A 6-year data set of daily counts of admissions to 79 acute care hospitals in Southern Ontario was analyzed in relation to concurrent measurements of air pollution and weather pooled over the same regions, using progressively more sophisticated statistical techniques. The diagnoses studied included a group of respiratory causes and two control diagnoses: accidents and gastrointestinal causes. The 6-year period (1979-1985) was subdivided into six 2-month "seasons" and the area of study was divided into three subregions. Bivariate correlations were found to be significant more often than expected due to chance for all three admissions variables, but accounting for the temporal variation within the 60-day seasons greatly reduced the significance of the control diagnoses. Twenty-four-hour averages for air quality were found to yield more significant associations than peak hourly concentrations. July-August was the only period not having important within-season temporal trends and also had the lowest daily counts for respiratory admissions. Based on a model which accounted for serial correlation, SO2, ozone, and sulfate aerosol were found to be significant predictors of respiratory admissions during July-August. Using cumulative lags increased the magnitude of the estimated response to about 20% of summer respiratory admissions, but no consistent relationships were found which could identify the "responsible" pollutant(s) with certainty. Average pollutant concentrations were generally within U.S. ambient standards.