On-line measurement of slurry viscosity was carried out to study the effect of viscosity on the cut (d(50)) size of hydrocyclone classifiers. As slurry viscosity increases, the settling rate of particle decreases, causing the d(50) size to become coarser. The viscometer set-up used a vibrating sphere viscometer and a specially designed slurry presentation device to avoid settling of solids during viscosity measurement. This set-up was mounted on a test rig for a 10.2 cm diameter hydrocyclone. Test samples were prepared from ground silica (80% passing 65 microns) and water. Both the solids content and temperature of the samples were varied to change their viscosities. Samples from overflow and underflow streams were collected at regular intervals, and data from these samples were used to calculate the d(50)(c) size. Temperature and viscosity were recorded simultaneously during these tests. From these data it was observed that d(50)(c) was proportional to the 0.35th power of the slurry viscosity. This relationship was then introduced in the existing Lynch and Rao model for siliceous material, to develop a modified model for hydrocyclone classification. This modified Lynch and Rao model incorporated a viscosity parameter from direct measurement, and predicted the cut size precisely when the viscosities of the slurries were altered by factors other than changing percent solids, such as temperature variation. This was not possible with the original Lynch and Rao model, which did not include any viscosity term. It was also determined that increasing slurry viscosity produced an increase in the bypass fraction, R(f). This effect was due to increased fluid drag in the hydrocyclone as the viscosity increased.