When a subject is walking blindfolded straight ahead towards a previously seen target, the brain must update an internal representation with respect to the environment. This study examines whether the information given by the vestibular system is necessary for this simple path integration task and gives a quantitative description of locomotor behaviour during the walk by comparing ten normal and seven bilateral labyrinthine-defective (LD) subjects. Each subject performed 20 blindfolded walks (EC) and ten walks with eyes open (EO) towards a target attached to the floor 4 m in front of them; these walks were made at different velocities. The positions of head, trunk and feet were recorded using a 3D motion analysis system. No significant difference was found between normal and LD groups in terms of the distance error of reaching the target, while LD subjects showed a larger lateral error. Path curvature, expressed as the standard deviation of the angle between the direction of one step and straight ahead, was found to be significantly larger for LD subjects in the EC condition, demonstrating their instability when walking without vision. Mean walking velocity was lower for LD subjects than for normal subjects in both EC and EO conditions. Both groups walked faster with eyes open; LD subjects increased their velocity by increasing step length, normal subjects by increasing step frequency. Head stabilisation in the frontal plane during locomotion was not significantly different between LD and normal subjects, whereas both head and trunk rotation were slightly larger in LD subjects during blindfolded walking. The results show that bilateral LD subjects are able to perform linear goal-directed locomotion towards memorised targets. Thus, the vestibular system does not appear to be necessary for active linear path integration.