The Survey of English Usage (see especially Quirk, 1968: 70 ff., 184 ff.) combines a corpus basis with data obtained by elicitation techniques which so far have taken the form of acceptability tests. Most of the latter type of work is done with aural (taped) input, the subjects making their responses in writing. The drawbacks in this process were acknowledged in Quirk & Svartvik (1966), hereafter QS, in connexion with the interpretation of responses to the instruction to make a sentence negative (pp. 95 f.), since a written response such as I haven't a car is ambiguous as between a negative sentence (I /haven't a càr#) and a denial of the corresponding positive sentence (I /hàven't a car#). It was therefore decided ‘to experiment with oral responses … through the use of a language laboratory’ (p. 104), in order both to collect data not available from the written responses and at the same time to preserve maximum similarity with the written-output experiment in having a group of subjects being tested simultaneously.