Skip to content
Mind Flaws

Self & Social Bias · Interactive Reading

Correspondence Bias

Can you tell what someone thinks from something they were told to say?

The experiment

Read the essay carefully. Then say what you think its author really believes.

Steve's essay

…

What happened?

Each person reads, at random, one of two essays: one for and one against capital punishment. Then they learn that the professor chose the side Steve had to defend. So the essay says almost nothing about what he thinks. Even so, in the study, people who read the pro essay thought Steve was in favor, and people who read the anti essay thought he was against. A single answer doesn’t show the effect: it shows up when the two groups are compared.

Why does your brain do this?

What a person does is right in front of you; the situation that led them to do it stays in the background. The mind first draws the most direct conclusion (“wrote in favor, so is in favor”) and only afterward tries to discount the pressure of the situation. That second step takes effort and tends to stop halfway.

When does this shortcut work well?

Most of the time, what people say and do really does match what they think and who they are. Reading a person through their behavior is a quick and often accurate way to predict what they’ll do next. And even when someone is given a task, the way they carry it out can say something about them.

Where does this show up in daily life?

Assuming the actor who plays the soap opera villain must be unpleasant in real life. Deciding the clerk was rude because they’re ill-mannered, without considering they might be at the end of a 12-hour shift. Judging a lawyer by the case they argue, or a classmate by the side they were assigned in a school debate. Before concluding something about someone, it helps to ask: what did the situation ask this person to do?

Nerd Mode

Open Nerd Mode

Edward Jones and Victor Harris (1967) asked college students to read essays for or against Fidel Castro and estimate the true opinion of the author. When they knew the author had chosen the side, they attributed the essay’s opinion to the author, which is reasonable. But they also did so, more weakly, when they knew the side had been assigned. The phenomenon became known as the correspondence bias, also called the fundamental attribution error: drawing conclusions about someone’s character from a behavior that the situation already explains.

Yuri Miyamoto and Shinobu Kitayama (2002) compared students in the United States and Japan using short, unconvincing essays about capital punishment, all with the side assigned. On a scale from 1 to 15, the Americans gave 10.8 to the author of the pro essay and 3.3 to the author of the anti essay; the Japanese, 9.3 and 7.0. The smaller gap among the Japanese suggested the bias depended on culture.

The Many Labs 2 project (Klein and colleagues, 2018) repeated that study with 7,197 participants, using the essays and professor information on this page. Those who read the pro essay gave 10.98 on average; those who read the anti essay, 4.45 (d = 1.82). The effect appeared in every sample and was similar in size in wealthy Western countries and elsewhere: the cultural difference from the original study did not replicate.

Some question the word “bias.” Drew Walker, Kevin Smith, and Edward Vul (2022) argue that an instruction rarely determines everything: someone strongly against capital punishment might refuse to write in favor or write a weak essay. So an assigned essay still carries some clue, and part of what was called an error may be a reasonable inference in an uncertain world. The debate is about how far people’s answers go beyond what that clue justifies.

This page uses the Many Labs 2 essays in the version for countries without capital punishment. The result compares your answer with that of the group that read the same essay, without saying that you personally fell for the shortcut: Steve may well agree with the side he was given.

Technical terms

  • Attribution
  • Heuristics

Discovery

Described by Edward E. Jones e Victor A. Harris in 1967.

Level of evidence

Robust evidence

Replicated in multi-lab studies or consistent meta-analyses.

Basis: Klein, R. A.; Vianello, M.; Hasselman, F.; Adams, B. G.; Adams, R. B.; et al. (2018). Many Labs 2: Investigating variation in replicability across samples and settings.

References

  1. Original studyJones, E. E.; Harris, V. A. (1967). The attribution of attitudes. Journal of Experimental Social Psychology, 3(1), 1–24.
  2. Original studyMiyamoto, Y.; Kitayama, S. (2002). Cultural variation in correspondence bias: The critical role of attitude diagnosticity of socially constrained behavior. Journal of Personality and Social Psychology, 83(5), 1239–1248.
  3. ReplicationKlein, R. A.; Vianello, M.; Hasselman, F.; Adams, B. G.; Adams, R. B.; et al. (2018). Many Labs 2: Investigating variation in replicability across samples and settings. Advances in Methods and Practices in Psychological Science, 1(4), 443–490.
  4. ReviewWalker, D.; Smith, K. A.; Vul, E. (2022). Reconsidering the “bias” in “the correspondence bias”. Decision, 9(3), 263–284.

Review

This text has not been reviewed by a volunteer yet. Want to help review it?

Found a mistake? Write to [email protected]