The use of animal models in certain types of psychobiological studies (for instance, the relationship between anxiety and depression) requires that the behavior measured is stable over time.The test-retest reliability of the elevated plus-maze indexes of anxiety and the immobility time in the behavioral despair were evaluated.The behavior of two groups of drug naive mice was measured on two occasions on the same test, 1 week apart, on the elevated plus-maze or on the behavioral despair and then the intraclass correlation coefficient and kappa were calculated.These behaviors showed a very low intraclass correlation coefficient (0.02 - 0.05) and low kappa (- 0.08 - 0.21) in the test-retest design, which suggest a poor reliability of these measures.These results suggest that the behavioral parameters of the elevated plus-maze and the behavioral despair are not stable and therefore they are possibly more related to state than trait characteristics. Therefore they appear to be not appropriate to evaluate trait characteristics which are supposed to be stable over time without treatment.