The dynamic performance of tractor-implement combinations is considered theoretically in terms of the changes in load, weight transfer, and forward speed arising from variations in implement working depth. Two tractors, one having a top-link sensing control system and the other having a lower-link sensing control system, were modified so that the driving wheels ran eccentrically, imparting a sinusoidal variation in working depth to mounted implements and field tests were carried out using mouldboard and chisel ploughs. The draught and vertical forces between the tractors and implements were recorded continuously during the test runs, together with implement working depth, tractor engine speed, and forward speed. Measured variations in draught correlated well with the variations predicted from laboratory measurements of the tractor implement-control system characteristics. It is concluded, therefore, that field performance can be predicted from standard test data if such measurements are incorporated. The theoretical analysis is used to suggest optimum parameters for implement-control systems. With fully-mounted implements, no significant difference was found between top- and lower-link sensing systems which were otherwise similar. Chisel ploughs were shown to be more difficult to control than mouldboard ploughs. Higher sensititivies would be required to retain control at higher working speeds and also for operating semi-mounted implements with lower-link sensing systems. Semi-mounted chisel ploughs are likely to be more easily controlled than fully mounted ones, but semi-mounted mouldboard ploughs would be slightly more difficult to control than fully mounted ones. A small dead-band in the response characteristics of implement-control systems is shown to have negligible effect on performance, and current maximum rates of lift are likely to be adequate for control purposes.