Abstract An experimental investigation has been conducted into the flow of a talc-based paste through various combinations of axisymmetric and non-axisymmetric ram extruder geometries. The paste flow pattern was found to be highly rate dependent for one non-axisymmetric case. The flows were modelled using the Benbow–Bridgwater approach, but this was incapable of predicting the observed flow patterns and under-predicted the extrusion pressure at low flow rates. Evidence suggested that significant liquid phase migration was occurring at low flow rates, with associated dewatering of paste in the barrel. The situation where significant liquid phase migration occurs is comparable to a drained experiment in soil mechanics. The simple, but little known, drainage criterion of Wroth and Houlsby was used to estimate the flow rates associated with drained and undrained conditions. The criterion successfully predicted the flow rate at which transition of the flow pattern was observed. Comparable experiments using a modelling clay, undrained over all flow rates, showed no transition in flow pattern. Furthermore, the drainage criterion successfully predicted the drainage state for three of four other paste flow data sets taken from the literature covering a variety of materials and geometries. It is concluded that: ( i) liquid phase migration can be a significant feature of paste extrusion which can dramatically alter the flow patterns in non-axisymmetric extruders, and ( ii) that the Wroth and Houlsby's criterion can successfully predict the drainage state of a process but is limited by the estimation of the coefficient of permeability.