A system based on principles of pattern recognition has been developed for identifying toxic and other volatile organic pollutants in complex environmental samples. It interprets the most commonly used monitoring data, mass spectral data, and produces a class designation, an estimate of the molecular weight and, if possible, a target pollutant identity. The system was developed and implemented with a very user friendly expert system which operates by asking questions of the user. The system was designed for samples with low concentrations of pollutants, e.g. ambient air samples. The mass spectra of 74 target and 31 other compounds were used to establish the five target and the one unknown classes. The target classes were nonhalobenzenes; chlorobenzenes; bromo- and bromochloro-alkanes/alkenes; mono- and dichloroalkanes/alkenes; and tri-, tetra- and pentachloro-alkanes/alkenes. The molecular weight is very important in eliminating possible identities for unknown pollutants. This system was extensively tested with mass spectra of potential air pollutants, randomly selected compounds, field gas chromatography/mass spectral air pollutant data and with 54,000 other spectra. Classification and identification accuracies from these tests were 90-100% and 96-100%, respectively. The resulting system runs on a personal computer and is very fast.