This research investigates using a manipulator to tactilely explore objects and environments when significant backlash affects its joint’s positions. A typical application is the exploration of rough environments, such as oil wells, where the harsh conditions dictate the use of tactile exploration. These conditions can result in large, unknown, and variable backlash in the manipulator’s transmissions, which strongly affects the measurement precision. Here, a method is developed to simultaneously map the unknown surface and identify the joint backlash. The robot probes the surface and uses its encoder readings to construct a partial map of the environment as a combination of geometric primitives. While the surface is built, the same data are also used to estimate backlash in the joints and to correct the surface measurements for backlash error. The effectiveness of the approach is demonstrated in simulation case studies and laboratory experiments.