SUMMARY The findings in the preceding paper on pages 773-778 are reviewed and the possible causes of error revealed by them are analysed.The disc test has been studied with a view to defining suitable culture media, and disc contents such that six can usually be placed on a 9 cm plate. Emphasis is placed on controlling the weight of the inoculum and on ensuring its uniform distribution. It is suggested that such cultures should be interpreted by comparison of inhibition zone sizes with those of a suitable control organism on the same medium.Indications and methods for tests in primary culture are discussed. Proposals are made for eliminating superfluous or inappropriate tests, thus limiting the number of antibiotics or other drugs with which tests need initially be done.There are now on record at least four studies in which the results obtained by numerous different laboratories in testing the antibiotic sensitivity of given cultures have been compared. Either pure cultures (Journal of Medical Laboratory Technology, 1960; College of Pathologists of Australia, 1968) or clinical specimens (Association of Clinical Pathologists, 1965) have been sent for testing by whatever method each laboratory favoured; in another study (Beaney, Goodwin, Jones, Winter, and Sippe, 1970) cultures were circulated for testing by a prescribed method. Comparison ofthefindings showed numerous discrepancies. When, for instance, 34 laboratories report an organism to be sensitive to a certain antibiotic and 26 find it resistant, one of these sets of reports must be wrong.The reason for an erroneous finding can only be either in the method used, in its execution, or in its interpretation. An inquiry into methods should therefore be of interest, and the most extensive yet in this country-and probably indeed anywhere-is reported in the preceding paper (Castle and Elstub, 1971 The second method, based largely on that originally described by Ericsson (1960), is proposed by a working group of WHO which has been engaged in an international collaborative study for nearly 10 years (Ericsson and Sherris, 1971). This report defines two dilution methods as well as a disc method. The latter is proposed as either a routine method or a reference method whereby the validity of an existing routine method can be tested. It uses large plates of Mueller-Hinton medium with 5 % sheep's blood, an inoculum yielding 'dense but not completely confluent growth' and high content discs (only those for five groups of antibiotics and for chloramphenicol are actually specified). Zones are measured, and can be translated into the minimum inhibitory concentration (MIC) by reference to regression lines (plots of zone diameters against MIC for organisms of differing degrees of sensitivity). This enables the organism to be placed in a category of sensitivity by reference to tables of 'break-points' (in terms of MIC) separating these categories. This process is simplified in instructions for this test issuedin Sweden kindly supplied to us by Dr Hans Ericsson. Scales are provi...