Based on human motor cortex, the effective spatial resolution of transcranial magnetic stimulation (TMS) is often described as 5-20mm, because small changes in TMS coil position can have large effects on motor evoked potentials (MEPs). MEPs are often studied at rest, with muscles relaxed. During muscle contraction and movement, corticospinal excitability is higher, thresholds for effective stimulation lower, and MEPs can be evoked from larger regions of scalp, so the effective spatial resolution of TMS is larger. We found that TMS over the supramarginal gyrus (SMG) impaired manual dexterity in the grooved pegboard task. It also resulted in short latency MEPs in hand muscles, despite the coil being 55mm away from the motor cortex hand area (M1). MEPs might be evoked either by a specific corticospinal connection from SMG, or a remote but direct electromagnetic stimulation of M1. To distinguish these alternatives, we mapped MEPs across the scalp during rest, isotonic contraction and manual dexterity tasks, and ran electric field simulations to model the expected M1 activation from 27 scalp locations and 4 coil orientations. We also systematically reviewed studies using TMS during movement. Across five experiments, TMS over SMG reliably evoked MEPs during hand movement. These MEPs were consistent with direct M1 stimulation and substantially decreased corticospinal thresholds during natural movement. Systematic review suggests that 54 published experiments may have suffered from similar motor activation confounds. Our results have implications for the assumed spatial resolution of TMS, and especially when TMS is presented within 55mm of the motor cortex.