An evaluation of soil-geosynthetic interface strength for different types of loading is important to the design of any anchorage detail of a reinforced soil structure or membrane-lined waste containment facility. The imposed loadings may be classified as static, repeated non-dynamic or cyclic, and dynamic. The test method best suited to model the anchorage behaviour is the pullout test. A large scale pullout apparatus was designed that accommodates a soil sample 1.30 m long x 0.64 m wide x 0.60 m thick. Samples of a uniformly-graded medium sand were prepared by air pluviation. A stress-controlled top boundary was used and tests performed for normal stresses in the range 4 to 30 kPa. Tests were performed on five types of geosynthetics: three geogrids, a smooth geomembrane and a textured geomembrane. A sophisticated electro-hydraulic control system was developed and two modes of testing were used to evaluate pullout resistance. The response of the geosynthetic is characterized by a non-linear variation of tensile force along the specimen. Consequently the profile of shear stress variation is non-linear and is dependent on the magnitude of pullout displacement: interpretation of the pullout test to obtain an interaction factor for design should account for this extensible behaviour. A generalized method is proposed for use with independent measurements of force and strain. The application of the generalized method is demonstrated: it describes very well the variation of interaction factor with pullout displacement and suggests a unique value that is independent of normal stress. Cyclic loading of the test specimen in most cases reveals that an interaction factor mobilized is equal to or slightly exceeds the value mobilized in corresponding DC test. A load ratio is defined as the ratio of the measured pullout load in cyclic pullout test to the corresponding DC test. A conceptual model is proposed that links a load ratio to stable and unstable behaviour in cyclic pullout, and identifies a threshold ratio above which an unstable behaviour results. The threshold ratio is observed to be influenced by the specimen characteristic, being >1 for a grid specimen with a relatively rigid bearing member, and =1 for all other test specimens except a few tests where a value <1 was observed. This implies that using a reduced value of interaction factor for dynamic loads in all cases is inappropriate, in that it does not properly describe the mobilized response.