One of the principal reasons that noisy measurements in the robot forward dynamics problem have not been analysed is the lack of an appropriate model. A dynamics formulation introduced by Rodriguez, J. R. (1981,I.E.E.E. J. Robotics Autom.,3, 30) and extended by Ko, K. S., and Read, R. K. (unpublished) provides a suitable framework for investigating the class of spatially distributed noise sources. The resulting solution of the forward dynamics problem has the form of a Kalman filter with smoothing for a spatially distributed system. Two measurement noise models are investigated: the uncorrelated white gaussian model and the correlated first-order markovian models. A modified Kalman filter solution is obtained in each case. Simulation experiments demonstrate the improvement in system response.