The Psycholinguistics of Morality

In my fellowship at Harvard University, I workied with Dr. Steven Pinker and Dr. Jesse Snedeker on an interdisciplinary project called The Psycholinguistics of Morality. Besides revealing the complexity of causal processing of morally-relevant action in language and thought, the project addressed specific concerns within and across disciplines. For linguistics, these studies aimed to inform the critical project of delineating intrinsic properties of the lexicon from extrinsic effects on language from world knowledge. For social-moral psychology, these studies brought new focus to the role of purely linguistic features in moral judgment of human behavior. And, by increasing understanding of the consequences of shifting causal representations in language and thought, this research addressed issues at the intersection of linguistics, cognitive science, and social-moral psychology, including the extent to which explicit causal models are alterable through interventions on implicit causality via general cognitive representations.


A long history of research in psycholinguistics has investigated verbs’ implicit causality biases — or, the tendency for verbs conveying agent-patient transitive events (“X verbed Y.”) to lead people to attribute causality to the sentence subject (agent, “X”), or sentence object (patient, “Y”). The extent to which patterns in implicit verb causality should be considered purely lexical features versus the result of richer inferences is debated.

In a series of studies carried out in collaboration with Josh Hartshorne, Tobi Gerstenberg, and Liane Young, I tracked implicit causality biases using a simple psycholinguistics task (e.g., participants selected a pronoun referring to “X” or “Y” to resolve: “X verbed Y because…”) and a range of morally relevant and neutral verbs. Results indicated that implicit causality for various verbs conveying harm and force (e.g., raped, killed, coerced) is predicted by (1) explicit judgments about agents and patients — i.e., ratings of patients’ capacities to control, allow and deserve events, and agents’ causal necessity and sufficiency; and, (2) participants’ own moral values — specifically, the extent to which participants endorse values that prohibit transgressions that fall outside the dyadic agent-patient framework.

Taken together, results indicate that moral values and generalized judgments about agency and patiency have the capacity to affect implicit verb causality, particularly for events involving harm and force. The implicit causality task promises to be a useful instrument for future research on motivate causal attribution.