Animacy-based predictions in language comprehension are robust: contextual cues modulate but do not nullify them.

Couldn׳t a humble coconut hurt a gardener? At least in the first instance, the brain seems to assume that it should not: we perceive inanimate entities such as coconuts as poor event instigators ("Actors"). Ideally, entities causing a change in another entity should be animate and this assumption not only influences event perception but also carries over to language comprehension. We present three auditory event-related brain potential (ERP) studies on the processing of inanimate and animate subjects and objects in simple transitive sentences in Tamil. ERP responses were measured at the second argument (event participant) in all three studies. Experiment 1 employed all possible animacy combinations of Actors and Undergoers (affected participants) in Actor- and Undergoer-initial verb-final orders. Experiments 2 and 3 employed a fairly novel context design that enabled us to compare ERPs evoked by identical auditory material to differing contextual expectations: Experiment 2 focussed on constructions in which an inanimate Actor acts upon an inanimate Undergoer, whereas Experiment 3 examined whether and how a preceding context modulates the prediction for an ideal Actor. Results showed an N400 effect when the prediction for an ideal (animate) Actor following an Undergoer was not met, thus further supporting the cross-linguistically robust nature of animacy preferences. In addition, though specific contextual cues that are indicative of a forthcoming non-ideal Actor may reduce this negativity in comparison to when such cues are not available, they nevertheless do not nullify it, suggesting that animacy-based predictions are stronger than contextual cues in online language comprehension.