Argyris & Schön's Double-Loop Learning

6 min read · SessionData

In 1978, Chris Argyris and Donald Schön published Organizational Learning: A Theory of Action Perspective. The book introduced a distinction that has shaped how we think about learning in organisations ever since: the difference between single-loop and double-loop learning.

The idea is easier to feel than to define. When something goes wrong and you fix the immediate problem, that is single-loop learning. When something goes wrong and you ask why your approach produced that problem in the first place, that is double-loop learning. The first corrects actions. The second revises the thinking behind them.

The model

Argyris and Schön described human action as flowing through three elements:

Governing variables — the values, beliefs, assumptions and frames that guide how we act. These are often tacit. We do not consciously consult them; they operate in the background, shaping what we notice and what we ignore.

Action strategies — the plans, designs and behaviours we choose, given our governing variables. If we believe that conflict is unproductive, our action strategy will avoid confrontation. If we believe that measurement is threatening, our action strategy will resist evaluation.

Consequences — the outcomes that result from our actions, both intended and unintended.

Single-loop learning

When consequences fall short of expectations, single-loop learning adjusts the action strategy. The thermostat analogy is Argyris's own: the room is too cold, so the heating turns on. The goal (temperature) is not questioned — only the action (heating on or off) changes.

In a training context, single-loop learning looks like this: participants did not engage, so we redesign the slides. Scores were low, so we slow the pace. Attendance dropped, so we shorten the session. Each adjustment stays within the existing frame. The programme's assumptions — its theory of what should change and why — remain untouched.

Double-loop learning

Double-loop learning feeds consequences back to the governing variables themselves. Not just "did the action work?" but "are we pursuing the right goal? Are our assumptions about how people learn actually correct? Is the theory of change behind this programme sound?"

This is harder. Governing variables are often deeply held. Questioning them feels like questioning competence, identity or authority. Argyris spent much of his career documenting how individuals and organisations systematically avoid double-loop learning — even when they believe they are practising it.

Defensive routines

Argyris identified a pervasive pattern he called defensive routines: behaviours that protect individuals and organisations from embarrassment or threat, at the cost of preventing learning.

A facilitator who avoids collecting honest feedback because it might reveal the session was ineffective is running a defensive routine. A manager who frames every programme as a success because failure would threaten the budget is running a defensive routine. An organisation that evaluates only at Kirkpatrick's Level 1 — reaction — because Levels 3 and 4 might surface uncomfortable truths is running a defensive routine.

The pattern is self-sealing. The routine prevents the information that would reveal the routine. Double-loop learning requires breaking through this — which is why psychological safety is not a nice-to-have but a structural prerequisite. People will not question governing variables in an environment where questioning is punished.

Espoused theory vs theory-in-use

Argyris drew a sharp distinction between what people say they believe (espoused theory) and what their behaviour actually reveals (theory-in-use). Most organisations espouse double-loop values: we welcome challenge, we learn from failure, we value honest feedback. But the theory-in-use — observable in how meetings run, how programmes are evaluated, how bad news travels — often tells a different story.

This gap is not hypocrisy. It is a structural feature of how humans manage complexity and threat. Closing it requires more than good intentions. It requires practices that surface the gap: structured reflection, external facilitation, data that cannot be explained away.

Kolb's experiential learning cycle provides one mechanism. The reflective observation stage — if done with sufficient honesty — is where the shift from single-loop to double-loop can happen. But only if the environment supports it. Reflection that stays safe, that does not touch governing variables, is single-loop reflection wearing a double-loop label.

Why it matters for programme design

A programme designed for single-loop learning will improve actions within existing frames. Useful, but limited. A programme designed for double-loop learning will surface and test the frames themselves. This is where genuine transformation occurs — and where most programmes fall short.

The implication is not that every programme needs to be double-loop. Compliance training, for example, operates deliberately at the single-loop level: here are the rules, here is how to follow them. The problem arises when programmes that need to produce behaviour change — Kirkpatrick Level 3 — are designed as if single-loop learning will be sufficient.

When Senge describes the learning organisation, he is describing an organisation that has made double-loop learning a collective capability, not just an individual one. The five disciplines — particularly mental models and systems thinking — are mechanisms for institutionalising the kind of reflection that Argyris argued most organisations instinctively avoid.

Argyris, C., & Schön, D. A. (1978). Organizational Learning: A Theory of Action Perspective. Reading, MA: Addison-Wesley.

Argyris, C. (1991). Teaching smart people how to learn. Harvard Business Review, 69(3), 99–109.