The goal of surveillance is to identify patterns of disease occurrence, detect disease outbreaks, develop clues about possible risk factors, find persons that need further investigation, and anticipate health service needs. Two sources of data are available for the purposes of surveillance: primary data, such as those arising from health surveys or local population-based registries; and secondary data, arising from large administrative databases. The effectiveness of any program to monitor the health of a community can be judged by the application of three "r's": right (accuracy), reasonable (cost) and rapid (speed). Programs using primary sources of data satisfy only one of the three "r's," that of accuracy. Programs using secondary sources sacrifice accuracy for speed and cost. The challenge for Canada in setting up surveillance for stroke is that there is a relatively small population unevenly distributed over a very large geographical area. To date, surveillance in Canada has consisted of a combination of individually initiated research projects and government sponsored programs. The main focus has been to tackle the issue of the accuracy of the large databases by validating the discharge codes listed in the provincial hospital discharge databases. Three provinces have carried out independent validation studies yielding remarkably similar results, and this lends confidence that hospital discharge databases will provide a means of carrying out surveillance for at least this one aspect of stroke. It is likely that any program for stroke surveillance in Canada will be multifaceted, involving the use of large computerized databases supplemented by hospital-based registries set up in a few highly motivated local centres. Stroke surveillance will best be accomplished by a joint effort between government and researchers to ensure that the end product is of high quality and will meet the needs of improving the health of Canadians.