A spectral line-by-line (LBL) method is developed for photon Monte Carlo simulations of radiation in participating media. The performance of the proposed method is compared with that of the stochastic full-spectrum k-distribution (FSK) method in both homogeneous and inhomogeneous media, and in both traditional continuum media and media represented by stochastic particle fields, which are frequently encountered in combustion simulations. By using random-number relations, the LBL method does its own spectral reordering. in effect similar to the FSK reordering. Both LBL and FSK methods result in the same level of statistical errors for similar numbers of photon bundles. It is shown that the FSK approach is superior to the LBL approach in both memory and CPU efficiency in all test cases. However. the CPU efficiency is not prominent when spectral calculations are not the dominant source of overall CPU-time cost. Due to a lack of correlation of absorption coefficients in inhomogeneous media, the FSK approach may produce substantial errors, which are avoided by the LBL method. The LBL database can be constructed once and for all, while the FSK database has to be reconstructed every time the reference state changes. Thus, the LBL method is more suitable for quickly-evolving media. (c) 2007 Elsevier Ltd. All rights reserved.