Differential Effects on Learning

This article was downloaded by: [McMaster University] On: 13 November 2011, At: 12:16 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Motor Behavior Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vjmb20

Reducing Knowledge of Results about Relative versus Absolute Timing: Differential Effects on Learning a

b

Gabriele Wulf , Timothy D. Lee & Richard A. Schmidt

c

a

Max-Planck Institute for Psychological Research, Munich, Germany

b

McMaster University, Hamilton, Ontario, Canada

c

Department of Psychology, University of California, Los Angeles, USA

Available online: 14 Jul 2010

To cite this article: Gabriele Wulf, Timothy D. Lee & Richard A. Schmidt (1994): Reducing Knowledge of Results about Relative versus Absolute Timing: Differential Effects on Learning, Journal of Motor Behavior, 26:4, 362-369 To link to this article: http://dx.doi.org/10.1080/00222895.1994.9941692

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Journal of Motor Behavior, 1994, Vol. 26, No. 4, 362-369

Downloaded by [McMaster University] at 12:16 13 November 2011

Reducing Knowledge of Results About Relative Versus Absolute Timing: Differential Effects on Learning Gabriele Wulf

Timothy D. Lee

Richard A. Schmidt

Max-Planck Institute for Psychological Research Munich, Germany

McMaster University Hamilton, Ontario, Canada

Department of Psychology University of California Los Angeles

two days-than giving K R after every trial (e.g., Winstein, 1988; Winstein & Schmidt, 1990; Wulf & Schmidt, 1989). Similarly, providing summary KR, where KR for a set of trials (e.g., five) is given only after this set has been completed, enhances motor learning compared with practice conditions with K R after every trial (e.g., Schmidt, Lange, & Young, 1990; Schmidt, Young, Swinnen, & Shapiro, 1989). Also, giving KR about the average of a set of trials facilitates learning, compared with an every-trial format (Young & Schmidt, 1992, Experiment 2). Finally, bandwidth KR-where KR is given only if the error exceeds a certain predetermined bandwidth (e.g., 10% deviation from the goa1)has been found to enhance retention and transfer performance (Lee & Carnahan, 1990; Sherwood, 1988). These findings contradict traditional views, according to which increased use of any KR variations (e.g., more frequent, more precise, or more immediate KR) should enhance learning (e.g., Adams, 1971; Bilodeau, Bilodeau, & Schumsky, 1959; Schmidt, 1975; Thorndike, 1927). The results of the recent studies have been explained in terms of the guidance hypothesis for K R (Salmoni, Schmidt, & Walter, 1984; Schmidt, 1991). According to this view, K R has positive effects, such as guiding the learner to the correct response, but it also has several negative effects for learning. For example, the learner might become too dependent on the extrinsic information and neglect the processing of intrinsic feedback. It

ABSTRACT. The purpose of the present experiment was to examine further earlier suggestions that a reduced relative frequency of knowledge of results (KR) can enhance the learning of generalized motor programs (GMPs) but at the same time degrade parameter learning, compared with giving KR after every trial (Wulf & Schmidt, 1989; Wulf, Schmidt, & Deubel, 1993). In contrast to these earlier studies, here KR was given separately for relative timing and absolute timing. Subjects practiced three movement patterns that required the same relative timing but different absolute movement times. KR was provided on 100% or 5P/o of the practice trials for relative timing or absolute timing, respectively. In retention and transfer tests, the groups that had had 500/0 KR about relative timing demonstrated more effective learning of the relative-timing structure, that is, GMP learning, than the groups that had had 100% KR about relative timing. The KR frequency had no effect on parameterization during retention; yet, when transfer to a task with a novel overall duration was required, the groups given lW/oKR about absolute timing were more accurate in parameterization than the groups provided with 50% KR about absolute timing. Thus, the reduced relative KR frequency enhanced GMP learning but had no beneficial effect, or even a degrading effect, on parameter learning. The differential effects of a reduced KR frequency on the learning of relative timing and absolute timing also provide additional support for the dissociation of GMP and parameterization processes. Key word: generalized motor program, knowledge of results, motor learning, movement parameters

S

everal recent studies have shown that practice conditions in which the knowledge of results (KR) is in some way limited are more effective for learning than conditions entailing more “useful” KR. For example, giving KR only on a portion of the practice trials, that is, reducing its relative frequency, has been found to produce more effective retention and transfer performance-especially if these tests are delayed by one or

Correspondence address: Gabriele WurJ; Max-PlanckInstitut fur psychologische Forschung, Leopoldstrasse 24, 80802 Munchen, Germany. E-mail: wuv@mpipf-muenchen. mpg de

362


Reducing KR About Relative Versus Absolute Timing is also possible that, because KR facilitates the retrieval of the next action, it may block important retrieval operations that are critical for long-term retention (Schmidt et al., 1989). Finally, frequent KR has been shown to make movement production very variable, presumably preventing the learning of a stable movement representation (Nicholson, 1992; Wulf & Schmidt, 1994). Whatever the exact reason, it is clear that, compared with 1 OO%-KR schedules, reducing the KR usefulness can facilitate motor learning. Most studies showing the effectiveness of reduced relative KR frequency have been concerned with the learning of single movements, but Wulf and colleagues (Wulf & Schmidt, 1989; Wulf, Schmidt, & Deubel, 1993) showed that reducing the relative frequency of KR can also enhance the learning of a class of movements that is presumably governed by a generalized motor program (GMP) (Schmidt, 1975,1985,1988).A GMP is a learned structure that governs a class of actions that share the same sequencing, relative timing, and relative forces. Yet, movements governed by a GMP can vary across super& cia1 dimensions, such as absolute time and absolute forces, by the assignment of movementparameters. Variable practice within a movement class (i.e., in parameter selection) leads to the development of a motor schema (Schmidt, 1975). Using a sequential timing task, Wulf and Schmidt (1989) had subjects practice three task versions with the same relative timing but different overall durations. They found that reducing the relative frequency of KR enhanced the learning of the relative-timing structure, compared with giving KR after every trial. The results were interpreted as showing that reduced KR enhances GMP learning. Although Wulf and Schmidt (1989) were not concerned with the learning of movement parameters, Wulf et al. (1993) examined the effects of a reduced KR frequency on parameterization learning, in addition to GMP learning. The movement patterning task used in the Wulf et al. (1993) study allowed the manipulation not only of movement time but also of movement amplitude. Errors in the GMP, and in time and amplitude parameterization, were separated by using a curve-fitting procedure. Wulf et al. (1993) again found that GMP learning was enhanced by the reduced KR frequency (63% KR), compared with 100% KR during practice. However, time parameterization (Experiment 1) and amplitude parameterization (Experiment 2) were degraded by a reduced relative frequency of KR (although these effects were not always statistically significant). The findingthat more frequent KR facilitated parameter learning can be interpreted in terms of schema theory (Schmidt, 1975). In this view, learning to parameterize effectively depends on the association of (a) movement outcomes (the information provided by KR) and (b) the parameters used to produce the movement. Trials without KR are thus neutral for learning, neither facilitating nor detracting from it. Therefore, increasing the relative December 1994, Vol. 26, No.4

number of no-KR trials, or reducing the relative frequency of KR, should enhance parameter learning, as found by Wulf et al. (1993). The purpose of the present study was to examine further the possible differential effects of reduced KR frequency on GMP and parameter learning. Showing that more frequent KR indeed degrades GMP learning while at the same time it facilitates parameter learning would increase our understanding of how these two kinds of movement proficiency are learned. Furthermore, such a dissociation would provide added confidence for the validity of the theoretical constructs of GMP and parameterization processes (Schmidt, 1975, 1985, 1988). Demonstrating that a given experimental variable (e.g., reduced KR frequency) enhances one construct (GMP) but degrades another (parameterization)-especially when measured on the same responses-would provide further evidence for the distinctivenessof these two types of motor control processes. In our earlier studies (Wulf & Schmidt, 1989; Wulf et al., 1993), the KR included information about both relative and absolute timing (and amplitude). In the Wulf and Schmidt (1989) experiments, KR consisted of the three actual movement-segment times (in relation to the goal segment times), whereas in the Wulf et al., 1993, experiments, KR consisted of the position-time curve of the subject-produced pattern (in relation to that of the goal pattern) as well as the root-mean-squared (RMS) error. In the present study, we used the same sequential timing task as Wulf and Schmidt (1989). Here, however, KR about relative timing or absolute timing was provided either on 100% or 50% of the trials, these factors being crossed in a 2 X 2 design. If frequent KR enhances parameter learning, the groups receiving 100% KR on absolute timing should show more effective retention and transfer performance with regard to absolute timing than the groups receiving 50% KR on absolute timing. On the other hand, the groups with 50% KR on relative timing were predicted to demonstrate more effective learning of the relative-timing structure than the groups with 1Wh KR on relative timing. Method Subjects

Students from the University of Munich ( N = 72) served as subjects. They were paid DM 20.00 for their services. Subjects were nake as to the purpose of the experiment, and none had prior experience with the apparatus. Apparatus The apparatus was similar to the one used by Wulf and Schmidt (1988,1989).It consisted of a wooden board (64 X 42 cm), upon which were mounted four buttons, each 3 cm in diameter (see Figure 1). The buttons were in a diamond-shaped pattern in which adjacent buttons were 363


G. Wulf, T. D. Lee, & R. A. Schmidt

FIGURE 1. Experimental apparatus.

18.5 cm apart. The buttons were keys modified from a computer keyboard and padded with foam. The buttons were interfaced with an IBM-compatible computer so that the timing of the button presses could be measured. A computer program that was developed for the purposes of this experiment was used for data collection.

Task The subject’s task was to depress the buttons in the prescribed sequence (see Figure 1) and to be as accurate as possible with regard to the relative goal movement times (MTs) for each of the three movement segments (Segment 1 was from Button 1 to Button 2, Segment 2 was from Button 2 to Button 3, and Segment 3 was from Button 3 to Button 4), as well as the overall MT. The relative timing was the same across all task versions, and only the absolute durations were different. The relative goal MTs (in percent) for the three segments of all task versions were 22.24.4-33.3. The total goal MTs (in ms) for the three practice task versions were 900 (Version A), 1,125 (Version B), and 1,350 (Version C). Thus, the absolute goal segment MTs were 200-400-300,250-500-375, and 300-600-450 ms, respectively. (Subjects were not explicitly told the absolute segment MTs, although they could, of course, calculate them; this was not necessary, as KR was also given about the relative segment times, as well as the total MT, and thus could be directly compared with the goal MTs.) Procedure Before each trial, the letter (A, B, or C) denoting the task version, as well as the relative segment goal MTs and total goal MT appeared on the screen for 4 s (e.g., “A 22.4-44.4-33.3 900”). During the following 4-s interval (5-s interval in transfer), the subject performed the move364

ment. If the movement was not completed within this time, or if the subject made an error in the button presses, a message appeared, and the trial was repeated. KR consisted of repeating the letter, task version, and relative MTs and total MT. The actual relative MTs (e.g., 20.145.5-34.4) and actual total MT (e.g., 1,032) were written underneath the respective goal times in a different color. The order of task versions was the same for all subjects (A, B, C, A, B, C, . . .). The experimental groups differed only with respect to the relative KR frequency in terms of relative timing and absolute timing (total MT). Subjects were randomly assigned to one of four groups of 18 subjects each: The 100-100, 100-50, 50-100, and 50-50 groups. The 100-100 group received KR about both relative and absolute timing after each trial (1000/0KR). The 100-50 group was provided with relative-timing KI? after each trial (loo%), but received KR about absolute timing only on 500/0of the trials. Specifically, KR was withheld on every other block of three trials (one trial each for Task Versions A, B, and C). The 50-100 condition had the reversed KR schedule; that is, subjects received 50% KR in terms of relative timing (with no KR being provided on every other three-trial block) and 100% KR in terms of absolute timing. Finally, the 50-50 group was given 50% KR about both relative and absolute timing; that is, this group received no KR at all on every second block. The experiment consisted of five phases: practice, immediate retention, and immediate transfer (all on the first day), and delayed retention and delayed transfer one day later. All subjects performed 108 trials during the practice phase (i.e., 36 trials on each task version) in the orders and with the KR schedules described above. Five minutes after the end of the practice phase, subjects performed the immediate retention test, which consisted of 12 no-KR trials on the three practice task versions (A, B, and C) presented in a random order (4 trials on each version). Subsequently, subjects performed 12 trials without KR on the novel task version D (immediate transfer) that had the same relative timing as the practice task versions (22.2-44.4-33.3) but a longer overall goal MT (1,575 ms). All subjects returned to the lab one day later and performed the delayed retention and transfer tests. These were identical to the immediate retention and transfer tests, respectively.

Statistical Analyses The dependent variables of interest were errors in relative timing, as a measure of the proficiency of the GMP, and errors in absolute timing, as a measure of proficiency in parameterization. We measured relative-timing performance by computing the sum of the absolute differences between the goal proportions and the actual proportions for each segment ( A E [prop]). To assess absolute-timing performance, we computed constant error (CE) by taking the signed difference between the overall goal MT and the actual overall MT. Because CE Journal of Motor Behavior

Reducing KR About Relative Versus Absolute Timing

has the potential problem that across subjects, positive and negative values tend to cancel out each other (e.g., Schutz, 1977), we used the ICE1 values for each subject in our analyses. The experimental design thus involved two crossed factors. One was relative timing KR frequency (100% or WO),hereafter abbreviated RT-KR. The second factor was absolute timing KR frequency, (1000/0or 50?h), hereafter abbreviated AT-KR. The design used a 2 X 2 X 6 (RT-KR X AT-KR X Block) design for practice, and a (RT-KR X AT-KR) analysis of variance (ANOVA) for each retention and transfer test.

the left panel of Figure 3. All groups reduced their errors in absolute timing over practice, F(5, 340) = 13.3, MSe = 205,404.6, p < .001. The 100-50 and 50-50 groups (50%KR about absolute timing) needed somewhat longer to approach the goal movement times than the groups with 100% KR (100-100 and SO-loo), but the interaction of AT-KR and block was not signnificant, Q5, 340) = 1.5, MSe = 22,312.4,~> .05. Also, none of the other main or interaction effects were significant, Fs(1, 68) and (5, 340), respectively, < 1.

Results

AE (prop.) All subjects performed 12 no-KR trials, with the three practice task versions (A, B, C) presented in a random order, 5 min after the end of practice. Errors in relative timing on this immediate retention test can be Seen in Figure 2. The groups with lW/o KR in relative timing during practice ( 100-1 00 and 100-50) had overall higher error scores than the groups with 50% relative-timingKR (50-100 and 50-50). The main effect of RT-KR was not statistically significant, however, F(1, 68) = 2.5, MSe = 116.3, p = .12. Also, neither the main effect of AT-KR nor the interaction were significant, Fs(1, 68) < 1.


Practice AE (Prop. 1 Relative-timing errors (our measure of GMP accuracy) for the four groups during practice can be seen in the left panel of Figure 2. All groups clearly reduced their AE (prop.) from the beginning to the end of practice, F(5, 340) = 32.4, MSe = 3 1 8 . 5 , ~< .001; the 100-100 group showed overall greater errors than the other three groups, who demonstrated similar performances. However, none of interaction effects, Fs < l, nor the main effects of RT-KR, F(1, 68) = 1.6, MSe = 191.2, or ATKR, F(1, 68) = 1.5, MSe = 177.9,~s> .05, were significant. ICE1 Errors in overall movement time (our measure of parameterization accuracy) during practice are shown in

20

Immediate Retention

ICE1 With regard to absolute timing, all groups showed similar performances during immediate retention; 1 CEsl were slightly higher than those reached at the end of practice (see Figure 3). There were no significant main or interaction effects, all Fs( 1,68) < 1.

-

-

.05. The interaction of RT-KR and AT-KR was not significant either, F(1, 68) < 1. ICE1 [CEsl in transfer tended to be higher than those in retention, probably in part because, the longer goal movement times were associated with larger errors in timing (e.g., Newell, Carlton, & Carlton, 1982; Schmidt, Zelaznik, Hawkins, Frank, & Quinn, 1979; Shenvood, 1986). In addition, the transfer task was “outside” the range of prior practice, requiring an extrapolation of movement parameters, which could also lead to increased errors compared with movement variations within the practice 366

range. The 50-100 group tended to have the lowest ICES[ in immediate transfer (see Figure 3), whereas the other three groups demonstrated similar performances. Neither the main effects of RT-KR, F( l , 68) = l .4,MSe = 74,597.3, p > .05, or AT-KR, F(1, 68) < 1 , nor the RTKR X AT-KR interaction, F(1,68) < 1, were significant.

Delayed Retention As temporary effects of the independent variables that are present during acquisition could still affect performance on immediate retention and transfer tests, we used retention and transfer tests that were delayed by one day to assess the relatively permanent, or learning, effects. The delayed tests were identical to the immediate tests in format. A E (prop.) On the second day, the 50% relative timing KR groups (50-1 00, 50-50) clearly outperformed the 100% relative timing KR groups (100-100, 100-50) in terms of the relative-timing pattern (see Figure 2), and this effect was significant, F(1, 68) = 4.5, MSe = 209.5, p < .05. That is, the GMP was learned more effectively with a relative frequency of KR that was reduced to 50%. There was no significant effect of AT-KR and no interaction, Fs( 1,68) < 1. ICE1 All four groups showed almost identical performances with regard to absolute timing (see Figure 3). There were no main or interaction effects; all Fs (1, 68) < 1. Journal of Motor Behavior

Reducing KR About Relative Versus Absolute Timing Delayed Transfer


AE (prop. ) As in the delayed retention test, the groups with 50% relative-timing KR (50-100, 50-50) were clearly more accurate in terms of relative timing than the groups with 100% relative-timing KR (100-100, 100-50; see Figure 2). The main effect of RT-KR was significant; F(1,68) = 7.7, MSe = 315.4,~< .01. Thus, the novel-task performance was enhanced by the 50%-KR frequency. Furthermore, errors in relative timing were generally of the same size as those in delayed retention. These results tend to support GMP theory (Schmidt, 1985, 1988), according to which movements with the same relative timing are governed by the same GMP. There was no effect of ATKR and no interaction, Fs(1, 68) < 1. ICE1

In absolute timing, the 50% relative timing KR groups (100-50,50-50) demonstrated considerably higher errors in absolute timing than the groups that had received 100%KR in absolute timing (100-100, 50-100; see Figure 3). That is, the reduced KR frequency in absolute timing degraded parameterization learning, at least when transfer to a novel task version was required. The main effect of AT-KR was significant, ZV, 68) = 5.3, MSe = 437,747.7, p < .05. There was no effect of RT-KR or of interaction, Fs( 1, 68) < 1. Discussion The aim of the present study was to further examine the notion that a reduced relative frequency of KR might enhance the learning of GMPs but at the same time degrade parameter learning (Wulf et al., 1993). Subjects practiced three movement patterns; all had the same relative timing (presumably requiring the same GMP), but they differed with regard to the overall MT (i.e., requiring different movement parameters; Schmidt, 1975,1985, 1988). KR was presented on either 100% or 50% of the trials for relative timing (percentage MTs for each segment) or absolute timing (total MT), respectively, yielding four groups of subjects (100-100, 100-50, 50-100, and 50-50). If reducing the KR relative frequency enhances GMP learning (as in the studies by Wulf & Schmidt, 1989, and Wulf et al., 1993), the 50-100 and 50-50 groups, with 50% KR in terms of relative timing, should demonstrate more accurate relative-timing performance in (delayed) retention and transfer than the 100-100 and 100-50 groups, who received lW/o KR about relative timing during practice. Furthermore, if a reduced KR frequency degrades parameter learning, compared with 100%KR (as in the study by Wulf et al., 1993), the 100-50 and 50-50 groups, with 50% KR in absolute timing, should show less effective absolutetiming performance in retention and transfer than the 100-100 and 50-100 groups that were provided with 100% KR in absolute timing. December 1994, Vol. 26,No. 4

There were no significant differences between groups during practice with regard to either relative-timing or absolute-timing performance. Also,in the immediate retention test, no significant group differences emerged, even though in relative timing the two groups with 100% relative-timing KR had larger errors, on average, than the groups with 50% KR. On the immediate transfer test, where a novel task with a new overall duration had to be performed, these group differences were clearer (even though not quite statistically significant); the 5OO!wKR groups were more effective in relative timing than the lW/o groups. There were no significant effects with regard to absolute timing in immediate transfer, however. On the second day, when the temporary effects of the independent variables presumably had dissipated, leaving only the relatively permanent, or learning, effects (Salmoni et al., 1984), the results were rather clear-cut. With regard to relative timing, in both delayed retention and transfer, 50% KR in terms of relative timing (50-100, 50-50) produced more effective learning than 100% KR (100-100,100-50). These results are in line with previous studies where a reduced relative frequency (with regard to overall performance) facilitated learning of a GMP (Wulf & Schmidt, 1989; Wulf et al., 1993). In addition, the present results showed that this effect is independent of whether or not KR about absolute timing is given at the same time. Other studies (e.g., Nicholson & Schmidt, 1991; Winstein & Schmidt, 1990) that found beneficial effects of a reduced KR relative frequency used global performance measures, thus confounding errors in the GMP and in parameterization. Our present results suggest that these effects were probably due to enhanced GMP learning. There are several possible reasons for the degrading effects of a frequent KR. For example, frequent feedback might prevent subjects from paying attention to their intrinsic (visual, acoustic, kinesthetic) feedback; to develop the capability to detect and correct errors (in no-KR retention and transfer) themselves, however, subjects must learn to interpret their intrinsic feedback. Also,frequent feedback could over-facilitate the planning of the next movement, thus preventing subjects from engaging in self-generated retrieval operations (e.g., Schmidt & Bjork, 1992; Wulf, 1991; Wulf & Schmidt, 1994). Finally, frequent feedback has been shown to produce excessive response variability during practice (e.g., Nicholson, 1992; Wulf & Schmidt, 1994), which could prevent the development of a stable movement representaiion (motor program). In the present study, 100% KR about relative timing (Groups 100-100, 100-50) also tended to produce higher variable error (W )in relative timing than 50% KR (Groups 50-100, -50); this effect failed to reach statistical significance, however, F(1, 68) = 2.9, p = .lo. With respect to parameterization learning, the effects of the relative KR frequency were different for retention (where the three practiced task versions had to be pro367


G.Wulf, T. D. Lee, & R. A. Schmidt duced) and transfer (where a novel total MT was required). Whereas all four groups showed almost identical performance in delayed retention, in delayed transfer the groups that had received 100% KR about absolute timing during practice were clearly more accurate in parameterizing the novel response than the groups that had received 50% KR. Thus, even though the KR frequency manipulation had no effect on parameterization in retention, more frequent KR was clearly more effective for parameterization when transfer to a novel movement version was required. The results from several experiments that have attempted to separate GMP and parameter learning are somewhat inconsistent with regard to the effects of a reduced KR frequency on parameter learning. Similar to Wulf et al. (1993), Winstein ( I 988, Experiment 2) found a trend for 100% KR tq enhance parameter learning, compared with 50% KR. However, reanalyses of the data in Wulf and Schmidt (1989) showed no effect of increased KR on CE measures of time-parameterization learning. Also, Wulf (1992) found no facilitation of amplitudeparameter learning by more frequent KR. Overall, even though a reduced KR frequency does not always degrade parameter learning reliably, there was never a case in these experiments where parameterization learning was enhanced by less frequent KR. Thus, these results are in line with the schema theory (Schmidt, 1975) assumption that motor schema development is a function of the number of times movement parameters and movement outcomes are being paired. According to schema theory, trials without KR (about parameterization) are neutral with regard to schema formation. Thus, reducing the relative frequency of KR should not enhance, and might even degrade, the capability to parameterize effectively. It may not be correct in general, however, that reduced feedback frequency never enhances parameterization learning. A recent unpublished study of ours (Wulf & Schmidt, 1992) showed gains in parameterization learning with reduced feedback. In that experiment, we used a task in which GMP learning was deliberately minimized by employing a pattern that had already been acquired by the start of the experiment, so that parameterization learning was probably the major accomplishment. One idea to account for this discrepancy is the notion that reduced feedback frequency operates to facilitate the learning of whatever requires the greatest improvement to meet the task goal. If the GMP is not well learned, reduced feedback frequency facilitates the learning of a new pattern. But if the GMP is well learned and/or the pattern is so simple that it can be approximated on the first trial, then reduced feedback frequency might facilitate the learning of parameterization. This set of findings, of course, cannot be predicted from schema abstraction theories (e.g., Schmidt, 1975), and would require completely new accounts of the learning process. The results of the present study, in our view, provide additional evidence for an empirical dissociation of the 368

theoretical constructs of GMP and parameterization processes (Schmidt, 1975, 1985, 1988). A major source of evidence for the separation of these two kinds of movement proficiency has been demonstrations that movement patterns can be scaled across various surface features (e.g., absolute time, absolute amplitude) although the fundamental movement structure remains essentially invariant (Schmidt, 1985; see also Gentner, 1987, for limitations of this view, and Heuer, 1988, for a rebuttal). Converging evidence involves the decrease in reaction time (RT) in choosing between left- and righthand actions that have the same versus different sequencing; same-sequencing choices presumably involve the same GMP, requiring less time for response organization than if sequencing is different (see Rosenbaum, 1991). Nevertheless, the supposition that GMP and parameterization processes were actually separable and distinct has often been viewed somewhat skeptically (e.g., Heuer & Schmidt, 1988). However, showing that the same experimental variable (i.e., reduced KR frequency) facilitates GMP learning, but has no effect on, or degrades parameter learning at the same time, provides additional evidence for the separation of these theoretical constructs, as well as for the psychological validity of GMP and parameterization processes. There are additional grounds for thinking that GMP and parameter processes are separable, such as their relative stability across a retention interval. Relative timing performances ( A E [prop.]), as opposed to absolute timing performances, were generally comparable for retention and transfer. In addition, errors in relative timing showed no decrement from Day 1 to Day 2, Fs( 1, 68) < 1, contrary to absolute timing performance, F(1,68) = 8.3, MSe = 81,974.1,~< .01, andF(1,68) = 4.8, MSe = 105,937.1, p < .05, for retention and transfer, respectively. That is, relative timing was fairly stable across the different task version, and did not suffer a retention loss over the delayed tests-contrary to absolute timing (see also Swinnen, Walter, Pauwels, Meugens, & Beirinckx, 1990; Wulf & Schmidt, 1988). These different characteristics of relative and absolute timing provide additional support for the distinction between GMP and parameterization processes. A GMP is supposed to be a highly stable memory structure, whereas parameterization apparently is not. Also, the same GMP is assumed to be used for all task versions, whereas the movement parameters are different. This is consistent with the notion that the group effects for the GMP are stable across task versions and days, whereas the effects for parameterization are not. In conclusion, the present results were clear in showing that GMP learning was enhanced by a 50% KR frequency (with regard to relative timing), compared with 100% KR. In contrast, the reduced relative KR frequency (with regard to absolute timing) degraded the capability to parameterize effectively when novel task performance was required. This finding is consistent with Journal of Motor Behavior

Reducing KR About Relative Versus Absolute Timing

the separation of GMP and parameterization processes long postulated by GMP and schema theories.


REFERENCES

Adams, J. A. (1971). A closed-loop theory of motor learning. Journal of Motor Behavior, 3, 111-150. Bilodeau, E. A., Bilodeau, I. M., & Schumsky, D. A. (1959). Some effects of introducing and withdrawing knowledge of results early and late in practice. Journal of Experimental Psychology, 58, 142-144. Gentner, D. R. (1987). Timing of skilled motor performance: Tests 0: the proportional duration model. Psychological Review, 94, 255-276. Heuer, H. (1988). Testing the invariance of relative timing: Comments on Gentner (1987). Psychological Review, 95, 552-557. Heuer, H., & Schmidt, R. A. (1988). Transfer of learning among motor patterns with different relative timing. Journal of Experimental Psychology: Human Perception and Performance, 14, 241-252. Lee, T. D., & Carnahan, H. (1990). Bandwidth knowledge of results and motor learning: More than just a relative frequency effect. Quarterly Journal of Experimental Psychology, 42A, 777-789. Newell, K. M., Carlton, L. G., & Carlton, M. J. (1982). The relationship of impulse to timing error. Journal of Motor Behavior, 14, 24-25. Nicholson, D. E. (1992). Information feedback disrupts performance stability. Unpublished doctoral dissertation, University of California, Los Angeles. Nicholson, D. E., & Schmidt, R. A. (1991). Scheduling information feedback to enhance training effectiveness. Proceedings of the Hwnan Factors Society 35th Annual Meeting (pp. 1400-1403). Santa Monica, CA. Human Factors Society. Rosenbaum, D. A. (1991). Human motor control. San Diego, CA: Academic Press. Salmoni, A. W., Schmidt, R. A., & Walter, C. B. (1984). Knowledge of results and motor learning: A review and critical appraisal. Psychological Bulletin, 95, 355-386. Schmidt, R. A. (1975). A schema theory of discrete motor skill learning. Psychological Review, 82, 225-260. Schmidt, R.A. (1985). The search for invariance in skilled movement behavior. Research Quarterly for Exercise and Sport, 56, 188-200. Schmidt, R. A. (1988). Motor control and learning: A behavioral emphasis (2nd ed.). Champaign, I L Human Kinetics. Schmidt, R. A. (1991). Frequent augmented feedback can degrade learning: Evidence and interpretations. In J. Requin & G. E. Stelmach (Ma), Tutorials in motor neuroscience (pp. 59-75). Dordrecht, The Netherlands: Kluwer. Schmidt, R. A., & Bjork, R. A. (1992). New conceptualizations of practice: Common principles in three paradigms suggest new concepts for training. Psychological Science, 3, 207-2 17. Schmidt, R. A., Lange, C., & Young, D. E. (1990). Optimizing summary knowledge of results for skill learning. Human Movement Science, 9, 325-348. Schmidt, R. A., Young, D. E., Swinnen, S.,& Shapiro, D. E. (1989). Summary knowledge of results for skill acquisition: Support for the guidance hypothesis. Journal of Experimen-

December 1994, Vol. 26, No. 4

tal Psychology: Learning, Memory, and Cognition, 15, 352-359. Schmidt, R. A., Zelaznik, H. N., Hawkins, B., Frank, J. S.,& Q u i ~ J., T. (1979). Motor output variability: A theory for the accuracy of rapid motor acts. Psychological Review, 86. 415-451. Schutz, R. (1977). Absolute, constant, and variable error: Problems and solutions. In D. Mood (Ed.), The measurement of change in physical education (pp. 82-100). Boulder, C O University of Colorado Press. Sherwood, D. E. (1986). Impulse characteristicsof rapid movement: Implications for impulse-variabilitymodels. Journal of Motor Behavior, IS, 188-214. Sherwood, D. E. (1988). Effects of bandwidth knowledge of results on movement consistency. Perceptual and Motor Skills, 66, 535-542. Swinnen, S.P., Walter, C. B., Pauwels, J. M., Meugens, P.F., & Beirinckx, M. B. (1990). The dissociation of interlimb constraints. Human Performance, 3, 187-215. Thorndike, E. L. (1927). The law of effect. American Journal O f PSychology, 39, 212-222. Winstein, C. J. (1988). Relativefiequency of informationfeedback in motor performance and learning. Unpublished doctoral dissertation, University of California, Los Angeles. Winstein, C. J., & Schmidt, R. A. (1990). Reduced frequency of knowledge of results enhances motor skill learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 677691. Wulf, G. (1991). A guidance explanation for contextual interference effects. Abstract Proceedings of the North American Society for the Psychology of Sport and Physical Activity (NASPSPA) (p. 129), Asilomar, CA. Wulf, G. (1992). The learning of generalized motor programs and motor schemata: Effects of KR relative frequency and contextual interference. Journal of Human Movement Studies, 23, 53-76. Wulf, G., & Schmidt, R. A. (1988). Variability in practice: Facilitation in retention and transfer through schema formation or context effects? Journal of Motor Behavior, 20, 133-149. Wulf, G.,& Schmidt, R. A. (1989). The learning of generalized motor programs: Reducing the relative frequency of knowledge of results enhances memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 748757. Wulf, G., & Schmidt, R. A. (1992). The learning of motorprograms and parameterization in simple reversal responses. Unpublished manuscript, Max-Planck Institute for Psychological Research, Munich, Germany. Wulf, G., & Schmidt, R. A. (1W). Feedback-induced variability and the learning of generalized motor programs. Journal of Motor Behavior, 26, 348-361. Wulf, G., Schmidt, R. A., & hubel, H. (1993). Reduced feedback frequency enhances generalized motor program learning but not parameterization learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 1134-1150. Young, D. E., & Schmidt, R. A. (1992). Augmented kinematic feedback for motor learning. Journal of Motor Behavior, 24, 261-273. Submitted July 27, 1992 Revised March 3, I993 Second revision July 2, I993

369