Literature loss has a dual meaning. In the research process, it refers to the difference between what exists in some published form and what is likely to be found in a literature search. On an institutional level, it refers to the declining ability of American research libraries to maintain comprehensive collections in the wake of the extraordinary growth and price inflation of scholarly literature since the I970s. Recent technological innovations in the library field provide the means not only to measure literature loss on an institutional level but also to alleviate it in one's own research work. An application of this technology to the field of anthropology is presented here in two parts: (i) an analysis of the growing gap in anthropology between book publication output and the aggregate holdings of 70 institutions in the Association of Research Libraries (ARL) and (2) a discussion of the relevance of the new technology to the closely related question of information retrieval in the research process. Underlying any analysis of literature loss is the problem of assessing the size of the universal stock of academic books in a given field. That problem has no definitive solution or any prospect of one. Moreover, even if accurate book-publication figures could be compiled worldwide, it would hardly make sense to consider them as realistic or worthy goals for research library collection development. The ancient ideal of the Library of Alexandria-to hold everything ever published-has been considered impossible and undesirable since the early part of this century. For the purpose of the analysis presented here-to assess the financial performance of American research libraries within reasonable bounds-book-publication output is derived from the Online Computer Library Center, Inc. (OCLC) Union Catalog (OLUC) bibliographic database. The decision to use this database recognized the utter infeasibility of attempting to locate the publication lists of hundreds of university and trade publishers, scholarly societies, research bodies, and the like, and involved the simplifying assumption that just about any book that appeared over the past decade in the Americas or in Europe would have shown up in the OCLC-OLUC system, which contains the shared cataloging of more than 4,800 libraries, including the Library of Congress, in 26 countries. That part of book output that is held in at least one of the 70 ARL institutions is derived from a different database, the OCLC/AMIGOS Collection Analysis Computer Disk (CACD), developed in I989, which records ARL holdings by year of publication. Figure i compares the aggregate holdings of 70 ARL institutions with total book output in all areas of anthropology for the period I978-87.2 Line A shows that output fluctuated from a high of I,II3 titles in I980 to a low of 934 titles in i982. On average, I,008 titles were published annually, with a standard deviation of 40. Line B shows that holdings in anthropology of the 70 ARL institutions fluctuated from a high of 694 for titles published in I980 to a low of 507 for titles published in I987. On average, 6o6 titles were acquired per year of publication, with a standard deviation of 56. Overall, ARL holdings in anthropology dropped from 66% of book output in the late I970S to 53% in the late I980s. Underlying the deteriorating position of American research libraries are certain inflationary tendencies built into the scholarly communications system. During the I978-87 period, the average price of a hardbound book on anthropology increased I67%, from $18.23 to $30.49 (Bentley I99I:404), and the cumulative inflation rate for institutional subscriptions to anthropology and sociology joumals increased 327%, rising to 45I% by I99I (Carpenter and Alexander I99I:59). Line C demarcates the mainstream literature, books held by at least io% of ARL institutions, from the gray