Despite the volume of research concerning granulation, the knowledge of its rate behaviour is still a matter of some debate. Here, granulation is explored by looking at the interaction between granule properties and rate behaviour; process conditions are perturbed to induce different rate behaviour. Granules of calcium carbonate powder (mass-mean size of approximately 40 mum) and Polyethylene Glycol (PEG) 1500 are made in a 10-l vertical axis, high-shear mixer with the PEG added in two different ways: by pouring on and melting in. The resulting granules are cooled and sieved on a fourth-root-of-two progression of sieves to obtain the granule size distribution. Granules in specific size ranges are analysed using a thermo-gravimetric technique, mercury porosimetry and the primary particles are liberated from the binder in order that their size distribution may be examined using a light-scattering particle size analyser. The results show marked differences between the two methods of binder addition. The granules made by pouring on the binder exhibit bimodal weight distributions that are unimodal by the end of the experiment, larger granules, faster growth, low air incorporation into granules and uniform binder and primary particle size distribution (PSD) by the end of an experiment. The granules made by melting the binder in situ exhibit bimodal weight distributions throughout an experiment, smaller granules, slower growth, higher incorporation of air into granules and nonuniform binder and primary PSD throughout. Both methods of binder addition show considerable heterogeneity in the properties of granules. We interpret the heterogeneity of binder distribution by means of the nucleation theories of Schaefer and Mathiesen [T. Schaefer, C. Mathiesen, int. J. Pharm., 139 (1996) 139] concluding that pour-on experiments follow the Schaefer and Mathiesen Immersion Hypothesis and that melt-in experiments follow the Schaefer and Mathiesen Dispersion Hypothesis. We put forward two competing hypotheses for the heterogeneity found in the distribution of primary particle sizes: the Preferential Nucleation Hypothesis and the Preferential Growth Hypothesis. We believe that the former applies to pour-on experiments and the latter to melt-in experiments.