The method commonly used for catheterization of the lumbar subarachnoid space in the rat implies inserting the catheter through the atlanto-occipital (A-O) membrane and moving the catheter caudally along the spinal cord. The method is associated with a considerable morbidity. A method for direct catheterization of the lumbar subarachnoid space was therefore developed. Major surgery was avoided by using a catheter-through-needle technique. Of 32 rats, none died. There were no signs of neurological disturbances, and all animals gained weight as normal the first week after implantation. Data from rats catheterized by the A-O method were used for comparison. Of 40 animals, 2 died, 11 showed signs of neurological disturbances, and the mean weight was reduced during the first week after catheterization. The two groups of animals showed different behavioural responses to intrathecal injection of N-methyl-D-aspartate (NMDA, 0.1-1.6 mM, 15 microliters) which is thought to stimulate afferent pathways mediating nociception. Animals with a lumbar catheter showed licking, biting and scratching behaviour in a dose-related manner for concentrations up to 1.6 mM. The animals with A-O catheters showed a maximum level of this behaviour already at 0.4 mM, while 0.5 mM induced convulsions. A possible explanation of this difference in response to NMDA could be a long-lasting pain state in the A-O group, caused by catheter-induced changes in the spinal cord and by the extensive surgery. It is concluded that the direct lumbar catheterization has several advantages compared to the A-O method, decreasing the suffering of the animals, the neurological disturbances and the interference with nociceptive functions of the spinal cord.