To examine postural stability of the head, slow, undetectable rotations of small amplitude were imposed about a vertical axis while human subjects maintained a stationary body. Six normal subjects were used. The rotations were imposed through an elastic linkage, and lasted 4 s. The amplitude of head rotation was small, approximately 0.002 rad. The imposed perturbations commenced from an unloaded resting position with the head facing forward, under four conditions (1) relaxed, eyes closed; (2) relaxed, eyes open; (3) still, eyes closed; and (4) still, eyes open. The terms "relaxed' and "still' refer to the prior instructions given to the subjects regarding how they were to hold their head. There was a near linear relationship between average torque and average head angle. The effective stiffness of the head on the neck was notably low, approximately 10 Nm rad-1. Two-way analysis of variance (ANOVA) demonstrated ability to increase mean stiffness between "relaxed' and "still' conditions by 51% (P < 0.02). Visual input did not change mean stiffness significantly. Therefore, for the rotations to have been imperceptible, either the visual shifts involved must have been imperceptible, or the eyes must have counterrotated.