SUGGESTIONS FOR FILE SIZE (A OR C)
The initial Medline search that creates files A and C is an important part of the Arrowsmith process of literature exploration. Arrowsmith extends the power of a Medline search by finding the words, phrases, and subject headings (the "B-list") that are common to two separate literatures that have no articles in common. But at the two extremes of very small and very large files, Arrowsmith cannot be of much help. Files that consist of no more than a few dozen records yield little or no output; moreover, the user can do better by simply reading the corresponding literature. At the other extreme, files as large as 30,000 or more records yield a B-list that is virtually unmanageable because it spans a substantial part of the entire working biomedical vocabulary. Ideally, each of the files, A and C, should define a community of authors who communicate with each other (as evidenced by the citation pattern), but with little or no communication across the A-C boundary. A set of articles as large as 30,000 records probably consists of a cluster of many more-or-less independent communities. The exploratory Medline searching that is preparatory to using Arrowsmith can be used to control both file size and the type of literature produced.
RECOMMENDED RANGE: ABOUT 100 - 6000 RECORDS for each file
Although it is not reasonable to suggest an optimal size of File A or C that fits all situations, experience with examples thus far analyzed indicates that the range of a few hundred up to five or six thousand records for each file is more promising than file sizes outside of that range.