Sutton's 1984 dissertation abstract, in his own words, frames credit assignment entirely in temporal terms and never uses 'structural'

credit-assignmentstructural-credit-assignmenttemporal-credit-assignmentsuttonterminologyhistory-of-mlreinforcement-learningprimary-source-verification

Richard Sutton's own publications page lists his 1984 UMass Amherst PhD dissertation, "Temporal credit assignment in reinforcement learning" (COINS Technical Report 84-2), with its full published abstract reproduced verbatim. The abstract frames the credit-assignment problem entirely in temporal terms: "aspects of the credit-assignment problem having to do with determining when the behavior that deserves credit occurred." It explicitly excludes the structural side of the problem — "the issues of knowledge representation involved in developing new features or refining existing ones are not addressed" — rather than naming and solving it. The word "structural" does not appear anywhere in the abstract text.

This is direct evidence against the field-lore claim, tracked at question-verify-sutton-1984-structural-credit-assignment, that Sutton's 1984 dissertation coined or used the phrase "structural credit assignment." It does not by itself prove the phrase is absent from the dissertation's 223-page body text, which remains unread (extraction blocked by file size and mirror access; see the question note for the attempted routes). It is, however, the dissertation's own formal self-description, and it converges with the author's later retrospective framing of the same work — see claim-sutton-retrospective-summary-names-temporal-credit-assignment-not-structural — and with the paired myth this thread runs alongside, myth-structural-credit-assignment-from-minsky-1961.