As the targets of scientific study, animals present a host of problems. They are complex, wet, individually variable, difficult to control, messy to take apart, and sometimes impossible to obtain. Models mathematic or robotic provide a way around these problems. Physically embodied robots, in particular, can help biologists model the dynamics of locomotor biomechanics, behavioral interactions, and even organic evolution. We use simple autonomous robots to test hypotheses about living and fossil fish. Specifically, we are interested in how fish control their swimming speed by varying the flexural stiffness of their vertebral column; how fish coordinate group behaviors; and what selection pressures may have been responsible for the evolution of the vertebral columns of the very first fish-like vertebrates, 500 million years ago.