The article describes laboratory tests carried out to assess the filtration characteristics of various fabrics when placed between a granular sub-base and a cohesive soil. An apparatus similar to that used by Snaith and Bell was used to test Terram 1000, Fibretex f2-b, G100 and Lotrak 45/45, all of which, except Lotrak 45/45, were non-woven. A loading period of 24 h was used for four types of cohesive soil in the first series of tests. Type 1 soil only was used in the second series with separate loading periods ranging from 12 h to 96 h. Results show that repeated loading without a fabric resulted in considerable migration of the subgrade into the sub-base. Introducing a fabric reduces the movement of the subgrade particles, the degree of reduction depending upon the properties of the fabric. The presence of a sand layer beneath the fabric assists in the formation of a filter. There was some evidence that the period of repetitive loading is more significant in subgrade migration than the number of load repetitions. (TRRL)