The problem of maintaining the consistency of state estimates with a decision maker's expectations, is considered. The discrepancy between the estimate and its expectation defines a jump in the state estimate. The jump makes the estimate consistent with its expectation. A method is described for generating corresponding covariance matrices for the Kalman filter in order to make the filter consistent with such a jump. The method is based on appropriate rank-one modifications to these convariances. It is shown that the method is able to alter the estimates in the precise direction indicated by the expectations. The corresponding alteration to the Kalman gain is also discussed. Convergence becomes evident if expectations become, through learning, increasingly accurate estimates of the corresponding state vector.