MONO91 Conclusion chapter Mary Elizabeth Stevens National Bureau of Standards any objective measure of comparative cost-benefit ratios may be obtained. Black reports, but without supporting data, that: "It has been estimated that the efficiency of KWIC indexing is about 76 per cent com- pared with about 82 per cent for conventional indexing or classification." 1/ White and Walsh report that: "From the limited experiment on methods of indexing the 1962 issues of the Abstracts of Computer Literature, the permuted title indexing retrieved only 52 percent of the information. This low percentage may be attributed to the changing and not yet uniformly standardized terminology existing in computer technology." 2/ KWIC indexes, because of their very currency, are fulfilling significant maintaining- awareness needs today. Improved titling practice, enforced by editorial rigor or contract- ual requirements or both, can improve their usefulness. They fill gaps in the bench scientist's or engineer's ability to know about what might be of interest to him, either because the material is not otherwise covered in normal secondary publication (e.g., con- ferences and proceedings of symposia, internal technical reports not produced[OCRerr]n Govern- ment contracts and therefore not announced and indexed by the cognizant agencies, and the like) or because the sheer bulk of the product of indexing-abstracting services in his field prevents his effective use of these services unless more specific access points are pro- vided. The claim that "something is better than nothing" is not without merit, 3/ even with all the problems of non-resolution of synonymity, homography, topical scatter, long blocks of entries under the sorting term, the even more significant disadvantages of author- bias towards his principle topic, the author's choice both of emphasis and terminology, and the like. Williams3 considering word-with-context indexes, whether limited to title only or to titles with readily available augmentation, makes the following comments: "Limitations and other troublesome features of the method have been obvious, but perhaps over obvious, in the light of its growing acceptance and of the basic validity of permitting a document to speak for itself, even in a much abstracted recapitulation. Wherever there are large and growing problems in maintaining publication schedules for established subject indexes, or wherever pressing needs develop for more fre- quent indexes, for rapid, low-cost cumulation, or for indexes in areas where suit- able indexing services are wanting, there no apology is needed for proposing that this method be considered and tried, as a precursor to `better' indexing, if not as a substitute. Its use may be of interest also in less troubled circumstances, in its own right, and because of common elements involved in its production and the pro- vision of other wanted products and functions (catalog records, current-awareness, lists, etc)." 4/ Returning to the question of whether automatic indexing is possible, it can be seen that, at least in the derivative indexing sense, it is not only possible but can be practically useful. To dismiss the evidence of automatic derivative indexing operations that are in production today by rigorous definition of what indexing is in effect anticipates both our 1/ Black, 1962 [65], p. 318. 2/ White and Walsh, 1963 [639], p. 346. 3/ See Veilleux, 1962 [624], p. 81: "Accepting the premise that partial control of in- formation satisfies more consumers than absence of control, perfection was traded for currency." 4/ T. M. Williams, private communication, dated January 43 1962. 176