Contamination of clinical specimens by Mycobacterium gordonae is a significant endemic problem in many laboratories. To investigate this problem, 84 cases at 1 hospital were retrospectively identified during 20 months. The overall rate of specimen contamination was 2. 4%, and 72 of the contaminated specimens were respiratory. A case-control comparison showed that the risk of respiratory specimen contamination was significantly increased if the specimen was expectorated (odds ratio [OR], 3.62; 95% confidence interval [CI], 1. 36-9.50) or if the patient consumed fluids within 2 days before specimen collection (OR, 8.92; 95% CI, 1.40-71.20). Cultures of tap water, ice, and iced drinking water all yielded M. gordonae at 10(-2)-10(0) cfu/mL. A culture survey of consenting patients showed contamination of 8 (24%) of 34 sputum specimens collected immediately after a tap water mouth rinse. These findings demonstrate that endemic specimen contamination arises from mycobacteria in hospital tap water and provide a foundation for control efforts.