A bit time estimator which uses adaptive filtering techniques is presented. The filter weights of an adaptive linear predictor are shown to provide a reliable estimate of the bit time T of a random binary square wave contaminated with additive white Gaussian noise, with little or no a priori information. The quality of this estimator is then evaluated via the least mean square algorithm, and a comparison is made between it and a more conventional estimator based on a zero crossing detector. This comparison shows that an adaptive estimator based on a linear predictor is generally superior. >