Following the lead of Bazemore and Walgrave (1999), this study examines harm and repair as the core concepts of restorative justice. Through observations of one program in action, the author analyzed how participants in a restorative justice initiative attempted to repair harm, including how they defined harm caused by criminal offenses, how they negotiated agreements with offenders to repair harm, and what difficulties they faced in fulfilling the central tenet of restorative justice. He analyzed videotapes of probationers in Vermont negotiating reparative contracts with victims, as well as of volunteers serving as community representatives. Of concern was how participants articulated the damage caused by the offenses and how they developed strategies to repair it. Difficulties encountered were also identified. The cases observed varied responses to similar offense categories, which is appropriate, because the same type of offense can have differing effects on victims and the community. Program efforts failed, not when they treated similar cases differently, but rather when they repaired harm in some cases but not in others. Vermont's community reparative boards typically negotiated reparative agreements that required apologies, restitution, and community service. This was true for 85 percent of the cases in this study. Most cases, therefore, fulfilled at least the thin criterion for restorative justice. Sixty-nine percent of the contracts contained restorative elements that were linked to the harm caused by the offense; only 19 percent of the cases, however, consistently linked harm with repair. Restorative justice was often ill-defined in the cases, particularly among practitioners. Harm was often ill-defined, thus making repair difficult. This failure to link harm to repair may lead to retributive contracts; e.g., using community service as "punishment" by assigning hours of labor arbitrarily without articulating how this labor makes amends for an identified harm. 3 tables and 50 references