Revisiting anti-locality effects: Evidence against prediction-based accounts

• Published in: Journal of memory and language Vol. 121; p. 104280
• Main Authors: Apurva, Husain, Samar
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.12.2021
• Subjects: Anti-locality effect; Parsing error; Processing in SOV languages; Shallow parsing; Working-memory; Working-memory; Parsing error; Anti-locality effect; Processing in SOV languages; Shallow parsing
• ISSN: 0749-596X; 1096-0821
• DOI: 10.1016/j.jml.2021.104280

•Robust prediction happens only when the preverbal syntactic environment is simple.•The role of working-memory on predictive processing in SOV languages is critical.•Anti-locality is driven by a shallow parsing mechanism, not predictive processing.•Shallow parsing is triggered with increased preverbal complexity (e.g., embeddings).•Theories of syntactic prediction need to be informed by working memory constraints. Anti-locality effects in Subject-Object-Verb (SOV) languages are characterised by a facilitation at the clause-final verb when the distance between the verb and its prior dependents is increased. These effects are understood to be driven by a robust prediction mechanism. In this work, we revisit this rationale by replicating key anti-locality experiments for Hindi. Through a series of sentence completion studies we demonstrate that neither are clause final verbal predictions always robust, nor do these predictions necessarily improve with increased distance. Despite the evidence from sentence completion studies, the self-paced reading studies reported in this work show a facilitation at the critical verb with increase in the noun–verb distance compared to when this distance was short. We suggest that the observed effects arise due to the formation of a shallow sentential representation where the required syntactic dependencies are not formed. In other words, the facilitatory effect is not driven by a robust prediction mechanism. This proposal is additionally supported by a lack of high comprehension accuracy in multiple experiments. Additionally, our work shows that robust verbal predictions are successfully made and these predictions are successfully maintained in structures involving fewer clausal embeddings, and simple verb argument structure. This suggests that when preverbal complexity increases prediction suffers. Together the results demonstrate the overarching influence of working-memory constraints on verbal prediction in an SOV language like Hindi.