Processing of stacked NPs in embedded sentences in Malayalam

Abstract

The paper examines the effect of stacked NPs in centre-embedded sentences in Malayalam and uses the experimental results to compare two theories of processing difficulty: Gibson (2000)’s Dependency Locality Theory (DLT) and Hale (2001)’s Surprisal Theory. Crucially, the study also looks at the definite vs. non-definite NP distinction in Malayalam in a stacked NP context. Non-definiteness in Malayalam is indicated by having the determiner oru before the noun; a non-definite NP in a centre-embedded sentence disrupts NP stacking. A self-paced reading task is run using single and double-embedded sentences in Malayalam with the embedded NP alternating between definite and non-definite conditions. The test is designed to determine if (i) stacked NPs in a centre-embedded sentence result in a processing difficulty (ii) having oru preceding an NP affects processing. The results showed that processing difficulty increases with the addition of each NP; NPs from the embedded clauses take longest to be read. When the embedded NP is preceded by oru, reaction time significantly drops at oru and the following NP, indicating that the determiner facilitated integration of theNP, disrupting NP stacking. The results were compared against the predictions of DLT and Surprisal models. We found that the anticipation based Surprisal account best accounted for the results for Malayalam.