In a series of reports published by LEAA, L. Paul Sutton presents the results of a statistical study of Federal criminal sentences. In the last report of the series, Sutton uses regression analysis to develop a formula for predicting sentences from information about the crime committed, the offender's personal history, and statistical information about the district court in which the offender was sentenced. Sutton describes the formula as an expression of national sentencing policy. He suggests that the formula, by revealing unarticulated policy that now informs the imposition of sentences, can serve as a starting point for developing explicit policy in the form of sentencing guidelines for the Federal courts. It is suggested that Sutton's analysis is grounded in one of the most common errors in the interpretation of statistical data: the unwarranted inference of a cause-and-effect relationship from a showing that phenomena are statistically associated. In his formulation, the agents of the causality are the judges; the unwarranted causal inference is that the predictor variables in the formula represent the criteria that judges use in arriving at sentencing decisions. Statistics alone cannot indicate the direction of a causal link. Sutton's method is, therefore, incapable of identifying policies or criteria that guide judges in imposing sentences. Development of a national sentencing policy should begin by asking judges what criteria they employ in making sentencing determinations. By talking with judges, polling them at various stages of development of the guidelines, and getting their comments on guideline drafts, analysts can effectively incorporate these criteria. Twenty footnotes are included.