We have investigated the ability of humans to stabilize their heads in space and assessed the influence of mental set and the relative importance of visual and vestibular cues. Ten normal subjects and 3 patients with bilateral vestibular loss were studied. Subjects were fixed firmly to the chair of a turntable facing a screen on which was projected a target spot. A 'gunsight' spot generated by a small projector fixed to the head provided feedback of head position. Four conditions were studied (1) Gunsight (GU): subjects were instructed to stabilize the head in space by superimposing the 'gunsight' spot on the fixed target spot while chair position was displaced according to a random pattern with a bandwidth from 0-1 Hz. (2) Imagined gunsight (IGU): identical to condition 1 except that the subject was blindfolded and so had to imagine the target position. (3) Mental arithmetic (MA): subjects did mental arithmetic while the chair was displaced. (4) Visual tracking (VT): subjects were instructed to track the target spot with the 'gunsight' spot while the chair was fixed and the target spot driven to follow the chair displacement trajectory used in conditions 1, 2 and 3. In GU normal subjects stabilized their head position extremely well (mean HEAD/CHAIR gain = 0.81). Significant stabilization was present in IGU although the gain (mean gain = 0.61) was reduced compared to GU. There was very little stabilization in MA (mean gain = 0.12). In VT, subjects tracked the target with about the same gain (mean gain = 0.68) as in IGU. By comparison, the vestibular patients could not perform IGU, for which their performance (mean gain = 0.08) was similar to MA (mean gain = 0.06). In GU (mean gain = 0.54), their performance was attributable to visual tracking (mean gain in VT = 0.50). For the frequency bandwidth in which subjects were tested, the results show that: When subjects were distracted by mental arithmetic, the contribution to head stability of the short latency cervico-collic (CCR) and vestibulo-collic (VCR) reflexes is negligible. As expected, vision plays an important role in stabilizing the head. Equally important are long latency stabilizing mechanisms whose onset times (140 ms) are shorter, but still comparable to that of vision. The latter mechanisms are of vestibular origin and their influence is under voluntary control so as to permit augmenting head stability compared to what it would be if vision acted alone.