Previous evaluation research has provided a great deal of empirical evidence that conferencing yields good outcomes for offenders and victims relative to other interventions (Schiff, 1999; Umbreit, 2001; Bazemore, 2000; Braithwaite, 2002). Yet this research has not been designed to test inferences about the relative effectiveness of various competing strategies used in conferencing practice. For example, many practitioners believe strongly that face-to-face preparation of victim and offender is essential to good conferencing outcomes while others do not. Some believe that it is important to have the victim speak first in the conference while others argue equally strongly for having the offender go first. Some believe that including only the victim and offender in conferences will yield the most effective outcomes for both, while others argue strongly that either supporters of the victim and offender, or representatives of the broader community, or both, provide the most effective mix of participants. There are numerous examples of such micro-level practice variations in the restorative conferencing movement, as well as middle-range policy choices that may have significant implications for conferencing outcomes. One can find practitioners who make strong logical arguments for a specific approach based on normative preference or strong beliefs about the effectiveness of this strategy. There is, however, little if any empirical evidence to support any of these positions.