Eight normal human subjects were asked to maintain monopodal equilibrium on a narrow beam (task 1) or bipodal equilibrium on an unstable rocking platform (task 2) for 5 s. Each task was performed under four experimental conditions: (1) in light, (2) in darkness, (3) in light while subject had to hold a full cup of water, and (4) as in 3, but with additional instructions to fix the gaze on the cup. The movements of the trunk and head in the frontal plane were recorded by means of a 50-Hz TV image analyzer that computed the coordinates of small reflective markers glued on the skin of the subjects. On the beam the trunk was inclined on the side of the supporting foot (13 +/- 9 degrees), on the rocking platform the mean trunk orientation during the tests was nearly vertical (2 +/- 7 degrees). Nevertheless, in both tasks the mean head position was the same and close to vertical: 1.5 +/- 4 degrees on the rocking platform and 1.5 +/- 5 degrees on the beam. For both tasks and all experimental conditions the head remained stabilized relative to vertical, despite large translations in the frontal plane. Standard deviations of head orientation from its mean value were 2.8 +/- 2 degrees for task 1 and 2 +/- 1.5 degrees for task 2. The changes of trunk orientation were significantly higher: 6.2 +/- 4.8 degrees and 4.5 +/- 4 degrees, respectively. The differences in angular stability of head and trunk, measured through the standard deviations of angular displacements, were especially pronounced in trials with large trunk movements. It was concluded that head angular stabilization, providing the central nervous system with necessary visual and vestibular references, is essential for effective dynamic postural control in the frontal plane during complex equilibrium tasks.