The ladder of inference

In plain terms

Someone misses a deadline. Within seconds you have decided they do not care about the project, and you are composing a sharp email. The ladder of inference is a way of seeing how you got there.

You started with a fact, a missed deadline. You picked that fact out of everything else you know about the person. You gave it a meaning, assumed a reason, drew a conclusion and acted on it. Each step felt obvious. Each step was also a choice, and someone else, with other facts in view, could have climbed a different ladder to a different place.

The tool does not say conclusions are wrong. It says they rest on steps that can be checked.

Where it came from

Harvard's Division of Continuing Education credits the ladder of inference to the late Chris Argyris, a former professor at Harvard Business School who worked on how people and organizations learn, and notes that it was elaborated in The Fifth Discipline Fieldbook: Strategies and Tools for Building a Learning Organization which Levene lists as by Peter Senge, Richard Ross, Bryan Smith, Charlotte Roberts and Art Kleiner (Levene, 2016; Senge et al., 1994). Argyris's own books on reasoning and defensiveness in organizations include Reasoning, Learning, and Action (1982) and Overcoming Organizational Defenses (1990).

Double-loop learning is another idea associated with Argyris.

The SAGE Encyclopedia of Action Research has an entry on the ladder of inference (SAGE, 2014). See action research.

How it works

One common description of the rungs, adapted from Levene (2016), from the bottom up:

1. Observable data and experiences. The pool of information available to you.
2. Selected data. The part you notice, usually what grabs your attention or seems significant, while the rest is ignored.
3. Interpretations. The meaning you give the selected data, drawing on personal and cultural meanings.
4. Assumptions. Drawn from those meanings.
5. Conclusions. Drawn from the interpretation.
6. Beliefs. Over time, conclusions inform beliefs.
7. Actions. Beliefs drive what you do.

Levene's example: a few memorable interactions with a few customers can lead you to generalize about the whole market. Beliefs built this way may rest on a faulty selection or interpretation of the data, and people climb these steps often without noticing (Levene, 2016).

Using the ladder means going back down. In a conversation, that sounds like questions about the lower rungs. Levene's examples include: What data are you basing your recommendation on? What assumptions are you making? Did anyone else arrive at a different conclusion? The goal is to bring reasoning into the open without putting anyone on the defensive (Levene, 2016).

What the research says

The Evidence Library holds no study of the ladder of inference. It is a conceptual tool from Argyris's work. That does not make it wrong. It means claims that it improves decisions or relationships go beyond what has been graded.

The closest graded evidence is on reflection after the fact. Debriefs, like the ladder, ask a team to go back over what happened and why.

• Tannenbaum and Cerasoli (2013) pooled 46 samples with 2,136 participants and found that debriefs were linked to about 25 percent better performance than comparison conditions (d = .67; .54 with outliers removed), with somewhat smaller effects in real work settings (d = .56) than in simulations (d = .70) (rated High after a full-text read). Many pooled studies were quasi-experimental.
• Keiser and Arthur (2021), in a meta-analysis read from its abstract only (rated Medium), reported that after-action reviews were effective overall and that objective performance review media was one of two features linked to a larger effect. This is an analogy, not a finding: putting objective records on the table is like starting from the ladder's bottom rung, observable data.

Neither study tests the ladder itself, and a debrief can be run without it.

Where the evidence is thin

1. No direct test. No study in the Evidence Library compares teams taught the ladder with teams that were not.
2. No measure of reasoning. No graded study on this page measures how people reason when they use the ladder. The debrief studies measure performance.
3. One description used. This page follows one description of the rungs (Levene, 2016). Other descriptions may name them differently, and none has been tested as a measured construct.

How it maps to the Unfrozen Loop

1. Week 0
   Before you start
2. 1
   Surface both ways

   A report that travels up a hierarchy can be a conclusion, not the data behind it. Asking "what did you see that led you there?" brings the lower rungs up with it.

3. 2
   Make sense together, top included

   When managers and the people above them disagree about a change, one side may be missing data the other has. Making sense together starts with putting the data on the table before the conclusions.

4. 3
   Start it close to the work
5. 4
   Review on a rhythm, then hold the pace
6. 5
   The middle’s own weather
7. Ring
   Safe to speak, and speaking changes something

Highlighted parts are the ones this idea speaks to. The mapping is my reading, not a tested finding. See the whole Loop and the evidence for each part.

Common misreadings

"It is a way to prove the other person wrong." The questions are meant to bring reasoning into the open, not to win (Levene, 2016). Used that way, the questions can put people on the defensive, which Levene says to avoid (Levene, 2016).

"Good thinkers stay at the bottom." People need conclusions to act. The point is to know which rung you are on and to be able to climb back down when a conclusion is challenged.

"It is only about other people's assumptions." The ladder worth examining can be your own, especially before you act on a strong conclusion about someone. Levene suggests managers start with their own thinking.

Using it well

1. Before a hard conversation, write out your own ladder. One line per rung: what you saw, what you picked out, what you think it means, what you assumed, what you concluded. Then ask what data you have not seen.
2. In meetings, ask for the data before the opinion. When a change is going badly, start with "What have people actually seen this week?" and collect that before anyone explains why.
3. Ask the question, not the accusation. "Can you help me understand your thinking?" and "What data are you basing your recommendation on?" are among Levene's suggested questions. Levene suggests a manager start by using the ladder on their own thinking before introducing it to the team.