Since the 1870s, the primary motor cortex (M1) has been known to have a somatotopic organization, with different regions of cortex participating in control of face, arm, and leg movements. Through the middle of the 20th century, it seemed possible that the principle of somatotopic organization extended to the detailed representation of different body parts within each of the three major representations. The arm region of M1, for example, was thought to contain a well-ordered, point-to-point representation of the movements or muscles of the thumb, index, middle, ring, and little fingers, the wrist, elbow, and shoulder, as conveyed by the iconic homunculus and simiusculus. In the last quarter of the 20th century, however, experimental evidence has accumulated indicating that within-limb somatotopy in M1 is not spatially discrete nor sequentially ordered. Rather, beneath gradual somatotopic gradients of representation, the representations of different smaller body parts or muscles each are distributed widely within the face, arm, or leg representation, such that the representations of any two smaller parts overlap extensively. Appreciation of this underlying organization will be essential to further understanding of the contribution to control of movement made by M1. Because no single experiment disproves a well-ordered within-limb somatotopic organization in M1, here I review the accumulated evidence, using a framework of six major features that constrain the somatotopic organization of M1: convergence of output, divergence of output, horizontal interconnections, distributed activation, effects of lesions, and ability to reorganize. Review of the classic experiments that led to development of the homunculus and simiusculus shows that these data too were consistent with distributed within-limb somatotopy. I conclude with speculations on what the constrained somatotopy of M1 might tell us about its contribution to control of movement.