Humic acids extracted from peats (P), brown coals (BC) and lignites (L), were characterized using different (chemical, H-1-nuclear magnetic resonance spectroscopy and differential thermal analysis) techniques. Fourteen variables were obtained from these analyses and only five were selected because uncorrelated in multiple partial correlation. The chosen variables were C concentration, aliphatic and aromatic components and the heat of reaction of the second exothermic peak. The multivariate discriminant analysis was performed on these variables and a discriminant function was obtained which was able to efficiently separate the P, BC and L. This function enables simple predictions on samples of unknown origin. The straightforward method proposed and the results obtained are discussed. (C) 2003 Elsevier Science Ltd. All rights reserved.