Archive | 2019
Comparison of Implicit and Explicit Learning Processes in a Probabilistic Task
Abstract
This experiment compared the performance with explicit (rule-application and rule-discovery) and implicit (nonrule-instructed) learning approaches on the performance of a probabilistic video game task requiring fine motor control. The task required visual tracking of a small ball of light and “catching” it by means of joystick manipulation. A general pattern of improvement with practice occurred for all conditions. All conditions showed use of predictive relations among stimulus events. However, task performance of the ruleapplication and rule-discovery conditions were inferior to the nonrule-instructed implicit condition, particularly during the early phases of rule acquisition and application. This pattern strongly suggests substantial performance costs associated with attempting to discover or apply probabilistic rules. Decrements are likely due to increased cognitive demands associated with attempting to remember and strategically apply provided probability rules or attempting to discover and apply potentially important and useful probability information from a complex visual display. Many real-world tasks as well as experimental laboratory tasks involve predictive relationships between stimulus events. Important issues surrounding these tasks concern whether individuals can learn to use such predictive relationships, and if one should be informed of the existence or the nature of such relationships in an effort to facilitate performance. The following realworld scenario nicely illustrates these issues. Imagine that you are a baseball batting coach, and you are aware of a specific predictive relationship involved in the pitching movements of the opposing lead pitcher for an upcoming contest. It seems that three out of four times, the pitcher makes a noticeable outward movement of the right elbow during the pitching move just prior to a right to left curve pitch. Knowing this information, would you tell your batters that a probability relationship exists, the exact nature of the probability relationship, e.g., 3 out of 4, or would you tell them nothing? The general issues raised in this scenario are central to the field of implicit and explicit learning research, which addresses basic research tasks such as visual target search and pattern sequence learning, as well as more applied 300 Gr ee n & Fl o w er s i n Pe r c eP t ua l a nd Mot or Sk i l lS 97 (2003) types of tasks such as learning grammar. Implicit learning has been defined as the process involved in the acquisition of abstract, unconscious knowledge about rule-governed covariations present in one’s stimulus environment, without conscious effort (Reber, 1989). In contrast, explicit learning has been characterized as a process similar to conscious problem-solving used for discovering and controlling task variables (Mathews, Buss, Stanley, BlanchardFields, Cho, & Bruhan, 1989), which gives rise to conscious, concrete knowledge of the regularities in one’s stimulus environment (Reber, 1989). Several studies have investigated individuals’ implicit acquisition and use of probability relationships, e.g., predictive covariations, rule-governed complexities, among stimulus events. A key characteristic of this research is that subjects were not provided verbal instructions concerning the existence or nature of the probability relationships. This research has addressed a variety of reaction time-type tasks involving social cognition (Lewicki, 1986), visual target search (Lewicki, Czyzewska, & Hoffman, 1987; Lewicki, Hill, & Bizot, 1988; Stadler, 1989), pattern sequence learning (Nissen & Bullemer, 1987), and event sequence learning (Reber & Millward, 1971; Millward & Reber, 1972). Also, complex rule-governed tasks involved in simulated production systems (Broadbent & Aston, 1978; Berry & Broadbent, 1984; Broadbent, FitzGerald, & Broadbent, 1986), and learning artificial grammar (Reber, 1967; Mathews, et al., 1989) have been addressed, as well as visual-motor continuous tracking tasks (Firth & Lang, 1979; Notterman & Tufano, 1980; Mather & Putchat, 1983). The results across this varied array of tasks have, in general, shown subjects implicitly, unconsciously acquire and subsequently use knowledge of predictive relationships between stimulus events. Other studies have investigated the use of specific probability information provided through explicit instruction (explicit rule-application). Here, subjects were not only told about the existence of a statistical contingency but were also told specifically what the contingency was and were encouraged to use it. Typical examples of such tasks are reaction time-type tasks involving lexical decisions (Neely, 1977), character classification (Taylor, 1977; Flowers, Nelson, Carson, & Larsen, 1984; Flowers, Reed, & Green, 1991), and detection (Posner & Snyder, 1975). The results associated with these tasks have, in general, shown facilitation of performance as a result of using probability information. Unlike the studies directed specifically toward implicit learning or explicit rule application, only a few studies have compared the utilization of implicit versus explicit processes. These comparison studies can be categorized as those contrasting implicit learning to explicit rule-application learning and those comparing implicit learning to explicit rule-discovery learning. In studies involving comparisons between explicit rule discovery and implicit learning, the implicit learning group typically receives no instructions as to specific probability relationships inherent in the task, while the rule discovimp l i c i t an d ex p l i c i t le ar n i nG pr o c es s e s i n a pr o b abi l i s ti c tas k 301 ery group is informed about the existence of probability relationships among stimulus events so as to encourage the discovery and application of the specific relationships. The tasks have involved learning artificial grammar (Reber, 1976; Brooks, 1978; Millward, 1981; Abrams, 1987), and paired-associate learning (Brooks, 1978). The results of these studies have shown both no difference between the noninstructed implicit group and the instructed explicit rule-discovery group (Millward, 1981; Abrams, 1987) and lower performance for the explicit rule-discovery group (Reber, 1976; Brooks, 1978). In studies comparing explicit rule-application learning to implicit learning, the rule-application group is verbally instructed as to the exact nature (degree) of the probability relationship, while the implicit group, as noted above, receives no instructions as to specific probability relationships. Results from two studies of event-sequence learning (Reber, 1966; Reber & Millward, 1968) yielded no effect of providing subjects with concrete knowledge of event probabilities. However, Green and Flowers (1991), in comparing an implicit group to a rule-application group, found that explicit verbal instructions concerning probability relationships among key stimulus features led to poorer overall performance for a probabilistic continuous fine motor task. From these studies, it is clear that subjects can implicitly learn probability information and can apply explicitly provided probability information. However, the findings concerning the utility of providing instructions about the existence (rule discovery) or specific nature of probability relationships (rule application) as compared to implicit learning are mixed and less clear. It appears that, in contrast to what common sense might suggest, providing subjects with explicit information about the existence of probability relationships, or providing subjects with explicit information as to the nature of the probability relationships is either not beneficial or potentially detrimental to task performance. It has been suggested that performance decrements observed in studies of rule discovery (Reber, 1976; Brooks, 1978) and the rule-application (Green & Flowers, 1991) comparative studies are due to differential demands placed upon subjects’ cognitive resources which result in performance costs for the explicitly instructed conditions (Green & Flowers, 1991). In other words, effort in trying to discovery and use, or remember and apply rules detract from cognitive resources needed for optimally performing the tasks. Further, it has been suggested that the degree of disruption is likely dependent on the type of instruction involved (Green & Flowers, 1991). However, there are no studies which specifically contrast explicit rule discovery to explicit rule application. Because the findings from studies which involve a comparison between implicit versus explicit learning are mixed and no direct comparison of rulediscovery and rule-application approaches has been made, the general pur302 Gr ee n & Fl o w er s i n Pe r c eP t ua l a nd Mot or Sk i l lS 97 (2003) pose of our study was to clarify whether explicit instructions are detrimental to skilled performance and to compare the relative quantitative and qualitative effects on performance of rule-application and rule-discovery procedures. In doing so, our study compared the performance effect of explicit rule-application instruction (e.g., instructions that Event “A” predicts Event “B” a given percentage of the time), explicit rule-discovery instruction (e.g., instructions that a predictive rule exists), and implicit learning (no instruction) on participants’ ability to learn and utilize correlations among visual events in a task that required the manipulation of a computer joystick to “catch” a ball of light as it “dropped” across a computer screen. Like many real-world tasks, the task involved predictive relationships between key stimulus features or characteristics, and incorporated a blend of discrete events and continuous visual and motor activity. We hypothesized that participants in all conditions would demonstrate general improvement with practice and would show application or use of the probabilistic relationships between stimulus fe