We determined the accuracy and reproducibility of whole-community fatty acid methyl ester (FAME) analysis with two model bacterial communities differing in composition by using the Microbial ID, Inc. (MIDI), system. The biomass, taxonomic structure, and expected MIDI-FAME profiles under a variety of environmental conditions were known for these model communities a priori. Not all members of each community could be detected in the composite profile because of lack of fatty acid 'signatures' in some isolates or because of variations (approximately fivefold) in fatty acid yield across taxa. MIDI- FAME profiles of replicate subsamples of a given community were similar in terms of fatty acid yield per unit of community dry weight and relative proportions of specific fatty acids. Principal-components analysis (PCA) of MIDI-FAME profiles resulted in a clear separation of the two different communities and a clustering of replicates of each community from two separate experiments on the first PCA axis. The first PCA axis accounted for 57.1% of the variance in the data and was correlated with fatty acids that varied significantly between communities and reflected the underlying community taxonomic structure. On the basis of our data, community fatty acid profiles can be used to assess the relative similarities and differences of microbial communities that differ in taxonomic composition. However, detailed interpretation of community fatty acid profiles in terms of biomass or community taxonomic composition must be viewed with caution until our knowledge of the quantitative and qualitative distribution of fatty acids over a wide variety of taxa and the effects of growth conditions on fatty acid profiles is more extensive.