SUMMARY Eight methods for the assay of chloramphenicol in clinical samples were compared with our own modification of a plate diffusion technique using Sarcina lutea and yeast extract agar. Six of the eight methods were less sensitive than originally reported, and five of them were considered unsuitable for use in clinical microbiology practice. The remaining three methods together with the S. lutea/yeast extract modification were used to assay chloramphenicol in 20 samples of serum. Twenty samples of cerebrospinal fluid were also assayed by the S. lutea/yeast extract method. Our results indicate that only the Bacillus subtilis (sensitivity 6-0 mg/l) and the S. lutea (sensitivity 2 5 mg/I) diffusion methods are suitable for use with clinical samples in routine practice. The problems of chloramphenicol toxicity, appropriate dosage regimens, and the need for assay of the drugs are considered.The liability of chloramphenicol to provoke toxic effects with fatal outcome 2 limits its usefulness, and it is generally reserved for the treatment of some cases of typhoid fever or other serious Gram-negative infections,3 finding application in paediatric, neurological, and neurosurgical practice.The toxic side effects are enumerated by Wade and Reynolds.4 While irreversible and sometimes fatal aplastic anaemia (incidence 1: 20 000-1:80 000) appears to be idiosyncratic and is, perhaps, genetically determined, depression of the bone marrow and the gray syndrome in premature and other newborn infants are dose related.5 6 Reversible depression of the marrow occurs usually when plasma chloramphenicol concentrations reach 25-35 mg/I, and the gray syndrome has been reported most frequently in infants given more than 50 mg/kg body weight daily. Reduced doses are advised for those with impaired renal or liver function, but there is little uniformity of view on appropriate dosage regimens for premature and other neonates. Recommendations taking account of prematurity have been made7-9 but it is unlikely that dosage schedules that are both therapeutic and relatively non-toxic can be determined unless plasma and tissue concentrations of the drug can be measured reliably and accurately for each individual.