In 1986, the U.S. Environmental Protection Agency (USEPA) recommended that Escherichia coli (E. coli) be used in place of fecal-coliform bacteria in State recreational water-quality standards as an indicator of fecal contamination. This announcement followed an epidemiological study in which E. coli concentration was shown to be a better predictor of swimming-associated gastrointestinal illness than fecal-coliform concentration. Water-resource managers from Ohio have decided to collect information specific to their waters and decide whether to use E. coli or fecal-coliform bacteria as the basis for State recreational water-quality standards. If one indicator is a better predictor of recreational water quality than the other and if the relation between the two indicators is variable, then the indicator providing the most accurate measure of recreational water quality should be used in water-quality standards. Water-quality studies of the variability of concentrations of E. coli to fecal-coliform bacteria have shown that (1) concentrations of the two indicators are positively correlated, (2) E. coli to fecal-coliform ratios differ considerably from site to site, and (3) the E. coli criteria recommended by USEPA may be more difficult to meet than current (1992) fecal-coliform standards. In this study, a statistical analysis was done on concentrations of E. coli and fecal-coliform bacteria in water samples collected by two government agencies in Ohio-- the U.S. Geological Survey (USGS) and the Ohio River Valley Water Sanitation Commission (ORSANCO). Data were organized initially into five data sets for statistical analysis: (1) Cuyahoga River, (2) Olentangy River, (3) Scioto River, (4) Ohio River at Anderson Ferry, and (5) Ohio River at Cincinnati Water Works and Tanners Creek. The USGS collected the data in sets 1, 2, and 3, whereas ORSANCO collected the data in sets 4 and 5. The relation of E. coli to fecal-coliform concentration was investigated by use of linear-regression analysis and analysis of covariance. Log-transformed E. coli and fecal-coliform concentrations were highly correlated in all data sets (r-values ranged from 0.929 to 0.984). Linear regression analysis on USGS and ORSANCO data sets showed that concentration of E. coli could be predicted from fecal-coliform concentration (coefficients of determination (R2) ranged from 0.863 to 0.970). Results of analysis of covariance (ANCOVA) indicated that the predictive equations among the three USGS data sets and two ORSANCO data sets were not significantly different and that the data could be pooled into two large data sets, one for USGS data and one for ORSANCO data. However, results of ANCOVA indicated that USGS and ORSANCO data could not be pooled into one large data set. Predictions of E. coli concentrations calculated for USGS And ORSANCO regression relations, based on fecal-coliform concentrations set to equal Ohio water-quality standards, further showed the differences in E. coli to fecal-coliform relations among data sets. For USGS data, a predicted geometric mean of 176 col/100 mL (number of colonies per 100 milliliters) was greater than the current geometric-mean E. coli standard for bathing water of 126 col/100mL. In contrast, for ORSANCO data, the predicted geometric mean of 101 col/100 mL was less than the current E. coli standard. The risk of illness associated with predicted E. coli concentrations for USGS and ORSANCO data was evaluated by use of the USEPA regression equation that predicts swimming-related gastroenteritis rates from E. coli concentrations.1 The predicted geometric-mean E. coli concentrations for bathing water of 176 col/100 mL for USGS data and 101 col/100 mL for ORSANCO data would allow 9.4 and 7.1 gastrointestinal illnesses per 1,000 swimmers, respectively. This prediction compares well with the illness rate of 8 individuals per 1,000 swimmers estimated by the USEPA for an E. coli concentration of 126 col/100 mL. Therefore, the