Limb, not touch location, is coded in 3D space
Reduced tactile-spatial accuracy with crossed versus uncrossed limbs is considered key evidence for automatic remapping of touch location from skin-based to 3D-spatial codes. Yet, typical tasks use only few stimuli, potentially confounding categorical choice of a touched limb with precise stimulus location. Over five experiments, participants judged whether two touches occurred on the same limb (anatomical) or on the same side of space (external-spatial). Crossing consistently impaired performance, but more under external-spatial than anatomical instructions - contrary to what automatic remapping predicts. Limb assignment errors occurred equally frequent for arm-leg as arm-arm pairs, implying confusion about limb identity, not spatial position. With mouse-based, precise localization, crossing again impaired limb assignment, mainly for stimuli distal to the crossing point, whereas precision was unaffected: localization scatter did not increase, and showed no bias toward the limb's usual side of space. Crossing effects therefore emerge from choosing a limb, not from localizing touch. We propose that they reflect limb assignment, with touch cueing the region of space it is habitually associated with, followed by selecting a limb whose habitual or current posture occupies that region, constrained by which limbs are task-relevant; precise localization is performed afterwards on the selected limb.