We have tested an approach to identify putative cancer stem cells that involves measurement of the infrared absorption spectrum of individual cells in an aqueous environment, and their subsequent classification using multivariate data analysis techniques. Two primary esophageal cell lines were characterized: the immortalized normal esophageal epithelial cell line, Het-1A, and the esophageal adenocarcinoma cell line, OE33. In addition, we also evaluated spheroids, reflecting stem-like cell populations, which were derived from each parent cell line when grown in serum-free media. As differences in cell size appeared to be a strong discriminating factor, a correction needs to be performed to allow a reliable classification based on infrared absorption spectra. We demonstrated that stem-like cells derived from Het-1A could easily be discriminated on the basis of absorbance differences in the 1000-1200 cm^-1 spectral interval, whereas this was not possible for OE33. Furthermore, we found that changes due to aging of OE33 cells in culture dominated the infrared absorption spectra and somewhat limited the potential of this approach to identify stem-like cell populations using this in vitro model system.