Kraft, Donald Harris. Ph.D., Purdue University, January 1971. The Journal Selection Problea In A University Library System. Major Professor: Thomas W. HilT Jr This dissertation is concerned with the problem of selecting the proper Journals to be acquired by a university library. A general model is formulated as a zero-one linear programing problea with an objective function that evaluates various selection policies in terms of the net worth of selecting specific Journals and rejecting others, journal usage and Journal productivity are two Important means of Measuring the worth of acquiring a specific Journal; and models are developed to describe both measures as dynamic, time-dependent variables. Journal usage is adopted as the better measure of Journal worth, using a modified Markovian approach to describe expected usage patterns over time. The constraints are primarily related to cost restrictions that arise due to budgetary controls. The relevant costs are those involved with the ordering, storage, and circulation of acquired Journals. The only other type of constraint deals with continuity in that once an item has been acquired, it reseing a permanent part of the library's collection. The optimitation of the Journal selection model is considered from two different approaches since its large size does not allow for a fit to one of the now existing computational algorithms for sero-one linear programing problems. A dynamic programing formulation allovi us to decompose the large problem into several smaller ones of the knapsack variety and generate a feasible solution if at least one exists. Special cases are found, some of which seem quite realistic In the light of practical library selection procedures, where the solution la also optimal. A Lagrangian formulation gives an upper bound on the optimal value of the objective function and also generates several near-optical solutions, scne of which nay be feasible. A Knapsack approach is then again taken to try to improve the best feasible solution at hand to yield a better, near-optimal or optimal, feasible solution. A small example problem, illustrating how the algorithms operate, is presented and discussed. The computational results show that a solution can be generated in a reasonable amount of time and computer storage space, for at least small and medium-sized problems.