The paper describes the development of a relatively robust estimator based on the Kalman filter and the useof this estimator to detect malfunctions in the instruments of a control loop. The technique involves tracking the states of the measurement and valve position of the working loop and monitoring to detect the development of bins caused by instrument malfunction. It is obviously desirable in such an application to UBC a relatively robust estimator. The Kalman filter is a powerful technique and a natural choice, but there are certain problems in using it, of which two are particularly important in this context. One is that the filter may not behave wall if there are errors in the a priori information. The other is that the methods for the determination of bias in the estimates tend to be cumbersome. The paper discusses the solutions available to these two problems and describes the development of an estimator which incorporates an adaptive method of Mehra (1070) which handles the a priori information problem and a partitioning method of t'nd (1969) which simplifies the bias estimation problem. ]t is a feature of the estimator that information generated by the former is used in the latter. The use of the estimator to detect instrument malfunctions in a laboratory level control rig is described.