The purpose of this study was to examine the effect of regulation of work rate, computer controlled versus controlled by the subject, on the relationship between work rate, freely chosen pedal rate (FCC) and gross efficiency. Eighteen male cyclists participated in the study. One group, freely cycling (FC) on a competition bike mounted on an electromagnetic roller, could use gearing and cadence to achieve each work rate. The other group (EC) was cycling on an ergometer which enables a constant work rate, independent of cadence. Subjects performed an increasing work rate protocol from 100 W up to exhaustion. We found a strong interaction between group and work rate on cadence (P < 0.001). In the FC group, work rate affected cadence (P < 0.001), increasing from 72 rpm at 100 W to 106 rpm at 350 W. For the EC group, no work rate effect was present (average FCC 92 rpm). Gross efficiency increased with work rate for both groups. The efficiency–cadence relationship was strongly affected by the protocol. At a given work rate, very similar efficiency values were obtained at highly different cadences. The discrepancy in the FCC-work rate relationship between the EC group and the FC group may be related to the manner in which one can regulate work rate. FCC depends not only on work rate but is also affected considerably by the manner in which the work rate can be controlled by cadence. This finding may have important implications for the interpretation of the preferred pedaling rate, especially how this is related to optimizing metabolic cost.