MONO91 Indexes Generated by Machine-Automatic Derivative Indexing chapter Mary Elizabeth Stevens National Bureau of Standards A major question in the area of user acceptability, however, is that of the adequacy of title alone to tell the searcher whether or not a specific document is relevant to his query or interest. A number of investigators, both documentalists and user-scientists, suggest that this is rarely the case. 1/ In fact, for many users, titles alone provide only a negative searching device--in an announcement bulletin or abstract journal the user's scanning of titles merely tells him whether or not he should read the abstract and then perhaps go on to the paper itself. It is for reasons of this type, in all probability, that Montgomery and Swanson found less effectiveness of titles on *re1evance[OCRerr]judgment tests than might be suggested by their more optimistic findings as to the success of machine procedures for replicating human subject heading assignments. Whereas they have claimed that about 90 percent of test items could have been as successfully indexed by machine as by manual procedures, (Montgomery and Swanson, 1962 [421]; Swanson, 1962 [584]), they have also reported that: 1'Comparison of title relevance judgment with judgment based on full text examina- tion indicates that titles are only about one-third effective (i.e., two-thirds of the relevant articles would be judged irrelevant) as the basis for estimating the relevance of the article to a given question". 2/ They go on to suggest, therefore, that ". . indexing should be based on more than titles and. . . a bibliographic citation system should present to the requester something more than titles. 1' 3/ Similarly, Jahoda reports in an analysis of 281 actual search requests at Esso Research and Engineering that only two-thirds could have been answered with a shallow index based on titles and major section headings of the documents and that answering the remainder of the requests would have required an index of considerable depth. 41 The obvious factors affecting the utility of titles as the source of indexing[OCRerr]searching clues include, first, the limitation of most titles to the principal subject matter, the main topic or topics of the document. The display of title context does to some extent provide for modifications of the topic to the special aspects treated, but it is of course obvious that a title cannot possibly provide clues to subject content not implied in the words of that title. In many cases, the potential user wants information contained in the paper, or even 1/ See, for example, Atherton and Yovich, reporting on evaluations by physicists of experimental citation indexing, 1962 [26], p. 22: "The reliance on titles of papers for retrieval purposes was not sufficient"; Levery. 1963 [359], p.235. "Titles are usually insufficient to furnish a correct index to the text"; Hocken, 1962 [274], p. 93: "The titles were not explicit enough"; Crane and Bernier, 1959 [145], p. 1053: "Li sts of titles can be prepared rapidly, but they are inadequately useful in selecting articles of interest, and they provide little or no directly usable [OCRerr] Dowell and Marshall, 1962, [159], p.324: "Frequently titles either lack sufficient detail or are in fact misleading"; Connolly, 1963 [136], p.35: "Most titles are inadequate as descriptions of the contents of papers." 2/ 3, 4' Montgomery and Swanson, 1962 [421], p.364. Ibid, p. 366. Jahoda, 1962 [298], p. 75. 61