Recent research into sentencing strategies by the Shinnars, Wilson, Van den Haag, and Ehrlich all assume that harsher sentencing will prevent more crime than is actually cleared by arrests. Evidence gathered by the Dangerous Offender Project in Columbus, testing the effectiveness of 18 different sentencing policies, suggests that such an assumption is in error. Using computer techniques, the study applied the 18 policies to all those offenders charged with violent crimes in 1973. The most severe policy would have led to a reduction of 6 percent of the violent crime in 1973 but would have required 5 year sentences for those convicted of serious nonviolent crimes and for first offenders as well. A look at homicide cases in Canada supports the project's evidence. A 5 to 10 percent reduction in crime might be achievable through incarceration, but it would require massive imprisonment of many who have never or will ever commit a violent crime. There is also the possibility that harsh sentences, especially when mandatory, decrease the likelihood of a conviction. Sentencing policies which would increase the certainty of apprehension and conviction may have a greater deterrent effect than those that would increase the severity of sentences. Footnotes, tabular data, and 21 references accompany the text. For related documents, see NCJ 76862 and 76864-5. (Author abstract modified)