Macrobenthic data obtained using three sampling schemes (0.06-sq m x 8-cm-deep sampling unit and 1.0- or 0.5-mm-mesh sieves, and 0.1-sq m x 8-cm-deep sampling unit and 1.0-mm-mesh sieve) previously identified as optimal or near-optimal for detecting differences between a reference and a moderately impacted station when animals were identified to species were reanalyzed at the genus, family, order, and phylum level to determine the taxonomic level sufficient to detect differences between the stations with t-tests on five measures of community structure. Taxonomically sufficient levels for number of taxa were family in 1.0-mm-mesh samples and species in 0.5-mm-mesh samples. Specific identification was usually required for a Dominance, Shannon's, 1 - Simpson's, and McIntosh's Index in 1.0- and 0.5-mm-mesh samples, suggesting limits to the utility of the taxonomic sufficiency concept when using those indices to detect moderate impacts. This and a previous study indicate that one could reliably (alpha = 0.05, 1 - beta about 0.80) detect moderate benthic impacts at the study site on number of taxa and five other measures of community structure with five to seven replicate 0.06-sq m x 8-cm-deep, 1.0-mm-mesh samples per station and identification to family only. Taxonomic sufficiency can vary depending upon the animal size fraction sampled and the measure used.