Prototype: corpus vs. elicitation

Geeraerts (1988) mentions two types of evidence for prototypicality, viz. usage and introspection. While the former makes it possible to determine what is ‘quantitatively more prominent’ (p. 215), the latter presumably gives access to the more salient kinds of usage. In his analysis of two Dutch verbs, 'vernielen' and 'vernietigen', Geeraerts comes to the conclusion that the two approaches give very similar results. The same sort of analysis will be applied to English periphrastic causative constructions. The aim is to find out whether usage and introspection necessarily point to the same direction in terms of prototypicality. The causative constructions investigated are those with 'cause', 'get', 'have' and 'make' followed by a non-finite complement (e.g. 'The explosion caused the temperature to rise' or 'Don’t make me laugh'). The usage approach is based on a subcorpus from the British National Corpus (5 million words of speech, 5 million words of writing), from which all causative constructions were extracted and analysed according to a number of semantic and syntactic parameters, such as animacy of the causer/causee, distinction between the causer and the causee, volitionality and transitivity of the effect, etc. For the introspective approach, elicitation tests were carried out among some 40 native speakers of English, who were asked to fill in causative constructions (causer, causee or effect) and make up one sentence for each causative verb. The resulting data were analysed according to the same parameters as the corpus data. The quantitatively more prominent constructions in the corpus are compared with the more salient constructions as they emerge from the introspective data. They are also compared to the definition of prototypical causative constructions given in works such as Lakoff (1987) or Kemmer & Verhagen (1994). It will be shown that the overlap between these different “prototypical” constructions is only partial. Thus, while inanimate causers with 'make' are slightly more frequent than animate ones in the corpus data, the elicitation tests show a reverse tendency, with a majority of animate causers, in accordance with the definition of prototypical causation given in e.g. Lakoff (1987: 54-55). Conversely, both usage and introspective evidence agree on the predominantly inanimate nature of the causee with the verb 'cause', which seems to contradict Kemmer & Verhagen’s claim that prototypical causees should be animate (1994: 129). The differences and similarities between the corpus and the elicitation data will be discussed, as well as the implications this has for the notion of prototypicality and the nature of the four verbs investigated.