Six different Bacillus cereus strains were selected from two different ecotypes: (1) three commonly used laboratory strains that are considered avirulent, and (2) three clinical isolates from meningitis patients. Screening of genomic DNA for the presence of genes encoding known toxins gave no candidate genes that were unambiguously able to distinguish between the two groups. However, the application of multivariate pattern-recognition methods to metabolite profiles derived from the different strains using 1H nuclear magnetic resonance spectroscopy (metabolomics) was able to classify the different profiles. The two different ecotypes were clearly separated on the basis of their metabolite profiles, showing that it is possible to use metabolomic methods to classify pathogens on the basis of their expressed physiology, even when it is not possible to infer a direct mechanistic link to specific virulence factors. This metabolomic approach could also have a wide range of possible applications in both general microbiology and microbial ecology for distinguishing and identifying different functional/physiological ecotypes of bacterial strains or species.