Lactose was melt agglomerated in a mixer torque rheometer with polyethylene glycol (PEG) 2000, 3000, 6000, 8000, 10 000, or 20 000 as meltable binder. A longer massing time caused an increase in mean torque until a maximum value after which the torque decreased. A smaller particle size of the PEG gave rise to a faster initial rise in mean torque. The higher viscosity of the PEG 20 000 resulted in a higher mean torque, whereas no clear difference in mean torque was obtained with the other PEGs. The binder concentration could be varied within a rather wide range without causing overwetting, the range being wider with PEG 3000 than with PEG 20 000. The mean torque values obtained were found to be related to the liquid saturation of the agglomerates. The reproducibility of the experiments was found to be very dependent on the experimental conditions, the highest binder viscosities and binder concentrations giving rise to a poor reproducibility. The results were compared with a few melt agglomeration experiments with PEG 3000 in a high shear mixer. The mixer torque rheometer was found not to be suitable for predicting melt agglomeration properties in the high shear mixer because of a marked difference in the shear forces in the two mixers.