Our goal is to describe a specific case of a general process gaining traction amongst biologists: testing biological hypotheses with biomimetic structures that operate in bioinspired robots. As an example, we present MARMT (mobile autonomous robot for mechanical testing), a surface-swimmer that undulates a submerged biomimetic tail to power cruising and accelerations. Our goal was to test the hypothesis that stiffness of the body controls swimming behavior and that both stiffness and behavior can be altered by changes in the morphology of the vertebral column. To test this hypothesis, we built biomimetic vertebral columns (BVC) outfitted with variable numbers of rigid ring centra; as the number of centra increased the axial length of the intervertebral joints decreased. Each kind of BVC was tested in dynamic bending to measure the structure's apparent stiffness as the storage and loss moduli. In addition, each kind of BVC was used as the axial skeleton in a tail that propelled MARMT. We varied MARMT's tail-beat frequency, lateral amplitude of the tail, and swimming behavior. MARMT's locomotor performance was measured using an on-board accelerometer and external video. As the number of vertebrae in the BVC of fixed length increased, so, too, did the BVC's storage modulus, the BVC's loss modulus, MARMT's mean speed during cruising, and MARMT's peak acceleration during a startle response. These results support the hypothesis that stiffness of the body controls swimming behavior and that both stiffness and behavior can be altered by changes in the morphology of the vertebral column.