During September and October 1984, a major meteorological and tracer study was conducted in Colorado's Brush Creek valley. The characteristics of atmospheric dispersion during the nighttime and morning transition periods are discussed in this paper. Tracer released near the valley floor did not reach the ridgetops (escape from the valley) during the nighttime but was confined to the valley, being carried in down-valley flows. After sunrise, with the onset of convective boundary-layer growth and initiation of upslope flows, the tracer within the valley was carried into the upper elevations of the valley atmosphere and ventilated from the valley. This was confirmed by the ridgetop tracer samplers and by a tracer mass budget applied to a valley atmosphere control volume. The ventilation rate of tracer from the valley atmosphere to the above-ridgetop flows was calculated from the tracer mass budget. A dimensionless form of the ventilation rate is proposed. The Gaussian plume equation adequately represented (16% average deviation) the average nighttime plume centerline concentration, out to 8 km from the release, when the plume was fully contained in down-valley flows. This agreement was attained by accounting for plume reflections from the valley sidewalls and using measured turbulence statistics in the calculation of the dispersion coefficients. Beyond 8 km down valley from the release, the Brush Creek valley merged with the Roan Creek valley and the two airstreams mixed, resulting in a sudden dilution of the tracer plume. The Gaussian plume equation was not valid after the two airstreams merged.