Translations as evidence for semantics: An illustration

Abstract

Though a corpus-based linguistics is potentially more empirical than the intuition-based kind, the mere use of a corpus does not guarantee empiricalness. A monolingual corpus is good for counting forms but it does not readily reveal their meanings. Since meanings are not directly observable, the monolingual corpus can only provide circumstantial evidence for them. Their (psychological) reality, however, can only be established through informant testing. But then, few linguists have the facilities to do more than pay lip-service to the use of informants. The translation corpus could provide a way out of the crisis, since translators, through the linguistic choices they make, inadvertently supply evidence of the meanings of the forms they are receiving and producing. The paper illustrates this by showing how data from a translation corpus can complement evidence from a monolingual corpus to prove that English forms like BE said to, BE considered to, BE reported to, etc., are turning into evidential auxiliaries.