Recognition-Primed Decision

Thinking Tools5.2 Selecting the Right Decision ToolsAlso known as RPD Model

Originated by Gary A. Klein, Roberta Calderwood, and Anne Clinton-Cirocco (Rapid Decision Making on the Fire Ground, 1986).

Describes how experts decide under time pressure: recognize the situation as a familiar type, picture the usual response playing out, and act if it works.

A flooding basement: recognize, picture the usual response, adjust, act

Water is rising in the plant room of an office building at 6 a.m. At each of the site manager's turns, pick the move; the right-hand column is the response run forward in the mind, which shows where the usual plan breaks before anyone acts on it. The count tallies the expert's moves. Illustrative.

Site manager

Picturing it play out

Site manager

Picturing it play out

Select a stage to read it, or play the sequence.Size it up. Clear water, rising about an inch every five minutes, the main switchgear on a plinth eight inches up. Pick the site manager's first move.Pick a move to see the reply.Twenty minutes later the list is good and the water is at the plinth.A comparison where speed matters: the situation outruns it.The cues fit: clean water, fast rise, a main valve upstream. The goal is clear too: keep water off the switchgear.Recognized as a type, with its goal and what to expect next.The type. The type brings its usual response: close the main valve, run both sump pumps, call the plumber.The usual response. Pick what happens before the plan is carried out.Pick a move to see the reply.Four minutes later water reaches the plant-room socket, the breaker trips, and both pumps stop.The first match, acted on unchecked: it fails in a way a moment's picture would have shown.Valve shut, pumps on, water still rising for a while... both pumps are on the plant-room circuit, and that socket is lower than the switchgear.The mental run finds the flaw before it costs anything.The run breaks. The plan is right in kind and wrong in one detail: the pumps' own power is the first thing the water reaches.Adjust. The picture broke on one point. Pick the next move.Pick a move to see the reply.The water keeps rising while the team debates hiring a pump truck.Throws away a response that was nearly right.Pictured again: the cable reaches, the generator carries both pumps, nothing else is on it. The run holds.Modified and simulated again: now it works.Run again. The second mental run finds no break, so the site manager acts.Act. The facilities director is not answering. Pick the move.Pick a move to see the reply.The director calls at 6:40. The water is over the plinth.A tested plan held back by a routine approval.Valve shut at 6:12, pumps on generator power at 6:20, water draining away by 7:00.Acting on a plan that survived the picture.The outcome. The tenants arrive to a damp plant room and a working building.Afterward. Pick what happens the next week.Pick a move to see the reply.The next site manager on duty has never heard that the pumps share the circuit.The case stays in one head.Two newer staff now know the type, its usual response, and the detail that breaks it.The case joins the team's experience, which is what recognition is built from.

Expert moves: 0 of 5Expert moves: 1 of 5Expert moves: 2 of 5Expert moves: 3 of 5Expert moves: 4 of 5Expert moves: 5 of 5

  1. Size it up (Site manager). Clear water, rising about an inch every five minutes, the main switchgear on a plinth eight inches up. Pick the site manager's first move.
    • Gather the team and list every option before choosing. Twenty minutes later the list is good and the water is at the plinth. A comparison where speed matters: the situation outruns it.
    • I've seen this four times: a burst main in the plant room. The cues fit: clean water, fast rise, a main valve upstream. The goal is clear too: keep water off the switchgear. Recognized as a type, with its goal and what to expect next.
  2. The type (Picturing it play out). The type brings its usual response: close the main valve, run both sump pumps, call the plumber.
  3. The usual response (Site manager). Pick what happens before the plan is carried out.
    • Do it now: valve shut, both pumps on. Four minutes later water reaches the plant-room socket, the breaker trips, and both pumps stop. The first match, acted on unchecked: it fails in a way a moment's picture would have shown.
    • Picture it playing out, step by step, before moving. Valve shut, pumps on, water still rising for a while... both pumps are on the plant-room circuit, and that socket is lower than the switchgear. The mental run finds the flaw before it costs anything.
  4. The run breaks (Picturing it play out). The plan is right in kind and wrong in one detail: the pumps' own power is the first thing the water reaches.
  5. Adjust (Site manager). The picture broke on one point. Pick the next move.
    • Drop the plan and work out all the options from scratch. The water keeps rising while the team debates hiring a pump truck. Throws away a response that was nearly right.
    • Keep the plan; power the pumps from the portable generator in the loading bay. Pictured again: the cable reaches, the generator carries both pumps, nothing else is on it. The run holds. Modified and simulated again: now it works.
  6. Run again (Picturing it play out). The second mental run finds no break, so the site manager acts.
  7. Act (Site manager). The facilities director is not answering. Pick the move.
    • Wait for the director to call back before switching power. The director calls at 6:40. The water is over the plinth. A tested plan held back by a routine approval.
    • Act, and leave the director a message while it runs. Valve shut at 6:12, pumps on generator power at 6:20, water draining away by 7:00. Acting on a plan that survived the picture.
  8. The outcome (Picturing it play out). The tenants arrive to a damp plant room and a working building.
  9. Afterward (Site manager). Pick what happens the next week.
    • Back to normal; it worked out. The next site manager on duty has never heard that the pumps share the circuit. The case stays in one head.
    • Walk the team through it, and add pump power to the flood checklist. Two newer staff now know the type, its usual response, and the detail that breaks it. The case joins the team's experience, which is what recognition is built from.

An example

At six in the morning a security guard called the site manager of a twelve-story office building: water in the plant room, rising. She had dealt with four burst mains in her career, and the picture was familiar before she reached the basement: clear water, rising about an inch every five minutes, the main switchgear on a plinth eight inches up.

The usual response came with the recognition: shut the main valve, run both sump pumps, call the plumber. Before doing it, she ran it through in her head, step by step, and the picture snagged. Both pumps were plugged into the plant-room circuit, and that socket was lower than the switchgear. The water would trip the breaker and kill the pumps within minutes.

She kept the plan and changed one piece: run the pumps from the portable generator in the loading bay. She pictured that too, found it held, and acted. The valve was shut by 6:12 and the water was draining away by seven. The following week the shared circuit went onto the flood checklist.

Why it works

Experienced people under time pressure rarely compare options. They recognize the situation as a kind they have seen, and the recognition arrives with the goals that matter, the cues to watch, what should happen next, and a response that usually works. That is why they are fast: the first option considered is usually a good one.

The check is the mental run. Before acting, the expert imagines carrying the response out in this particular situation, and the imagined run catches the detail that makes this case different from the usual one. If the run breaks, the response is patched and run again; only when nothing familiar fits does a deliberate comparison begin. The quality of the whole process rests on the stock of experience, which is why reviewing cases afterward matters.

The thinking behind it

Gary Klein Gary Klein, a research psychologist and a pioneer of naturalistic decision making, studied how fireground commanders and other experts decide under time pressure, and set his findings out in Sources of Power (1998), published by MIT Press. His model holds that experienced people recognize a situation as typical, take the first workable option it suggests, and test it by imagining it played out, without comparing options side by side.

Roberta Calderwood Roberta Calderwood, a decision researcher, was a coauthor with Klein and Anne Clinton-Cirocco of Rapid Decision Making on the Fire Ground (1986), in the proceedings of the Human Factors Society's thirtieth annual meeting. That study of fireground commanders first stated the model: experienced commanders rarely compared options; they recognized the situation as typical and mentally simulated the first workable course of action.

Anne Clinton-Cirocco Anne Clinton-Cirocco was the third author, with Gary Klein and Roberta Calderwood, of the 1986 study of fireground commanders that first set out the recognition-primed decision model. The commanders it studied recognized a situation as typical and acted on the first workable option, without comparing alternatives, the finding that ran against the classic advice to list options and weigh them before choosing.

How to run it

  1. Size up the situation. Ask what kind of situation this is, and note the cues that make it look like that kind.
  2. Name what the type brings with it. Say the goal that matters most, what you expect to happen next, and the response that usually works.
  3. Picture the response playing out. Walk through carrying it out here, step by step, looking for the point where it would fail.
  4. Patch it if the picture breaks. Change the failing piece, keep the rest, and picture it again.
  5. Act when the picture holds. Carry it out without waiting for a fuller analysis, and watch whether what you expected happens.
  6. Compare options only when nothing fits. If the situation matches no type you know, or the expectations keep failing, slow down and weigh alternatives.
  7. Review the case afterward. Talk the team through the cues, the response, and the detail that broke it, so the case becomes shared experience.

Where it goes wrong

The mental run gets skipped. Under pressure, the first match feels so right that it goes straight to action, and the one detail that makes this case different goes unnoticed. Make the picture a habit, even when it takes only seconds.

The pattern does not match the experience. Recognition is only as good as the cases behind it, and a manager with one kind of experience will see that kind everywhere. When the cues do not quite fit, or what you expected does not happen, treat it as a new situation.

The lesson stays in one head. Experts are fast because of the cases they have lived, and an unshared case is lost when that person leaves. Review incidents as a team and write the deciding cue into the checklist.

When to use something else

  • OODA Loop. When the situation keeps changing in response to an opponent, and the cycle of observing and acting repeats.
  • Premortem. When there is time to plan, and the team should imagine the plan failing before committing to it.
  • Weighted Decision Matrix. When nothing familiar fits and there is time to compare options against agreed criteria.

Where it sits in SOLVE eX

Go to the source

  • Rapid Decision Making on the Fire Ground, Proceedings of the Human Factors Society 30th Annual Meeting, 1986

Back to the catalog