MONO91 Indexes Compiled by Machine chapter Mary Elizabeth Stevens National Bureau of Standards the Harvard Computation Laboratory (Salton, 1961 [512], 1962 [513], 1963 [514] and [515]), are concerned with citations as a basis for grouping and categorizing sets of related documents. Early examples of citation indexes that have been produced include the precedents in the fields of statistics and information theory listed by Tukey. 1/ Tukey also refers to early experimentation involving manually manipulated card files by J. L. Hodges, Jr., Charles H. Kraft, and William H. Kruskal. [OCRerr], Goodman (1963 [235]) describes the use of Termatrex cards showing for each item other items cited by it. Examples of machine-compiled citation indexes, however, are those of Garfield and Sher in the field of genetics (1963 [546]), Lipetz's experimental index to the citations in the proceedings of the two United Nations conferences on the peaceful uses of atomic energy, (1961 [364], 1960[365]), and the citation indexto references listed inthe tiShort Papers" submitted for the 1963 Annual Meeting of the American Documentation Institute (Luhn, 1963 [OCRerr]377]). As of January, 1964, the first five volumes of Science Citation Index are available from the Institute for Scientific Information. These volumes are reported to have 2, 250, 000 lines of copy representing the computer-compiled citation trails for 102,000 articles published in 1961. Al Preliminary evaluations of the citation indexing principle have, as noted previously, been carried out in an American Institute of Physics project supported by the National Science Foundation. One experiment involved the selection of a single paper from the December 1, 1961 issue of The Physical Review and the tracing of references and citations through that journal for the period y956 tQ 1960 A bibliography of 64 papers was pro- duced as a result. This was then evaluated by a nuclear physicist, who found that the titles alone were an insufficient basis for judging whether or not these papers should all have been in[OCRerr]cluded, and who commented critically that there was no way of knowing if all the papers really relevant to the subject of the test paper had indeed been found. A further check by search of the subject index did in fact reveal six pertinent papers which had been missed by the citation indexing technique. A second experiment at the American Institute of Physics involved application of Kessler's "coupling strength" criteria to 41 of the 64 papers selected in the first experiment, the remainder being excluded because they shared no references with any other paper. The resultant groupings of presumably highly related papers were also evaluated by a subject matter specialist, who found them relevant to each other but the selection incomplete. Atherton and Yovich, reporting these A. I. P. experiments, con- cluded that: 1'More work will have to be done before the usefulness of citation indexing can be accurately determined." 11 21 3/ 4/ Tukey, 1962 [611], pp. 23-24. Ibid. p. 24. See news note, Special Libraries, Jan. 1964, p. 58. Atherton and Yovich, 1962 [26], p. 22. 35