Partial reanalysis: A large-scale replication testing plausibility and transitivity bias

Abstract

Partial reanalysis occurs when readers of a garden path (GP) sentence fail to fully reanalyze it, resulting in a persistent misinterpretation. We investigate partial reanalysis in two large-scale single-trial experiments (n=4,560). Experiment 1 used sentences with optionally transitive verbs and a simple post-verbal NP, manipulating the plausibility of the initial interpretation and the transitivity bias of the verb; Experiment 2 used reflexive or alternating unaccusative verbs, which eliminate the need for an additional NP. We find a main effect of Sentence Type, with lower accuracy for GP sentences, and a main effect of Plausibility, with plausible sentences showing lower accuracy than implausible ones, but no interaction between the two - indicating that the semantic consideration is independent of the syntactic manipulation. Transitivity bias interacts with Sentence Type - the accuracy advantage of non-GP over GP sentences is reduced for more transitively biased verbs - suggesting that highly transitive verbs attach the NP as their object regardless of its position. Our results replicate the existence of partial reanalysis in GP sentences with a simple post-verbal NP and support good-enough approaches to sentence processing.

Samuel Joseph Amouyal

PhD candidate @ TAU, Research scientist @ LegionSecurity

I am a PhD candidate in computer science interested in the intersection between Natural Language Processing (NLP) and other fields (psycholinguistics, economy, game-theory, literature …)