Selection was practised for improved feed efficiency (gain/feed intake) of mice on two alternative feeding regimes. In one set of lines animals were fed ad libitum, in the other set they were individually fed a fixed amount of feed (about 10% below the control ad libitum intake) which was not changed over generations. For each treatment, a pair of replicate lines (E) were selected on efficiency from 3-5 weeks of age for 8 generations and another pair (L) from 5-7 weeks for 7 generations. A control line was maintained for both E and L lines. In terminal generations mice from each line were tested on each feeding regime, and carcasses of ad libitum fed mice were analysed.The realized heritability (within families) for efficiency averaged 13%, without much variation over treatments. In the E lines efficiency increased by about 18% of the control mean and in the L lines by about 60%, although absolute changes were small, and responses were similar on the two feeding regimes. Weights at the start of test decreased in the E lines and increased in the L lines; weights at the end of test increased in both.When tested on the alternative regimes, no interactions were detected for live weights, weight gains or efficiency; selection under fixed intake led to the same increase in appetite as did that under ad libitum.There were no interactions for carcass composition. Selection for efficiency led to an increase in fatness on both selection regimes and both weight ranges.