1. Jumping spiders turn accurately towards moving objects even in the absence of normal visual feed-back. The leg movements made during such turns were studied by cinematography to determine the way in which the retinal location of the stimulus specifies the angle through which the spider turns.
2. In ordinary walking the pattern of stepping is one of alternating tetrapods, similar to that described by Wilson (1967) in tarantulas. Backward walking is very similar except that powerstrokes are protractions rather than retractions of the legs.
3. The stepping pattern during turning is like that of walking, except that the legs on the side towards which the turn is directed walk backwards while those on the other walk forwards. The phase relations of the legs, and the relative durations of power and returnstrokes are the same as in walking.
4. When successive turns are made in the same direction, the stepping pattern continues across the interval between turns (Fig. 3); the pattern is thus continuous in space, but not in time. At the end of one turn each leg stops abruptly at whatever phase of its step has been reached, and resumes the step at the same phase when the next turn begins. Legs in returnstrokes are depressed at the end of a turn, but are elevated and resume the returnstroke after the interval. There is no single resting posture that the legs adopt when stationary.
5. When successive turns are made in opposite directions the legs reverse direction but do not change their stroke: protraction powerstrokes become retraction power-strokes and vice versa.
6. Turns may be executed over at least a ten-fold range of angular velocities (120-1200°/sec). Within the course of a single turn the angular velocity may change several times. Turning velocity is not related to size of turn made.
7. Changes in turning rate are caused by proportional changes in rate of stepping. Step amplitude (angle turned during a step) remains virtually constant at about 75° over the whole velocity range.
8. The spiders can turn accurately when forced to move loads with moments of inertia at least 375 times greater than their own bodies, and accuracy of turning is only slightly reduced when the inertial load is 900 times greater. A large inertial load decreases the upper limit to the velocity of turning attainable. In spite of this decrease in velocity the spider still performs more work in turning the greater load. Stepping rate is reduced by increased load, but not step amplitude.
9. A turn made with a large inertial load ends in a damped oscillation: the spider overshoots its final position and returns to it. The termination of the turn has the characteristics of a suddenly imposed resistance reflex, not a cessation of motor activity.
10. It is argued that conclusions 4-8 above cannot be explained by existing models of arthropod locomotion based on purely endogenous rhythm generators. The constraints on a neural model capable of producing the stepping movements seen during turning are listed, and a model is proposed in which the alternating activity of the motoneurones is driven by proprioceptive feed-back, and only facilitated by central ‘commands’.
11. The size of a turn is specified before its execution by the position on the retina at which the stimulus appears. It is proposed that this retinal instruction is conveyed to the legs as the number of steps that must be taken. One of the eight legs steps, on average, after every 9°, and this angle is within the observed accuracy of turning (S.D. 16°); thus if the number of steps to be made were specified, and counted during the turn, the turn could be terminated at the appropriate moment when that number had been reached. Such a mechanism assumes constancy of step amplitude, and all existing evidence indicates that step amplitude is the only constant feature of the leg movements, under a variety of conditions.