A computational fluid dynamic model has been developed for a modified Denver flotation cell used in laboratory batch tests. The model incorporates fundamental aspects of bubble–particle interactions including bubble–particle collision, attachment and detachment. Model predictions are compared against carefully planned laboratory experiments with narrowly-sized spherical ballotini. Flotation results for different impeller speeds have been obtained for a constant air flow into the cell. The particles have been made hydrophobic by methylation with trimethylchlorosilane. For a given particle size, the results from the model and experimental work indicate that there is an optimum stirring speed that produces a good compromise between attachment and detachment rates in the cell for flotation. For less hydrophobic particles, a lower stirring speed is beneficial in having a longer contact time because of the longer induction time required for attachment.