In condition monitoring practice, one of the primary concerns of maintenance managers is how long the item monitored can survive given condition information obtained to date. This relates to the concept of the condition residual time where the survival time is not only dependent upon the age of the item monitored, but also upon the condition information obtained. Once such a probability density function of the condition residual time is available, a consequencial decision model can be readily established to recommend a ‘best’ maintenance policy based upon all information available to date. This paper reports on a study using the monitored vibration signals to predict the residual life of a set of rolling element bearings on the basis of a chosen distribution. A set of complete life data of six identical bearings along with the history of their monitored vibration signals is available to us. The data were obtained from a laboratory fatigue experiment which was conducted under an identical condition. We use stochastic filtering to predict the residual life distribution given the monitored condition monitoring history to date. As the life data are available, we can compare them with the prediction. The predicted results are satisfactory and provide a basis for further studies. It should be pointed out that although the model itself is developed for the bearings concerned, it can be generalized to modelling general condition‐based maintenance decison making provided similar conditions are met.