3D dipole tracing with 1 ms step of visual evoked potentials recorded from 40 electrodes was performed under exposition of crosses in 5 healthy human subjects. The data on dipole displacement were compared with prediction of the simulation study on distortion of dipole localization by the signal filtration in the low-frequency band. These predictions were experimentally confirmed: the effect depends on the degree of filtration (0.1 or 0.5 Hz) and on the latency of EP waves. Localization and strength of P1 dipoles were not changed under filtration, while for later components--N2 and especially P3--they changed significantly. For the improvement of these distortions time constant of the amplification tract must be some times longer than the time of the dipole activity.

The viscoelastic properties of the human arm were measured by means of short force perturbations during fast reaching movements in two orthogonal directions. A linear spring model with time delay described the neuromuscular system of the human arm. The obtained viscoelastic parameters ensured movement stability in spite of the time delay of 50 ms. The stiffness and viscosity ellipses appeared to be predominantly orthogonal to the movement direction, which reduced the effect of force perturbation in the direction orthogonal to the reaching movement. Thus, it can be argued that the viscoelastic properties of the neuromuscular system of the human arm are adjusted to the direction of movement according to a "path preserving" strategy, which minimizes the deviation of the movement path from a straight line, when exposed to an unexpected external force.

We are proposing a human arm model that consists of three rigid segments with seven degrees of freedom. The shoulder joint was modeled as a ball-and-socket joint and the elbow and wrist joints were modelled as skew-oblique joints. Optimal parameters for this model were calculated on the base of in vivo recordings with a spatial tracking system. The criterion of optimality was defined as the minimum of the mean-square deviation between the experimentally obtained sensor positions and orientations and their positions and orientations calculated by solving the direct kinematics problem. The minimal value of the direct kinematics error was found to be 0.5-0.6cm for sensor positions and 5-7 degrees for sensor orientations. We are proposing that these values serve as the assessment for the accuracy of the arm model.