ABSTRACT Stanford Research Institute carried out an experimental program in St. Louis during the period August through October 1971@ for the further development and testing of a practical@ multipurpose urban diffusion model (APRAC-IA) for carbon monoxide to determine whether earlier (1970) SRI findings in San Jose@ California could be generalized to larger cities with highly developed urban cores. Two adjacent downtown street canyons were instrumented to obtain measurements of CO concentration at 30 points and winds at eight. The instrumented canyons are at right angles to each other and have aspect (height-to-width) ratios of 1.5 and 2@0; these are significantly larger than the value of 0.7 of the site studied earlier in San Jose. Wind@ temperature@ and CO were also measured to a height of 130 m above the site on a television tower@ Helicopter and van-borne instrumentation was used to supplement data collected with the automated street-canyon instrumentation system. The data collected generally confirm the findings of the earlier studies@ and only minor revisions of the model were required to improve the specification of atmospheric stability and small-scale street-canyon effects. The distribution of CO in the street canyon indicates the @ presence of a single-cell@ helical circulation in the deep St@ Louis street canyons under cross-street@ roof-level flow conditions; the same pattern was found in the shallow San Jose canyon. Carbon monoxide measurements made along the street to within 10 m of the intersection indicate that the street-effects formulation should be applicable in this region. The model was applied using only routinely available meteorological and traffic data. Concentrations were calculated for four locations in the canyons and two at roof level. These calculations were compared with about 600 hour-averaged observations for each location. The observed concentrations of CO were simulated with root-mean-square errors of 3-4 ppm. This is half the uncertainty that had been encountered when the model was applied to St@ Louis prior to refinements of the stability and diffusion formulations and without the street-effects model. Linear regression (calibration) would reduce the differences by an additional 1 ppm. Median and 90-percentile concentrations were specified within 2-3 ppm by the current model; these errors would be halved by the use of calibrated values@ It is felt that the APRAC-IA model is now suitable for practical applications.