Fishes commonly use their lateral line system to detect moving bodies such as prey and predators. A remarkable case is the Mexican blind cavefish Astyanax fasciatus who evolved the ability to detect non-moving obstacles. The swimming body of A. fasciatus generates fluid disturbances, whose alteration by an obstacle can be sensed by the fish's lateral line system. It is generally accepted that these alterations can provide information on the distance to the obstacle. We observed that A. fasciatus swimming in an unfamiliar environment open and close their mouths at high frequency (0.7-4.5 Hz), in order to generate suction flows. We hypothesized that repeated mouth suction generate a hydrodynamic velocity field, whose alterations by an obstacle induce pressure gradients in the neuromasts of the lateral line, and corresponding strong lateral line stimuli. We observed that the frequency and rates of mouth opening events varied with the fish's distance to obstacles, a hallmark of pulse-based navigation mechanisms such as echolocation. We formulated a mathematical model of this hitherto unrecognized mechanism of obstacle detection and parameterized it experimentally. This model suggests that suction flows induce lateral line stimuli that are weakly dependent on the fish's speed, and may be an order of magnitude stronger than the correspondent stimuli induced by the fish's gliding body. We illustrate that A. fasciatus can navigate non-visually using a combination of two deeply ancestral and highly conserved mechanisms of ray-finned fishes: the mechanism of sensing water motion by the lateral line system and the mechanism of generating water motion by mouth suction.