A new method of estimating temperature, soot volume fraction, and gas species concentrations from spectral radiation intensity measurements is reported. The spectral radiation intensities emitted from a one-dimensional McKenna burner is measured at multiple wavelengths using a high-speed mid-infrared spectrometer. The spectrometer obtains the spectral radiation intensities from 1.3 to 4.8 mum at 1320 Hz. Radiation from the soot, H2O, and CO2 gas hands are used to estimate soot temperature, soot volume fraction, gas temperature, and mole fractions Of CO2 and H2O gas. Because of high nonlinearity between the gas; temperature and the emissivity of the gas molecules, temperature, and gas concentration estimation requires a robust nonlinear inversion algorithm. A linearized radiative transfer equation (LRTE) in conjunction with the maximum likelihood estimation (MLE) method is developed to deconvolute the spectral radiation intensities for temperatures and concentrations. The LRTE-MLE method is first verified using synthetic dataset. For the synthetic dataset, the LRTE-MLE method always converges to the exact solution from any reasonable initial guess. The LRTE-MLE method is found to be insensitive to random errors in radiation intensity measurements. However, the estimated gas temperature is sensitive to any potential errors in the wavelength calibration. The LRTE-MLE method is then applied to measurements from a one-dimensional premixed laminar flame. The estimated gas temperature and concentrations obtained using the LRTE-MLE method are very close to those obtained with thin filament pyrometry (TFP) and theoretical calculations. The computational cost for the LRTE-MLE method was found to be minimal when compared to other nonlinear methods. (C) 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.