Determination of a direction of travel is a necessary component of successful navigation, and various species appear to use the geometric shape (global geometric cues) of an environment to determine direction. Yet, debate remains concerning which objective shape parameter is responsible for spatial reorientation via global geometric cues. For example, the principal axis of space, which runs through the centroid and approximate length of the space, and the medial axis of space, a trunk and branch system that fills the shape, have each been suggested as a basis to explain global spatial reorientation. As the principal- and medial-axis accounts appear to have substantial theoretical implications regarding the nature of shape perception, spatial memory, and the underlying psychological representations of space, it appears critical to empirically differentiate between these global geometric accounts. The present experiment explicitly placed predictions from the principal- and medial-axis-based accounts of global spatial reorientation in conflict for theoretical diagnostic purposes. We used a standard reorientation paradigm in which human participants first reoriented in a rectangular environment; subsequent testing in a critical I-shaped enclosure allowed dissociation of the principal- or medial-axis-based accounts. We show that reorientation in the I-shaped enclosure was consistent with the principal-axis account and inconsistent with the medial-axis account. We suggest that the use of the principal axis for spatial reorientation is a relatively simple and efficient way to establish directionality that would be advantageous over a more complex and less efficient medial-axis-based account.