Twenty-nine patients with idiopathic spasmodic torticollis (ST) and matched normal control subjects were asked to align a target line to perceived earth vertical [visual vertical (VV)]. Settings were made against a whole-field random-dot background that was either stationary or rotating around the line of sight, and subjects performed the task upright and lying horizontally on their sides. Normal subjects were tested both head upright and after assuming a voluntary head tilt. Patients with ST set the VV close to true upright with a minimal deviation toward tilt of the head in contrast to normal subjects assuming a head tilt who set the VV in the opposite direction to the head tilt (Müller "E" effect). Settings against the spinning disk were biased in the direction of rotation similarly for both subject groups. Settings made against static or spinning disk performed when subjects lay horizontally were tilted in the direction of recumbence (Aubert "A" effect) similarly for both subject groups. When attempting to set the target line parallel to the long axis of the face, patients with head tilt set the line to upright, whereas normal subjects correctly estimated their tilts. One hypothesis offered to explain these results is that the patients referenced only their upright trunk for vertical and did not make use of neck proprioceptive or vestibular signals of head tilt so that all settings were made as if the trunk and head were upright. Alternatively, patients may have used only otolith signals as the reference for upright, and these are recalibrated in ST patients with head tilt. The pathological deviation becomes accepted as upright posture, and VVs and facial orientation are estimated as if the head were upright. In either case the findings indicate abnormal processing of the perception of visual verticality in ST patients.