An analysis is presented of a sampled-data parameter-adaptive control system. The accuracy in the region of optimum is seen to be adversely affected by fluctuations in the measured performance index owing to short-term smoothing. A normalised performance index is proposed to reduce these fluctuations. This index is largely insensitive to fluctuations in the control-system input. Its use also ensures that the gain of the parameter-adjusting loop does not depend on the amplitude of the control-system input.The parameter-adjusting loop, a stepping system, is incorporated in a digital computer. Provision is made for a dead space in the parameter loop to reduce the likelihood of false steps near the optimum parameter setting. A steady-state solution of the parameter loop equations indicates the dependence of parameter loop accuracy on step length, dead space and the variance of the performance-index fluctuation.