Ingest: Circling Back — Clearing Up Myths About the Deming Cycle

Type: kb/sources/types/ingest-report.md

Source: circling-back-clearing-up-myths-about-the-deming-cycle.md Captured: 2026-07-22 From: https://qicentral.rcpch.ac.uk/wp-content/uploads/sites/9/2021/10/Moen-and-Norman-on-Deming-circling-back-ph556x.pdf

Classification

Genre: conceptual-essay -- a 2010 Quality Progress (ASQ) article that argues a historical framing (the PDSA cycle is the scientific method carried into industry) and corrects three myths about PDCA vs. PDSA. It carries references but no methodology, data, or experiment, so it is a framing/historical-position essay rather than a scientific-paper; the snapshot's capture-time genre was corrected to conceptual-essay during ingest. Domains: discovery-lifecycle, improvement-loops, scientific-method Author: Ronald D. Moen (ASQ Deming Medal, 2002; assisted Deming at 70 of his four-day seminars, 1982-1993; managed Deming's monthly GM visits) and Clifford L. Norman, both partners at Associates in Process Improvement. First-hand proximity to Deming and the primary-letter evidence give unusual authority on the PDSA/PDCA distinction; the same proximity is a source of partisanship (see Limitations).

Summary

Moen and Norman trace the plan-do-study-act (PDSA) cycle as a lineage of the scientific method reaching industry: Galilean designed experiment and Baconian induction, then Peirce/James/Dewey/Lewis pragmatism, then Shewhart's specification-production-inspection cycle (1939), Deming's four-step "wheel" presented in Japan in 1950, its Japanese recasting into PDCA (1951, later extended by Ishikawa), and Deming's own return to and renaming of the cycle as the Shewhart Cycle for Learning and Improvement — PDSA (1986, 1993). Their central corrective claims: Deming did not create PDCA and disowned it ("Be sure to call it PDSA, not the corruption PDCA"); PDCA and PDSA "seem related only through the scientific method"; and the 1991 Moen/Nolan/Provost addition made the cycle require prediction and associated theory, so that the "study" step supplies the deductive-inductive interplay learning demands. Read it to get the primary-source history and the argued distinction between a management/learning loop (PDSA) and a local problem-solving loop (PDCA).

Connections Found

No casebook-style notes exist for this lineage: a body search for pdsa|deming|shewhart|plan-do across notes, reference, and the external-system collections returned zero hits, so this source is a new anchor for the discovery and self-improving-systems areas rather than a participant in an existing map. Its durable role is an applied, decades-old external witness for three existing kb/notes/ claims that connect discovery treats as candidates for an evidenced-by link into this snapshot. The single relationship most likely to bear on the source's role: it is a worked industrial instantiation of the conjecture-consequence-test core that discovery-lifecycle adapts from Peirce — the paper's own "study = deductive-inductive interplay" language maps directly onto that definition's derive-consequences and test phases.

Extractable Value

  1. PDSA as an applied-lineage witness for the discovery lifecycle -- the paper documents an independent industrial tradition arriving at the same conjecture → derive-consequences → test structure the discovery-lifecycle definition draws from Peirce, complementing that note's philosophical (Peirce-SEP) evidence with a practitioner-tradition instance. [quick-win]
  2. The generalization step is a distinct judgment (external corroboration) -- Moen/Nolan/Provost: it is "not enough to determine that a change resulted in improvement during a particular test"; you must predict improvement "under the different conditions you will face in the future." This is near-verbatim, decades-old corroboration of an-accepted-edit-verifies-the-change-not-the-rule from an unrelated field. [quick-win]
  3. PDSA as a fourth improvement-loop reference model -- plan≈search, do/study≈evaluation-that-can-reject, act≈operative retention, parallel to the MAPE-K model that a-proposal-selection-loop-requires-search-evaluation-and-retention already treats; lets that note observe the quality-management tradition independently converged on the same generate/evaluate/retain decomposition. [quick-win]
  4. "Check" vs. "study" — a naming failure mode -- Deming rejected PDCA partly because "check" means "to hold back," losing the learning intent; a concrete case where a single mislabeled step degrades a method, relevant to KB claims that titles/labels carry commitments. [just-a-reference]
  5. Prediction-before-test as the mark of learning -- the 1991 revision makes an explicit prediction the precondition for the study step to yield knowledge (compare observed data to a stated prediction); reusable framing for distinguishing a genuine test from a confirmation. [experiment]
  6. A four-tradition improvement-loop cluster -- with Ashby (cybernetics), Weyns/MAPE-K (self-adaptive software), the self-adaptive-systems survey, Gödel machines, and Zinkevich (online learning) already in kb/sources/, PDSA adds the quality-management tradition, strengthening a future synthesis that improvement-loop reference models recur across independent fields. [deep-dive]

Limitations (our opinion)

This is editorial opinion. As a conceptual essay the weakness is what is not argued rather than what is not tested. (1) Partisanship: Moen is a Deming protege and co-author of the PDSA revision he credits; the "PDCA is a corruption" thesis rests heavily on Deming's own letters and roundtable remarks — primary but self-interested evidence, with no comparably sourced defense of PDCA's utility. (2) The lineage is narrated as a clean linear descent (Galileo → Bacon → pragmatism → Shewhart → Deming → PDSA); this is reasoning by curated genealogy, and the causal links (e.g., Lewis's "enormous influence" on Shewhart and Deming) are asserted, not demonstrated. (3) The core "PDSA is the scientific method in industry" claim is hard to vary but also hard to falsify as stated — almost any iterative improvement loop can be described this way, so it risks conflating naming the cycle with explaining why it works. For KB use the durable, well-sourced content is the specific prediction-before-test requirement and the PDCA/PDSA distinction; the grand lineage narrative should be cited as the authors' framing, not as settled history. Note also the snapshot is a degraded PDF-to-markdown capture with heavy word-splitting — verify any quotation against the reprint before promoting it.

Add an evidenced-by link from discovery-lifecycle to this snapshot (the strongest of the three reverse edges), citing the paper's "study = deductive-inductive interplay" documentation of an applied instantiation of the conjecture-consequence-test core. The two weaker reverse edges (an-accepted-edit-verifies-the-change-not-the-rule, a-proposal-selection-loop-requires-search-evaluation-and-retention) can follow when those notes are next revised.