We tested an energy intake rate maximisation hypothesis to predict preferences. In experiment 1, five groups of INRA401 dry ewes (G1-G5) grazed down conterminal mono-cultures of Lolium perenne (L)-Festuca arundinacea (F), 0.50:0.50 by area, at a stocking rate of 92 ewes ha(-1). Both plant species were 220-mm high at the beginning of grazing. Time spent grazing either L or F was measured daily from dawn to dusk in groups G1, G2, G4 and G5, composed of six ewes, using automated cameras. In experiment 2, intake-rate from L and F was measured over two periods (P1 and P2) on a group of 8 dry INRA401 ewes (G3), using the short-term double weighing technique. The four ewes that grazed L in P1 grazed F in P2 and vice versa. Sward height at the beginning of grazing was 220-mm high for both plant species in P1, and 195 and 156 mm for L and F, respectively, in P2. Time spent grazing either L or F was measured during P2. During the first 5 days of the grazing down, ewes spent on average 71% of their grazing time on L. A marked switch to F occurred on day 6 on average. Thereafter, there was no more preference for either species. At the time of the switch and thereafter, mean sward height and green leaf mass were, respectively, 25-30 mm and 50% lower for L than for F, whereas they were similar at the beginning of the depletion. From the switch to F onwards, the decrease in preference for L was mediated via an increase in the time spent grazing F, rather than a decrease in the time spent grazing L. For a given sward height, DM and DOM intake rates (DMIR and DOMIR) were lower from L than from F in both periods. Ratio of DOM intake rate between L and F (DOMIRL/DOMIRF) at the beginning of grazing was 0.74 and 1.08 on average for P1 and P2, respectively. Two days before the switch, it decreased to 0.59 and 0.70 in P1 and P2, respectively. After the switch, it remained at 0.60 and 0.73 on average in P1 and P2, respectively. These results demonstrate that diet selection cannot simply be explained by a strategy that maximises energy intake rate. We observed a marked interindividual variability in preferences. The degree of preference for L was repeatable between years (P < 0.05) in all but one ewe. (c) 2005 Elsevier B.V. All rights reserved.