A direct infrared spectral mixture interpretation technique, PAWMI (program for automated waste mixture interpretation) was described. PAWMI is an automated rule based artificial intelligence program designed to identify major components in a complex mixture by interpreting the infrared spectrum of the mixture. The method has been developed to identify 62 of the most commonly identified organic compounds on hazardous waste sites. The program is easily adapted for adding additional components to the interpretation list. PAWMI can be operated on site in a mobile lab during a remedial action project by trained personnel. Other applications include screening incoming organic waste at RCRA disposal facilities or interpreting gas phase spectra obtained during industrial hygiene air monitoring. Two problems were encountered during operation. The first, a need for a technique to automatically select peaks in spectra, especially where base lines are nonlinear, has been solved by PUSHSUB, which selects peaks by transforming the first 256 data points right of the centerburst from the original 16,384 data point sample interferogram into a threshold curve. The second problem, false positive results occurring because many of the compounds of interest on rememdial action sites are structurally similar molecules, was solved through the use of PAIRSPLUS, a softwater program for limiting spectral similarity. The quantitative limit of detection of minor compounds in mixtures tested with PAWMI ranged from 0.4 to 15 percent, depending on the spectral similarity of components of the mixtures. The sensitivity of the method was 77 percent. False positives represented 13 percent of the total positive results reported.