Journal of the American Society for Information Science
40(3):152, 1989

Also See:
Don Swanson, "Verification of Results That Logically Related Noninteractive Literatures Are Potential Sources of New Knowledge" JASIS, 40(3):152, 1989

Don Swanson, "Response: Absence of Citations Can Be Valuable Clue," JASIS, 40(3):152, 1989

Letters to the Editor

Verification of Results That Logically Related Noninteractive Literatures
Are Potential Sources of New Knowledge

Don Swanson has written a very stimulating essay on the analysis of scientific literature to assist in problem solving [1]. This is a good example of how information science methods might be used to directly contribute plausible scientific hypotheses to be tested in the laboratory. However, Swanson finds both the techniques of citation analysis and analysis of indexing terms of little use in revealing the supposed logical connection between two concepts: the ingestion of "fish oil" and the alleviation of Raynaud�s disease.

However, we think he has limited the exercise too narrowly in bibliometric terms. He wants to see a one-step link, like direct citation, or a two-step link, like co-citation or bibliographic coupling, between literatures on the two topics. What may really be the case is an indirect, multistep path. We have found in our own work that co-citation clusters separated by a path of several cocitation links in one year (e.g. AIDS and human T-cell leukemia virus work), can merge together in a subsequent year when the causal connection between the virus and AIDS is discovered. In Swanson�s case, perhaps the fish oil and Raynaud�s groups are separated by a third group of papers on various aspects of blood. This third group serves to tie the first two literatures together logically, by a multistep chain of reasoning, not a direct inference. Using his notation, instead of 1) A causes B, and 2) B causes C, we would have 1) A causes B1, 2) B1 causes B2, and 3) B2 causes C.

This will of course expand the number of potential hypotheses due to the number of possible pathways connecting two concept nodes. But such diversity might be a good thing. To find such indirect, multistep paths requires sophisticated computer analyses of the underlying citation data files. Perhaps the new parallel pro cessors will make these kinds of knowledge analyses and hypoth esis generation more feasible. It would certainly open up many new and exciting opportunities for information scientists who are interested in making a direct contribution to the advancement of science.

Henry Small
Eugene Garfield
Institute for Scientific Information

1.  back to text Swanson, Don R., "Two Medical Literatures that are Logically but not Bibliographically Connected." Journal of the American Society for Information Science. 38(4):228�233; 1987.