A non-verotoxin-producing isolate of Escherichia coli O157:H7 was inoculated at final concentrations of 10(3) or 10(6) ml-1 into natural non-carbonated mineral water (MW), sterile natural mineral water (SMW) and sterile distilled deionized water (SDDW) and stored at 15 degrees C for 10 weeks. Samples were examined every 7 d for the presence of E. coli O157:H7 using a resuscitative/selective agar procedure. The MW samples were also plated onto a non-selective agar, R2A, to enumerate E. coli O157:H7 and the autochthonous flora. There was a significant difference in the survival of E. coli O157:H7 (10(3) ml-1 inoculum) between the MW and the SDDW at time periods 0, 7, 14 (P < 0.005) 21, 28, 35 (P < 0.001) and 42 d (P < 0.05) and between the MW and the SMW at time periods 7, (P < 0.05) 14, 21 (P < 0.005) 28 (P < 0.01) and 35 d (P < 0.05), with the pathogen surviving longest in the MW samples. In contrast, at 10(6) ml-1, no significant differences in the survival of E. coli O157:H7 were observed between the water types. The presence of E. coli O157:H7 (10(3) ml-1) in the MW samples did not have an antagonistic effect on the recovery of the autochthonous flora. Transmission electron microscopy analysis demonstrated that the E. coli O157:H7 cells lyse during storage, releasing their contents into the surrounding environment. These substances may have been utilized by the autochthonous flora and thereby explain why the numbers of flora recovered from the inoculated MW samples were higher than those recovered from the uninoculated samples.