Researchers have learned some things in the course of assessing imprisonment and the crime rate. They no longer use cross-sectional data sets because it is impossible to separate simultaneous effects. Also, researchers no longer use national time series data and ratio variables because they produce inflated estimates. Better research methods have improved the validity and narrowed the scope of recent estimates. Most studies show that doubling current U.S. prison capacity would reduce index crime rates by 20 to 40 percent. Nonetheless, some problems persist that relate to simultaneity, specification error, and difficulties in comparing among States. In the case of simultaneity, prison affects crime and crime affects prison and it is difficult to isolate one effect from the other. Specification error may involve left out variables, while difficulties in comparing among States may be related to the fact that States use prisons very differently. Perhaps most important, the range of estimates of the effect of imprisonment on crime falls in an awkward spot. At the low end, further prison construction is probably not cost-effective. At the high end, further prison construction is very likely cost-effective. Further research is recommended on the cost-effectiveness of alternatives to prison, especially in the context of scarce resources. An appendix lists variables included in crime equations. 109 references, 7 footnotes, 4 tables, and 2 figures