In order to study the effect of rearing juvenile halibut in different social environments, individually tagged juvenile halibut were size-graded into two size classes (Large, L, and Small, S) with ungraded fish as control. After ca 6 weeks, the two sizegraded groups were again graded into two size classes creating four experimental groups: Large of the Large (LL), Small of the Large (SL), Large of the Small (LS), and Small of the Small (SS). Grading (overall mean of the four grading groups) improved growth rate by 10% compared with ungraded controls, but the effect was also significantly affected by social environments, because in the latter half of the experiment overall growth was improved by 11 and 12% in the two groups with larger size variation (i.e. SL and LS, respectively) compared with the two other groups (i.e. LL and SS). Significant size rank correlations were maintained during the experiment, these were higher in the ungraded (Control) group and the SS and LL groups than in the SL and LS groups. Further, the degree of mean rank position changes varied between the experimental groups and was higher in the SL (20.7) and LS (25.6) groups than in the Control (10.5), LL (15.1), and SS (15.4) groups. This could possibly indicate a stronger social hierarchy in the last three groups. Growth rate differences may be the product of different degrees of interactions among individuals, and based on the higher overall growth rates in the groups with larger size variation (i.e. SL, LS) it is concluded that juvenile halibut should not be too intensively size graded. Imsland and Jenssen have contributed equally.