Open data · CC BY 4.0 what each one claims, and what replicated no ads · no tracking
Bias Atlascognitive biases

Moral typecasting

also known as — Moral agent-patient asymmetry · Agent-patient typecasting
People sort others into those who act and those who are acted upon, and once someone is a victim they are seen as less capable of doing wrong.
Does it replicate? Replicated Across 8,002 people the asymmetry came back slightly larger than the original, at d = 0.95.
VerdictReplicated
First published2009
FieldSocial and self
Sources3
Last checked2026-09-21
What it claims

What does Moral typecasting mean?

Gray and Wegner proposed that moral perception runs on two roles rather than one dimension. A moral agent does things and can be blamed; a moral patient has things done to them and can be harmed. The claim that makes this a bias is that the roles are exclusive: casting someone as one makes it harder to see them as the other, independently of the facts.

The test is symmetrical by design. The same accident, with the same shattered glasses and the same cut leg, is shown with the older party knocking the tray over or with the older party being cut. The question is whether responsibility follows the act or follows the casting.

Verdict

A repeat found the effect again. The other 163.

The field
41 of the 170 entries in social and self share this verdict.
Does it replicate?

Has Moral typecasting been retested?

Replicated Across 8,002 people the asymmetry came back slightly larger than the original, at d = 0.95.

In the original studyd = 0.80 (95% CI 0.31 to 1.29)
Pooled across the replication, aggregated = 0.95 (95% CI 0.91 to 1.00)
no effect Original Original: d = 0.80, 95% CI 0.31 to 1.29 Replication Replication: d = 0.95, 95% CI 0.91 to 1.00
The same scale for both, so the distance between the two points is the change the repeat found. Hover a point for its interval.

Many Labs 2 ran the glasses vignette in Slate 2, which 61 of the project's 125 samples administered. Its materials put the pair in a picture and made the younger party a baby. The adult was rated 5.41 for responsibility and the baby 3.77, giving d = 0.95 with a 95% confidence interval from 0.91 to 1.00. The original was 0.80, interval 0.31 to 1.29, an estimate Many Labs 2 derived from the original test statistic.

The replication estimate sits inside that interval and is far more precise. Responsibility was the only measure in the confirmatory test, but the protocol planned the other two as secondary analyses and both came back. The adult was rated as acting more intentionally, d = 0.50 with an interval from 0.46 to 0.55, and as feeling less pain when harmed, d = 0.62 with an interval from 0.57 to 0.66. Each is smaller than the estimate Many Labs 2 gives for the original on the same measure.

Read off Klein, R. A., Vianello, M., Hasselman, F., 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.

Examples

What are some examples of Moral typecasting?

The collision in a doorwayillustration

A small child and a grown woman walk into each other and a tray goes over. Whichever of them ends up cut, onlookers put the responsibility on the adult, so the role each has been cast in is doing work the description does not support.

The neighbour who has had a bad yearillustration

After a run of misfortunes the street has rallied round, that neighbour starts a boundary dispute. Others find it oddly hard to see them as the one in the wrong, because they have been filed as someone things happen to.

The person who always organises thingsillustration

Whoever runs the committee and chases every decision is assumed to be behind whatever has happened. When they are the one harmed, it takes longer for anybody to notice, because the other role is already occupied.

Cases marked as illustrations describe nobody in particular. They are there to make the pattern recognisable, not to report that a particular thing happened.

The experiments

What experiments is Moral typecasting based on?

The design

The paper reports seven studies. Study 1a is the one with a replication record. Sixty-nine people, recruited in dining halls, shopping centres and train stations, saw pictures of Sam and Roger, each of whom was either 5 or 25 years old. They read one sentence, in which Sam pushes a tray of glasses off a table and a shard cuts into Roger's leg.

They then rated Sam's responsibility, how intentional his behaviour was, and how much pain Roger felt, each on a seven-point scale. The paper prints test statistics and nothing else for this study: t(68) = 3.32 for responsibility, t(68) = 2.20 for intentionality and t(68) = 3.49 for pain. It gives no means, no standard deviations and no effect sizes for Study 1a.

The figures usually quoted for it, 5.29 against 3.86 and d = 0.80 for responsibility, 4.05 against 3.07 and d = 0.53 for intentionality, and 4.63 against 5.76 and d = 0.85 for pain, come from Many Labs 2's account of the study rather than from the paper. Responsibility was the measure carried forward for replication.

2 primary research1 replication and reanalysis

Origin

Who first described Moral typecasting, and when?

Year2009
WhoKurt Gray and Daniel Wegner
WhereJournal of Personality and Social Psychology

The paper is "Moral typecasting: Divergent perceptions of moral agents and moral patients", and it sits inside their broader dyadic account of moral cognition.

The claim, and the replication this entry reads its figures from.

Where it runs out

When does Moral typecasting not apply?

The confirmatory test rested on responsibility alone. Intentionality and the pain judgement, which are what make the finding about typecasting rather than about responsibility alone, were secondary analyses, and both shrank. The replication also widened the contrast it inherited. Gray and Wegner set a child of five against an adult of twenty-five; Many Labs 2 set a baby against a grown man.

Either way the two roles are confounded with almost everything else that separates the old from the young: capability, foresight, physical control. The vignette shows that responsibility tracks something other than the act; it does not isolate moral agency as the something. The effect also varied across the samples that ran it, with I2 of 73 per cent, and the test comparing WEIRD with less WEIRD samples returned p = .01.

Commonly misread as

How is Moral typecasting commonly misread?

The finding is sometimes read as showing that people excuse victims, which overstates it. What was measured is a difference in attributed responsibility between two parties in a described accident, and the more mundane explanation is that adults are expected to be careful around small children. The original study is widely described, by Many Labs 2 among others, as pitting a man against a baby. The paper's own targets are a five-year-old and a twenty-five-year-old. Treating this experiment as strong evidence for the dyadic theory of morality also runs ahead of what one vignette can carry, however well it replicates.

Sources

Sources for Moral typecasting

  1. 1Gray, K., & Wegner, D. M. (2009). Moral typecasting: Divergent perceptions of moral agents and moral patients. Journal of Personality and Social Psychology, 96(3), 505-520. Read in full as a PDF on 21 September 2026, from the Wayback Machine's copy of Daniel Wegner's Harvard page. The live file at scholar.harvard.edu answers with an Akamai access denial to every user agent tried, and OpenAlex and Unpaywall both record the article as closed.primary
  2. 2Klein, R. A., Vianello, M., Hasselman, F., 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. Read in full as the published version, from Radboud University's repository, on 21 September 2026, after journals.sagepub.com answered 403. Effect 22 in Slate 2 carries the figures used here, and Table 3 carries the heterogeneity test.replication
  3. 3Klein, R. A., Vianello, M., Hasselman, F., et al. Many Labs 2 protocol, ML2_-_Protocol.pdf, deposited on the Open Science Framework. Section 22 read on 21 September 2026. It carries the replication materials, the analysis plan naming responsibility as the confirmatory measure, and the list of known differences from the original, which mentions online administration but not the change of target.method

Every claim on this page was held against these sources on 2026-09-21. Nothing here is written from memory. Written and checked by Krishna Chagti; how entries are written and corrected. Found a mistake? Report it, with the sentence and a source.

About Moral typecasting

Is Moral typecasting real?

Across 8,002 people the asymmetry came back slightly larger than the original, at d = 0.95. The effect sizes from the original study and from the replication are printed side by side above, so the comparison is visible rather than asserted.