The linear regression method of Okubo and Ebbesmeyer (1976) has been commonly used to compute velocity gradients and residual motion. However, Okubo and Ebbesmeyer did not properly address the question of how estimates of velocity gradients and residual velocities are biased by using a cluster consisting of a small number of drogues. Since this problem has caused some confusion in the oceanographic literature, we will address it here.