MONO91 Indexes Generated by Machine-Automatic Derivative Indexing chapter Mary Elizabeth Stevens National Bureau of Standards Bernier and Crane who report that for the non-organic chemistry items covered by Chemical Abstracts, 34 percent of the entries can be derived from the titles. 1/ With respect to the Biological Sciences Communications Project studies, Shilling reports as follows: `1Titles of scientific articles are being utilized at present in a great many ways under the general assumption that there is a positive correlation between the title and the content of the article. A study was undertaken to analyze the accuracy of titles in describing the content of biomedical articles. It was conducted in two parts. In part one, a group of scientists were asked to predict the content of selected scientific articles, in their area of interest, from the title, the author1s name, and the name of the journal in which it appeared. The results of the first phase of the study on the first trial journal were so diverse as to make analysis impossible, and this part of the study was not pursued further. From this small segment of the study it appears that scientists are deluding themselves when they search by title only and then decide what they wish to read. "In the other half of this experiment, the article without title, author1s name, or journal name was sent to 20 scientists, selected as experts in the scientific field of the article, who were asked to write a meaningful title. Fifty articles were used, five from each of ten selected biomedical journals. From this part of the study it is apparent that if the article is in a field which is relatively well standardized and has an accepted vocabulary, it is possible for a group of titlists to agree remarkably well on an appropriate title. However, if the article is loosely organized, contains more than one subject, or is in a specialty in which there is no standard vocabulary, then titling scientists fail to agree to a rather alarming extent. _ Other studies involving the question of usefulness of titles alone for indexing purposes include those of Doyle, Lane, Montgomery and Swanson, O'Connor, Ruhl, Swanson, and White and Walsh, among others. Doyle checked the retrieval loss likely to result from the synonymity-scatter problem for a permuted title index compiled in 1958 to the internal reports of the System Development Corporation. He found, for example, that for 12 direct references to McGuire Air Force Base, there were one to "New York Air Defense Sector", two to "New York Sector", ten to "NYADS" and five to "N.Y. Sector". 1/ 2/ 3/ Bernier and Crane, 1962 [56], p.120. Shilling, 1963 [551], pp.205-206. Doyle,[OCRerr]96[OCRerr] [166], p. 11. 58