The competition of memory and expectation in resolving long-distance dependencies: psycholinguistic evidence from Persian complex predicates

In the literature of sentence processing, there are two seemingly contradicting accounts for resolving long-distance dependencies: (1) memory-based account according to which delaying the appearance of a verb in a noun-verb dependency tends to increase the processing difficulty at the position of verb leading to the socalled locality effect (Gibson, 2000), and (2) expectation-based account, based on which the verb becomes even more predictable and easier to process due to this delay leading to the so-called surprisal effect as the reader is highly expecting the verb (Levy, 2008). Levy and colleagues (2013) argue that the verb-medial languages tend to exhibit the general patterns predicted by memory-based theories, while verb-final languages tend to exhibit the general pattern predicated by expectation-based theories. We aimed to investigate these two accounts on Complex Predicates in Persian as a verb-final language in which multiple-word verbs (i.e., complex predicates) are more widely-used as compared the one-word verbs. To this purpose, four psycholinguistic experiments (two masked self-paced reading studies and two eyetracking studies) with four pre-tests (two sentence-completion studies, one acceptability rating study, and one corpus study) with a total of 279 participants were conducted in different homogenous groups of native Persian speakers with no history of cognitive disorders and normal or corrected-to-normal eye-sight. In our experimental design (2x2), we opposed complex predicate conditions to simple predicate conditions and manipulated the stimuli by two factors of distance (i.e.,short conditions versus long conditions) and predictability (i.e., strong predictability versus weak predictability). The intervener in the first self-paced reading study and the first eye-tracking study was a short prepositional phrase in the short conditions and a prepositional phrase plus a relative clause in the long conditions. In the second self-paced reading and the second eye-tracking studies, the intervener consisted of the same short prepositional phrase in the short conditions whereas it was a long uninterrupted prepositional phrase in the long conditions. We decided to keep th esame type of intervener in the long conditions of the second design in order to find out if the type of intervener affects the processing difficulty.