Thirty-two clinical specimens submitted to the laboratory during a 12-month period from July 1980 to June 1981 were reported to be culture-positive for Mycobacterium gordonae, an organism generally considered to be a slow-growing saprophyte with natural habitats which include soil and water. Only seven similar isolates had been recovered in the preceding 4 1/2 year period. The discordance between clinical findings and the mycobacterial cultures suggested extrinsic contamination of the specimens. Contamination in the laboratory was believed unlikely because: 1) clinical samples obtained in an aseptic manner were never contaminated; 2) various surveillance cultures of reagents and deionized water used in the laboratory were negative; and 3) substitution of deionized water with sterile water did not control the outbreak. Extensive hospital-wide cultures of water sources implicated the use of ice and ice water from contaminated ice machines as the source of this pseudoepidemic. Cleaning of the ice machines resulted in a sharp decrease in the number of M. gordonae isolates. Pseudoinfection by M. gordonae from improperly maintained ice machines has not been reported before.