Experiments studying non-resonant error fields have been conducted in the National Spherical Torus Experiment (NSTX) using a set of six midplane error field correction (EFC) coils. When scanning the amplitude and phase of an applied n = 3 field, an asymmetric response in the pulse length and plasma rotation has been observed; this indicates that there is an intrinsic n = 3 error field. By studying this asymmetry in plasmas with varying levels of plasma current, toroidal field and elongation, it has been concluded that the main vertical field coil is the source of the error field. Measurements of the coil shape indicate that the coil has a significant n = 3 distortion. The amplitude and phase of the applied n = 3 field, which is calculated to cancel this intrinsic error field in vacuum, are close to the experimentally derived optimal correction. Modeling of the neoclassical toroidal viscosity (NTV) also shows that the total NTV torque is minimized for the EFC coils' current and phase determined to be optimal in the experiment. Experiments have also determined that n = 2 error fields are small, consistent with the calculated error field from the distorted vertical field coils.