Phenobarbitone levels in the blood do not always provide a reliable guide in the management of the poisoned patient. They may rise in response to active treatment though the patient may be improving clinically. Some reasons for this are considered, with illustrations from three cases. It is concluded that the poor correlation of phenobarbitone levels with a patient's condition is due not to inadequate analytical techniques but to the poor representation of the "whole body" situation by a single blood specimen. The usefulness of analysing such specimens is questioned.