1. A random bred population of mice was subjected to inbreeding and the changes of litter size, measured as the number of live young in first litters, were followed.

2. The mean litter size declined at a rate of 0·56 young per 10% increase of the inbreeding coefficient.

3. Selection for large litters within the lines during the inbreeding did not effectively reduce the rate of decline.

4. Out of twenty lines at the beginning of the inbreeding seventeen were lost by the time the inbreeding coefficient reached 76%. Two more were lost later and one survived indefinitely. The three lines that survived longest started at a level below the mean and did not decline in litter size. The one that survived indefinitely reached 99% inbreeding without dropping below the non-inbred control.

5. The three lines surviving at 81% inbreeding were crossed and the litters produced by the crossbred progeny were larger than the non-inbred control by about two young per litter. This gain from heterosis is attributable to selection among the lines on their performance as inbreds. A second and third cycle of inbreeding and crossing yielded no further progress, and the level of the first cross was never regained. This is attributable to the ineffectiveness of the selection applied and to the previous restrictions of the population size.

6. The behaviour of the lines in the inbreeding and crossing point to simple dominance rather than over-dominance at the loci causing variation of litter size.

7. This experiment suggests that, as a means of improvement of farm animals, cyclical inbreeding and crossing does not look very hopeful.