The bigger picture

Do people process moral issues differently in their native and non-native languages? So far, many studies ( Driver, 2022 ; Costa et al., 2014 ) have suggested bilinguals process morality differently in their L1 versus L2, which might be accounted by the foreign langauge distance effects. In addition, some studies (Dorfman et al., 2024) found that L2 proficiency level is related to the moral processing. This is probably because processing moral information in a less proficient langauge consumes more cognitive resources, leaving less room for moral processing.

The study: How does the brain respond to moral violations?

Before asking how moral processing changes across languages, we first need to understand what happens when people encounter a moral violation in their native language. This study examined whether the brain treats morally objectionable behavior as simply another kind of meaning-related inconsistency or recruits partly distinct processes.

The project has now been published in Brain Sciences:

Meng, J., Zhang, D., Zhong, Y., Xu, X., & Kaan, E. (2026). Comparing Brain Responses to Moral and Semantic Violations. Brain Sciences, 16(4), 375.

Why compare moral and semantic violations?

Consider the difference between reading about someone doing something morally wrong and reading a sentence that simply does not make sense. Both events conflict with what we know about the world, but a moral violation may additionally engage emotional and evaluative processes.

To separate these possibilities, we compared three kinds of short texts:

Moral violation: The action was plausible but morally objectionable.
Semantic violation: The action was morally irrelevant but did not make sense in the situation.
Neutral control: The action was both plausible and morally acceptable.

The critical target sentence—and, importantly, the critical word—was identical across the three conditions. Only the preceding context changed. This design allowed us to attribute differences in brain activity to the meaning established by the context rather than to differences between the target words themselves.

Example of the three experimental conditions. The context changes whether the same critical expression is morally objectionable, semantically implausible, or neutral.

How we measured processing in real time

We recorded electroencephalography (EEG) while monolingual English speakers read the texts for comprehension. EEG measures small changes in electrical activity at the scalp. When the signal is aligned to the onset of a particular word and averaged across trials, it produces an event-related potential, or ERP. ERPs allow us to examine how language processing unfolds from one fraction of a second to the next.

Thirty-eight adults participated, and 33 datasets remained after applying the study’s EEG data-quality criteria. Participants first read a context sentence at their own pace. The target sentence then appeared in short segments while EEG was recorded. They answered occasional comprehension questions rather than making moral judgments during reading. This was important because it let us study relatively natural comprehension without directing participants to evaluate every action explicitly.

We examined three time windows after the critical word appeared:

150–250 ms: an early interval associated with attention to salient information;
300–500 ms: the conventional N400 interval, often associated with predicting, retrieving, and integrating meaning;
500–900 ms: the Late Positive Component (LPC), often associated with sustained evaluation and reanalysis.

What we found

The clearest difference appeared in the brain’s positive-going response. Moral violations elicited a long-lasting posterior positivity beginning at approximately 300 ms. Semantic violations elicited a later positivity that was descriptively more frontal, although the difference in scalp distribution was not statistically conclusive.

In the 500–900 ms window, both violation conditions produced a larger LPC than the neutral condition, and the moral violations produced a larger positivity than the semantic violations. We found no reliable difference among the conditions in the earlier 150–250 ms window.

ERP responses at the critical word. Moral violations are shown in red, semantic violations in blue, and neutral controls in black. The highlighted region marks the late positive time window.

One result was especially informative: across individual items, moral acceptability explained the size of the posterior LPC even after plausibility was taken into account. Plausibility did not provide the same additional explanatory value once moral acceptability was included. This pattern suggests that the LPC was not merely responding to something unusual or difficult to understand; it was particularly sensitive to the moral status of the event.

Why it matters

The findings indicate that moral and semantic violations are not processed in exactly the same way. Although both require readers to evaluate an unexpected continuation, moral violations engage a sustained posterior response that begins earlier and is more strongly related to moral acceptability. In other words, the brain appears to distinguish between an event that does not make sense and an event that violates a social or moral norm.

This monolingual study provides a baseline for the broader question motivating the project: how might these processes change when bilinguals encounter moral information in a native versus a non-native language? Establishing the timing and character of moral processing in native-language comprehension is a necessary step toward understanding how language experience, proficiency, emotion, and cognitive demands may shape moral evaluation.

Read the paper and explore the materials

Questions or ideas about this work are always welcome. Please feel free to contact me.