MONO91 Indexes Generated by Machine-Automatic Derivative Indexing chapter Mary Elizabeth Stevens National Bureau of Standards It is also of interest to note, moreover, that the very existence of machine-generated permuted title indexes should greatly increase the likelihood that authors will use better and more useful titles. 1/ At a seminar on word and vocabulary byproducts of permuted title index[OCRerr]ng held at Biological Abstracts headquarters on October 8, 1962, Rigby of Meteorological and Geoastrophysical Abstracts reported informally that as of that time there was already discernible improvement in titles covered by their KWIC index. In the same year (1962), Tukey similarly stated that: "Chemjcal Titles has been heavily enough used to affect the construction of titles of papers on chemical subjects. " 2/ Instructions to authors of the previously mentioned 11Short Papers" 3/ for the A. D. I. 1963 Annual Meeting specified that at least six significant words should be included in their titles and nearly all authors did in fact comply. Two of the "Short Papers" are specifically directed to the topic of improvements that authors can make in writing their titles (Brandenberg, 1963 [80]; Kennedy, 1963 [312]). Instructions of this type can be effectively used for situations where all authors are under the same administrative control, as in the internal reports prepared in a single organization. This type of situation, incidentally, is one for which KWIC proponents are often most enthusiastic (Kennedy, 1962 [310]; Black, 1962 [65]; Linder, 1960 [362]). Finally, there is considerable promise that pressures brought to bear by journal editors of the publications of professional societies, notably the American Institute of Chemical Engineers and other cooperating member societies of the Engineers Joint Council, will result in improved adequacy of titles and thereby increased effectiveness of title word indexes. Certain other disadvantages of KWIC indexing techniques, however, relate specif- ically to operational problems and requirements in the machine production of these indexes. There is, first, the problem of the amount of context that is usually displayed--that is, the question of line length--and the related problems of title truncation and wrap-around. As Kennedy notes: "Progressive shifting of the title to bring a given word to the indexing column frequently causes portions of the title to exceed the line space available, first at the right margin, then the left, or even both simultaneously. " 4/ A case in point is the perhaps apocryphal "EROTIC TENDENCIES AMONG TRAPPIST MONKS" where "ATHEBOSCL" had been dropped off at the left. For multi-column KWIC indexes, in particular, where the line length is typically 58-60 characters, "much of the relevance is lost because the reader sees the wrong slice of the title". 5/ The Bell Laboratories KWIC index, 6/ Chemical-Biological Activities, 7/ 1/ 2/ Tukey, 1962 [611], pp. 9-10. 3/ Luhn, 1963 [376] and [377]. 4/ 5/ 6/ 7/ See for example, Black, 1962 [65], Kennedy, 1961 [311], p. 117. Brandenberg, 1963 [80], p. 57. Kennedy, 1961 [311], p. 118. Figures 4 and 5. p. 317; Youden, 1963 [658], p. 332. 63