Changes in the use of literature with time - obsolescence revisited

Gupta's analysis of citations to the 1983 volume of Physical Review (1990), to which he applied corrections for growth, showed a decrease in citation which fitted an exponential model. He warned against assuming that this result necessarily reflects use.

There has been one big change since 1974. At that date, the literature was still growing fast. Since then, growth rates have slowed substantially; not only is the annual net increase in the number of serials quite small, but the number of articles per serial has grown very slowly in most subjects--in some not at all (Archibald & Line, 1991). The need to apply corrections to synchronous data on citations received by articles published in the last twenty years should therefore be rather less, at least in some subjects. In fact, such synchronous studies should show slower citation or use decay, another hypothesis that deserves testing.

More attention has been devoted to the social sciences in recent years. In addition to the articles already mentioned (Line & Carter, 1974; Line, 1981; Leavy, 1983) there are the papers by Rao (1974) and Oromaner (1977), the latter being an interesting, though limited, diachronous study. The humanities have received less attention; two examples (there may be others) are the studies by Longyear (1977) on musicology and Heisey (1988) on the Dead Sea Scrolls.

Highly "productive" journals (in terms of citations received) in desalination proved to have shorter active lives than unproductive journals, but journals between the extremes of productivity exhibited no pattern (Wallace, 1986). Bottle and Gong (1987) found that the content typology of articles in biochemistry affected their aging, the median citation ages ranging widely--from 2.9 (for physical and chemical properties of substances) to 9.3 (for studies on living organisms--cell or greater level). McCain and Turner (1989) analyzed eleven highly cited articles in molecular genetics, four of them aging slowly and seven quickly, and related these patterns of aging to their contents. In a wide-ranging citation study of the social sciences (Line, 1981), big differences were found between different disciplines; for example, the proportion of pre-1964 citations drawn from serials to sociology was 48 percent and to environmental planning 34 percent.

Citations to journals show a faster decay rate than citations to books, both in the social sciences (Line, 1981) and at least in some science subjects (e.g., phycology [Musib, 1988]). It is highly probable that this is a general phenomenon. It is certainly to be expected since, in general, articles are intended to report research at the frontiers and therefore date more rapidly than books, which consolidate knowledge.

Several studies have looked at the use of articles rather than journals over time. There are possible practical applications of this kind of study (see later discussion), but it is also of some theoretical interest, not least because it qualifies ideas that journals obsolesce in some regular sort of way. Walsh's (1977) dissertation on articles in physics followed up on a much larger scale two small studies (Line et al., 1972; Line, 1974), which showed that articles that were least cited in the first three years became less and less cited, while the most cited, far from "obsolescing," attracted more and more citations. Walsh found an increasing concentration of citations on a limited number of articles as time passed. A further study (Line, 1984), also in physics, provided additional confirmation. Aversa (1985) studied 400 very highly cited papers in science and discovered that they fell into two groups: one group declined in citedness after the first two or three years, the other after the sixth or seventh year. For patents, however, Noma and Olivastro (1985) found that highly and lowly cited papers "obsolesced" at the same rate.