engleski

Who is likely to be re-mentioned? A cross- linguistic study of implicit causality bias

Certain interpersonal verbs have been identified as showing a strong bias that mirrors the reader's expectations about who is likely to cause the event or state described by the verb. This phenomenon is known as an implicit causality (IC) bias, and it has been studied extensively in psycholinguistics, where it is assumed to play a crucial role in determining one’s preferences when resolving reference ambiguity (Garvey & Caramazza 1974). IC bias shifts the direction of causality in a subsequent clause to one of the verb arguments (subject or object) in the preceding clause. For example, when presented with sentences fragments such as Mary amazed Ann because she... or Mary scolded Ann because she…, one can predict which referent is likely to be re- mentioned. According to the verb bias, in the first sentence, the referent is Mary, while in the second one the referent is Ann. Several accounts on IC bias exist, e.g. the world knowledge account, the arbitrary semantic tag account and the semantic structure account (Hartshorne & Snedeker 2013). A growing body of research stems from the semantic structure account, which specifies IC bias by relating it to semantic classes and thematic roles of the verb’s arguments. The most widely adapted taxonomy was proposed by Rudolph and Försterling (1997 ; originally from Brown & Fish 1983), who divided IC verbs into two major semantic classes: action (behavioural) and state (mental) verbs. Four pairs of thematic roles are further distinguished: agent-patient, agent-evocator, stimulus-experiencer and experiencer-stimulus. Since semantic structure account allows for cross-linguistic comparisons suitable for testing the generalizability or variability of IC bias (review: Hartshorne et al. 2012), it is an approach taken in the current study, as well. Two main motives drove this study. First, although there is overall evidence of IC bias consistency across languages, differences may emerge regarding its direction and strength. Second, most of the available data on IC bias still comes from English (Ngo & Kaiser 2020). Taking these two facts into consideration, the purpose of the present study was to examine IC bias in another Indo-European language, Croatian. The specific aim was to examine the direction and strength of IC bias in Croatian and to compare the obtained values to those obtained in English. The comparisons are also observed with respect to the semantic class. A web-based questionnaire in a completion-task form was sent to University undergraduate and graduate students (N=30), all native speakers of Croatian. They were instructed to complete 82 sentence fragments such as John frightened Peter because he… [Ivan je uplašio Petra jer je…] and to report who they referred to in each sentence. The stimuli were based on the former study in English (Ferstl et al. 2011), as to allow for valid cross-linguistic comparisons. Only verbs with IC bias higher than 70% were included. Sentence fragments were additionally controlled for relevant and language-specific features. Referents were controlled for gender, sentence fragments for length and tense, and verbs were controlled for frequency and semantic class (51 state and 31 action verbs). IC bias was calculated using the percentage- based formula reported in Ferstl et al. (2011). This is a study in progress, but the preliminary results show that IC bias scores are almost evenly distributed across continuations, with 79% of subject bias (SD = 8.98) and 81% of object bias (SD = 10.28). There is also a 99% match rate in the direction of causality between Croatian and English. A Mann-Whitney U-test was used to analyse the difference in IC bias score between the two languages, and the results show no significant difference neither in the overall IC bias score, nor in the subject or object bias score (all p > 0.05). However, a significant difference is found in the strength of IC bias for action verbs (p < 0.02), although not for state verbs (p < 0.33). The obtained preferences for IC verbs in Croatian replicate and extend previous findings. IC bias seems to generalize to interpersonal verbs in Croatian as well, but its strength somewhat varies across semantic classes. Since this is an ongoing study, the relation between thematic roles and the strength of IC bias is to be discussed further.

implicit causality, semantic structure account, verb taxonomy, Croatian

nije evidentirano