Citation and the Construction of Facts

Paula Treichler, a discourse analyst and cultural critic, has questioned the realist presumption that the HIV hypothesis became fact simply "because it's true." With reference to Gallo and his close associates, Treichler has argued: "By repeatedly citing each other's work, a small group of scientists quickly established a dense citation

network, thus gaining early (if ultimately only partial) control over nomenclature, publication, invitation to conferences, and history."^[6]

Yet Treichler provided neither evidence nor elaboration to support this claim. To consider the acceptance of Gallo's hypothesis more carefully, I performed a content analysis of those scientific articles published in seven major journals in 1984, 1985, and 1986 that cited Gallo's crucial 1984 article published in Science (the one specifically advancing the causal argument).^[7] Within the pages of these journals during the period 1984–1986, there were 244 articles that cited Gallo, ranging from 19 in Nature to 66 in Science . These were the pivotal years, as evidenced by the citation trend in all scientific publications: Gallo's article was cited 116 times in 1984 and 410 times in 1985; this dropped off slightly to 384 citations in 1986 and then more sharply to 341 citations in 1987 and 284 citations in 1988. Like most influential scientific articles, Gallo's enjoyed a brief period of heavy use during which more and more people found reason to invoke it—after which point the claims enunciated within it presumably became too much a part of common knowledge for citation to be necessary.^[8]

In my content analysis I investigated four factors: first, whether scientists and clinicians referenced Gallo's article in order to support qualified claims about the causal role of the virus ("HIV is the virus widely believed to cause AIDS") or unqualified claims ("HTLV-III, the etiological agent in AIDS, …"); second, whether qualified and unqualified claims were made more (or less) frequently over time; third, whether articles authored or coauthored by Gallo or any of his 1984 coauthors were more (or less) likely to make unqualified claims than those written by other authors; and fourth, whether the articles that made unqualified claims were more likely to cite later articles in conjunction with Gallo's in order to back up those claims. (For details on how this analysis was carried out, see the Methodological Appendix.)

Table 1 shows a striking shift over the three-year period. Only 3 percent of the 1984 articles that cited Gallo (2 out of 59) did so in conjunction with unqualified claims about the virus being the cause. Fifty-eight percent of them cited Gallo to support qualified claims, while 34 percent cited Gallo for other reasons and without making etiological claims. By the following year, the percent citing Gallo to make unqualified claims had jumped to 25 percent (26 out of 106). And by 1986, 62 percent (49 out of 79) of the articles cited Gallo to make unqualified claims, and only 22 percent did so in conjunction with qualified claims.^[9]

A glance at some of the more authoritative sources within the sample reveals the trend over time. A review article by the Council on Scientific Affairs of the AMA, published in October 1984, spent nearly two pages reviewing various etiological hypotheses and then concluded: "At the present time, the retrovirus etiology of AIDS is the hypothesis for which there is the most convincing evidence." However,

"even though the time may be close at hand when the definitive cause of AIDS is established, there is yet much to be explained." The authors noted several questions: "Why has AIDS been largely confined to certain groups while the majority of the population does not appear to be at much risk of acquiring the disease? Why does frank AIDS develop in only a small percentage of patients with the lymphadenopathy syndrome?"^[10] A year later, several prominent epidemiologists from the CDC wrote an article for Science reviewing the literature on the epidemiology of AIDS, noting in the opening paragraph that, in 1984, "a retrovirus variously termed lymphadenopathy-associated virus (LAV), human T-lymphotropic virus type III (HTLV-III), or AIDS-associated retrovirus (ARV) was isolated and shown to be the cause of AIDS [emphasis added]."^[11] Similarly, a group of investigators with the Division of Virology of the Food and Drug Administration (FDA) wrote in the Annals of Internal Medicine in late 1985: "The role of the human T-lymphotropic virus type III (HTLV-III) in the development of the acquired immunodeficiency syndrome (AIDS) is firmly established"—citing only the three "discovery" papers published, respectively, by Montagnier in 1983, Gallo in 1984, and Levy in 1984.^[12]

Expressions of doubt or skepticism—let alone support for other hypotheses—were extraordinarily rare throghout this period from 1984 to 1986. The Lancet 's initial editorials reporting on the Heckler press conference and on Gallo's articles in Science criticized the public fanfare and argued for a sober assessment of the facts. The journal noted that the finding of a virus "is no proof of causality," given the possibility "that these viruses are simply passengers—yet another opportunistic infection to which these patients are susceptible."^[13] "Nevertheless," the editorial went on to say, the fact that two laboratories had isolated the viruses from AIDS patients and that the viruses had been identified in many of the risk groups and rarely in controls "[lends] credence to our prejudice that viruses such as these are likely to be the guilty party."

One article in the New England Journal , published in August 1984, noted in passing that "further evidence is necessary to substantiate [the] possibility" of a retroviral cause.^[14] More serious doubts tended to be voiced only in short letters to journals. Arthur Ammann of UCSF Medical School called for a "cautious appraisal" in a September 1984 letter to JAMA . Arguing that "there is no historical precedent for believing that a single infectious agent is capable of abolishing a normal immune system," Ammann suggested that AIDS might conceivably be

the consequence of "multiple infections" involving HTLV-III acting synergistically with other viruses, such as CMV. Before concluding that HTLV-III is the sole cause, he cautioned, "a great deal more evidence must be obtained."^[15] But interventions such as these failed to gather any allies, and they largely vanished from the pages of the prominent medical and scientific journals after 1984.

What role did Gallo play in pushing the argument that his virus was the cause of AIDS? Articles authored or coauthored by Gallo or his coauthors certainly accounted for a substantial part of this literature—about a third of all publications in the sample for 1984 and 1985 and a quarter of them for 1986. (Of course, in many of these cases where the names of Gallo and/or his coauthors appear somewhere in the middle of a lengthy list, Gallo's lab most likely played a marginal role, perhaps no more than supplying viral samples.) But an interesting finding, revealed in table 2, is that, in terms of the pattern of making qualified or unqualified claims, there are no statistically significant differences^[16] between the papers influenced by Gallo's group and those with other authorship. As Treichler has asserted, Gallo's group may have strengthened the credibility of their claim simply by publishing so many papers about AIDS and referencing their 1984 paper at every turn. But in referencing it, they were no more or less likely to make either qualified or unqualified etiological claims than other scientific researchers who cited it during the same years. They promoted their product—but they did not, in any crude or blatant sense, oversell it in terms of its causal implications.

On the other hand, Gallo and his collaborators were less circumspect than Montagnier's group, who, as late as March 1985, wrote cautiously in JAMA: "Although the precise role of LAV in the pathogenesis of AIDS remains to be confirmed, the present report, together with the recent description of HTLV-III and AIDS-related virus, provides growing evidence in favor of a T-lymphotropic retrovirus as the causative agent in AIDS."^[17] And in November 1985, Montagnier reminded his readers that "cofactors"—factors in addition to the primary cause, such as antigens or foreign proteins—might often work together with LAV to cause AIDS.^[18] Montagnier, in other words, sought to incorporate some of the claims contained in earlier hypotheses, such as the immune overload hypothesis, within a new explanation that gave star billing to LAV.

A remarkable aspect of the change that took place over the three-year period is that the trend is observable even among those papers that

do not cite any subsequent research. Table 3 shows all papers in the sample from 1985 and 1986 that made a causal claim when citing Gallo's paper (as opposed to citing him for some other reason). These papers (ninety in 1985 and sixty-nine in 1986) are divided into two groups: those that cited Gallo alone or Gallo in conjunction with other previous or roughly contemporaneous "discovery" papers, such as work by Montagnier and Levy; and those that cited Gallo in conjunction with at least some later research. One might expect that papers also citing later research would be more likely to make unqualified claims. While that apparently is not the case in 1985, it seems to be in 1986. (The numbers are too small for the results to be statistically significant, but twelve of fourteen such articles made unqualified claims.)^[19] However, two points deserve note. First, only a minority of

the articles that made unqualified claims about the virus being the cause (five out of twenty-six in 1985 and twelve out of forty-nine in 1986) pointed to any later research when making such claims. In both years, most articles that made unqualified claims referenced the early set of papers only. Second, even among those articles that failed to cite any subsequent research, the original trend is apparent: between 1985 and 1986 the ratio of qualified to unqualified claims shifts markedly.^[20] These articles would seem to follow Gallo's own lead, gradually reinterpreting the significance of the early papers by imputing to them a stronger causal claim than even their authors made at the time of their publication.

To be sure, one might argue that people who failed to cite the subsequent research have been influenced by it nonetheless. Perhaps when researchers cited the early papers as having "proven" that HIV causes AIDS, they intended these citations to serve as a kind of shorthand or synecdoche that symbolically represented the larger body of published scientific literature. For this reason alone, it makes sense to attend to the other research findings published in this time period that supported the HIV hypothesis. However, many of the subsequent articles

that were co-cited with Gallo's provide little additional evidence. The two most commonly cited articles (each cited three times by articles in the sample making unqualified claims)—Samuel Broder and Gallo, published in the New England Journal in November 1984, and Flossie Wong-Staal and Gallo, published in Nature in October 1985^[21] —are review articles summarizing the literature on the "HTLV family" of viruses. The only additional etiological evidence that Broder and Gallo provided appeared in a brief mention of unpublished data from Gallo's lab showing (1) that HTLV-III had been found in semen and saliva in addition to blood, and (2) that HTLV-III had, by this point, been isolated from over 95 persons and could be recovered over 90 percent of the time from patients with "pre-AIDS." These latter data from Gallo's lab were eventually published by S. Zaki Salahuddin et al. in the Proceedings of the National Academy of Sciences ,^[22] though in this paper the Gallo group also reported that they still were unable to isolate the virus from half of the patients with full-blown AIDS. The Wong-Staal and Gallo paper is essentially the same in terms of its contribution to the question of causality. The authors simply reviewed the original Science articles and the additional findings published by Salahuddin et al. in the Proceedings .

Both review articles pointed to the Salahuddin article for the new evidence, and that paper was also cited on its own by other authors in the sample. But turn to that study and the citation pathway neatly reverses direction. The authors declared that "it is clear that HTLV-III is causatively involved in the immune disorder AIDS," and they explicitly advised the reader to refer back to Broder and Gallo for a summary of the evidence.^[23] Findings such as these certainly support Treichler's claim—that Gallo and his close associates established a network of citations that served to create the impression of greater certainty than Gallo's own data warranted. In circular fashion, each article points to a different one as having provided the definitive proof; the buck stops nowhere. Still, one can find places where Gallo did a better job of summarizing the existing evidence. For example, in a less frequently cited article in the journal Blood , Wong-Staal and Gallo reviewed two types of evidence in addition to the serology studies.^[24] First, in an unpublished study from Gallo's lab, out of ten blood donors whose recipients later developed AIDS, ten tested positive for HTLV-III antibodies. Second, an "animal model" was provided when chimpanzees were deliberately infected with HTLV-III.^[25] The chimps "mounted an immune response to the virus, and some

developed enlarged lymph nodes analogous to the lymphadenopathy syndrome that often precedes AIDS." According to Wong-Staal and Gallo, these various results, accompanied by the knowledge that HTLV-III kills T cells in the test tube, "unambiguously establish HTLV-III as the etiologic agent of AIDS."^[26]