Clinical periodontal research often involves the treatment of patients who have disease at several sites in the mouth. Should the units of analysis be the patients, the sites, or a combination of both? These strategies are explored and a worked example given.