In this article we discuss two divergent accounts of non-human animals as analog models of human biomedical phenomena. Using a classical account of analogical reasoning, toxicologists and teratologists claim that if the model and subject modeled are substantially similar, then test results in non-human animals are likely applicable to humans . However, the same toxicologists report that different species often react very differently to the same chemical stimuli. The best way to understand their findings is to abandon the classical view of analogical - i.e., linear - reasoning, and replace it with a version informed by chaos theory. We briefly outline the current understanding of chaos, and trace its implications for toxicology and teratology.