This study is based on the analysis of transcripts of 137 audio-recorded tests using a Conversation Analysis (CA) methodology. The institutional aim of standardisation in relation to assessment is shown to be the key principle underlying the organisation of the interaction. Overall, the vast majority of examiners conform to the instructions; in cases where they do not do so, they often give an advantage to some candidates. The overall organisation of the interaction is highly constrained, although there are some differences in the different parts of the test. The organisation of repair has a number of distinctive characteristics in that it is conducted according to strictly specified rules, in which the examiners have been briefed and trained.
Speaking test interaction is an institutional variety of interaction with three sub-varieties. It is very different to ordinary conversation, has some similarities with some sub-varieties of L2 classroom interaction and some similarities with interaction in universities.
A number of recommendations are made in relation to examiner training, instructions and test design.