We report a new method of estimating the completeness of cancer registration, in which the proportions of unregistered patients are derived from the time distributions of three probabilities, each of which can be directly estimated from the registry's own data--the probabilities of survival, of registration of the cancer during the patient's life, and of the mention of cancer on the death certificate of a cancer patient who dies. This method allows completeness to be assessed routinely by factors such as age, sex, geographical area and tumour type.