A fundamental question in the study of goal-directed limb control is the relative importance of central planning processes that take place before a movement begins, and corrective processes based on the utilization of response-produced feedback that occurs during movement execution. In the study of discrete aiming movements, researchers have examined the importance of visual feedback by eliminating vision of the moving limb either upon movement initiation or at various points in the movement trajectory. These studies have shown that unless the movement is extremely rapid, participants generally benefit from the availability of visual response-produced feedback during the movement. However, there is less agreement on when during a discrete aiming movement vision is most useful.

Woodworth (1899), and more recently Beggs and Howarth (1970) and Carlton (1981b), have proposed dual component models of limb control in which the initial portion of the movement is ballistic and controlled centrally with visual guidance only being important when the limb approaches the target area. These models are consistent with experiments which show that visual feedback has a greater impact on target-aiming accuracy when it is provided later in the movement trajectory. The dual component model of limb control is also bolstered by kinematic data which indicate that the final stages of an aiming trajectory are generally less stereotyped, and often involve second and third accelerations. However, the apparent pre-eminence of vision late in a trajectory may be related more to the acuity of the visual system to detect error than the necessity for central control of the initial trajectory. When pointing errors are large, such as in a situation of prismatic distortion, they are detected and corrected early in the movement. Also when visual feedback is provided very late in the movement it is not useful because the nervous system simply does not have time to use it. For example, in Meyer, Abrams, Kornblum, Wright and Smith (1988) ‘optimized dual submovement model’, a corrective submovement based on visual feedback about the primary submovement endpoint directs the limb over the final stages of the movement. Presumably once the corrective movement is initiated, vision is no longer useful. Thus, vision plays its role at the completion of the primary, but before the initiation of the secondary/corrective movement.

One of the problems associated with examining the role of vision during the execution of discrete aiming movements is the precision that can be achieved in conditions without visual feedback. That is, unless vision is eliminated several seconds before movement initiation, or distorted to artificially induce large errors, the performer generally exhibits reliable, but proportionally small advantages for target-aiming error when vision of the limb movement is available. For example, when comparing radial error in a full vision condition to a condition in which visual feedback was eliminated upon movement initiation (movement TIME=150–300 ms), Elliott and Allard (1985) found a 1.7 mm difference for a 20 cm movement. These small error differences in extreme visual circumstances (i.e., less than 1% of the movement amplitude) make it understandable why researchers have often failed to find visual influences early in a discrete trajectory where tiny errors would be even more difficult to detect.

Certainly the movements we make in everyday life are typically more complex than attempting to hit a single target with the index finger or a hand-held stylus. In using a calculator, telephone or in operating various other pieces of equipment we are often required to make a series of movements that vary in spatial-temporal difficulty. Ideally in a sequential task, the performer attempts to achieve sufficient accuracy while minimizing temporal and energy costs.

In the context of this study, we decided to examine the visual control of aiming in a more difficult and visually demanding situation. Rather than examine single discrete movements, we borrowed a task from researchers interested in response programming. Specifically, subjects were required to perform a two-movement sequence in which we varied the size of the second target, but not the first target in the sequence. In reaction time experiments, it has been demonstrated that the size of the second target influences the time to initiate the movement sequence, suggesting that the second movement is at least partially prepared prior to movement initiation. This of course is consistent with the idea that it takes longer to assemble a more complex motor program.

Initially we were not concerned with the relation between movement complexity and reaction time, but rather with how the difficulty of the second movement might influence the characteristics of the initial movement trajectory. In single discrete movements, we know that the velocity profiles of low index of difficulty movements (e.g., relatively small amplitude movements to large targets) are almost symmetric with the limb spending proportionally similar amounts of time before and after peak velocity. As the accuracy demands of the task increase, movement times increase due primarily to an increase in time after peak velocity. Presumably this extra time after peak velocity is required to visually direct the limb to the more demanding target. In a two-component movement sequence, where the accuracy demands associated with the second target affect the trajectory of the first movement, it is unlikely that the two components are prepared sequentially in two discrete programming phases. One explanation for the between-phase trajectory effects would involve the performer using on-line visual information about the positioning of the aiming limb and the second target during the first movement. Evidence for this position would be compatible with the notion that visual feedback can play a role quite early in the control of a goal-directed aiming movement. A second possibility is that the two movements are prepared together in advance of movement initiation.

The purpose of Experiment 1 was to determine whether the first and second phase of a two-component movement are independent. Experiment 2 was designed to determine if the interaction between the two movement phases was the result of a common preparation phase prior to the movement sequence initiation, on-line preparation of movement to the second target during movement to the first target, or some combination of these two types of controls.

In this study, participants made rapid two-component aiming movements away from the body. Although target size of the first target was always the same, on different trial blocks the second target was either larger, smaller or the same size as the first target. We also manipulated the availability of vision when the aiming limb was in contact with the first target. Our expectation was that removing vision at this point would negatively affect the second phase of the movement if the two movement components were controlled independently and sequentially. The failure of this manipulation to affect performance would suggest that the second movement was prepared prior to the termination of the first movement, either prior to movement initiation or while the first movement was in progress. In this experiment, participants had as much time as they needed to prepare their movements.