In order to determine the rate and degree of redistribution of morphine within the cerebrospinal fluid (CSF), and whether it was affected by the site of and volume of the injection, morphine was given to 23 elderly patients undergoing thoracotomy - in 10 ml saline in the lumbar epidural interspace (n = 5), in 10 ml saline in the thoracic epidural interspace (n = 5), in 2 ml saline in the thoracic interspace (n = 8) and in 10 ml saline in the lumbar epidural interspace (n = 5). The plasma concentration of morphine in all patients was comparable and was much lower than in the CSF. The CSF morphine concentration, measured as the area under the CSF concentration curve (AUC), the maximal CSF concentration (Cmax) and the time to reach maximal CSF concentration (tmax), varied between the four groups. The variation was related to the site of the injection; the AUC and Cmax were lower and tmax appeared later after thoracic than lumbar injection. Lumbar CSF morphine concentrations were further reduced by thoracic epidural injection of morphine in a small as compared to a large volume. The permeability of the dura to morphine was not influenced by the volume used. The results show that morphine is not homogeneously distributed within the CSF. The availability of morphine to CSF from the epidural space is not altered by the injection volume, but the drug remains more localized in CSF after epidural injection of morphine in a small volume.(ABSTRACT TRUNCATED AT 250 WORDS)