In many man-machine systems an operator must re peatedly predict which one of a group of increas ing or decreasing functions will be most critical at some time in the future, or which one will reach a critical value first. The present study was designed to investigate factors bearing on human performance in such a sequential decision making task. Each subject prepared graphs of each of four separately identified sequences of numbers presented to him four at a time, one for each graph. Each of the four number sequences increased linearly, but at different rates, and each increasing sequence was obscured by random variations. Both the rate of increase and the variability of the sequences were experimentally controlled. The subjects' task was to predict, as early in the sequence as possible, which one of the four sequences would attain the largest value at the end of 100 presentations. TE SUB JECTS' PERFORMANCE WAS BEST WHEN THE DIFFERENCE BETWEEN THE RATES OF INCREASE OF THE FOUR NUMBER SERIES WAS LARGE AND WHEN THE VARIABILITY WAS SMALL; HOWEVER, INCREASING THE DIFFERENCE BETWEEN THE RATES OF INCREASE OF THE FOUR SEQUENCES IM PROVED PERFORMANCE MORE THAN DID REDUCING THE VARIABILITY. (Author).