The transient effects of increases in the average sentence length are studied. Initially it is assumed that the changes in sentence structure are implemented immediately, and that the offender population is invariant with time. The results show that, in a wide variety of examples, the prison population does not rise monotonically from the old steady state to the new one. Rather, it increases to a short term peak, and subsequently declines to the long term value. Two types of changes are reviewed in sentencing policy. Under the Type I change, those receiving positive sentences are given longer sentences on average, and those receiving suspended sentences are unaffected. Under the Type II change, those receiving suspended sentences are given positive sentences, and those receiving positive sentences continue to receive terms of the same length. The ratio p of the initial increase to the final increase varies with the type of change in sentence structure and the offender population to which it is applied. For a homogeneous offender class, p is, on average, 1.14 for a Type I change, and 1.19 for a Type II change; for a subgroup of high rate offenders, it is, on average, 1.22 and 1.30 respectively. Finally, the model is used to study the Pennsylvania prison population. From data about the recent increases in the prison population, the changes that must have taken place in sentencing policies are developed. Future levels of the Pennsylvania prison population are then projected, based on the transient effects of these changes. If there are no more changes in sentence structure and criminal behavior in Pennsylvania, then, given our conjectures of the changes in sentencing policies, the increase in the Pennsylvania prison population will soon cease and the level will decline after 1992. 7 references, 20 figures, 7 tables. (Author abstract modified)