Stream habitat characteristics were measured at twenty sites in agricultural areas of eastern Wisconsin by the U.S. Geological Survey in May and June, 1993 as part of the National Water-Quality Assessment Program Western Lake Michigan Drainages study unit. These "benchmark" stream sites were selected for study to represent standards of reference for comparison to other streams in similar physical settings that appear to be more detrimentally affected by agriculture. The agricultural benchmark streams were selected from four physical settings, or relatively homogeneous units (RHU's), that differ in bedrock type and texture of surficial deposits. Habitat characteristics at streams in these four physical settings are described and compared to each other, and a habitat classification scheme was used to rank the quality of habitat in these streams. Additional aquatic information was collected along with the habitat data: water-quality data and population surveys of fish, invertebrates, and algae. Habitat data were collected at three levels: drainage basin, stream segment between major tributaries (length from 1 to 14 km), and stream reach (approximately 150m). Results of statistical analyses show that, in general, most correlations are among basin-level habitat characteristics. Few correlations were observed among reach- and basin-level characteristics. Principal components analysis (PCA) on basin-level data resulted in principal components that reflect RHU, land use or latitude, and basin size. Groupings of habitat characteristics at the reach level are less clearly attributed to some outside environmental factor. Streams that have undergone habitat restoration for fisheries group closely together on PCA ordination plots. Less than half of the habitat characteristics were found to be significantly different between one RHU and the other three. Characteristics that differed between RHU's were mainly at the basin level but also included some reach-level characteristics. Stream-habitat characteristics were classified according to the Michigan Department of Environmental Quality, Great Lakes Environmental Assessment Section (GLEAS) Procedure 51. No relation was found between GLEAS scores and RHU or the percentage of agricultural land in the drainage basins above the benchmark-stream sites. GLEAS scores were varied in each RHU. Streams with high GLEAS scores (rated good or excellent) can be considered benchmark or reference streams as far as habitat is concerned. Of the 20 streams sampled, 16 met this criterion.