連動結構：比較用的概念和跨語言通則

The Serial Verb Construction: Comparative Concept and Cross-linguistic Generalizations

1970年代以來，在非洲、大洋洲的語言以及世界上許多其他語言中已有關連動結構的探討。以往研究中的主要問題是：對連動結構的界定不清晰且過於寬鬆。因此，有一些語言學家對於找到統一的連動結構的跨語言概念感到絕望。Foley (2010:107)曾寫到：「連動結構真的有放諸四海皆準的定義嗎？很可能沒有」以往關於連動結構研究的這些問題實質上都是源於混淆了比較用的概念(Haspelmath 2010)和自然類。連動結構常被看作自然類（普遍範疇），以致其他語言中具有不同性質的現象也被誤認為連動結構。這不可避免地導致對此概念的理解過於廣泛而不準確。並且，此類「典型結構」無法證偽。本研究基於比較語言學和理論語言學的文獻，而非系統性的語言樣本。本研究提出連動結構的狹義定義，並且概括出十項關於連動結構的通則。

Since the 1970s, serial verb constructions (SVCs) have been discussed widely in African, Oceanic and many other languages throughout the world. This article gives an overview of the most important generalizations about SVCs that have been proposed and that do seem to hold if a sufficiently restrictive definition of the concept is adopted. The main problem with the earlier comparative literature is that the notion of an SVC has not been delimited clearly, and/or has been formulated in much too wide terms. As a result, some linguists have despaired of finding a coherent cross-linguistic concept of SVC. For example, one scholar asked ‘Are there any universal defining properties of serial verb constructions? Probably not.... These problems can be seen as a result of the confusion between comparative concepts and natural kinds: Serial verb constructions have (most often implicitly) been regarded as natural kinds (universal categories), so that phenomena in additional languages were regarded as SVCs even when they had somewhat different properties. This procedure inevitably leads to a fuzzy and very broad understanding of the concept, with a prototype (or ‘canonical’) structure that does not allow falsifiable claims. Here I propose a narrow definition of SVC and formulate 10 universals that are apparently true of all serial verb constructions in this narrow sense. The claim that these are universally true of (narrowly defined) serial verb constructions is based on a thorough reading of the comparative and theoretical literature, not on a systematic sample of language-- the latter would not have been practical, because SVCs are rarely described in sufficient detail in descriptive grammars. No attempt is made at explaining these generalizations in the present article, but I claim that we finally have a good idea of what it is that needs to be explained in a general way.