The gut contents of larvae at 48, 96 and 144 hah were significantly dependent on the feeding level (Pb0.0001) but independent from fish density (P of 0.1491, 0.0634 and 0.1766) and from the interaction between the two variables (P of 0.1235, 0.3420 and 0.0680). At all sampling dates, the gut contents of larvae that received 1 RFL were
close to those predicted by the model that was used for calculating the reference feeding level (Fig. 4). By contrast, larvae that were offered larger amounts of food ate more than predicted. At 48 hah, the gut contents of fish fed 3 to 81 RFL were about three times higher than predicted by the model. Differences between predicted and observed food intakes at 3, 9 and 81 RFL persisted in older fish, because fish grew faster in this experiment than in the feeding trials thatwere used to elaborate the feeding schedule. At almost all fish densities, the coefficient of variation of food intake was reduced when the feeding level was increased (Fig. 4).