This study reports data extending work by Marsh and colleagues on the "big-fish-little-pond effect" (BFLPE). The BFLPE hypothesizes that it is better for academic self-concept to be a big fish in a little pond (gifted student in regular reference group) than to be a small fish in a big pond (gifted student in gifted reference group). The BFLPE effect was examined with respect to academic self-concept, test anxiety, and school grades in a sample of 1020 gifted Israeli children participating in two different educational programs: (a) special homogeneous classes for the gifted and (b) regular mixed-ability classes. The central hypothesis, deduced from social comparison and reference group theory, was that academically talented students enrolled in special gifted classes will perceive their academic ability and chances for academic success less favorably compared to students in regular mixed-ability classes. These negative self-perceptions, in turn, will serve to deflate students' academic self-concept, elevate their levels of evaluative anxiety, and result in depressed school grades. A path-analytic model linking reference group, academic self-concept, evaluative anxiety, and school performance, was employed to test this conceptualization. Overall, the data lend additional support to reference group theory, with the big-fish-little-pond effect supported for all three variables tested. In addition, academic self-concept and test anxiety were observed to mediate the effects of reference group on school grades. Copyright 1999 Academic Press.