Key pecking by pigeons was maintained by arithmetic progressive-ratio schedules of food delivery. Successive conditions arranged different step sizes, and each condition remained in effect until behavior appeared stable. Each session continued until a period of time passed in which no key pecks were recorded (the break-point criterion); both a 5-min and a 15-min criterion were tested across a range of step sizes. Average breaking points (i.e, the largest ratio completed) were relatively unaffected by step-size magnitude, whereas the average number of ratios completed and average response rates generally declined across increasing step sizes. Within sessions, preratio pauses were relatively short and fairly constant in duration as the ratio increased; pause durations increased rapidly near the end of a session. The relation between the average number of completed ratios and step size was described well by a power function [y = b(xa), in which y represents the average number of completed ratios, x represents the step size, and a and b are fitted parameters]. Increasing the break-point criterion from 5 to 15 min resulted in increased values of b, whereas parameter a was relatively unaffected and was close to -1 (consistent with the lack of effect of step size on breaking point). This function also provided an excellent description of data drawn from previous reports.