To address the need for a fast and sensitive method for the detection of bacterial contamination in solutions, we evaluated the use of Fourier-transform near infrared (FT-NIR) spectroscopy and multivariate pattern recognition techniques. The complex cellular composition of bacteria yields FT-NIR vibrational transitions (overtone and combination bands) that might be used for identification and sub-typing. Bacterial suspensions (E. coli HB101, E. coli ATCC43888, E. coli 1224, Pseudomonas aeruginosa, Bacillus cereus and Listeria innocua) were filtered to concentrate the cells and eliminate the matrix which has a strong NIR signal. FT-NIR measurements were done by using a diffuse reflection-integrating sphere. The use of a simple membrane filtration procedure to produce a thin, uniform bacterial film generated reproducible FT-NIR spectra that can be used for rapid discrimination among closely related strains. Transformation of the spectra with second derivatives resolved specific FT-NIR features in the information-rich spectral region of 5000-4000 cm-1 to allow principal components analysis to group the samples into different tight clusters. The use of Anodisc membranes gave the more reproducible results. This methodology appears promising for the rapid evaluation of potential bacterial contamination in liquids with minimal sample manipulation.