In order to develop predictive process models and to enhance process understanding in high shear granulation, there is an ongoing search for non-intrusive methods for measuring the wet mass velocities in the mixer. In this study, a high speed CCD camera is used in combination with software for particle image velocimetry (PIV) calculations to obtain information about the wet mass velocities. The focus has been on obtaining good spatial and angular resolution for the velocities along the glass bowl wall. In a Jenike shear cell, both internal and wall frictional properties have been measured and together with velocity data, this information is used for prediction of the impeller torque. It has been shown that the near wall velocities are strongly dependent on the coefficient of wall friction, which decreases during liquid addition. The decrease in the coefficient of wall friction results in increased wet mass velocities close to the bowl wall. It is also found that the wet mass velocity has a strong angular dependence, resulting in a high frequency pulsing bed behaviour which cannot be detected by visual inspection. The predictive impeller torque model developed by Knight et al. [2001. Prediction of impeller torque in high shear powder mixers. Chemical Engineering Science 56, 4457-4471] has been generalized to account for cohesive materials and with frictional and velocity data, the level of the impeller torque is well predicted. However, the model is based on crude assumptions regarding the velocity distribution and hence, it cannot capture the dynamics in the measured torque curve satisfactorily. (C) 2007 Elsevier Ltd. All rights reserved.