The potential of near infrared spectroscopy in conjunction with partial least squares regression to predict chemical composition of various wood species including softwoods and hardwoods was examined. Hot-water-soluble extractive, pentosan and cellulose content of various wood species were predicted with high coefficient of determination between the predicted and measured values, the ratio of performance to deviation, range error ratio, and low root mean square error of cross validation for cross-validation and root mean square error of prediction for test set validation. Hot-water-soluble extractive and cellulose content models were only suitable for quality control analysis, but pentosan content model had an excellent fit with the data and could be used in any application. All the results indicate that Fourier transform near infrared spectroscopy could be applied to predict the chemical composition of various wood species. (C) 2013 Elsevier Ltd. All rights reserved.