This study conducted to determine representation impact on information items retrieval in terms of precision and recall performance and overlap used the INSPEC "Computers and Control Abstracts" loaded on DIATOM, an online retrieval system based on DIALOG, as the database to be searched. Sixty-nine users provided 84 queries which were searched for high recall by intermediaries under each of seven representations: title only, abstract only, descriptors, identifiers, title and abstract, stemmed title and abstract, and the descriptor and identifier fields. Copies of the retrieved citations and abstracts were sent to users for judging relevance. Then the seven representations were tested using a latin square design on the 84 queries. Measures of recall, precision, and total retrieval of citations were analyzed using standard analysis of variance computations; the performance measures and overlaps findings are presented in detail. The results confirm earlier observations that there is relatively little difference in performance among the representations and relatively little overlap. Plans for observations and findings replication of the first phase and theory development for Phase II are described. Eleven tables, 19 references, and five appendices are provided. (RBF)