AbstractInappropriate land use practices, pollutants, exploitation, and overpopulation have simplified stream habitats and degraded water quality worldwide. Management agencies are now being tasked to ameliorate impacts and restore stream “health,” yet there is a dearth of rigorous scientific methods and theory on which to base sound restoration design and monitoring. Despite this, many localized restoration projects are being constructed to stabilize erosion and enhance habitat heterogeneity in streams. Many restoration attempts adopt the paradigm that increasing habitat heterogeneity will lead to restoration of biotic diversity, yet there have been few studies that have manipulated variation of a physical parameter independent of the mean to isolate the effects of heterogeneity per se. We conducted a field experiment to mimic restoration of habitat heterogeneity in a shallow. stony stream. By using an experimental approach rather than a detailed assessment of existing restoration work, we were able to control the starting conditions of replicate riffles so that organism responses could be unambiguously attributed to the heterogeneity treatments. We successfully manipulated the variability of streambed particle sizes and consequently near‐bed flow characteristics of entire riffles. These factors define axes of habitat heterogeneity at scales relevant to the resident macroinvertebrate fauna. Despite this, we were unable to distinguish differences in community structure between high and low habitat heterogeneity treatments. Power analysis indicated that macroinvertebrate populations were more sensitive to individual site conditions at each riffle than to the heterogeneity treatments, suggesting that increasing habitat heterogeneity may be an ineffective technique if the restoration goals are to promote macroinvertebrate recovery in denuded streams. With extremely high variability between replicate riffles, monitoring programs for localized restoration projects or point source impacts are unlikely to detect gradual shifts in community structure until the differences between the reference and treatment sites are extreme. Innovative measurement of other parameters, such as ecosystem function variables (e.g., production, respiration, decomposition), may be more appropriate indicators of change at local scales.