Contribution
A teaching method (role-playing a failed executive who asks for help) and a proposal to treat it as a quasi-experiment testing the theory of action. For this record, its value is a compact first-hand statement of that theory.
Key insight
An organization can correct errors indefinitely without ever being able to question the values that define what counts as an error. The reasoning that protects those values from embarrassment also hides that they are being protected.
Assumptions
- Theories-in-use can be inferred from observed conversation and scored reliably (Table 1; a 1965 scoring procedure is cited for inter-rater agreement).
- Model I is near-universal, “regardless of gender, race, culture, education, wealth, and type of organization”, on the strength of Argyris’s earlier studies.
Key results
- Definitions of single- and double-loop learning (introduction).
- The Model I governing variables and their consequences: defensiveness, misunderstanding, self-fulfilling and self-sealing processes (Part II).
- The four-step logic of organizational defensive routines (Part II).
- Model II’s governing values and productive reasoning (Part II).
- The Andy case. The CEOs gave abstract, untestable advice, attributed closedness to Andy without testing it, and blamed him as he blamed his board (Part I).
- Hypotheses for a quasi-experiment, including the claim that “one exception serves to disconfirm the hypothesis” (Part III).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Single- and double-loop learning are distinct | strong | definitional |
| C2 | Model I is the near-universal theory-in-use | weak | asserted from the author’s earlier work; not shown here |
| C3 | Organizational defensive routines and Model I sustain each other, so neither changes alone | moderate | argued; illustrated by the workshop |
| C4 | Participants reproduce the errors they diagnose | moderate | one transcript, analysed by the facilitator who ran it |
| C5 | Egalitarian structures drift back to unilateral control | weak | 1970s alternative schools and communes, cited without data |
Method
Facilitated role-play of a published case (“Andy”; Ciampa & Watkins 1999) with 34 CEOs, recorded and transcribed. The facilitator intervenes when the dialogue becomes self-reinforcing, then has participants reflect on their feelings and attributions.
Concepts
- governing values (variables) — what an action strategy is designed to keep within acceptable ranges, such as control or winning.
- theory-in-use — the design that actually produces action, as opposed to the espoused theory.
- skilled incompetence — counterproductive action produced fluently, without awareness.
- organizational defensive routine — a practice that prevents embarrassment and the discovery of its causes.
- productive reasoning — explicit premises and inferences, conclusions testable by independent logic.
Connections
Restates Argyris & Schön (1974, 1996) and Argyris, Putnam & Smith (1985). It uses Campbell & Stanley’s quasi-experimental design and White’s competence concept. The article-length statement usually cited, the 1977 HBR paper (LIT-739), is unread.
Bearing on the record
Source of THEORY-145. For the owner’s essay:
- §8.3 describes double-loop learning: the essay’s “procedures for reconsidering the standards through which policies acquire authority” are Argyris’s change in governing values.
- §8.5 says recursive revision does not guarantee normative adequacy. Argyris explains a specific failure: the revision procedure itself runs on Model I, so the values it should question are undiscussable.
- §8.2 (Longino) names uptake of criticism as a norm. Argyris describes the mechanism that defeats uptake, criticism that is neither illustrated nor testable, met with attributions that are never tested.
- Rasmussen (THEORY-144) has defences eroding unobserved. Argyris adds that the observation can be blocked socially as well as technically: the inconsistency is made undiscussable.
Limitations
The evidence is the author’s own workshop, interpreted by him, and claims of universality rest on earlier work not presented. The proposed quasi-experiment is not reported as run. Argyris’s statement that one exception disconfirms makes the theory testable in principle; no test is given here.
Open questions
Whether Model I is as universal as claimed, measured by raters independent of the theory’s authors, and whether Model II learning, once acquired, persists in organizations whose routines remain Model I.