Fitt's empirical result is stated and its information theoretic interpretation briefly discussed. An alternate derivation from a model assuming continuous velocity control of hand position is shown to fit the motion time data equally well. Detailed studies of hand motion trajectories in Fitts' reciprocal tapping task have confirmed the exponential target approach predicted by this model but also revealed systematic fluctuations that it cannot explain. A further alternate model is therefore presented, based on intermittent feedback with target approach by a sequence of discrete positional corrective motion “impulses”. The second model also predicts motion times in accord with Fitts' Law. Detailed studies of target aimed translational hand and rotary wrist motions have shown that the intermittent corrective impulses are of approximately Gaussian integral form, with minimum s.d. on the order of 30 msec. for wrist rotation, and recurrence rate about 10/s. Mechanical and physiological interactions capable of explaining this corrective impulse trajectory are discussed, and a model based on balance of forces between agonist and antagonist muscles is briefly developed. The interaction between external (visual) and internal (proprioceptive or kinaesthetic) feedback channels determining impulse amplitude is discussed in the light of results obtained using force disturbance and with amplified, attenuated or delayed visual feedback. It seems that an internal “secondary positional reference” must be postulated to explain results obtained when S is deprived of visual feedback.

No abstract

A human subject making a sequence of choice-responses is considered as a channel transmitting information. Earlier work suggests that the rate of transmission is limited, and so that response time is proportional to the “entropy” of the source of signals. Entropy is reduced by unbalance in the relative frequency of the possible signals according to the formula, sum (p log p). Unbalance should therefore reduce average response time This prediction is tested in a card-sorting task. The subjects sorted playing-cards into classes in various ways; times taken were proportional to calculated entropy-per-card. Departures from the expected results occurred and were found to be due to differences in perceptual difficulty of discriminations. Some incidental results are mentioned.

Human performance at a typical pursuit tracking task has been analysed by information theory. Without preview of the course, the channel-capacity was found to be about 4 bits per sec. With preview performance ceased to be perfect at about 2 bits per sec. but continued to improve up to at least 8 bits per sec., at about half the ideal slope. The findings are discussed in relation to the information-capacities measured in other tasks, and it is suggested that with full preview the rate should reach a maximum of about 10 bits per sec. Without preview the rate appears to be lowered by the need to offset response lag by predicting course behaviour.

No abstract