The goal of this project was to find ways of enhancing the efficiency of searching machine readable data bases. Ways are sought to transfer to the computer some of the tasks that are normally performed by the user, i.e., to further automate information retrieval. Four experiments were conducted to test the feasibility of a sequential processing hypothesis: a multi-step search process using Boolean search as the first step and subject term clustering as the second. The multi-step processing can be further strengthened by incorporating some semantic information into statistical string processing by the use of a new method of Automatic Term Classification (ATC). The results suggest an organization for information retrieval systems of the future in which several processing techniques are used during a single retrieval. Charts, tables, figures, and statistical data for the experiments are included. Appendices include all symbols used during the experiment; probability of term match formulas, computer programs used in the experiments; and sample mappings of selected words. The data bases used were selected files of Chemical Abstracts Services CACon and Engineering Index COMPENDEX. (Author/JPF)