The Assumption Nobody Questions
Moral principles are supposed to be universal. You believe killing is wrong in English and in Mandarin, in your living room and in a courtroom. The language wrapped around a dilemma should be packaging, nothing more. Philosophy treats moral reasoning as operating above the surface features of language. Common sense agrees. Both are wrong.
The Experiment That Broke the Assumption
Albert Costa at Universitat Pompeu Fabra and Boaz Keysar at the University of Chicago recruited 725 volunteers and presented them with a scenario that moral philosophers have argued over for decades: a runaway trolley is heading toward five people, and the only way to stop it is to push a large man off a footbridge into its path, killing him but saving the five (Costa et al., 2014). Half the participants read the dilemma in their native language. Half read it in a foreign tongue they had learned later in life.
Among native-language participants, fewer than 20% said they would push the man. Among foreign-language participants, roughly 50% said yes, despite facing the same dilemma, the same logic, and the same five lives versus one. The only variable was the language doing the thinking.
The effect was symmetric: native Spanish speakers tested in English showed it, and native English speakers tested in Spanish showed it equally. The shift was not about any specific language; it was about using one that was not the one you grew up speaking.
Feeling Less, Not Thinking More
The obvious first explanation is that processing a foreign language is harder, so people slow down, deliberate more, and arrive at the coldly rational answer. Sayuri Hayakawa tested this directly in six experiments using a process-dissociation technique that separately measures two moral impulses: the deontological pull (rules like "do not kill") and the utilitarian push (maximize welfare). If the foreign-language effect worked by boosting deliberation, utilitarian responding should increase, but it didn't (Hayakawa et al., 2017).
What changed was the other side: deontological responding decreased. People were not thinking more clearly; they were feeling less intensely. The foreign language did not amplify the calculator; it muffled the alarm.
This makes sense when you consider how languages are acquired: you learn your native language embedded in punishment and praise, love and fury, the full voltage of childhood, while a foreign language learned in a classroom carries vocabulary lists and grammar drills. The semantic content is identical, but the emotional residue is not.
Catherine Harris demonstrated this by measuring skin conductance in Turkish-English bilinguals (Harris et al., 2006): reprimands like "Shame on you!" triggered powerful physiological reactions only in Turkish, with participants reporting they "heard" the phrases in the voices of their parents, which suggests your native language is not just a communication tool but a container for emotional memory where moral intuitions live.
The Meta-Analytic Confirmation
Circi, Gatti, Russo, and Vecchi conducted the first meta-analysis of the foreign-language effect on decision-making in Psychonomic Bulletin & Review (Circi et al., 2021). Across 17 studies and 38 experiments, the pooled effect was Hedges' g = 0.22 (95% CI: 0.14–0.30, p < .0001). The fail-safe number was 1,175. Egger's test found no publication bias (p = .32). Small to medium, but durable.
A critical moderator emerged: linguistic similarity. When native and foreign languages came from different branches, such as Spanish and English or Korean and English, the effect was robust at g = 0.30, but when they shared a branch, such as German and English or Norwegian and Swedish, it vanished entirely (g = 0.06, p = .28), which strongly suggests that closely related languages share enough cultural and phonological overlap to preserve the emotional charge that normally dampens in a truly foreign tongue.
What This Means at Scale
Roughly 1.35 billion people speak English as a non-native language. The European Parliament conducts debates in 24 official languages, but working documents circulate primarily in English and French, languages foreign to most of the 705 Members. If a Hedges' g of 0.30 operates whenever a legislator reads a policy brief in English rather than Finnish or Greek, the EU's regulatory apparatus is filtered through emotional dampening that no one voted for and no one audits. The same logic applies to the United Nations, NATO, and the boardrooms of every multinational corporation that deliberates in a lingua franca. Nobody chose to blunt their moral intuitions; they just chose to speak English.
The Strongest Case Against
The most pointed criticism comes from Brouwer (2019) and Dylman & Champoux-Larsson (2020), who failed to replicate the effect in highly proficient early bilinguals and same-branch language pairs, raising the possibility that the foreign-language effect is a transitional artifact of incomplete language acquisition rather than a fundamental property of non-native processing. A person who has spoken English for 30 years in an English-speaking country may no longer experience it as emotionally distant, and the effect would vanish. This boundary condition limits generalizability: the effect may be strongest in people still struggling with the language and weakest in fluent professionals at international negotiating tables, which is precisely where it would matter most.
The meta-analysis partially addresses this: proficiency as a moderator did not reach significance in the meta-regression, but the authors note proficiency data were averaged across entire samples, which may have buried individual-level variation, so the question remains genuinely open.
What We Didn't Prove
Every study uses hypothetical dilemmas rather than real moral decisions, so whether someone who says they would push a man off a bridge would actually do so is unknown, and Bostyn et al. (2018) found that hypothetical and real-world moral behavior can diverge substantially, which means the laboratory effect may overstate or understate the real-world impact.
The meta-analytic effect size of g = 0.22 is small to medium, meaning the foreign language tips people at the margin rather than overriding deep convictions. The 50% utilitarian rate in the original Costa study still means half the foreign-language group refused.
All studies used sacrificial trolley-style dilemmas, so whether the effect extends to medical triage, military rules of engagement, or corporate ethics remains untested. Heterogeneity across studies was moderate (I² = 64.74%), meaning substantial variation is not fully explained by the moderators tested.
The Bottom Line
The language you think in changes what you think is right. Not because it changes the logic, but because your native language carries emotional freight that a foreign language does not. Across 38 experiments, thousands of participants, and a dozen countries, the effect holds: switch the language, and the moral calculus shifts. The finding is meta-analyzed, bias-free, and mechanistically coherent.
What You Can Do
If you make important decisions in a non-native language: When a policy proposal or ethical judgment feels unusually cold, consider whether the language is dampening an emotional signal that would normally inform your reasoning. Translate the key question into your native language and check whether your answer changes.
If you work in international organizations: Audit whether key ethical decisions are routinely made in a lingua franca native to few participants. Emotional distance from a non-native language is a systematic bias to account for, not a defect to overcome.
If you teach ethics or run training programs: Present moral scenarios in participants' native languages when possible. A foreign language may produce more utilitarian responses that reflect reduced emotional engagement, not better moral reasoning.
If you are bilingual and curious: Try the footbridge dilemma in both your languages. The shift is not a sign that one answer is right. It is a window into how deeply language has shaped the moral intuitions you thought were yours alone.