The backpropagation citation gap is Merton's Matthew Effect, not his obliteration by incorporation

sociology-of-sciencerobert-mertonobliteration-by-incorporationmatthew-effectbackpropagationcredit-assignmentcitation-analysis

Merton's own definition of "obliteration by incorporation" (OBI), read directly from his 1988 paper (claim-merton-1988-obi-passage-footnoted-aside-defers-to-earlier-texts), names a specific mechanism: a source becomes anonymous because the finding itself became universal, uncontested common knowledge — the standard examples (Merton's and Garfield's alike) are pi, DNA, "self-fulfilling prophecy," cases where nobody is credited because everybody already assumes the fact. This does not match the shape of the backpropagation-gap as the vault has documented it. The gap has named parties on both sides: Seppo Linnainmaa (1970) and Paul Werbos (1974), the earlier and obscure independent derivers (claim-linnainmaa-priority-not-paternity); David Rumelhart, Geoffrey Hinton, and Ronald Williams (1986), the later and famous rediscoverers who became the canonical citation (claim-rhw-1986-reference-list-four-works). Credit did not dissolve into anonymity; it transferred to a specific, credited, more eminent name.

That is Merton's own Matthew Effect, read in the same 1988 paper: "eminent scientists get disproportionately great credit for their contributions to science" (coined 1968, restated here). Merton adds that demonstrating this skew requires the diversely credited contributions to be of much the same quality — a condition he says is "approximated in cases of full collaboration and in cases of independent multiple discoveries" — exactly the multiple-independent-discovery structure claim-linnainmaa-priority-not-paternity and myth-werbos-invented-backpropagation already document for backpropagation. The backprop gap is a citation-credit-skew story with a winner and named losers, not a story of a source becoming nobody's in particular; on Merton's own terms, the Matthew Effect names it, and OBI does not.

This is a third, distinct instance of OBI failing to fit a vault cluster it superficially resembles. observation-obliteration-by-incorporation-and-citational-narrowing-are-a-false-friend already found OBI a false friend to the Song Jian cluster's "citational narrowing" — there, OBI erases winners while that cluster erases dissenters, a mismatch of valence. Here the mismatch runs deeper: OBI is the wrong mechanism type altogether for this case — erosion into anonymity, applied to a process that was actually eminence capturing credit from named obscurity.