Self-controlled concurrent feedback and the education of attention towards perceptual invariants

Abstract

The present study investigates the effects of different types of concurrent feedback on the acquisition of perceptual-motor skills. Twenty participants walked through virtual corridors in which rhythmically opening and closing sliding doors were placed. The participants aimed to adjust their walking speed so as to cross the doors when the doors were close to their maximal aperture width. The highest level of performance was achieved by learners who practiced the task with unambiguous self-controlled concurrent feedback, which is to say, by learners who could request that feedback at wish. Practice with imposed rather than self-controlled feedback and practice without concurrent feedback were shown to be less effective. Finally, the way in which the self-controlled concurrent feedback was presented was also found to be of paramount importance; if the feedback is ambiguous, it may even prevent participants from learning the task. Clearly, unambiguous self-controlled feedback can give rise to higher levels of performance than other feedback conditions (compared to imposed schedule) but, depending on the way it is presented, the feedback can also prevent the participants from learning the task.

In the discussion it is argued that unambiguous self-controlled concurrent feedback allows learners to more rapidly educate their attention towards more useful perceptual invariants and to calibrate the relation between perceptual invariants and action parameters.

Introduction

Imagine a common perceptual-motor skill, such as cycling, kicking a soccer ball, returning a tennis serve, or driving a car, or yet more common activities such as walking or maintaining postural balance. Most of these skills are learned, and for many skills, the learning never ends. Even the most experienced perceivers and actors can often further improve, or further adapt to changes in the environment, or to changes in their own bodies. It is easy to appreciate the crucial importance of learning in our everyday life (Crossman, 1964, Kottke et al., 1978). Given, then, the fact that learning processes can also be long and costly, it is not surprising that the challenging possibility to facilitate learning has motivated considerable bodies of research, especially in the more applied areas (e.g., Starkes & Ericsson, 2003).

One of the pivotal results of research on learning is that feedback (or instructions) can be decisive for the acquisition of expertise. The converging evidence indicates that feedback is most beneficial if directed towards the environmental consequences of the movement to be produced, rather than towards the movement itself (see Wulf, 2007, for a review). A possible explanation of this finding is that an external focus of attention (i.e., towards the consequences of the movement) allows the movement to be executed on an automatic basis, without more cognitive interferences. The present article therefore focuses on feedback about the environmental consequences of actions. Moreover, we focus on a type of feedback given during the execution of the action (i.e., concurrent feedback), and which learners can ask for at wish (i.e., self-controlled feedback).

Let us first review a few more general findings concerning perceptual-motor functioning and learning. Self-movements and/or movements of objects relative to agents typically result in transformations of ambient energy arrays (e.g., the ambient optic array). In many cases, the perceptual substrate of goal-directed movements has been shown to reside in such transformations (Gibson, 1958, Warren, 2006, Warren et al., 2001). There is also a large body of evidence indicating that learning perceptual-motor skills goes together, among other things, with a shift in which perceptual variable is used to perform the task, from variables weakly correlated to the event under consideration to sources of information highly correlated or even specific to that event, which is to say, to perceptual invariants (e.g., Fajen and Devaney, 2006, Jacobs et al., 2001, Smith et al., 2001).

In the study of Smith et al. (2001), observers were asked to release a pendulum at an appropriate time in order to create a collision between an approaching ball and a sphere at the end of the pendulum. In this situation, several optic flow variables are related to the temporal closeness of the ball. Whereas the expansion rate of the ball is merely correlated with time to contact, tau, defined as a combination of optical size and expansion rate, specifies time to contact (Lee, 1976). In the course of practice, participants in the study of Smith et al. abandoned the use of lower-order variables (i.e., expansion rate) to the benefit of higher-order variables (i.e., tau). Gibson referred to this tuning toward the more useful perceptual invariants as the education of attention (Gibson, 1966, Gibson, 1979). With this understanding of learning, our question becomes: What kind of procedures might allow learners to get attuned more quickly to the relevant properties of the perceptual flow?

Two types of procedures are generally distinguished: explicit and implicit ones. In explicit procedures, performers are asked to focus their attention on specific visual cues. Observers typically aim to predict a forthcoming event, such as the place of arrival of a ball or the direction of a stroke, while several key visual cues are described or highlighted (e.g., the racquet position or speed in the tennis serve; Farrow, Chivers, Hardingham, & Sachse, 1998). Implicit procedures do not explicitly describe or highlight specific cues (Magill, 1998). These procedures start from the assumption that the information that allows one to perceive an event correctly need not necessarily be located at a specific place (e.g., at the position of the racquet or the position of the feet). Implicit procedures also assume that reliance on informational variables can be achieved implicitly, which is to say, without learners being aware of the informational basis of their actions. The available evidence indicates a superiority of implicit procedures over explicit ones, especially under performance pressure (e.g., Koedijker et al., 2007, Liao and Masters, 2001, Masters, 1992). Moreover, implicit procedures seem to be better suited for the discovery of higher-order perceptual invariants, because these invariants might be difficult to describe explicitly or highlight.

Implicit procedures, also labelled guided discovery procedures, can rely on instructions or on the use of feedback protocols. Williams, Ward, Knowles, and Smeeton (2002) tested whether a guided discovery protocol based on instructions improves anticipation skills in tennis. The attention of observers was directed towards potentially informative areas. For example, rather than being informed about the relation of hip orientation and shot outcome, observers were encouraged to focus their attention to the midriff region. The precise relationship between hip orientation and shot outcome hence remained to be discovered by the observers. This procedure was found to enhance anticipation skills as much as an explicit procedure. Although the implicit procedure used by Williams et al. was less prescriptive than the explicit procedure tested in the same study, the instructions still led observers to focus on specific areas. Feedback-based procedures, in which such instructions are not given, are in this sense yet more implicit.

Developing a feedback-based procedure was the focus of a study by Jacobs et al. (2001). Participants in this study were shown simulations of colliding balls, and they were asked to judge the relative mass of the balls. As in the study of Smith et al. (2001), the judgments of novice perceivers were based on lower-order variables but, after practice with feedback, perceivers progressed to the use of a higher-order variable that specifies relative mass. In addition, Jacobs et al. found that perceptual learning proceeds faster if practice is organized so that judgments based on non-optimal sources of information are systematically erroneous. Such practice conditions can be achieved through the selection of stimuli (i.e., collisions) for which the use of non-optimal sources of information leads to particularly inappropriate judgments, and hence to negative feedback. The negative feedback associated with the use of lower-order variables leads perceivers to change in variable use and to discover higher-order perceptual variables. Jacobs and Michaels (2007) later modeled these findings, showing more precisely how the structure of the errors made by observers might guide the learning process.

Recently, Camachon, Jacobs, Huet, Buekers, and Montagne (2007) tried to facilitate a feedback-driven shift in the use of perceptual variables using a concurrent feedback procedure. Camachon et al. used a perceptual-motor task that was previously studied by Buekers et al., 1999, Montagne et al., 2003, and Camachon, Montagne, Buekers, and Laurent (2004). In these studies, observers walked through virtual corridors and adjusted their walking speed in order to pass through a pair of rhythmically opening and closing sliding doors. This same task has also been studied in a real-world environment (Cinelli & Patla, 2008; cf. Cinelli, Patla, & Allard, 2008). The task is of particular interest in the domain of learning because of the strong spatio-temporal constraints that are imposed on the participants, and because of the relatively low levels of performance observed at the beginning of practice, at least for the slightly faster moving doors in the virtual environments.

The adjustments in walking speed that are required to successfully cross the sliding doors depend on the spatiotemporal vicinity of the doors as well as on the cyclical characteristics of the oscillation of the doors. Hence, observers have to discover an informational variable that allows them to integrate these two processes. Camachon et al. (2007) hypothesized that concurrent feedback might help them to discover such an informational variable. The concurrent feedback used by these authors, provided during the completion of the task, informed about the consequences of the participants’ current walking speed if the walking speed would remain constant. More precisely, a gauge presented at the left-hand side of the visual scene indicated whether observers would arrive early or late at the door crossing. In spite of this, however, the experiment of Camachon et al. did not reveal an advantage of concurrent feedback as compared to a learning regimen without concurrent feedback.

A possible drawback of the protocol used by Camachon et al. (2007) is that the concurrent feedback was imposed on the observers, who could not control the moment and frequency of the feedback. To date, several studies have revealed an advantage of self-controlled feedback procedures over classical ones for the learning of various perceptual-motor skills, including throwing (Janelle et al., 1997, Janelle et al., 1995), skiing (Wulf & Toole, 1999), and sequential timing (Chiviacowski and Wulf, 2002, Chiviacowski and Wulf, 2005). The results of the self-controlled feedback studies also show that learners spontaneously opt for a fading schedule: During the acquisition process they call less and less for the feedback (e.g., Janelle et al., 1997, Wulf and Toole, 1999). This spontaneous strategy prevents the learner from the so-called guidance effect, or feedback dependence (cf. Salmoni, Schmidt, & Walter, 1984). Taken together, these studies provide convincing evidence in favor of the beneficial role of self-controlled feedback designs.

The present study is a continuation of the Camachon et al. (2007) study, with the notable difference that a self-controlled feedback protocol is applied. Thus, participants in the study adjusted their walking speed in a virtual reality environment in order to pass through sliding doors. During the execution of the action, participants had the opportunity to solicit concurrent feedback whenever they wished. The concurrent feedback informed participants about the current relation to the environment. In visualizing the current relation during the trial, participants might be able to more readily access the sources of information that specify this relation. With this study we aim to discover the extent to which self-controlled concurrent feedback helps learners to converge on the perceptual regularities that are required to control the action. We hypothesize that the self-controlled feedback helps the learner to acquire a high level of perceptual-motor expertise.

A second aim of the study is to find out to what extent the way in which concurrent feedback is provided affects the learning process. The concurrent feedback was provided to the participants in the form of a gauge (as in Camachon et al., 2007) or in the form of ghost doors. The gauge was placed on the left-hand side of the visual scene (look ahead to Fig. 2A). The ghost doors were placed in the virtual corridors at the position at which participants would be, in the absence of changes in walking speed, at the moment of maximal door aperture nearest to the door crossing (look ahead to Fig. 2B). This condition was designed to provide the concurrent feedback in a more natural way than the gauge: The ghost doors are fully embedded in the visual scene. As a consequence we hypothesize that the ghost doors lead to faster learning than the gauge.