In order to make the earliest possible prediction of the type of woman likely to develop pregnancy-induced hypertension (PIH), one hundred and ninety-six primigravidas underwent a roll-over test (ROT) during wk 28-32 of their pregnancy. Blood pressure (BP) readings were taken with a standard 12 cm cuff as well as with cuffs adapted to various arm circumferences. We found that the prediction rate of ROT readings with a standard 12 cm cuff was relatively low (38.5%) as compared with Gant's study (94%) (Amer. J. Obstet. Gynec., 120 (1974) 1). When a suitably sized cuff was used, the prediction rate dropped (to 14.7%). Data analysis at term for the whole population of this study shows that, by measuring with a standard 12 cm cuff, 10.2% of the women were found to have PIH, whereas measuring with a suitable cuff showed PIH in only 2.55% of the cases (1:4 ratio). We suggest that the low prediction rates in this and other studies demonstrate that the ROT test is not sufficiently reliable as a tool for predicting which women are liable to develop PIH, but there is definitely enough in it to predict which group will not develop PIH (in this study 89-93%). It has been recommended that ROT be considered only as a test of possible reliability. It should be done according to proper criteria for BP measuring, and a repeat ROT should be considered after several days before starting any kind of treatment.