Unless examiners are carefully selected, trained, and monitored, examinations may become haphazard. This is perhaps most true of oral or viva voce (“viva”) examinations, which can generate marks unrelated to competence. To help other bodies to short circuit some years of experiment in connection with the oral component of the Royal College of General Practitioners' membership examination (MRCGP), this paper describes the selection, training, guidance, and monitoring arrangements that have been developed.

The oral or viva voce examination (“viva”)—a general non-patient based encounter between a candidate and one or more examiners—has held an important place in medicine for centuries.1 Tradition aside, it is used for its flexibility, its apparent fidelity (much medicine concerns oral encounters over issues of diagnosis and management), and its potential for testing higher order cognitive skills.

Unfortunately oral examinations are prone to many errors.2 These include errors relating to halo effects (a judgment of one attribute influences judgments of others); errors of central tendency (judgments cluster in the middle); so called errors of logic (mistakes); a general tendency towards leniency; and errors of contrast (judgments of a candidate are influenced by impressions of preceding candidates). Oral examinations tend to test at a low taxonomic level—for example, factual knowledge rather than problem solving.3 Scores are related to irrelevant attributes of candidates such as appearance or confidence.4 Agreement between examiners is often poor.4 It is, moreover, difficult to establish in any formal way how valid an oral examination is.5

Largely abandoned in North America, oral examinations are still widely used in undergraduate and postgraduate examinations in the United Kingdom. In 1990, 19 out of 27 medical schools used vivas in their final qualifying examinations, 11 as a major assessment method.6

The membership examination of the Royal College of General Practitioners (MRCGP) …