It is our contention that there are two types of visual regulation associated with goal-directed movement. Very early online control is associated with a comparison of dynamic information from the limb about velocity and direction with an internal representation of the expected sensory consequences of the movement. Ideally, this internal model guides the movement to a successful conclusion. However, when there is a mismatch between what is expected and what actually occurs, the discrepancy (error) drives corrective processes designed to reduce the difference to none. Because under normal circumstances the eyes fixate on the target before movement initiation (Binsted & Elliott, 1999), early control probably depends primarily on visual information from the periphery of the retina, which is designed for this type of dynamic information pickup (Paillard, 1982).
The second type of visual regulation occurs near movement termination and involves the use of foveal vision to compare the limb position with the target position. This is the type of regulation associated with Woodworth’s original two-component model as well as its more recent variants (e.g., Elliott et al., 2004; Meyer et al., 1988). By its nature, this type of error-reducing regulation, which involves two afferent sources (visual and proprioceptive information from the limb and visual information from the target), is discrete and thus more likely to manifest as identifiable discontinuities in the trajectory.
In a series of recent experiments, we directly tested our ideas about early and late visual regulation by perturbing the movement environment in order to affect either one or both types of visual regulation. Our work was designed to discover the extent to which the two modes of online control operate independently or covary to determine not only the movement speed and accuracy but also the characteristics of the reaching or aiming trajectories (Grierson & Elliott, 2009). This was accomplished by introducing perturbations influencing the perceived position of the target and the perceived velocity of the limb. The idea was that the former manipulation affects late control while the latter influences early control.
Participants made rapid aiming movements away from the body and toward a small target that was defined by the intersection of three black lines. Depending on the experimental condition, perturbations were made (1) to the target, (2) to the visual background against which the aiming movement was performed, or (3) to both upon movement initiation. Following Mendoza, Hansen, Glazebrook, Keetch, and Elliott (2005) and Mendoza, Elliott, Meegan, Lyons, and Welsh (2006), we used the arrows associated with the Müller-Lyer illusion to affect perceived target position. Compared with the control condition, an arrows-in configuration introduced at movement initiation typically resulted in target undershooting, while a shift to an arrows-out configuration had the opposite effect (Mendoza et al., 2006). We also looked to see whether this Müller-Lyer type of perturbation resulted in a discrete adjustment to the limb movement late in the trajectory, which is the result consistent with the notion of late regulation based on limb and target information (e.g., Elliott et al., 2001; Woodworth, 1899); the results are discussed shortly.
Our second type of perturbation was adapted from the work by Proteau and Masson (1997) and involved introducing a moving background. Texture elements over which the limb traveled on its way to the target moved in either the same or the opposite direction of the moving limb. This manipulation made the limb appear to move more slowly or faster than intended and thus created a mismatch between the perceived velocity and the expected velocity (i.e., internal model) of the limb. These perturbations typically resulted in velocity adjustments that either extended or truncated the movement (Proteau & Masson, 1997). Of interest was whether these velocity adjustments, which are based on a comparison between current movement and feed-forward information about movement expectations, interact with adjustments based on late information about the relative positions of the limb and target.
The two types of perturbations had the expected effects when introduced alone. The Müller-Lyer type of perturbation influenced end-point error (resulted in relative undershooting for the arrows-in condition and overshooting for the arrows-out condition), and this bias was apparent only after peak deceleration, a finding indicating that late adjustments were made to the movement trajectory due to a misperception of target position. As noted in the work of Proteau and Masson (1997), the background manipulation was more robust when the background was moving in the opposite direction of the limb; participants terminated the movements early and demonstrated negative constant error relative to their performance during the other two background conditions. More importantly for our hypothesis about early regulation, this bias was already evident at peak deceleration.
These findings were also observed when the two types of perturbations were introduced together. Moreover, the condition for which the background moved in the same direction as the limb also elicited an effect. The kinematic analysis again indicated that the perturbation involving the moving background influenced early limb regulation more than it influenced late regulation, while the opposite was true for the conditions involving the Müller-Lyer illusion. Once again our findings supported the idea that there are two types of online limb control: (1) early continuous or pseudocontinuous regulation associated with dynamic information from the limb and (2) late discrete regulation tied to the perceived relative positions of the limb and target. Interesting for both performance and kinematic measures is the observation that the effects of our two manipulations were always additive and never interactive. This suggests at least some independence between these two modes of control (Sternberg, 1969).
The Two-Component Model Revisited
It appears that Woodworth’s model has stood the test of time. It is certainly clear that goal-directed limb movement consists of two different phases—an initial movement phase that is more dependent on planning processes completed before the movement and a late phase that involves discrete feedback-based regulation. In contrast to Woodworth’s original ideas and the more recent versions of the two-component model (e.g., Beggs & Howarth, 1970; Carlton, 1981; Elliott et al., 2004; Meyer et al., 1988), the initial movement phase appears to be susceptible to online regulation. That is, the initial impulse is not completely ballistic. Rather, the online regulation involves feed-forward processes, which are associated with the expected sensory consequence of the movement, as well as early dynamic information for the limb. When these two sources of information match, the movement unfolds as planned and can be viewed as ballistic. However, if there is a mismatch between what was planned and what was expected, graded adjustments can be made to the muscular forces used to propel and arrest the limb. These adjustments bring the central (i.e., feed-forward) and peripheral (i.e., feedback) sources of information into harmony.
Because this type of early control is driven by a movement representation that is formed well before movement initiation, there is very little time lag between the movement onset and the comparison processes that provide the basis for regulation. Thus this early control may explain the very short estimates of visual processing time required for the regulation of goal-directed aiming (e.g., Carlton, 1992; Elliott & Allard, 1985; Zelaznik et al., 1983; c.f., Woodworth, 1899; Keele & Posner, 1968). It may also explain why practice results in rather robust reductions in the end-point variability associated with the initial impulse (Elliott et al., 2004). As well, practice leads to early corrective processes that begin earlier (Hansen et al., 2005) and are more error reducing (Khan, Franks, & Goodman, 1998). At least some of the changes associated with practice could be due to the development of a more stable and precise representation of the movement being performed (Proteau, 1992). Alternatively, the comparison process could become more streamlined (Elliott, Chua, Pollock, & Lyons, 1995; Elliott et al., 2001). Issues associated with practice and online control are dealt with in more detail in chapters 2 and 15.
In our current work, we are exploring online control by introducing perturbations that create behavioral dissociations between early and late regulatory processes. As well as using visual illusions to affect the perceived velocity of the limb and the position of the target, we are physically perturbing the velocity of the limb, the movement of the target, or both (Grierson & Elliott, 2008). Once again, movement kinematics provides us with information about where in the unfolding movement these hypothesized control processes operate.
In addition to perturbing the movement environment, it is possible to disrupt the neural systems responsible for online control by using techniques such as transcranial magnetic stimulation (TMS). Following Desmurget and colleagues (1999), we expect the late (target-associated) correction to be sensitive to superior parietal stimulation, while we expect the internal model of anticipated feedback (against which actual feedback is compared) to be susceptible to premotor simulation of the left hemisphere at, or even before, movement initiation. Certainly some careful pilot work is needed to isolate the specific cortical areas and time course for magnetic stimulation.
Another way to study limb regulation processes is to examine groups of people who have difficultly with some aspect of limb control. Our research group has a long history of examining atypical goal-directed behavior in children and adults with Down syndrome (Elliott, Welsh, Lyons, Hansen, & Wu, 2006; Hodges et al., 1995). Following our recent work on anticipatory awareness of goal-directed action in adults with Down syndrome (Obhi et al., 2007), we have suggested that some of the clumsiness associated with perceptual–motor performance in this group may be the result of an inability to form or maintain an internal representation of goal-directed behavior against which to compare feedback. This makes persons with Down syndrome overly dependent on late afferent sources of information for limb control. If this is the case, the aiming behavior of persons with Down syndrome should be affected by perturbations that affect late but not early visual regulation. The challenge in this work is identifying methods of motor skill instruction and practice that minimize the effect of these deficiencies in information processing.