Reference frames for coding touch location depend on the task.

The position of gaze (eye plus head position) relative to body is known to alter the perceived locations of sensory targets. This effect suggests that perceptual space is at least partially coded in a gaze-centered reference frame. However, the direction of the effects reported has not been consistent. Here, we investigate the cause of a discrepancy between reported directions of shift in tactile localization related to head position. We demonstrate that head eccentricity can cause errors in touch localization in either the same or opposite direction as the head is turned depending on the procedure used. When head position is held eccentric during both the presentation of a touch and the response, there is a shift in the direction opposite to the head. When the head is returned to center before reporting, the shift is in the same direction as head eccentricity. We rule out a number of possible explanations for the difference and conclude that when the head is moved between a touch and response the touch is coded in a predominantly gaze-centered reference frame, whereas when the head remains stationary a predominantly body-centered reference frame is used. The mechanism underlying these displacements in perceived location is proposed to involve an underestimated gaze signal. We propose a model demonstrating how this single neural error could cause localization errors in either direction depending on whether the gaze or body midline is used as a reference. This model may be useful in explaining gaze-related localization errors in other modalities.