The quartz content of each of the 61 thermal coals used in power stations in India has been determined using Fourier Transform Infra-Red (FTIR) Spectroscopy. It has been observed that quartz is abundant in the thermal coals and its proportion varies from 5 to 20% by wt. The abrasion index (AI), a measure of abrasion caused by coals, has been determined for each coal according to the procedure laid down in Indian Standard IS: 9949-1986. The data generated on abrasion together with ash and quartz percentages of the coals studied have been subjected to regression and correlation analysis. Positive correlations have been found between AI and quartz content and between AI and ash yield, but the correlation between AI and ash (A) and quartz (Q) percentages has been observed to be the most significant (R(2) = 0.86). The linear regression model AI = 1.00A + 135Q thus developed has the ability to predict AI of the thermal coals within +/- 10 mg/kg at 95.5% confidence level. Results of application of the model to predicting abrasion of a limited number of foreign coals with different origins have been found to be encouraging. Integration of other variables like the size and the shape of the abrading particles along with other physical properties of coal, like the bulk density and the grindability, with the model, in addition to the variables already considered, has been suggested for improved prediction. (C) 2010 Elsevier B.V. All rights reserved.