Compensation in intercropping was examined in a field experiment at Norwich, U.K., by simulating failure of one component at different stages of growth. All two-crop combinations of mustard, oat, and bean were examined in comparison with their sole crops. In the absence of failure, only the oat-bean combination gave an advantage over sole cropping (LER = 1.39). For all combinations, if failure of either component occurred at the early or medium stages of growth, or if the less competitive component failed late, the surviving component compensated and intercropping was higher yielding than sole cropping. However, if the more competitive component failed late, the suppressed surviving component was unable to compensate sufficiently and sole cropping was higher yielding. Compensation was dependent on the nature of the surviving component but was not simply a reflection of its competitive ability in the absence of failure.