题元层级之实质与描写

The Nature and Construction of Thematic Hierarchy

题元层级被认为是一种有吸引力的理论架构,但有关题元层级的构成及其角色的排序等问题还存在许多不同的观点,这些争议主要源自对题元层级实质的忽视和描写方法的不当。从实质上来说,题元层级是对某特定事件结构所反映的局部性经验事实的概括陈述,故将属于不同事件类型的题元角色组合进一个层级进行排序是缺乏理据的。描写题元层级的有效方式是采用基于事件参与者的相对突显关系表征方法,如Dowty的原型角色理论。要正确理解和应用题元层级必须弄清以下两个问题:第一,题元层级背后所揭示的实质是什么,表达的概括是什么?第二,题元层级对语义角色进行排序的理据是否充分?只有弄清楚了这两个问题,才能在用于解释不同现象、基于不同理据而提出的层级之间进行有意义的选择和比较,否则,只会造成对题元层级理解的混乱和误用。

Thematic hierarchies （THs） have been widely used to account for regularities in the linking of thematic and grammatical roles and adopted by proponents of a range of theoretical frameworks. But over them there have been controversies, most of which derive from the failure in understanding the nature a TH reflects or in ranking thematic roles properly. Essentially, a TH is the statement of a local empirical generalization （Levin ＆ Rappaport Hovav, 2005）, so different THs reflect different generalizations and have different motivations for ranking thematic roles. To explore the nature of a TH, Fillmore＇s （1968） Subject Hierarchy ＂agent 〉instrument 〉objective （or patient in Levin Rappaport Hovav, 2005）＂ is taken as an example. In fact,the generalization it captures, i.e. ＂the structural prominence of causes over patients （Levin ＆ Rappaport Hovav, 2005） ,＂ is valid only for causatives, in which a causal chain must be formed among arguments. In causatives, theremust be an entity undergoing the conversion from the entity acted on （Ground） into the entity that acts （Figure） （Frawley, 1992）, such as the INSTRUMENT shuibeng （water pump） in sentence ＂Tamen yong shuibeng chou wushui （They are pumping the sewage）. ＂ The conversion process involves the transmission of force and therefore endues shuibeng with the semantic feature of [＋force], which makes shuibeng possess some animate feature. A force-dynamic structure is thus formed in the event, tamen （they） →shuibeng （water pump） →wushui （sewage）. Only under this condition, can the INSTRUMENT shuibeng become the subject if there is no AGENT, as in the sentence ＂Shuibeng zheng-chou-zhe wushui （The sewage pump is working now）. ＂ On the other hand, the INSTRUMENT danjia （stretcher） in the sentence ＂Tamen yong danjia tai shangyuan （They are carrying the injured on a stretcher）＂ doesn＇t undergo the conversion from Ground to Figure, so no causal chain is formed in the second event： tamen （they）→ danjia （stretcher） shangyuan （the injured）, and consequently tai （carry） can not constitute a causative event. Hence the sentence ＂ ＊ Danjia zheng-tai-zhe shangyuan （ ＊ The stretcher is carrying the injured）＂ is unacceptable. If a TH is simply the statement of a local empirical generalization, there is no underlying motivation for ranking the thematic roles of different event types in a hierarchy, like Chen＇s （1994） Subject Hierarchy： agent 〉 experiencer 〉 instrument 〉possessor 〉 location 〉 theme 〉 patient. Otherwise, a false prediction about the linking of thematic and grammatical roles might arise. For example, according to Chen＇s Hierarchy, the sentence ＂ ＊Danjia zheng-tai zhe shangyuan （＊ The stretcher is carrying the injured）＂ is acceptable but in fact it is not. The contradiction results from the fact that Chen＇s Hierarchy mixes types of event and does not respect the fact that ＂agent 〉 instrument 〉objective＂ is only the local empirical generalization about a canonical causative. With the locality nature of THs, the framework for a particular TH must be local, too. Event structures are adequate frameworks for this purpose, but they should he defined relative to a verb because if the ranking of thematic roles is independent of the event structure defined by the verb, the failure of the TH to predict the result of argument selection might he unavoidable. Of the event-based construction of THs, Dowty＇s （1991） proto-role theory, ＂a reflection of the salience of event participants （Levin ＆ Rappaport Hovav, 2005）,＂ is the right approach to the representation of a TH. The reasons are as follows. （a） it describes the essentially relational nature of thematic roles because what semantic roles define is a type of relations rather than categories, and relations can only be understood in an event~ （b） it can more explicitly reflect shared nature or subtle differences between thematic roles, for example, the shared nature of [＋change of state] that both EXPERIENCER and PATIENT have; （c） it can avoid the difficulty in identifying and defining discrete roles or the problems caused by granularity of thematic roles. For example, there is no need to identify a new role ＂FORCE＂ to describe argument Yunshi （a meteorite） in the sentence ＂Yunshi ba wuding za-le yigedong （A meteorite smashed a hole in the roof）＂ as Yuan （2008） suggests, if Yunshi （a meteorite） is defined in terms of a set of entailments of AGENT and INSTRUMENT. { [＋ volition], [- ＋ cause], [＋ movement] } ; （d） it has been empirically proved to have psychological validity. Requirements for a proper understanding and application of THs should thus be to determine the nature a TH reflects, the generalization it captures, and whether there is any underlyingmotivation for the ranking of thematic roles.