MONO91 Indexes Compiled by Machine chapter Mary Elizabeth Stevens National Bureau of Standards necessitate some other form of retrieval. This is the proportion which is involved with the problem of generics, where a term in one system subsumes two of another ---and vice-versa. An additional problem evolves in attempting to reconcile two different subject concepts, one, the subject heading which usually has a single access point and one, the uniterm or descriptor which has multiple access through coordination. Thus the practicality of a system made up of many units supplying information indexed differently, using as a basis for retrieval a table of equivalents, is questionable." 1/ Moreover, the results of tests of inter-indexer consistency rates within the same agency were not encouraging. Thus Painter further concludes: "Tne study, in combining the results of the equivalency analysis and the consistency of indexing within each system and an equivalency of only 30 percent within the broadest system, a table of equivalents is at present of little value in either a manual or a machine system. In order to apply a table of equivalents efficiently, both a high degree of consistency and a high degree of equivalency is essential." 2/ She therefore stresses that the possibilities for conversion by machine techniques from one indexing set to an equivalent set for another vocabulary are adversely affected by the generally poor rates of inter-indexer consistency. With reference both to the Datatrol Studies 3/ and to corroborative findings of her own, she states: "The value of equivalency studies and most particularly the table of equivalents presuppose the consistency of indexing. Convertibility between systems is thus dependent on the consistency of indexing. Without consistency, the vocabularies as units are not sound; equivalencies cannot be drawn or effectively used for convertibility." 4/ 1/ 2/ 3/ 4/ Painter, 1963 [460], p. 104. Ibid, p. ix. Hammond, 1962 [250]; Hammond and Rosenborg, 1962 [252]. Painter, 1963, [460]. p. 109. Note that these estimates of nter-indexer con- sistency may be quite optimistic, as discussed on pp. 157-l6Oof this report. 39