A ‘Russet Burbank’ potato crop was subjected to 14 irrigation treatments in a rainshelter experiment at Lincoln, New Zealand in the 1987–88 growing season. In 12 treatments, plots were trickle irrigated and covered automatically by the shelter whenever rain fell, so that precise responses of yield and quality to water deficit timing and severity could be obtained. These treatments ranged from overwatering to withholding water to develop maximum potential deficits of up to 138 mm. The water stress treatments were applied either ac tuber initiation, early tuber bulking, or late tuber bulking (assessed by sampling spare plants). Except when fit was raining, these plots were exposed to prevailing climatic conditions. The two remaining treatments, which were outside the rainshelter, checked for direct effects of the shelter on the crop. One treatment was irrigated to achieve the same deficits as one of the treatments inside the rainshelter, and the other was not irrigated. Highest yields of process grade tubers (63 t/ha} were obtained from the overwatered plots. Among the drought-stressed treatments, the highest yields were 58 t/ha of process grade tubers, obtained from plots subjected to drought during tuber initiation; these had fewer bit larger tubers. Drought during late tuber bulking reduced process grade yields by 20%, with both tuber size and tuber number being reduced by up to 20%. Non-irrigated rainfed plots outside the shelter had low yields and small tubers because the season had half the average rainfall. The levels of water deficit imposed had no effect on tuber processing quality.