With the basic features of the social stratification system in chemistry sketched out, it is now possible to use the findings of the 1993 SRP pilot project (Bensman 1996) to demonstrate how this social stratification system integrates with the scientific journal system in chemistry. Two of the major findings of this pilot project were the following. First, there was high correlation—0.72—between the LSU chemistry faculty ratings of journals and the total ISI citations to these journals. This high correlation was even more striking given the fact the LSU chemistry faculty was not a large sample randomly selected but a small, local group of 25 professors, and it strongly suggests that the LSU chemistry faculty were responding to an extremely powerful variable.
Second, the library market for chemistry journals is bifurcated, with costs concentrating in commercial, largely foreign journals, and scientific value (as measured by LSU faculty ratings and total ISI citations) concentrating in journals of U.S. associations. The data collected by the NRC show that there was a high correlation between peer ratings of the quality of the program faculty with total ISI citations to their publications and that the citations concentrated on the publications of faculty mainly at the traditional elite research institutions. It seems reasonable to deduce that the journals of U.S. associations are highly cited because the scientific elite (located largely at the traditional elite research institutions) is publishing in them. Therefore it appears that the LSU chemistry faculty—as any other freely forming band of primates—was responding to the imperatives of their social stratification system in rating journals.
The ST journal system is thus in many respects an external manifestation of the underlying social stratification system of science and technology. As such, the ST journal system displays many of the same characteristics as its underlying social stratification system. Strong evidence for this was discovered in the tests done to check the validity and stability of the faculty scores constructed for journals from the survey data gathered in the SRP. Such evidence will be presented below. For the time being, we will restrict ourselves to analyzing two major phenomena of the ST journal system: (1) the stability of citation patterns at the upper end of the distribution, and (2) "uncitedness," better defined as the zero citation class.
Similar to the academic social system, the scholarly journal system is remarkably stable at the top. This is particularly true of the citation patterns. In the initial exploratory studies at ISI leading to the creation of the JCRs, it was found that of the 206 SCI journals most highly cited in 1969, 169 or 82.0% remained in the top 206 most highly cited in 1974 (Garfield 1976, 609). This finding was corroborated in a study done at the British Library Lending Division (BLLD) (Line 1984). Here there was found an 83% overlap between the 200 SCI journals with the most citations in 1975 and 1980 as well as a 78% overlap between the 200 SSCI journals with the most citations in 1977 and 1982. The stability extended further down the ranks, and for the top 500 journals there was an 87% overlap between the SCI journals and an 80% overlap between the SSCI journals in the same 5-year comparison period.
Despite the consistency in these findings, Garfield (1996) felt compelled to return to this topic in order to dispel the myth of an ever-rising tide of scientific journals. Garfield demonstrated the stability of the concentration of citations over time by pointing out that in both 1989 and 1994 a mere 500 SCI journals accounted for more than 70% of all citations. He then utilized data from the SCI JCRs to show that 48 of the 50 science journals most highly cited in 1989 were still among the 50 most highly cited science journals in 1994. It should be pointed out that the above studies were done on a global basis without breaking down the journals into subject sets, thereby indicating not only the stability of citation patterns within disciplines but also among disciplines. This corroborates the finding of Metz and Litchfield (1988) with respect to the stability of the subject patterns in library use at VPI.
In comparison with the stability of the top of the journal citation distributions, the zero citation class—or "uncitedness," as it is known in the literature—is a much more complex phenomenon. Similar to the zero class in library use or any other distribution, it is extremely difficult to determine the extent of uncitedness, because in essence one is measuring a phenomenon that is not manifesting itself and consequently one does not know what truly belongs in this class or what might be there for other reasons. Price (1986, 107) made one of the first attempts to assess this phenomenon, estimating that 10% of all published papers would never be cited. Stern (1990,193–94) reviewed 8 studies of uncitedness published from 1974 to 1983, and the uncitedness rate found by these studies ranged from 0.45% in chemistry to 49.9% in pharmaceutical literature–the latter being so high that Stern considered it an outlier.
Uncitedness attracted public attention when the journal Science reported the results of an ISI study indicating that 55% of the scholarly papers published between 1981 and 1985 had received no citations in the five years after they were published (Hamilton 1990). The controversy that swirled around the reporting of this "fact" reveals the importance of proper set definition in library and information science. In a follow-up study done at ISI and again reported in Science (Hamilton 1991), the rate of uncitedness was found to be 47.4% in the hard sciences including engineering and medicine, 74.7% in the social sciences, and a stunning 98% in the arts and humanities.
However, it turned out that included in these studies were such "papers" as meeting abstracts, editorials, obituaries, letters, etc., and, when these were excluded, the rates of uncitedness dropped to 22.4% in the hard sciences, 48.0% in the social sciences, and 93.1% in the arts and humanities–still nothing to stand up and cheer about. A further breakdown by nationality in the hard sciences revealed that U.S. authors had an uncitedness rate of 14.7% and that foreign authors had a rate of 28.0% (Pendlebury 1991).
The zero citation class has not been studied to any great extent, and much research remains to be done on it. However, two general features can be described. The first pertains to set definition. Not only does the rate of uncitedness vary from discipline to discipline—as is evident from the findings described above—but it also varies within disciplines and between disciplines on the basis of whether the intention of the publication is basic research or application. Narin, Pinski, and Gee (1976) found that in biomedicine basic research journals and disciplines have a significantly higher citation "influence"–i.e., receive more citations than they give–than clinical journals and disciplines. In the measures constructed by Schwarz (1997) on the basis of ISI data, the uncitedness rate in engineering is 1.6 times the average for the physical sciences, while the uncitedness in library science is 1.5 times the average of its subject group, the social sciences.
The second general feature relates to the operation of the social stratification system of scholarship. Garfield (1991) noted that a paper is more likely to be cited if it is published in a highly cited journal than in lowly cited one, and Pendlebury (1991) points out that articles published in prestigious journals such as Science are almost never left uncited. In his study of the Journal of the American Chemical Society, Ghosh (1974) found that only 1 of 222 test papers (0.45%) published in 1965 remained uncited in the following six years. Ghosh (1975) did another study of uncitedness, this time with Nature, and his data show that of the 75 lead articles published in 1965, only 3% or 4% remained uncited through 1970. In their study of uncitedness in sociology, Bott and Hargens (1991) found that the average citation levels of journals were highly correlated with other measures of journal stature, and their figures showed that the percentage of uncited articles rose in almost inverse lock step with the drop in the average citation rate.
Revelations by ISI of the extent of the zero citation class confirmed the worst suspicions some people have about scientists and scientific publishing. These suspicions were once tersely summarized by Broad and Wade (1982, 221–22) with respect to what they describe as "the ocean of unread and unreadable articles of which the scientific literature is so largely composed":
In the first Science article reporting ISI's findings on uncitedness, references were made to the need to pad resumes, the struggle for grants and promotions, etc., and MIT biology professor Richard Young was quoted as doubting whether the scientific enterprise would suffer if the bottom 80% of the literature would just vanish (Hamilton 1990). More ominously, in a Newsweek article subtitled "Does the Country Really Need All those Scientists?" Begley (1991) made reference to the ISI findings and questioned the need for constantly increasing the federal science budget. In the same article the declaration that "scientists . . . with their belief in their God-given right to taxpayer dollars are little more than 'welfare queens in white coats'" was attributed to the materials scientist Rustum Roy, professor at Penn State.
However, Seglen (1992) was perhaps correct when he ascribed the skewed pattern of citations to the operation of a basic probability distribution. He theorized that the skewness implies that there will always be a large fraction of uncited articles that will be impossible to eliminate for statistical reasons. The zero class is probably a necessary part of the cost of doing science and technology; the only question may well be how best to handle this class from the economic point of view.
Having thus set forth the theoretical principles underlying the functioning of the ST serials system, it is now time to demonstrate how this theory can be put into practice to evaluate the ST serials holdings of an actual university.