Referential form and implicit causality (Poster/Discussion,International)
2017, September 7-9
 Duygu Özge, Joshua Hartshorne, Jesse Snedeker,FLE
Abstract: How do listeners interpret speaker’s choice of anaphoric form? According to Centering Theory, speakers use reduced forms when both the anaphoric from and the antecedent are the topics of consecutive sentences.[1] This prediction has been a fixture in a number of theories,[2] but was recently challenged in a cross-linguistic comprehension study focusing on implicit causality (IC).[3] The study found stimulus bias, regardless of whether the pronoun is dropped, both in the stimulus-experiencer verbs (verbs where the subject is the stimulus causing a psychological state in the object –the experiencer; frighten-type) and in the experiencer-stimulus verbs (verbs where the experiencer has the subject role and the stimulus has the object role; fear-type). Yet, anaphoric form was not manipulated within language in this study. A recent study on Italian IC verbs found a stimulus bias regardless of anaphoric form.[4] Japanese speakers, however, showed a stimulus-bias for frighten-type verbs and an experiencer-bias in fear-type verbs in zero-pronoun condition in explanation continuations, whereas they showed a stimulus bias for both conditions in overt pronoun and free prompt conditions,[5] which is puzzling. Despite systematically manipulating the anaphoric form, Italian and Japanese IC studies were based on sentence-continuation. We test comprehension systematically manipulating anaphoric form and verb-type within a language. We investigate whether experiencer or stimulus is selected as the antecedent of an anaphor in IC verbs in Turkish. We also test an intuition-based hypothesis that Turkish IC verbs result in an experiencer bias.[6] If true, this would challenge the assumption that IC biases are consistent across languages.[7,8] However, this intuition was based on examples with (+) human experiencer and (-) animate stimulus. If there is an experiencer bias, [6] we expect the subject to be selected as the antecedent of the ambiguous pronoun in fear-type verbs and the object to be selected in frighten-type verbs; if there is a stimulus bias, [3, 4] we expect the object to be selected in the fear-type verbs and the subject to be selected in frighten-type verbs. In two rating studies,[8] participants read 24 sentences with frighten-type and fear-type verbs. Each sentence had a clause conjoined with ‘because’ that had an ambiguous anaphor (Study-1:full-pronoun; Study-2:zero-pronoun) ending with a non-word adjectival predicate (Bahar dazzles Ceren a lot because she is extremely dax). Participants chose the referent for this non-word (i.e.,Who is dax?). There was a greater subject preference in frighten-type verbs compared to fear-type verbs both in Study-1 (full-pronoun) (p<.001) and in Study-2 (zero-pronoun) (p<.001). Pronoun omission significantly increased the subject preference only in fear-type verbs (p<.001). The anaphor was not resolved towards the experiencer when the referents were controlled for animacy.[c.f.,6] Our findings corroborate with a cross-linguistic tendency where frighten-type verbs are strongly biased towards the stimulus regardless of form. [4,7] However, different from previous studies we found that fear-type verbs are more malleable and their causality biases change in line with the form of the anaphor. This pattern challenges all current theories of anaphora. [1] Grosz, Joshi, & Weinstein (1995). [2] Rohde & Kehler (2013). [3] Hartshorne, Sudo, & Uruwashi (2013). [4] Fedele & Kaiser (2015). [5] Ueno & Kehler (2011; 2016). [6] Turan (1998). [7] Hartshorne et al. (2013). [8] Hartshorne & Snedeker (2013).
Architectures and Mechanisms of Language Processing (AMLaP 2017)