A prototype for guideline systems is the Minnesota system, which took effect in 1980. This system uses a sentencing grid that produces a specific sentence when applied to a specific factual situation. These guidelines allow some disparity, but they are based on the notion of equal treatment of factually identical cases. However, sentencing reformers have overlooked Jerome Frank's statement, made more than 50 years ago, that sentencing decisions result from the substantive rules as applied to the facts of individual cases. The Guggenheim Sentencing Project at Yale Law School, an ongoing research project focusing on sentencing in six jurisdictions, has shown ways in which the procedures that develop facts for sentencers are often unreliable and erratic. For example, facts developed at trial are often more detailed than those presented during plea bargaining, which is the main way guilty pleas occur. In addition, jurisdictions differ substantially in who conducts the initial interviews of witnesses, victims, and police officers, and these differences can affect information development. Moreover, plea negotiations can involve either sentence bargaining or charge bargaining, each of which can result in different facts being presented. Furthermore, presentence investigation reports often omit information, contain erroneous information, or are slanted. Finally, wide variations exist in procedures used in sentencing hearings. Better identification of the facts relevant to the guideline systems and the use of habeas corpus actions can help improve accuracy. Further research should focus on the nature and extent of factual errors. 116 footnotes.