To determine the impact of a surgeon's diagnosis of surgical wound infections on infection rates, during a 6-month period we prospectively examined patients undergoing surgical wound surveillance for any of four services (orthopedic surgery, general surgery, neurosurgery, or cardiovascular surgery). Criteria were judged as standardized if the infection control practitioner observed pus, redness, or drainage associated with positive culture or if a diagnosis of deep-seated infection was made. Surgeon's diagnosis was judged as a nonstandardized criterion. Using the Centers for Disease Control's criteria, we identified 113 surgical wound infections in 3024 patients undergoing surgical procedures in the four services. Of these, 95 (84%) met objective criteria (pus observed in 53%; drainage, redness, and positive culture in 20%; and deep-seated infection in 11%). In 18 patients (16%), the nonstandardized criterion alone was used for diagnosis. There was wide variation in use of the nonstandardized criterion, ranging from 5% of orthopedic infections to 21% of cardiovascular surgery infections and 40% of neurosurgical infections. For individual surgeons with at least one wound infection, the range of surgeon's diagnosis was up to 67%. We conclude that a surgeon's diagnosis can have a major impact on surgical wound infection rates; this impact is not borne equally among surgical services or individual surgeons.