Greenberg's 2009 BMJ citation-network study names "citation transmutation" — a hedge dropped at the point of citation, hardening a hypothesis into unqualified fact

citation-analysiscitation-distortionhedge-erosionattribution-driftbibliometricsevidence-based-medicinefable-hook

Steven Greenberg's "How Citation Distortions Create Unfounded Authority: Analysis of a Citation Network" (BMJ 339:b2680, 2009) traced a complete citation network of 242 papers and 675 citations built around a single biomedical claim — that β-amyloid is produced by and injures muscle in inclusion body myositis — and named several distinct, defined subtypes of citation distortion found in that network. Among them is "citation transmutation": a claim first stated with a hedge ("may represent early changes of IBM") is cited by a later paper as unqualified fact, with the hedge dropped silently at the point of citation rather than through any new evidence. Greenberg's own framing: "This subclaim had transformed from hypothesis to 'fact' through citation alone, a process that might be called citation transmutation." He also names "amplification" — expansion of a belief system through citations to papers containing no data on the claim at all — and groups both under a broader category of "invention," the production of apparent fact via citation rather than via data.

This is a named, peer-reviewed mechanism for the shape the vault had already documented without an external name: Schmidhuber's own hedged "perhaps the first DL systems" claim about Ivakhnenko's GMDH networks loses its qualifier as it travels through popular restatements and Wikipedia echoes, hardening into "deep learning was invented in the USSR in 1965." Greenberg's vocabulary explicitly separates citation transmutation from citation bias (selectively ignoring contrary evidence) — the hedge isn't suppressed, it's simply not carried forward.