Abstract. Cloud chamber investigations into ice nucleation by mineral particles were compared with results from cold-stage droplet freezing experiments. Kaolinite, NX-illite, and K-feldspar were examined, and K-feldspar was revealed to be the most ice-active mineral particle sample, in agreement with recent cold-stage studies. The ice nucleation efficiencies, as quantified using the ice-active surface site density method, were found to be in agreement with previous studies for the lower temperatures; however, at higher temperatures the efficiency was between a factor of 10 and 1000 higher than those inferred from cold-stage experiments. Numerical process modelling of cloud formation during the experiments, using the cold-stage-derived parameterisations to initiate the ice phase, revealed the cold-stage-derived parameterisations to consistently underpredict the number of ice crystals relative to that observed. We suggest the reason for the underestimation of ice in the model is that the slope of the cold-stage-derived ice-active surface site density vs. temperature curves are too steep, which results in an underestimation of the number of ice crystals at higher temperatures during the expansion. These ice crystals suppress further freezing due to the Bergeron–Findeison process. A coagulation model was used to investigate the idea that the mineral particles coagulate in suspension. This model suggests that coagulation during the experiments may be sufficient to significantly remove the particles for the suspension by sedimentation or reduce the total particle surface area available for ice nucleation to take place. Aggregation was confirmed to take place in mineral suspensions using dynamic light-scattering measurements. However, it is not proven that aggregation of the mineral particles is able to reduce the surface area available for ice nucleation. The implication is that the mineral particles may be more important at nucleating ice at high temperatures than previously thought.