The biological effect of water abstraction from unregulated streams in New South Wales, Australia, was assessed with data on macroinvertebrate assemblages in riffles (445 samples) and pool edges (718 samples) obtained from a national assessment of river condition in 1995 to 2000. We used limiting environmental difference (LED) modeling to predict the macroinvertebrate assemblage expected for each sample in the absence of water abstraction and estimated effect by comparing predicted and observed assemblages for sites with upstream abstraction. We found a high likelihood of effect on macroinvertebrates in 30% of riffles with upstream abstraction and 22% of edges. Complex combinations of biological traits appeared to determine the sensitivities and tolerances of individual taxa to water abstraction. For example, rheophilous families generally were rarer than predicted in apparently affected riffles, where thermophilic families seemed tolerant of the effects of water abstraction, and free-swimming families were less common than predicted in apparently affected edges, where families with tegument respiration appeared unaffected by water withdrawals. The trait response was clearer for the riffle fauna than for the edge fauna, perhaps because the physical and chemical effects of abstraction are more consistent for riffles than for edges. Our approach can support management planning by identifying taxa that are most at risk from water abstraction and sites where apparent effects are most evident.