Children with leukemia receive CNS therapy to improve long-term survival. Neurotoxic effects, such as cognitive impairment, have been associated with this therapy. A rat model was developed to determine which agent, or combination of agents, in CNS therapy causes neurotoxicity. The agents examined were cranial irradiation (1000 cGy), methotrexate (2 or 4 mg/kg, intraperitoneally), and prednisolone (18 or 36 mg/kg, intraperitoneally). Young Sprague-Dawley rats were exposed to each agent alone or to two- or three-agent combinations. Each therapy had matched controls that received sham radiation and/or intraperitoneal saline. Subsequent to exposure, spontaneous behavior was tested using a computer pattern recognition system, which recorded and classified behavior in a novel environment. Behavioral initiations, total times, and time structures were compared in therapy and control groups. Combined rather than single-agent therapies had more behavioral effects, and these were dose- and sex-dependent. Synergistic interactions between agents caused behavioral deficits, and components of the combination determined the abnormality. Some combinations interacted antagonistically, and thus mitigated behavioral deficits. Prednisolone was clearly pivotal to behavioral outcome. A low prednisolone dose antagonized methotrexate preventing deficits, whereas a higher prednisolone dose altered behavior by enhancing effects of methotrexate and radiation. These findings emphasize that steroids are important in agent interactions. Their role in morbidity associated with leukemia treatment protocols may be equally important as that of methotrexate and cranial irradiation.