Benthic habitat and macrobenthos recovery were measured over an 11-y period beginning in January 2002 following the closure of a green-lipped mussel farm in the Marlborough Sounds, New Zealand. Habitats and abundance of eight invertebrate species were compared under two types of farming structures (production backbones and warps) used to grow mussels and secure the farm to the seabed, adjacent areas unaffected by farming served as the control reference. The mean percent of the seabed covered by mussel shells under backbones declined from 38% to 0% after 11 y whereas deposition of shells beneath warps was low (<4%) and resembled that at reference sites after only 3 y. One month following farm closure, mean species densities beneath warps, where the upper confidence limit of mean percent shell cover was 7%, did not differ from reference sites. By contrast, the mean densities of two macrobenthic species beneath backbones exceeded that at reference sites whereas the densities of three species were less than that at reference sites. The mean densities of the remaining three species did not differ (P > 0.05) between backbones, warps, and reference sites. The mean abundance of invertebrates was significantly and positively (three species) or negatively (three species) related to the percent shell cover on the seabed. Recovery of four epibenthic species was rapid with mean densities beneath backbones similar to reference sites after 1- to 2-y following farm closure. Densities of the fifth species resembled that at reference sites after 10 y whereas the remaining three species did not differ from controls. Multivariate analyses showed that epibenthic invertebrate communities beneath warps were similar to reference sites throughout the entire 11-y study period. Communities beneath the two backbones differed initially from reference sites but were similar to reference sites 6- and 8-y postfarm closure. After establishing the importance of percent shell cover as an indicator of effects on epibenthic species, data from multiple mussel farms in the Marlborough Sounds were analyzed. These analyses showed that the mean percent of mussel shells beneath backbones (x̄ = 29.5%, range = 0%–100%), exceeded that beneath warps (x̄ = 2.3%), and at reference sites (x̄ = 0.08%). Analyses also showed that the mean maximum distance of mussel shell deposition was 13.5 m distance (range = 1.5–26.5 m) away from farms and was positively (P < 0.01) related to the depth of water where farms were located. Based on multiple mussel farms, mean mussel shell cover declined to below a possible impact threshold of 7% by 7.5 m distance for farms in <12 m depth and 18.5 m distance for farms located in deeper water (>20 m depth). Taken together, the effects of mussel farm on epibenthic species abundance as well as overall community structure showed relatively localized impacts, that recovery occurred within 1–10 y following farm removal and that percent shell cover appears to be a useful indicator of effects of mussel farming on the epibenthos.