The references Wikipedia repeats
How often one citation key recurs across the corpus
Wikipedia's references are mostly singletons: a citation key is carried by one article's capture and never seen again. The few keys that recur by the hundred or the thousand are not papers at all — they are the reference catalogs that whole editorial beats share: national censuses, music chart books, protected-building registries, eponym dictionaries.
Source: K6-citeworks receipt (A8 repeat-citation distribution and A4 top works) over the citation spine, via the dataset page.
The works every article reaches for
The twelve most-carried keys make the point plainly. Each is a catalog that answers a recurring editorial need — charts, censuses, registries, lexicons — rather than a source about any single subject. None is a journal article.
| # | Work | Medium | Streams |
|---|
How to read the grain
The unit here is the canonical deduplication key: 45.0 million citation records collapse to 36 276 286 keys once equivalent strings of author, title, and publisher canonicalize together. Recurrence is counted in capture streams — the distinct passes of the dump that carried the key — which treats rows captured before occurrence counting began the same as rows captured after.
Two honest caveats. Occurrence counts on the spine are capture-era-weighted (zero on legacy rows), so this figure uses streams throughout. And none of this is a scholarly citation count: no reference-list parsing of who-cites-whom happened — a key carried by 6 399 streams is carried by 6 399 slices of the dump, which is a floor on how widely it is cited, not an exact tally.
The full distribution, the top sixty works, and the receipted shares live in the figure's data payload, described on the dataset page.