Reagents may improve crease recovery (CR) in fabrics woven from celluosic fibres by forming cross-links between cellulose chains. For linen improvements in CR performance are accompanied by a decline in resistance to abrasion (AR) producing a detrimental effect on competitiveness. Conventional measurements of CR and AR are inefficient, require controlled environment and provide no chemical or structural information on the effect or efficiency of the treatments. Thermogravimetric analysis was performed on linen samples treated with a cross-linking agent (DMU). Multivariate analysis was used to establish calibration models relating discrete and continuous data to the required parameters. These models were then validated by prediction. It was demonstrated that the analysis was able to detect the DMU and predict the CR and AR parameters. The best equations were obtained using a heating rate of 20 degrees C min(-1) and modelling using continuous data rather than the discrete measurements. (c) 2007 Elsevier B. V. All rights reserved.