One hundred and one fiberoptic pressure transducers (59 subdural and 42 ventricular) were studied in 86 patients (some in whom more than one device had been inserted). Only four complications occurred: two transient cerebrospinal fluid leaks after removal and two clinically insignificant intracerebral hematomas. No intracranial infections could be attributed to the devices. Technical problems occurred 23 times, with 11 devices ceasing to function before removal, seven becoming displaced, and five microventricular catheters failing to enter the ventricles. Zero-drift readings were obtained for 83 devices at the time of removal (median 66 hours after insertion, range 2 hours-13 days). There was a clear negative bias in the readings (median -3), with a wide range of values (-12 to +14 mm Hg; interquartile range -6 to -1) that was apparent even in the first 3 days of use. There was no important relationship between zero drift and any recorded variable. It is concluded that zero drift of fiberoptic pressure transducers is a significant problem and that undue reliance should not be placed on intracranial pressure readings from these devices in isolation from other clinical and radiological information.