In recent years, polyethylene (PE) has found increasing use in applications involving impact and erosion. This paper describes a detailed study of the properties of PE subjected to solid particle impact. Flat discs of the material were eroded by sieved sand (300-600 pm) accelerated by using an air blast rig in which the important variables of velocity, angle and mass flux rate are accurately controllable and measurable. Scanning electron microscopy of lightly eroded specimens enabled four basic crater types to be identified: smooth, ploughed, cut, and dented. The proportions of each were established over a range of angles. Long time erosion experiments were conducted in which the flux rate for each angle was adjusted to keep the number of impacts per unit time constant. The dimensionless erosion parameter, e (mass lost per unit mass of erodent that has struck) was computed by using the rate of mass loss when steady-state erosion had been established. Most erosion was found to occur at an angle of 20—30°, the mass loss becoming zero at around 80°. An analysis by D. R. Andrews is presented, showing that the flux rates used in these experiments are well below those needed to cause wear by thermal mechanisms, and this was confirmed by changing the flux rate: mass loss increased in proportion. Macroscopic particles were used to model sand grain impacts, spheres for rounded particles and square plates for sharp ones. A range of techniques was used in this study including high-speed photography (framing speed of 5 x 104s-1), scanning electron microscopy, and moire methods (both in-plane and out-of-plane). A deformation map was constructed for steel sphere im pacts giving the type of crater to be expected at a given angle and speed. It was observed that sand grains required much lower speeds at a given angle to produce a given crater type. High-speed photography enabled mass-loss mechanisms for single-particle impact to be identified. These were the drawing-out of filaments and the machining-out of chips. Quantitative data on kinetic energy losses were obtained, and these, combined with moire methods that gave the sizes of deformed zones, enabled an estimate of the temperature rise per impact to be made (25 K).