1. A highly selective medium forH. pertussishas been made by incorporating 4:4'-diamidinodiphenylamine dihydrochloride (M & B 938), penicillin and sodium fluoride in a partially defined base.2. M & B 938 has an interesting selectivity which is much influenced by its environment, especially by the ionic composition of the medium. No pattern has emerged from a study of the effects of eighty salts and many other substances on its activity; but the gain in selectivity forH. pertussisproduced by cysteine and magnesium malonate has been usefully exploited.3. In a strict comparison with Bordet-Gengou the new medium yielded 28% more positive cultures from per-nasal swabs and 75% more from post-nasal swabs. With the new medium the incidence of serious rival growth from any part of the nasopharynx has been about 4% and, in contrast with Bordet-Gengou, postnasal swabs have yielded as many positive cultures as per-nasal swabs.4. The new medium is simple to make, stable, reliable, economical of material and time and suitable for routine use with per-nasal, post-nasal or supralaryngeal swabs, sputum or vomit.5. The new medium growsH. parapertussispoorly; but a highly selective medium forH. parapertussisis easily made by adding streptomycin and penicillin to Bordet-Gengou.6. Whatever the medium, the nature and number of swabs are important factors affecting the isolation ofH. pertussis:with the new medium positive diagnoses from per- and post-nasal swabs taken at the same time were about 40% more than from either swab alone.7. A statistical analysis strongly supports the opinion of Bradford & Slavin (1940) that the relatively poor results with post-nasal swabs on Bordet-Gengou medium are due to the overgrowth ofH. pertussisby rivals rather than to the absence ofH. pertussisfrom the pharynx. It also indicates that the expectation of isolatingH. pertussison Bordet–Gengou varies directly with the numbers present and inversely with the numbers of rivals, especially of diphtheroids and streptococci.8. Addition of 32μg. stilbamidine di-isethionate and 0·25 unit penicillin/ml. to Bordet–Gengou makes an effective selective medium for oral streptococci related to, or of, groups K and L, but not forH. pertussis.I wish to thank my clinical colleagues, especially Dr N. O. Richards and Dr A. Bogdan, for kindly sending swabs from their patients, Professor R. J. V. Pulvertaft for his continuous support, Dr J. A. Fraser Roberts for advice and help with the statistics, Dr A. J. Ewins and Dr H. J. Barber, May and Baker Ltd., for gifts of aromatic diamidines, Dr H. King for gifts of alkyl diamidines and Mr M. Breach, F.I.M.L.T., and Mr A. Drake, F.I.M.L.T., for their help with the preparation of sera, media and routine cultures. I am indebted also to the Central Research Fund of London University and to the Governors of Westminster Hospital for grants for the purchase of materials.