The task was to turn a crank so as to keep a pointer aligned with a target which moved irregularly over a 120[degree] sector on the circumference of a dial. Any one of four crank-turning rates could be required for accurate pursuit, ranging from Rate 1 (slowest) to Rate 4 (fastest). By use of five matched groups, scores made at Rate 1 without training were compared to scores made at Rate 1 after training at Rates 2, 3, or 4; and scores made at Rates 2 or 4 without training were compared to scores made at those rates after training at Rate 1. Significant positive transfer was obtained from Rates 2, 3, and 4 to Rate 1; and from Rate 1 to Rate 2. Relative amount of transfer correlated positively with inter-task response similarity. Greatest transfer was obtained between tasks not differing in difficulty, but more transfer resulted when going from a difficult to an easy task than vice versa. (Rates 1 and 2 were equally difficult, but faster response rates were increasingly more difficult.) The results contradict the Wylie hypothesis of transfer and are consistent with,.but not completely explained by, a response generalization theory.