A DNA macroarray was developed and evaluated for its potential to distinguish variants of the dinitrogenase reductase (nifH) gene. Diverse nifH gene fragments amplified from a clone library were spotted onto nylon membranes. Amplified, biotinylated nifH fragments from individual clones or a natural picoplankton community were hybridized to the array and detected by chemiluminescence. A hybridization test with six individual targets mixed in equal proportions resulted in comparable relative signal intensities for the corresponding probes (standard deviation, 14%). When the targets were mixed in unequal concentrations, there was a predictable, but nonlinear, relationship between target concentration and relative signal intensity. Results implied a detection limit of roughly 13 pg of target ml(-1), a half-saturation of signal at 0.26 ng ml(-1), and a dynamic range of about 2 orders of magnitude. The threshold for cross-hybridization varied between 78 and 88% sequence identity. Hybridization patterns were reproducible with significant correlations between signal intensities of duplicate probes (r = 0.98, P < 0.0001, n = 88). A mixed nifH target amplified from a natural Chesapeake Bay water sample hybridized strongly to 6 of 88 total probes and weakly to 17 additional probes. The natural community results were well simulated (r = 0.941, P < 0.0001, n = 88) by hybridizing a defined mixture of six individual targets corresponding to the strongly hybridizing probes. Our results indicate that macroarray hybridization can be a highly reproducible, semiquantitative method for assessing the diversity of functional genes represented in mixed pools of PCR products amplified from the environment.