Harms tangent screen, a subjective measurement method of three-dimensional binocular alignment, was compared with search coil recording. Twenty-three patients with unilateral trochlear nerve palsy were measured in nine gaze positions. The two methods correlated best for the horizontal gaze deviation, the vertical gaze deviation, and the vertical incomitance, but there was no correlation for the results of torsional incomitance. Using Harms tangent screen, torsional deviation underestimated the torsional incomitance measured by the search coils. Therefore, central torsional fusional mechanisms or alignment error in the Harms tangent screen are assumed.