Garner, J. P. 2005. Stereotypies and other abnormal repetitive behaviors: Potential impact on validity, reliability, and replicability of scientific outcomes. ILAR Journal 46(2), 106-117.
Normal behavior plays a key role in facilitating homeostasis, especially by allowing the animal to control and modify its environment. Captive environments may interfere with these behavioral responses, and the resulting stress may alter many physiological parameters. Abnormal behaviors indicate that an animal is unable to adjust behaviorally to the captive environment and, hence, may be expressing abnormal physiology. Therefore, captive environments may affect the following aspects of an experiment: validity, by introducing abnormal animals into experiments; reliability, by increasing interindividual variation through the introduction of such individuals; and replicability, by altering the number and type of such individuals between laboratories. Thus, far from increasing variability, enrichment may actually improve validity, reliability, and replicability by reducing the number of abnormal animals introduced into experiments. In this article, the specific example of abnormal repetitive behaviors (ARBs) is explored. ARBs in captive animals appear to involve the same mechanisms as ARBs in human psychiatry, which reflect underlying abnormalities of brain function. ARBs are also correlated with a wide range of behavioral changes that affect experimental outcomes. Thus, ARBs in laboratory animals may compromise validity, reliability, and replicability, especially in behavioral experiments; and enrichments that prevent ARB may enhance validity, reliability, and replicability. Although many links in this argument have been tested experimentally, key issues still remain in the interpretation of these data. In particular, it is currently unclear (1) whether or not the differences in brain function seen in animals performing ARB are abnormal, (2) which common behavioral paradigms are affected by ARB, and (3) whether enrichment does indeed improve the quality of behavioral data. Ongoing and future work addressing these issues is outlined.