Retention and Motor Memory
One of the most frequently studied theoretical issues in psychology—an issue that people often disagree about—concerns memory. Is memory a result of some processing of an event, or does memory refer to the processing itself? Are there different types of memory, such as memories for movements, for sensations, for smells, and the like, or is there just one memory, whose retention characteristics are a product of the nature and type of processing that is conducted? Questions such as these are hotly debated topics. For example, a scan of the chapters in Byrne (2008) reveals an extremely wide diversity of topics, studied at many different levels of analysis. For the most part, these topics are beyond our present purposes. Rather, we present some of the evidence about the retention (this section) and retention loss (next section) of motor skills.
Retention of Skill for Continuous Tasks
That many motor skills are nearly never forgotten is almost a cliché. Examples such as swimming and riding a bicycle, in which performance after many years of no intervening practice is nearly as proficient as it was originally, are frequently cited. Ideas about such examples, though, are seldom based on acceptable experimental methods; fortunately, many laboratory examples of these situations have been studied, and these results seem to say the same thing.
Although many studies could be cited to illustrate the point, we consider a representative study with long retention intervals by Fleishman and Parker (1962). They used a three-dimensional compensatory tracking task (the Mashburn task, figure 2.5b, p. 32), with movements of the hands in forward–backward and left–right dimensions and movement of the feet in a left–right dimension. Subjects practiced in sessions for 17 days, and then separate groups performed retests after either 9, 14, or 24 months.
The scores for practice and retention tests are shown in figure 14.4, where scores for all three retention groups have been averaged together in the practice session. After the different retention intervals, the various groups were nearly equivalent, and none had shown any appreciable losses in proficiency even after two years of layoff. Some tendency was seen for the two-year group to have slightly less proficiency than the groups with shorter retention intervals, but the differences were very small and the losses were regained completely in three sessions. These small differences are not very meaningful when one compares the retention-test performance to the level of performance at the start of practice. Certainly, this continuous task was retained nearly perfectly for two years.
Other studies, using different continuous tasks, have shown very similar effects. Meyers (1967), using the Bachman ladder climb task, demonstrated nearly no loss in performance for retention intervals of up to 12 weeks. Ryan (1962), using the pursuit rotor and stabilometer tasks, found nearly no retention losses after retention intervals of 21 days; later, he found only small losses in performance on the stabilometer task with retention intervals of up to one year (Ryan, 1965). There are many other examples, and the generalization continues to hold. Continuous motor tasks are extremely well retained over very long retention intervals, just as the cliché about the bicycle would have us believe.
Retention of Skill for Discrete Tasks
While there is ample evidence of nearly complete retention of continuous skills, the picture appears to be quite different for discrete skills. Consider an example by Neumann and Ammons (1957). The subject sat in front of a large display with eight pairs of switches arranged in an inside and an outside circle of eight switches each. The subject was to turn the inner switch “on” and then discover which switch in the outer circle was paired with it; a buzzer sounded when the correct match was made. Subjects learned the task to a criterion of two consecutive errorless trials, and then retention intervals of 1 min, 20 min, two days, seven weeks, and one year were imposed for different groups of subjects.
The main findings are presented in figure 14.5. Some losses in performance appeared after only 20 min, and the losses became progressively greater as the length of the retention interval increased. In fact, after one year, the performance was actually less correct than the initial performance in practice had been, suggesting that the forgetting was essentially complete. However, notice that in all cases the improvements during the retention trials were more rapid than in the original-practice session (as indicated by comparing the slopes of the relearning and practice session curves), indicating that some memory for the skill was retained, which facilitated performance in these relearning trials.
Continuous Versus Discrete Tasks
Why is there such a large difference in the retention characteristics of continuous and discrete skills, with continuous tasks having nearly perfect retention and discrete tasks having such poor retention? A number of hypotheses have been proposed to explain these differences, and they are discussed next.
One hypothesis is that verbal–cognitive components are somehow more quickly forgotten than motor components; because discrete tasks seem to have a heavier emphasis on verbal–cognitive elements (learning which switch in the inner circle is paired with which switch in the outer circle in the Neumann & Ammons study, for example), there is more loss for the discrete tasks over time. Ideas similar to this have generated considerable interest among neuropsychologists who study differences in the retention characteristics of various tasks (e.g., see “Retention of Motor Skills in Amnesia”).
However, while it is true that most of the discrete tasks that have been studied in retention situations seem highly verbal-cognitive (e.g., Schendel & Hagman, 1982), there is no reason that discrete tasks must be so. Certainly, one can think of many discrete tasks that have relatively little reliance on verbal–cognitive abilities (e.g., throwing, striking, pole-vaulting). What would be the retention characteristics of a discrete task that was highly “motor” in nature? Lersten (1969) used an arm movement task (the rho task) in which a circular and a linear movement component had to be performed as quickly as possible. He found approximately 80% loss (of the original amount of improvement) in the circular phase, and a 30% loss for the linear component, with retention intervals of one year. Similarly, Martin (1970) used a task in which the subjects moved the hand over two barriers and then returned to a starting switch as quickly as possible, finding approximately 50% retention loss over a four-month retention interval. The large amount of loss in retention for discrete skills that can be considered “mostly motor” is similar to the loss experienced by Neumann and Ammons’ subjects (figure 14.5), suggesting that there is more to these effects in retention than just the “motorness” of the tasks.
Amount of Practice
One of the major factors determining absolute retention is the amount of original practice, with retention increasing as the amount of original practice increases. In tracking, for example, there are many instances within a trial lasting 30 s in duration in which the pointer and track become separated, with each instance requiring a separate adjustment. Thus, a single “trial” may require many separate “discrete” actions. Contrast this situation to that for discrete tasks, for which a trial typically consists of a single adjustment or action. It stands to reason, therefore, that with the same number of learning trials, the continuous task receives far more practice than the discrete task. The extra amount of practice, according to this hypothesis, leads to increased retention, since it is well known that absolute retention is directly related to the amount of original practice.
What is a “Trial”?
Another notion, related to the one just presented, is that the definition of trial is quite arbitrary; a trial can refer to both a 200 ms reaction-time (RT) performance and a 2 min duration performance on a tracking task. This poses a problem for defining the amount of original practice for the task, and it is also a problem in connection with the retention test. Remember, the level of absolute retention is measured in terms of the performance on the first few “trials” of retention-test performance. If a “trial” is a 2 min performance, there could be a great deal of relearning occurring within a trial for the continuous task, with no relearning within a trial for a rapid discrete performance. So the initial movements within the first trial for the continuous task could show considerable retention loss, but the experimenter might not detect it because the error in the initial performance would be “averaged” with the later portions of the trial on which performance was more proficient. Because this could not occur for the discrete task, it is possible that the amount of forgetting is typically underestimated for the continuous task and not for the discrete task, making the two kinds of tasks appear to be different in their retention characteristics when they might otherwise not be. Fleishman and Parker (1962) found a great deal of improvement within a continuous-task trial, as might be suspected.
Retention of Generalized Motor Programs Versus Parameters
Another possible difference in the forgetting of continuous and discrete tasks is that researchers might be examining different characteristics of the task. Evidence of this was found in a study by Swinnen (1988), who had subjects learn an elbow flexion–extension–flexion task with a goal movement time (MT) of 650 ms. Following 60 trials of practice (with knowledge of results, KR), no-KR retention tests were given after intervals from 10 min to five months. Swinnen analyzed separately the retention of absolute timing (related to the movement parameter) and relative timing (related to the generalized motor program, GMP) and found that absolute timing decayed rapidly, supporting much of the research in this area for discrete tasks. In contrast, the GMP information suffered no loss in relative timing accuracy. These findings suggest that at least some information from learning discrete tasks is retained quite well. Moreover, these findings make sense from a schema theory view (Schmidt, 1975b). One has no need to retain parameter information over long periods of time, because that information is used only briefly to update the schema. In contrast, schema theory suggests that the retention characteristics of GMPs are quite strong so that the invariant features of the action can be recalled and parameterized as needed. Certainly, much more work could be done to explore the ideas introduced in Swinnen’s experiments.