lűdtke lower math self-concepts et al. (2005) East German students had significantly lower academic self-concepts than did their West German counterparts Tymms (2001) Investigations of feelings Data from the PIPS Project which Multivariate multi-level models Positive feelings were predicted by toward math, reading, and tracks the academic progress of Class as a unit for modeling BFLPE individual achievement but negatively school of 7-year-old pupils pupils from the age of 4 to 11 Feelings toward math, reading, and predicted by average class attainment school were measured by students’ (effect sizes=−0.06 to −0.11) responses to a series of statements that reflect a hybrid a attitude and self-concept measures Educ Psychol Rev (2008) 20:283–317 Table 1 (continued) Publication Research focus Data sources/ setting/sample Design and data analysis Results Marsh et al. Juxtaposition of Big-fish- 7,997 students, 44 high schools, A measure of perceived school status Higher school-average achievements (2000) little-pond effect (BFLPE), 4 years in Hong Kong to estimate “reflected glory effect” led to lower academic self-concepts and “Reflected glory effect” (contrast effect) Longitudinal multi-level path models; Higher perceived school status had a academic self-concept was measured counterbalancing positive, “reflected- using a Chinese adaptation of the glory effect” on self-concept academic self-concept subscale of (assimilation effect) the SDQ-II Assimilation effect is weaker than contrast effect based on s

The Big-Fish-Little-Pond Effect: What Do We Know and Where Do We Go from Here? PDF

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue