Natural loss of hydrocarbons was often low from mussel (Mytilus trossulus) beds (which were typically not cleaned after the Exxon Valdez oil spill), thus this habitat remained a long-term source of oil. Consequently, experimental restoration of nine contaminated beds was attempted in 1994; mussels were removed, contaminated surface sediment was replaced (33 metric tons), and original mussels were returned. Hydrocarbon concentrations and mussel populations were monitored for 5 years thereafter. Post-restoration mussel population fluctuations were indistinguishable from regional changes. Increased short-term oil loss was apparent, but long-term (5 year) improvement was equivocal and difficult to distinguish from natural losses. By 1999, oil concentrations in mussels were typically at baseline levels in restored and oiled reference beds; concentrations in replaced sediment were elevated in one third of restored beds, indicating recontamination from underlying or surrounding sediment. Our results suggest mussel relocation is feasible but suggest oil might more effectively be removed from sediment mechanically or chemically than manually.