Hierarchically structured thesauri--including MeSH--were studied to test the assumption of an inverse relationship between term specificity and the number of postings in online databases. It was suggested that this assumption holds only for peripheral terms and that the opposite is true for the central terms of a discipline. The Environment tree structure of MeSH was used to test the "peripheral" hypothesis, which was supported at a .05 level of significance, but the scattergram on the Endocrine Diseases tree, which was run to test the "central" hypothesis, had an even better level, .02. The commonly held belief that specific terms are added to a thesaurus when more general ones get too many postings was also tested. The relationship between specificity and dates was supported in MeSH (.015). The corollary--that terms with later dates have fewer postings--was supported at a .001 level of significance. The results for MeSH were much more significant than for other hierarchically structured thesauri, which indicates that thesaurus development at NLM involves interaction with MEDLINE to preclude an excessive number of postings to any one term.