We determined factors that affect responses of bacterial isolates and model bacterial communities to the 95 carbon substrates in Biolog microliter plates. For isolates and communities of three to six bacterial strains, substrate oxidation rates were typically nonlinear and were delayed by dilution of the inoculum. When inoculum density was controlled, patterns of positive and negative responses exhibited by microbial communities to each of the carbon sources were reproducible. Rates and extents of substrate oxidation by the communities were also reproducible but were not simply the sum of those exhibited by community members when tested separately. Replicates of the same model community clustered when analyzed by principal- components analysis (PCA), and model communities with different compositions were clearly separated un the first PCA axis, which accounted for >60% of the dataset variation. PCA discrimination among different model communities depended on the extent to which specific substrates were oxidized. However, the substrates interpreted by PCA to be most significant in distinguishing the communities changed with reading time, reflecting the nonlinearity of substrate oxidation rates. Although whole-community substrate utilization profiles were reproducible signatures for a given community, the extent of oxidation of specific substrates and the numbers or activities of microorganisms using those substrates in a given community were not correlated. Replicate soil samples varied significantly in the rate and extent of oxidation of seven tested substrates, suggesting microscale heterogeneity in composition of the soil microbial community.