Experimental approach of the prosodic component in the linguistic input of gardenpath sentences in Brazilian Portuguese [In Portuguese]

This study investigates what part the prosodic component plays in linguistic processing. Can prosodiccues be computed early in the linguistic process and influence the syntactic processing? Some studiesin American English (Kjelgaard&Speer, 1999; DeDe 2010) and German (Steinhaueret al 1999) haveshown that yes, furthermore, prosodic elements like intonational phrases can modify the syntacticchain during processing. Our work is based on the premises that, during the perceptive processing, anearly activation of the prosodic component in the linguistic input can lead the sentence’s syntacticstructure.There have been studies in Brazilian Portuguese that show how prosody influences parsing(Louren o-Gomes 2008; Magalh es& Maia 2006), however, none that tested the effects of prosodicconstituents’ organization in online tasks. Therefore, a self paced listening test, with garden-pathsentences, was conducted in Brazilian Portuguese. 60 subjects listened to 252 experimental items,presented in two prosodic conditions and three syntactic versions (EC, LC and SW), as follows:Cooperating prosody – Syntactic and prosodic boundaries are the same(1) Early Closure Sentence (EC): While Mary was mending // the sock / fell off her lap.(2) Late Closure Sentence (LC): While Mary was mending / the sock // it fell off her lap.(3) Semantic Weak Sentence (SW): While Mary was mending // the sky / turned grey and dark.Conflicting prosody – Syntactic and prosodic boundaries are not the same(4) (EC): While Mary was mending / the sock // fell off her lap.(5) (LC): While Mary was mending // the sock / it fell off her lap.(6) (SW): While Mary was mending / the sky // turned grey and dark.Reaction times (RT) were measured when subjects listened to both conditions. The greater reactiontimes were registered while subjects listened to items from the conflicting prosody condition. Weperformed factorial ANOVA (3x2) comparing the average RT (normalized by z-score) in the threesyntactic versions tested and in both prosodic conditions. Prosody was significative in our analyses bysubjects and by items, considering subjects and items as random factors: F1(1,59)=45,697 p<0,001;F2(2,246)=34,395 p<0,001, and we did not find any statistic differences between syntactic versionsEC, LC e SW: F1(1,59)<0,001 p=1; F2(2,246)<0,001 p=1.Based on these results, we consider that in Brazilian Portuguese, the prosodic difference between thetwo conditions, conflicting prosody and cooperating prosody, in garden-path sentences is detected bythe listener. Moreover, it can affect the chosen parser for a given syntactic st