Searching for patterns in the MeSH vocabulary.

NLM revises its MeSH vocabulary annually to reflect changes in biomedical literature and the health sciences community. This study tested two hypotheses about NLM's MeSH vocabulary. The first is that new terms are added to MeSH when their broader terms have an increased number of postings. One examination compared the number of postings for the broader terms of new and existing terms in the current MEDLINE file; the other compared them over time. No significant statistical difference was found in either case. A second hypothesis--that there is a relationship between the patterns of MEDLINE indexing and searching and the organization of the MeSH tree structure--was tested by comparing the distribution of searched terms in the MeSH trees with the distribution of all terms. It was found that certain trees are searched more often than could be predicted by the overall term distribution, while others are searched less frequently than expected. In summary, (1) new terms cannot be predicted by the increase in postings of existing terms, and (2) searchers' and indexers' use of the terms' tree structure does not correlate with the terms' distribution in the MeSH trees.