: Background: Medical Subject Headings (MeSH) are a hierarchical taxonomy of over 42 000 descriptors designed to classify scientific literature; it is hierarchical with generic high order headings and specific low order headings. Over 1,000 resources in the Primary Care Electronic Library (PCEL – www.pcel.info) were classified with MeSH.

Methods: Each of the entries or resources in the primary care digital library was assigned up to five MeSH terms. We compared whether the most generic or specific MeSH term ascribed to each resource best predicted user preferences.

Results: over the four month period analysed statistically significant differences were found for resources according to specific key MeSH terms they were classified by. This result was not repeated for generic key MeSH terms. Conclusions: Analysis of the use of specific MeSH terms reveals user preferences that would have otherwise remained obscured. These preferences are not found if more generic MeSH terms are analysed.