Colloidal aggregation processes arising at different electrolyte concentrations were studied by means of experiments and confronted with theoretical predictions of different kinetic aggregation models. For this purpose, aqueous dispersions of relatively large polystyrene microspheres were chosen as experimental systems. Aggregation was induced by adding KBr electrolyte to the initially stable particle dispersions. During the aggregation processes, the cluster-size distribution was monitored by means of single cluster light scattering. Analyzing the time evolution of the monomer concentration, we found that the processes arising even at moderate electrolyte concentrations cannot be described by pure time-independent irreversible aggregation models. Hence, alternative models such as time-dependent irreversible aggregation and several reversible aggregation models were also tested. The model that considers a time-dependent sticking probability was found to fit the data quite satisfactorily. Nevertheless, the fitted was so slow that it seems not very likely to find such a behavior in real systems. The aggregation-fragmentation models reported in the literature were unable to reproduce the experimental observations. Hence, a more realistic reversible aggregation model was developed. This model accounts also for reenforced or double bonds between the constituent particles. The corresponding fit improved significantly and reached the same quality as the time-dependent model. Moreover, the obtained fitting parameters were in qualitative agreement with the DLVO predictions and so, reversible aggregation seems to be a more reasonable explanation for the experimental data than time-dependent irreversible aggregation. However, no definite statement on the possible secondary bond fragmentation mechanism may be made since both the applied shear stress in the measuring cell and thermal fluctuations can cause weaker bonds to break. (C) 2004 American Institute of Physics.