Animacy Effect

The second research question examines whether core UEs are more affected by the animacy of the verb argument than non-core ones. The results of the present study provide an affirmative answer to the question. The overpassivization tendency of core UEs are affected by animacy types more strongly than non-core ones. When the subject of the target sentence is animate, core UEs had a higher correct response rate than non-core ones. It indicates that core UEs are less likely to be passivized than non-core ones. Upper-intermediate learners stop passivizing core UEs with animate subjects, and their performance is already native-like. When the target sentence is non-animate, core UEs are more likely to be passivized than peripheral ones, and intermediate learners still tend to passivize core UEs. Advanced learners did not passivize core UEs more frequently than non-core ones.

The results of the present study also show that L2 learners of the three groups are more likely to passivize UEs with inanimate subjects than with animate subjects. When the subject is animate, UEs, especially non-core ones, cause problems to low-intermediate learners but not to

- 160 -

upper-intermediate and advanced learners. When the subject is inanimate, L2 learners, even advanced level learners, tend to passivize UEs.

Previous studies show that animacy types play a significant role in L2 acquisition of UAs (Chung, 2014, Ju, 2000). UAs with animate subjects are less likely to be passivized than with inanimate subjects. The present study indicates that animacy types not only influence the overpassivization tendency of UAs but also of UEs. As discussed in 4.4.1, animacy is found to affect the overpassivization tendency of UAs. UAs with animate subjects had higher correct response rates than UAs with inanimate subjects. However, advanced learners had overcome the animacy effects on UAs, and they did not tend to passivize UAs with non-animate subjects.

The effects of animacy on UEs, however, last even for advanced learners. Advanced level learners are still prone to passivize UEs with inanimate subjects more frequently than UEs with animate ones. It is indicated that animacy types affect the tendency of overpassivization of UEs to a greater extent than UAs because advanced L2 learners had overcome the animacy effects on UAs but not on UEs.

The results are partially in consistency with the findings of Vernice and Sorace (2018), who examine the animacy effects on the processing of Italian intransitive verbs. By monitoring Italian native participants’ eye-movement when they processed sentences involving UAs and UEs with animate and inanimate subjects, their study finds that the animacy effect affected the processing of the subject argument of UEs to a greater extent than the processing of the subject argument of UAs. That is, the reading time of UEs with inanimate subjects is significantly longer than UEs with animate subjects, while there is no significant difference in the reading time between UAs with inanimate subjects and UAs with animate ones. The results confirmed

- 161 -

that the subject argument of UAs and UEs are not processed in the same way in Italian.

Agentivity is a primary feature of unergativity, and it plays an important role in deciding a strong preference for animate agents in UEs but not for UAs.

There might be two reasons responsible for the different animacy effects on UAs and UEs.

First, the argument of UAs and UEs have different thematic properties. The argument of UAs is usually non-agentive, which bears the thematic role of patient or theme, while the argument of UEs is typically agentive, which bears the role of agent (Dowty, 1991; Van Valin, 1990). An agent is the voluntary initiator of some action (Fillmore, 1968; Jackendoff, 1990) that tends to be typically animate. Inanimate entities generally lack volition and initiation. On the assumption that the arguments of UEs are agents and prototypical agents tend to be animate (Comrie, 1989; Dowty, 1991), the argument of UEs typically requires an animate entity while the argument of UAs does not. Second, the lexical properties underlying the UA-UE distinction are different. As suggested by the SIH, agentivity is a fundamental property of unergativity, while telicity is a primary feature of unaccusativity (Sorace, 2000, 2004, 2011). UAs typically denote change of location or change of state. UEs typically involve actions that are performed by a prototypically agentive argument, and agentive arguments tend to be animate (Dowty, 1991;

Van Valin, 1990). Therefore, agentive UEs tend to show a stronger preference for animate entities while non-agentive UAs do not exhibit the preference. The violation of animacy expectation has a stronger effect on UEs than on UAs.

Furthermore, the study finds that animacy types have a stronger effect on core UEs than peripheral ones. Agentivity is assumed to be the primary feature that distinguishes core UEs from peripheral ones. Recall that maximally agentive verbs at the bottom of the UE end of the

- 162 -

hierarchy are core verbs denoting controlled processes. It is predicted that core verbs that are strongly agentive would indicate a stronger preference with animate subjects than non-agentive peripheral UEs. The results of the study indicate that the position of the verb along the SIH continuum influences the overpassivization tendency of UEs. When the argument of the verb is animate, core UEs are less likely to be passivized than non-core ones; when the argument of the verb is non-animate, core UEs are more likely to be passivized than non-core ones. The experimental results confirm that the animacy types influence core UEs to a greater extent than peripheral ones.

The results of the present study are also in line with the findings of Vernice and Sorace (2018). In Italian, core UEs are categorical in selecting auxiliary ‘have’ as a perfective marker, whereas peripheral UEs are more variable in selecting ‘be’ or ‘have’ as a perfective marker.

Regression data of UEs indicate that inanimate subjects caused more regression in core UEs in comparison with non-core ones. According to Vernice and Sorace (2018), the explanation for the regression data might be due to the ‘semantic (or thematic)’ accounts (Trueswell et al., 1994;

Ferretti et al., 2001), which suggests that the comprehension system attempts to set up a link between the animacy of subject and the thematic and aspectual features of the verb. When the argument of the verb is inanimate, the comprehension system forces reanalysis, as indicated by more regressions, because the animacy of the subject does not match the verb’s semantic entailment like agentivity. The argument of UEs is typically an agent, and agents are prototypically animate entities (Bock and Warren, 1985; Dowty, 1991). Core UEs are assumed to be more agentive than peripheral ones. Therefore, the preference of animate subjects is modulated in a gradient fashion, with core UEs with non-animate subjects signaled more

- 163 -

regressions to previous regions compared to non-core ones.

Returning to the data of the present study, it seems that L2 learners did not tend to passivize core UEs when the animacy of subjects in the target sentence inherently fits the verb’s thematic entailment. That is, when the animate subject of the target sentence matches the high agentivity entailment of core UEs, L2 learners know the grammatical form of core UEs. However, when the subject in the target sentence appears to be non-animate, L2 learners would rapidly assign the thematic role of a patient or theme to the argument in order to fit the semantic entailment of the verb (Trueswell et al., 1994). In other words, L2 learners would treat the subject argument of core UEs as the promoted subject of passives, which bears the thematic role of patient. Thus, they tend to passivize core UEs more frequently than non-core ones when the subject argument is inanimate.

What is interesting in the present study is that causation types do not play a significant role in the tendency to over-passivize UEs. In other words, there is no significant difference in the correct response rates between UEs in externally caused events and UEs in internally caused events. UEs in externally caused events are not more likely to be passivized than UEs in internally caused events. UEs, unlike UAs denoting events that typically involve a change of state, usually denote controlled or uncontrolled processes. Core UEs are agentive and volitional processes, whereas peripheral/non-core UEs are less agentive and volitional (Dowty, 1991; Van Valin, 1990). Levin and Rappaport Hovav (1995) assume that agentive verbs are always internally caused. Therefore, for core UEs, there is no necessity to conceptualize an agent for the event, for the verb itself denotes high agentivity. To sum up, the most important factor that could influence L2 learners’ tendency to passivize UEs is animacy types and the pattern of

- 164 -

influence varies depending on the position of the verb along the hierarchy (the core-peripheral distinction).