Quantitative structureproperty relationship techniques were applied to develop a predictive method for autoignition temperatures of a wide range of organic molecules, ioeludmg hydrocarbons, alcohols, phenols, ethers, aldehydes, ketones, acids, amines, esters and halogenated compounds. Multivariate linear regression models in terms of easily available molecular descriptors or intrinsic molecular properties such as critical pressure, parachor, atomic charges, etc were p r o m . Principal component analysis on the set of descriptors employed uncovered the significant contributions of the polarity-related factors to auto-ignition temperatures. For the majority of the 250 compounds over an auto-ignition temperature range of 170-6WC, good agreement between observed and calculated autoignition temperatures was confirmed. The method could be useful for assessing the flammability of new compounds.