ABSTRACT
The Survey of English Usage 1 has never been solely a corpus-based project. From its inception in 1959, the double orientation was made clear (Quirk 1968, Ch. 7). A corpus representative of the repertoire of the mature, educated native speaker would be subjected to analysis in terms of total accountability. And there would be ‘techniques for eliciting the required features’ to establish the internalised rules for unused sentence. A schema representing the two facets of the Survey's work appears as Figure 11.1. But just as the two branches of elicitation work, ‘performance’ and ‘judgment’, are in a deliberate reciprocal relation (see Greenbaum and Quirk 1970, pp. 3ff.), so there is a purposive relation between corpus and elicitation. Not only do we turn to elicitation for information on features that chance to be of fairly rare occurrence: we do so when corpus instances reveal a variation that needs to be explained.
