New requirements by several regulatory agencies for testing the psychotoxic potential of new drugs, chemicals, and environmental contaminants raise unique problems. In order to assess intra- and interlaboratory reliability of behavioral tests a model animal maze learning procedure was designed and run in 3 cooperating laboratories. Uniform procedures were written and identical mazes were constructed. Normal control animals of identical age and sex, but of different strains, were used by the participants. A positive control group of neurologically impaired rats was run by one laboratory. Significant differences in test results among the laboratories were found. Data obtained from the positive control animals (mean errors=28.3) indicated a learning impairment statistically significant compared to the negative control data (mean errors=12.7) from any of the participating laboratories. Based on the results of this study, a reasonable standard of interlaboratory reliability in behavioral testing appears an attainable goal.