Acquisition of two-way avoidance by mice was slower with a light CS than with a buzzer CS, with punishment of intertrial responses than without punishment, and with a short CS-CS interval than with a long CS-CS interval (30 vs. 60 sec). Light-cued avoidance was little affected by shock level (.35-1.5 mA), whereas mice trained with the buzzer CS learned faster at 1.5 mA. Animals required to move away from light or toward light showed comparable rates of acquisition. Other CS, US, and apparatus variables (directionality of cue, maximal shock duration, and presence vs. absence of a central partition in the shuttlebox) interacted in a complex fashion with those already mentioned. This resulted in widely differing performances in what may superficially appear to be different versions of the same task. The differences in mouse and rat responses to some of the variables can contribute to an understanding of the interactions between organismic and test factors and the relative explanatory value of alternative avoidance models.