Douglas Wahlsten and Emily Marcotte

Department of Biological Sciences and Great Lakes Institute for Environmental Research, University of Windsor, Windsor, ON, Canada

Two labs at different sites rarely use identical apparatus and run a test in the very same way. There are reasons to worry that this lack of test standardization may contribute to failures to replicate genetic experiments on mice across labs. We first address this question by examining results of inbred surveys on behavior done recently in four different labs and comparing modern and classical data sets collected 30 to 50 years apart. For certain kinds of behavioral tests, there is a great similarity of strain data across sites and decades, whereas anxiety-related behavior shows remarkable differences between results in different labs. Thus, for some tests, standardization would probably not make much difference, but for others it would probably enhance replicability. We then analyze published data in 100 reports of mice learning a water escape task and ask whether variations in task parameters in this large sample are related to differences in the rate of learning to locate a submerged platform. Multiple regression analysis of this diverse literature points to several aspects of the task that are clearly important. We believe that standardization of this task would be beneficial, and we offer suggestions about the best choices of parameters.