I generated the analysis in this book from a range of sources that I treated as primary data: first, published instances of claims-making found in scientific and medical journal articles, mass media news reports, articles in the gay and lesbian press, activist documents and publications, and government documents; second, extensive interviews with more than thirty of the principals in the story (researchers, activists, and government officials); and third, conferences, meetings, forums, demonstrations, and other public events that I attended. My fundamental analytical strategy has been to bring into critical juxtaposition contemporaneous records from different "social worlds." The sources that I have considered document the positions that actors have taken and the claims they have advanced; these sources are also, in themselves, important arenas of struggle. My assumption is that each of these different sources engages in a different calculus of credibility—that, for example, the New York Times version of reality is not the same as the Gay Community News version. Bringing them into common focus reveals the different regimes of credibility assessment, just as it exposes the stratification of credibility that inevitably becomes manifest when different social worlds collide. Of course, published sources tell only part of the story—sometimes, in their linearity and smoothness, finished documents conceal the story—and it is in that regard that interviews with the participants in events
have helped me to describe those events in a more satisfactory way and to provide needed context.
In some cases, as in my reconstruction of the controversy surrounding Peter Duesberg, my strategy has tended toward the exhaustive, and I have unearthed sources from less influential publications as well as better-known ones. In other cases, the enormity of the source materials available has necessitated certain strategic choices. With regard to scientific and medical journals, popular impressions coincide with evidence from citation analyses in suggesting that Science, Nature , the New England Journal of Medicine, Lancet , the Journal of the American Medical Association (JAMA) , and the Annals of Internal Medicine are the most generally significant, influential, and credible in the world of biomedical research. In studying debates in the mass media, I have relied almost entirely on influential print sources, such as the New York Times , the Los Angeles Times , and the Washington Post , as well as Time, Newsweek , and, where appropriate to the story, the British press. My material from the gay press derives mainly, though not entirely, from those newspapers and magazines with some fair degree of national circulation, particularly the Advocate (California), Gay Community News (Boston), and the New York Native , but also a number of local newspapers and magazines. In terms of treatment publications (which I cite heavily in part two), I make most use of AIDS Treatment News (San Francisco), PI Perspectives (San Francisco), Treatment Issues (New York City), the ACT UP/New York Treatment & Data Digest , and TAGline (New York City).
Symmetry and the Study of Scientific Controversies
How can scientific controversies best be reconstructed? One of the guiding principles of the sociology of scientific knowledge is the so-called principle of symmetry, which proposes that the researcher invoke the same types of causes or apply the same conceptual tools to explain "true" and "false" beliefs. This principle represents a crucial break with a more traditional approach, which begins by accepting the views of dominant scientists and then sets itself the task of explaining why the other side in the controversy (creationists, say, or anti-fluoridationists) might be so deluded as to persist in its errors. In the traditional approach, "social explanations are selectively applied
to the side without authoritative scientific backing." By contrast, contemporary sociologists of scientific knowledge insist on employing a common conceptual apparatus to explore the knowledge claims and social factors on both (or all) sides of a controversy. In my view, a symmetric analysis is not necessarily a "neutral" analysis, either in intent or in effect. But it is, potentially, the most fair-minded way to approach knowledge controversies, and one that requires the investigator to bend over backwards to consider the arguments of scientific "underdogs."
Archeology and Genealogy
Without suggesting any strict allegiance, I would also characterize my method as having affinities with modes of investigation employed in the work of Michel Foucault. First, my analysis is "archeological" in that it seeks to focus attention on the conditions of possibility of different forms of knowledge in different places at different times. Such an analysis concerns itself with a recovery of the immanent rules of what is sayable and unsayable, thinkable and unthinkable. Second, my account is "genealogical" insofar as it aims to disrupt any assumptions of a smooth path of development of knowledge about AIDS over time. A genealogical analysis rejects teleological accounts in order to emphasize shifts and discontinuities. By the logic of genealogy, contemporary knowledge about AIDS cannot be inferred in some automatic fashion from earlier moments in the epidemic, nor was early thinking about the epidemic predictive of a pathway to the present. Rather, without denying the links from one moment to the next, we must be attentive to gaps, breaks, and transmutations in the trajectory of knowledge development.
Finally, the notion of genealogy is consistent with my "democratic" approach to claims-making—my strategy of attending to the claims of an activist reported in the gay press in the same way that I note those of a famous scientist writing in the pages of Nature . Genealogical research, in Foucault's words, "[entertains] the claims to attention of local, discontinuous, disqualified, illegitimate knowledges against the claims of a unitary body of theory which would filter, hierarchise and order them in the name of some true knowledge and some arbitrary idea of what constitutes a science and its objects." The recovery of these "subjugated," "local," or "situated" knowledges and their
juxtaposition with formal, accredited knowledge is essential to a full understanding of the relations of power and the formation of knowledge in a given society at a given time.
The remainder of this appendix provides details concerning the design of the two content analyses reported in chapter 2—the analysis of scientific journal articles that cited Gallo's 1984 paper (see tables 1, 2, and 3) and the analysis of references to causation in the New York Times (see table 4).
Scientific Journal Articles
I selected key scientific and medical journals for the period 1984–1986 using the Science Citation Index Journal Citation Reports , which rank journals annually based on the number of citations to the articles they publish. Three highly prominent general science journals (Science, Nature , and the Proceedings of the National Academy of Sciences ) and two highly prominent medical journals (the New England Journal of Medicine and Lancet ) ranked within the top ten for all three years. Two other prominent medical journals (the Annals of Internal Medicine and Journal of the American Medical Association ) ranked only within the top fifty but ranked near the top of the subset of journals designated as "Medicine, General and Internal"; moreover, these journals account for a high percentage of publications specifically about AIDS, according to several studies. Therefore, I selected these seven publications for the content analysis.
Next, I identified all articles published in these journals in 1984, 1985, or 1986 that cited Gallo's paper, using the 1984–1987 issues of the Science Citation Index . I retrieved and photocopied the selected articles and then screened them for "meatiness" using an algorithm developed for this purpose by Garfield. This algorithm selects for communications (articles, letters, and so on) reporting substantive research while eliminating items likely to have minimal impact. Items failing to achieve the threshold level identified by Garfield (such as short letters) I excluded from consideration. This yielded a total population of 244 articles, which I included in the content analysis.
I then coded each article for three pieces of information:
1. In the sentence containing the citation of Gallo's article, what kind of causal claim is made?
Explicit unqualified reference to the virus as the cause (These include, for example, "The cause of AIDS has been found to be HTLV-III"; "HIV, the virus that causes AIDS …"; "HIV is the primary etiological agent in AIDS"; and so on.)
Implicit unqualified reference to the virus as the cause (The virus is referred to as "the AIDS virus.")
Qualified reference to the virus as the cause (These include "HIV, the putative agent in AIDS …"; "HIV is believed to cause AIDS"; "The bulk of the evidence suggests [or strongly suggests] that AIDS is caused by HTLV-III"; or the use of words such as "etiologically linked," "associated," and so on.)
Implicit reference to the possibility that the virus may not be the cause (I included this coding possibility for the sake of logical completeness, but no articles were coded as such; therefore I excluded it in my subsequent analysis.)
Explicit reference to the possibility that the virus may not be the cause (These include references to the lack of evidence, to other hypotheses, to the need for cofactors, and so on.)
Article not cited in conjunction with a causal claim (Some authors cited the Gallo paper simply to establish a different point about HTLV-III.)
2. Are any of the thirteen coauthors of the Gallo article included among the coauthors of the article in question?
3. Are additional articles cited to support the causal claim? (Here I distinguished between articles that cited only Gallo or that cited Gallo along with earlier or roughly concurrent articles [1983 and 1984] by the Gallo, Montagnier, or Levy groups and those articles that also cited subsequent articles by any authors [1985 and 1986].)
I then tabulated data by journal. However, due to the small numbers, I have reported results only for the full population (the seven journals combined).
New York Times Articles
I performed a database search to identify and retrieve all articles from July 1, 1984, to December 31, 1986, that included the word "AIDS" and any of the following terms: "HIV," "HTLV," "LAV," and "virus." I excluded letters to the editor, indexes, and substantively irrelevant articles. For the second half of 1984, I subjected the entire population of articles to content analysis. Due to the explosion in reporting about AIDS after 1984, for subsequent six-month time periods, I selected samples of the full article population. Samples consisted of every fifth article, in chronological order, beginning with a randomly selected number from one to five. This procedure yielded a total sample of 115 articles.
I then coded these articles according to causal claims made anywhere in the body of the article:
Only unqualified references to the virus as cause (The virus is identified, without qualifiers, as the cause of AIDS, one or more times in the article; no qualified references appear anywhere in the article.)
Only qualified references to the virus as cause (Qualified references ["Most scientists believe that AIDS is caused by a virus called HTLV-III"; "HIV, the virus believed to cause AIDS"; and so on] appear one or more times in the article; no unqualified references appear anywhere in the article.)
Mixed qualified and unqualified references (Both qualified and unqualified references appear in the article.)
Implicit claim that the virus is the cause (This was usually indicated by references to the "AIDS virus" in the absence of any other qualified or unqualified causal claims.)
Explicit references to the possibility that the virus may not be the cause (This included articles about alternative etiological hypotheses.)
No causal claim (No qualified or unqualified references and no use of the phrase "AIDS virus"; this proved to characterize only a very small percentage of the articles.)
On the basis of this coding procedure, I then tabulated and reported data for each of the time periods studied.