Humans and other animals can measure distances nonvisually by legged locomotion. Experiments typically employ an outbound measure (M) and an inbound report (R) phase. Previous research has found distance reproduction to be maximally accurate, when gait symmetry and speed of M and R are of like kind: Successful human odometry manifests at the level of the M-R system. In the present work, M was an experimenter-set distance produced by a blindfolded participant using a primary gait (walk, run). R was always by walk. Fast and slow versions of walk and run were adopted by participants, such that when M was fast R was slow, and vice versa. Distance was underestimated when M was slower than R and overestimated when M was faster than R. However, the pattern of participant-adopted velocities indicated that it was the instructions, not the speed as such, that yielded the pattern of results. The results are interpretable through a dynamical perspective and indicate speed is an imperfection parameter acting on the attractors of the M-R system.