The micro-scale abrasive wear test is attractive for coated substrates because it is simple, only small samples are required, and the specific wear rates for both coating and substrate kappa(c) and kappa(s) can be determined simultaneously. This paper reviews and critically discusses the methods available for data analysis in this test and proposes some new approaches. The wear volumes of the coating and the substrate can be described by two parameters chosen from among the inner and outer crater diameters, the coating thickness, and the penetration depth. The inner crater diameter can usually be measured more accurately than the outer crater diameter since it is more clearly defined. It is recommended to obtain an accurate value for coating thickness, e.g. by creating and measuring a sharply defined crater, and then to calculate the wear volumes in terms of the inner crater diameter and the coating thickness. In general the errors in kappa(c) and kappa(s) are determined by the ratio of the specific wear rates kappa(c)/kappa(s), the non-dimensional parameter a(2)/Rt (where a is the inner crater diameter, R is the ball radius and t is the coating thickness), and the relative measurement errors of the inner crater diameter and the coating thickness. As these relative errors decrease, the errors in both kappa(c) and kappa(s) decrease. In addition a decrease in kappa(c)/kappa(s) and/or a(2) /Rt will decrease the error in K-c, but increase that in K-s. For a typical case where R= 12.5 mm, t= 3 +/- 0.15 mum, and a = 1 +/- 0.01 mm, the errors in kappa(c) and kappa(s) will be < 50% for 0.08 < kappa(c)/kappa(s) < 10. A lower limit to the inner crater diameter is proposed, determined by the ball radius and the abrasive particle size, in order to achieve reasonable accuracy in the data. A new method is proposed for plotting the experimental results, termed the double intercept method, which provides a clear graphical representation of the data and usually gives reliable values for kappa(c) and kappa(s). However, for the analysis of typical experimental data to obtain values for the specific wear rates another method, termed the KVH plot, is shown to be somewhat more consistently accurate. Detailed guidelines are proposed for analysing the data by this method. (C) 2003 Elsevier B.V. All rights reserved.