AFTER HUMANS LEARN a unimanual motor skill, they

437 Training With an Upper-Limb Prosthetic Simulator to Enhance Transfer of Skill Across Limbs Douglas L. Weeks, PhD, Stephen A. Wallace, PhD, David I. Anderson, PhD ABSTRACT. Weeks DL, Wallace SA, Anderson DI. Training with an upper-limb prosthetic simulator to enhance transfer of skill across limbs. Arch Phys Med Rehabil 2003;84:437-43. Objective: To examine what effect bilateral transfer of movement across limbs may have in a person s ability to learn use of an upper-limb prosthetic simulator. Design: Randomized trial. Setting: University laboratory. Participants: Able-bodied subjects randomly assigned to 1 of 3 groups. Interventions: Subjects performed 3 different tasks that required manipulation of objects with the simulator. Group 1 practiced with the simulator on the preferred limb and then transferred it to the nonpreferred limb; group 2 practiced with the simulator on the nonpreferred limb and then transferred it to the preferred limb; group 3 was a control group. Groups 1 and 2 underwent pretest trials, acquisition practice, posttest trials, and a 24-hour retention test; the control group followed the same design with the exception of acquisition practice. Main Outcome Measures: Elapsed time (1) from a signal to move until movement began (initiation time) and (2) from the beginning of movement to task completion (movement time). Results: Compared with the controls, groups 1 and 2 significantly reduced initiation time across all tasks from pretest to posttest (P.003) and from pretest to retention test (P.029). Groups 1 and 2 did not differ from each other. Movement time did not differ among the groups in the posttest. However, groups 1 and 2 significantly (P.026) reduced movement time across all tasks from pretest to retention test compared with the control group. Groups 1 and 2 did not differ from each other. Conclusion: The effects of bilateral transfer were evident for initiation time immediately on transfer, and this learning effect persisted to the retention test. The ability to execute movement, represented by movement time, occurred after consolidation of learning was complete. Cross-limb training with a prosthetic simulator may be useful for persons with recent unilateral upper-extremity amputation who are learning to use a prosthesis. Key Words: Amputation; Psychomotor performance; Rehabilitation. 2003 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation From the Department of Physical Therapy, Regis University, Denver, CO (Weeks); and the Department of Kinesiology, San Francisco State University, San Francisco, CA (Wallace, Anderson). Supported by the National Institute on Disability and Rehabilitation Research (grant no. H1336000024-01), the National Center for Research Resources, and the Office of Research on Minority Health, National Institutes of Health (grant no. 5 P20 RR11805). No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated. Reprint requests to Douglas L. Weeks, PhD, Dept of Physical Therapy, Regis University, 3333 Regis Blvd, Denver, CO 80221, e-mail: dweeks@regis.edu. 0003-9993/03/8403-7122$30.00/0 doi:10.1053/apmr.2003.50014 AFTER HUMANS LEARN a unimanual motor skill, they typically perform it with their preferred limb. However, situations may arise in which the nonpreferred limb may have to perform the task. The proficiency with which the nonpreferred limb performs the task gives rise to this question: Is the functional competency that is established through unimanual practice specific to the limb that was trained, or can the ability to produce the learned response be generalized so that the unpracticed contralateral limb is able to produce the response proficiently? The latter possibility has long been supported by research showing that performance by an unpracticed limb improves after the contralateral limb has acquired the action. Improvement in the unpracticed limb after practice with the contralateral limb has been termed bilateral transfer or intermanual transfer of skill. 1 Typically, researchers studying bilateral transfer assess pretest to posttest gains in an unpracticed limb as a function of intervening practice with the opposite limb. In laboratory experiments, bilateral transfer has been reported to occur in tasks such as mirror tracing, 1 normal and mirror-image writing, 2 rapid continuous tapping, 3 continuous rotary pursuit tracking, 4,5 drawing of nonmeaningful figures, 6 anticipatory timing, 7 and linear positioning capability. 7 Transfer of functional competence across limbs has implications for rehabilitation in persons whose unilateral upperextremity amputation leaves their preferred limb unavailable to perform a well-learned unimanual task. In these people, rehabilitation typically centers on training with a prosthetic device for the residual limb. Before a patient begins to use the prosthesis, or as an adjunct to practice with the prosthesis, he/she may be fitted with a prosthetic simulator on the intact limb. Because the simulator mimics the control functions of the actual prosthetic device, the person with amputation has the opportunity to gain functional competency that may be transferred to the residual limb. Practice soon after amputation may be designed to promote bilateral transfer to the residual limb when use of the actual prosthetic device is initiated. Beginning prosthetic training as soon as possible after amputation has been suggested as a means to enhance acceptance of the device and to improve the probability that the prosthetic device will be used skillfully in daily functional use. 8,9 A question examined in previous investigations is whether the amount of transfer is greater from the preferred side to the nonpreferred side or vice versa. Although it is possible for the amount of transfer to be symmetric (ie, not favoring 1 direction over the other), most investigators examining direction of transfer have reported asymmetrical transfer. Several researchers 2,10 have reported the amount of transfer to be greater in the nonpreferred to preferred direction, whereas others 3,7,11 have reported an advantage for the preferred to nonpreferred direction. It seems that when the task was relatively simple or not completely novel (ie, a derivation of the task was previously learned), greater transfer of proficiency occurred from the nonpreferred to the preferred limb. 2,10 However, when the task was complex and novel, it seems that transfer was greater from the preferred to nonpreferred limb. 3,7,11 This study examined bilateral transfer for several complex tasks. In it, subjects learned to perform prehension skills with

438 TRAINING WITH A PROSTHETIC SIMULATOR, Weeks a simulator that mimicked an upper-extremity prosthesis. Subjects performed 3 tasks in a 3-dimensional workspace. The tasks we selected, requiring manipulation of a variety of objects, were not an exhaustive set of prehension skills, but they were representative functional skills that would enable us to examine the hypotheses. Subjects were told to perform as quickly as possible while achieving the goal of the task. We had 3 outcome measures: (1) the time from movement initiation to task achievement (movement time), (2) time to get a movement under way (initiation time), and (3) the frequency of errors in performing the tasks (error rate). Measure 2 was recorded so we could determine whether reduced initiation time was a function of practice or transfer. Because of the tenuous nature of the question about transfer symmetry, we investigated both directions of transfer: preferred to nonpreferred limb and nonpreferred to preferred. Clarification of the symmetry question has practical consequences. Understanding whether transfer is symmetrical or asymmetrical enables practitioners to set realistic rehabilitation goals according to whether the amputated limb is the patient s preferred or nonpreferred limb. To ensure an adequate sample size, individuals without amputations were used as subjects. To assess the amount of skill that is retained, 12 studies of motor learning commonly use a retention test that is separated in time from the original practice phase. In this study, subjects performed a retention test the day after their initial practice to determine the amount of bilateral transfer they retained after they practiced with their contralateral arm. We hypothesized that subjects would show a positive performance gain when using the simulator on the unpracticed limb after having extensive practice with the simulator on the opposite limb. This gain would be evident in initiation time, movement time, and error rate in an immediate posttest exercise and on the retention test. Because the task of learning to control the simulator was complex and novel, we also hypothesized that transfer would be greater in the preferred to nonpreferred direction. METHODS Participants Forty-two right-handed adults (21 men: mean age, 22.4y; 21 women: mean age, 22.7y) volunteered to participate. All reported freedom from known neurologic or upper-extremity musculoskeletal problems that might influence performance. Subjects were randomly assigned to 1 of 3 groups, each composed of 7 men and 7 women. Group 1, the nonpreferred transfer group, practiced with the preferred limb before they performed with the nonpreferred limb. Group 2, the preferred transfer group, practiced with the nonpreferred limb before performing with the preferred limb. Group 3 was a control group that performed a pretest and then posttest protocol without intervening practice between tests. All subjects signed an informed consent document before participation. The consent form and experimental procedures were approved by the local institutional review board. Apparatus The prosthetic simulator was designed to mimic a prosthetic device for an upper-extremity amputation that saved enough of the residual limb to permit forearm supination and pronation (fig 1). The simulator featured a figure-8 harness fitted around the shoulder contralateral to the prosthesis. The harness was attached to a cable that ran across the back and across the upper arm of the limb on which the prosthesis was worn. The cable Fig 1. The prosthetic simulator donned by a subject. inserted into the proximal portion of the simulator and ran the length of the simulator to interface with a split-hook (5XA) terminal device identical to that of a regular prosthesis. The simulators for right and left arms were identical. With the simulator placed atop a table in a position that permitted the anterior aspect of the forearm to be in contact with the tabletop, the terminal device was oriented in the vertical plane with its tips pointing downward. The terminal device was a voluntaryopening device; that is, subjects opened it by adjusting the tension of the cable with relative motions of the torso, shoulders, and arm that was fit with the simulator. A lightweight fiberglass molded sleeve represented the forearm portion of the prosthesis. To don the simulator, the subject first placed the contralateral limb through the harness and then slipped the forearm into the fiberglass sleeve. This configuration enabled the subject s hand to grasp a small handle just proximal to the terminal device. Thus, the device resembled, as closely as possible, an actual prosthesis in both fit and function. A task board enabled the subjects to perform a variety of prehension tasks with the simulator. Each task was initiated from a common start position, that is, at a microswitch button located at midline and 20cm from the near edge of a table at which subjects were seated (fig 1). Before starting a task, subjects depressed the button with the simulator and were prompted to initiate a trial with an auditory tone. Three tasks (toggle, aiming, prehension) were studied that required manipulation of various objects at different points in the workspace. The toggle-switch task (fig 2) required a 25-cm forward and 20-cm upward movement to grasp and flip the paddle of a small toggle switch upward. Subjects were informed that they were to grasp the paddle with the terminal device, not just flip it without a grasp. To successfully grasp the paddle with the terminal device, the subject had to perform a pronation movement of the simulator. The fine-aiming task (fig 3) required a 20-cm forward movement to place a metallic stylus held in the terminal device into a hole that was.80cm in diameter and flush with the surface of the task board. The stylus was.70cm in diameter and 10cm in length. The experimenter placed the stylus into the terminal device before the trial started so that 7cm of the stylus extended downward from the terminal device. When the stylus was in place in the hole, subjects pressed it downward to activate a microswitch located 2cm under the surface of the task board. This task required precision manipulation of a small object. A prehension task (fig

TRAINING WITH A PROSTHETIC SIMULATOR, Weeks 439 Fig 2. Toggle-switch task, with the subject pressing the switch. 4) required the subject to reach 20cm laterally and 10cm forward from the start position to grasp a 200-g metallic cylinder that was 4.2cm high and 2.8cm in diameter. Once grasped, the object was transported to the opposite side of the task board and placed within a 3.5-cm-diameter target well with a lip 1.5cm high. The subject used the cylinder to depress a.50-cm button centered within the well, thereby triggering a microswitch mounted underneath the target well. The cylinder was covered with fine-grain sandpaper (180 grit) to increase the friction between the terminal device and the cylinder. This task required great control of the simulator to grip the cylinder combined with a large trajectory movement to transport the object. The start button and the microswitches associated with the 3 tasks were interfaced with millisecond clocks so that the initiation time and movement time were determined. Initiation time was an indicator of time to perceive the signal and then initiate a response, whereas movement time was an indicator of movement execution proficiency. Procedure Before being seated at the task table, subjects were provided with assistance donning the simulator on the arm that would Fig 4. Prehension task. (A) Subject is in the start position. (B) Trial has begun, and subject has reached laterally to grasp the cylinder. (C) Subject has transported the cylinder and placed it into the target well. Fig 3. Aiming task, with the subject placing the dowel in the target hole. perform the bulk of practice (preferred or nonpreferred, depending on group). While standing, subjects viewed a model on videotape wearing the simulator. The model showed several basic control motions used to operate the cable to open and close the terminal device. The demonstrated motions included humeral flexion combined with an elbow extension motion, a bilateral shoulder-shrug motion, and a pronation-supination motion. Subjects passively watched the model perform the motions once, then watched the model a second time while

440 TRAINING WITH A PROSTHETIC SIMULATOR, Weeks Dependent Measure Table 1: Means for Errors, Initiation Time, and Movement Time for Each Task per Selected Phase Phase Task Toggle Aiming Prehension Initiation time (ms) Pretest 399 82 413 140 450 138 Movement time (ms) Acquisition 370 116 360 104 392 91 Movement time (ms) Pretest 1922 500 1595 492 3869 902 Movement time (ms) Acquisition 1473 260 1347 241 2927 296 Movement time (% change) Posttest 20 15 12 21 24 13 Movement time (% change) Retention test 25 23 17 17 27 13 Error frequency Pretest.17.37.64 1.14.74 1.34 Error frequency Retention test.12.33.36.93.48.97 NOTE. Values are mean standard deviation (SD). Values are collapsed across groups. concurrently imitating the motions with the model. The simulator was then switched to the opposite arm, with practice of the control motions disallowed. The subject was seated at the task table and instructed to maintain an upright seated posture and to avoid awkward body motions to control the device. Each subject performed each task, with task order counterbalanced within groups. The experimenter verbally explained each task before the subject s first trial of that task. For all tasks, subjects were told to perform as quickly as possible while achieving the goal of the task. No specific instructions were given to react to the start tone as soon as possible, even though the influence of bilateral transfer on initiation time was assessed. The testing protocol for each task for groups 1 and 2 was as follows: (1) pretest trial: subjects performed 5 pretest trials with the transfer arm (preferred or nonpreferred, depending on group); (2) acquisition trial: after completing the pretest maneuvers, subjects donned the simulator on the other arm and practiced each task 30 times; and (3) posttest trial: subjects then donned the simulator on the transfer arm and performed 5 more trials with that arm. In the control group, half the subjects performed the tasks with the right arm and half with the left arm. All completed the pretest and posttest trials, and in lieu of the acquisition trial, the control group viewed a 15-minute video about American history in the 1950s. To determine the degree of skill retention, subjects in each group returned the next day to perform 5 more trials. In this retention test, subjects undertook each task with the arm that was used in the previous day s pretest trial. Any trial in which the subject did not achieve the goal of the task was called an error trial. Subjects repeated these trials until they achieved a full complement of successful trials. The frequency of error trials was recorded for each task. For example, on the aiming task, if the stylus was dropped or if it slipped within the terminal device, the trial was considered an error and was repeated. In the toggle task, if the terminal device slipped off the paddle before the switch was thrown, the trial was considered an error and was repeated. In the prehension task, if the subject knocked over the cylinder during grasping or dropped it during transport, the trial was considered an error and was repeated. Data Analysis For each task, initiation time and movement time for each trial were blocked into groups of 5, with means calculated for each 5-trial block. Likewise, the average number of errors per 5-trial block was calculated. We conducted separate analyses on block means for initiation time, movement time, and errors in each phase. Block scores for the pretest were subjected to a 3 2 3 analysis of variance (ANOVA), that is, group (nonpreferred, preferred, control) by gender by task (toggle, aiming, prehension), with repeated measures on task. To assess bilateral transfer, we calculated the percentage change in block scores for initiation time and movement time in the posttest trial and retention test relative to the pretest block scores. Percentage of change scores were subjected to 3 2 3 ANO- VAs (group by gender by task). Finally, to determine whether groups 1 and 2 performed similarly during skill acquisition, block scores for initiation time, movement time, and errors were subjected to a 2 2 6 3 ANOVA (group by gender by block by task), with repeated measures on block and task. In all analyses, main effects were further analyzed with Newman- Keuls post hoc tests. Significant interactions were further analyzed with simple main effect tests. All analyses assumed a type I error rate of P.05. RESULTS Initiation Time Analyses The ANOVA for initiation time in the pretest indicated no significant differences among tasks (table 1), between genders, or among the 3 groups (table 2). No interactions were significant. Thus, random assignment initially equated groups relative to initiation time. The ANOVA for percentage change in initiation time from pretest to posttest trials indicated no significant differences between genders or among tasks. However, a group main effect was detected (F 2,36 6.923, P.003). Newman-Keuls analyses showed that the reduction in initiation time for groups 1 and 2 was significantly greater than for the control group (table 2). Groups 1 and 2 did not differ from each other. As was the case for the posttest trials, the ANOVA on percentage of change in initiation time from pretest trial to retention test indicated no significant differences between genders or among tasks. However, a group main effect was once again detected (F 2,36 3.918, P.029). Newman-Keuls analyses again showed a significantly greater reduction in initiation time for groups 1 and 2 than for the control group (table 2). Groups 1 and 2 did not differ from each other. Thus, the bilateral transfer benefit for initiation time was retained after the initial practice phase. The ANOVA comparing acquisition initiation time for groups 1 and 2 indicated no significant differences for the groups or genders or across blocks. A significant task main effect was detected (F 2,48 3.262, P.047). Newman-Keuls analyses indicated initiation time was significantly longer for the prehension task than for the aiming task (table 1). Movement Time Analyses The ANOVA for movement time in the pretest trials indicated no significant differences among groups (table 2). How-

TRAINING WITH A PROSTHETIC SIMULATOR, Weeks 441 Dependent Measure Table 2: Means for Initiation Time and Movement Time for Each Group per Phase Phase Group Control Group 1 Group 2 Initiation time (ms) Pretest 419 117 427 99 417 143 Movement time (ms) Pretest 2413 529 2580 666 2393 680 Initiation time (% change) Posttest 0.25 29* 12 15 16 14 Initiation time (% change) Retention 2 25* 10 20 15 16 Movement time (% change) Posttest 16 13 19 16 21 18 Movement time (% change) Retention 16 20* 26 14 26 15 NOTE. Values are mean SD. For percentage of change means, positive values indicate a net reduction in time (ie, improvement) relative to pretest means. *Significant difference compared with other levels within the variable. ever, a significant (F 1,36 13.106, P.001) main effect was detected for gender. The mean movement time for men (2204ms) was significantly faster than for women (2720ms). A significant (F 2,72 271.48, P.001) main effect was also shown for task. Newman-Keuls analysis on task means indicated that movement time for each task differed significantly from that for the other tasks (table 1). Thus, random assignment initially equated groups relative to movement time, even though a gender difference existed. Inconsequentially, the time to perform the various tasks also varied. The ANOVA for percentage of change in movement time from pretest to posttest trials indicated no significant differences among the 3 groups (table 1). However, we detected a significant (F 1,36 4.937, P.033) main effect for gender. The mean percentage change in movement time for women (22%) was significantly greater than for men (15%). A significant (F 12,72 6.529, P.002) main effect for task was also shown. Newman-Keuls analysis on task means indicated that the percentage reduction in movement time was significantly greater for the toggle and prehension tasks than for the aiming task (table 1). The ANOVA on percentage of change in movement time from pretest to retention test indicated significant differences among the 3 groups (F 2,36 4.05, P.026). Newman-Keuls analysis on task means indicated that the percentage of reduction in movement time was significantly greater for both transfer groups than for the control group (table 2). Transfer groups did not differ. A significant (F 1,36 8.952, P.005) main effect for gender was detected. The mean percentage of change in movement time for women (28%) was significantly greater than for men (18%). We also found a significant (F 2,72 3.316, P.042) main effect for task. Newman-Keuls analysis on task means indicated that the percentage of reduction in movement time was significantly greater for the prehension task than for the aiming task (table 1). The ANOVA comparing acquisition movement times for the 2 transfer groups indicated no significant differences among the groups. A significant (F 1,24 4.674, P.041) main effect was detected for gender. The mean movement time for women (2035ms) was significantly longer than for men (1796ms). The block main effect was significant (F 5,120 7.41, P.001). Means for blocks 1 to 6 were 2189, 1928, 1853, 1815, 1838, and 1820ms, respectively. The mean movement time for block 1 was significantly longer than the means for all other blocks. The mean movement time for block 2 was significantly longer than the mean movement time for blocks 4, 5, and 6. Means for blocks 3, 4, 5, and 6 did not differ significantly. Trend analysis indicated a significant (t 3 4.66, P.016, R 2.935) quadratic trend for block means. Also, we detected a significant (F 2,48 256.91, P.001) main effect for task in the acquisition movement time data for the transfer groups. Newman-Keuls analyses indicated that movement time was significantly longer for the prehension task than for the aiming or toggle tasks (table 1). Thus, although groups 1 and 2 did not differ in performance across blocks, women performed more slowly as a group than men. Furthermore, the block main effect indicated that performance improved as a function of practice. Analysis of Errors The ANOVA on the frequency of errors in the pretest trials indicated only a main effect for task (F 2,72 3.973, P.023). Newman-Keuls analyses showed that subjects committed significantly fewer errors in the toggle task than in the other 2 tasks (table 1). In the pretest trials, 9 of 14 subjects in the control group committed errors on at least 1 task, whereas 10 of 14 subjects in each of the transfer groups committed errors on at least 1 task. Thus, random assignment equated the initial performance of the groups relative to the frequency of errors. However, the tasks that involved transport of an object, as opposed to simple manipulation of a fixed object, resulted in more errors. The ANOVA for the frequency of errors between groups 1 and 2 in acquisition indicated no significant differences. The ANOVA on the frequency of errors in the posttest trials indicated that the task by group interaction was significant (F 4,72 3.628, P.009). Simple main effect tests showed that the group that transferred to their preferred arm committed significantly more errors in the aiming task (mean errors,.64) than the group that transferred to their nonpreferred arm (mean errors,.07) in the posttest trials. The preferred group also committed significantly more errors in the aiming task (mean errors,.64) than in the toggle task (mean errors, 0). Thus, transfer from a nonpreferred to preferred direction was problematic for performing the aiming task with precision. In contrast, transferring from a preferred to nonpreferred direction resulted in accurate performance on all 3 tasks. The ANOVA for the frequency of errors in the retention test indicated no significant difference in error frequency among the 3 groups or between genders. Five of 14 subjects in the control group and group 1 committed errors on at least 1 task, whereas 6 of 14 subjects in group 2 committed errors on at least 1 task. A task main effect was detected (F 2,72 3.454, P.037). Newman-Keuls analyses showed that subjects committed significantly fewer errors in the toggle task than in the prehension task (table 1). DISCUSSION As hypothesized, practice with the contralateral arm, whether preferred or nonpreferred, positively benefited bilateral transfer with respect to the time used to initiate the move-

442 TRAINING WITH A PROSTHETIC SIMULATOR, Weeks ments. Both transfer groups significantly reduced initiation time from pretest to posttest trials and from pretest to retention test compared with the control group. Because the subjects who received substantial practice did not reduce movement time more appreciably than the control group on transfer in the posttest, we deduced that an immediate bilateral transfer benefit was absent. However, groups 1 and 2 performed significantly more rapidly than the control group in the retention tests, indicating that the bilateral transfer effect for movement time emerged a relatively long time after initial transfer. This finding supports the idea that movement execution in bilateral transfer may not be complete until sufficient time has passed for changes within the nervous system to occur. This plasticity phenomenon is referred to as consolidation of memory. 13 Experimental attempts to estimate the amount of time required for consolidation to occur suggest that it is complete within 4 to 6 hours of the conclusion of practice on a task to be learned. 13 Consolidation for the ability to get the movement under way may have been complete at the posttest trial, as evidenced by initiation time results. However, once the movement was initiated, the ability to guide it in an efficient manner by using response-produced feedback did not show a benefit immediately on transfer. Instead, transfer of ability to use response-produced feedback may not occur until the subject has made the necessary neural connections, that is, until the cortical areas responsible for making motor adjustments on the basis of sensory information are completely modified. This delayed bilateral transfer effect for movement execution has clinical implications: transfer benefits should not be expected immediately on transfer but instead may become evident later, after a period of no practice. The clinician must be cautious when inferring learning at the end of a bout of practice because the full effects of consolidation, and the benefits to performance that consolidation affords, will not be realized immediately. Women had a greater degree of bilateral transfer than men with respect to movement time in the posttest trial and the retention test. The greater percentage of improvement in women is important clinically. Men, who may initially move more rapidly than their female counterparts, may have a smaller window for improvement, and failure to see substantial improvement may be discouraging. The clinician should be aware of the need to encourage extended practice even when current performance seems to have reached a plateau. Men may also have to engage in more practice than women to maximize bilateral transfer. With respect to the temporal qualities of bilateral transfer, acquiring functional competence with a prosthetic device to transfer this competence to the opposite limb does not seem to favor one direction of transfer over the other. The lack of a direction-of-transfer effect may lie in the role of the opposite arm and shoulder in controlling the terminal device. Whether the preferred arm was the simulator arm, or the arm contributing to cable tension through the harness, it actively contributed to skill production. The bilateral contribution to skillful use of the simulator may have nullified a direction-of-transfer effect. This finding has implications for the rehabilitation environment: transfer of learning with respect to improvement of temporal qualities of the movement may not be differentially affected by whether the patient s preferred or nonpreferred arm is used to build initial proficiency with the simulator. The prosthetic simulator may be equally effective for influencing a patient s transfer of skill to the residual limb, whether or not his/her preferred or nonpreferred arm was amputated. Movement accuracy, however, was influenced by direction of transfer in the posttest trial for 1 task the aiming task. Subjects who transferred from a nonpreferred to preferred direction committed significantly more errors in performing this task than subjects who transferred in the opposite direction. In a rehabilitation environment, this may indicate that persons with a preferred-limb amputation may need relatively more practice performing fine-aiming tasks with a simulator on the nonpreferred limb. This practice will reduce errors on transfer and let the patient take full advantage of the bilateral transfer effect. Errors were not problematic for either direction of transfer in the tasks that required manipulation of a fixed object or grasp and transport with a large trajectory movement. Although precision of performance was not influenced by direction of transfer or by whether practice was minimal (as in the control group) or substantial (as in groups 1 and 2), the type of task did have a bearing on how frequently errors were committed in the retention trial. Errors were more frequent in the grasp and transport task. In a rehabilitation setting, day-today retention of accuracy may be slow to develop for these types of tasks. Better accuracy retention may be obtained by having the patient practice more object transport tasks than tasks in which they manipulate fixed objects. Of theoretical importance is this question: What was learned through practice that enabled bilateral transfer to occur? To move competently, the motor system must acquire the ability to generalize beyond the movement instances it experienced in structured practice. Because the ability to generalize skill to an unpracticed limb was evident, what was learned seemed to be independent of the practiced limb. Instead, the memory representation formed through practice may have contained strategy-related information and/or abstracted information about controlling the prosthetic device that was unrelated to the specific arm that was practiced. In this way, the unpracticed limb would not need to relearn this information to perform. Thus, movement parameters for an extensively practiced action may be stored independently of information about the effector typically used to perform the task. In this study, generalization of learning was quite flexible in that the bilateral transfer effect for movement execution and initiation did not depend on gender, type of task, or direction of transfer. CONCLUSION Unilateral training with a prosthetic simulator enhanced the ability of the unpracticed limb to operate the terminal device of a prosthesis in a temporally and spatially skillful manner. This finding has implications for rehabilitation of the individual with a unilateral upper-extremity amputation. The time required to learn how to use a prosthetic device may be decreased by using a bilateral transfer-training program. This type of training could allow a person with a recent amputation to practice functional skills with a simulator on the sound arm so that, once the individual is fit with the actual prosthetic device, he/she may expect partial proficiency because of skill transfer from previous practice with the simulator. Acknowledgments: We thank Robert Radocy, Lawrence Carlton, and Maurice LeBlanc for their help in designing and constructing the simulator. References 1. Cook TW. Studies in cross education. 1. Mirror tracing the starshaped maze. J Exp Psychol 1933;16:144-60. 2. Freeman GL. Studies in the psychophysiology of transfer: III. Bilateral practice effects in normal and mirror writing. J Psychol 1938;5:285-9. 3. Laszlo JI, Baguley RA, Bairstow PJ. Bilateral transfer in tapping skill in the absence of peripheral information. J Mot Behav 1970; 2:261-72.

TRAINING WITH A PROSTHETIC SIMULATOR, Weeks 443 4. Hicks RE, Gualtieri CT, Schroeder SR. Cognitive and motor components of bilateral transfer. Am J Psychol 1983;96:223-8. 5. Milisen R, Van Riper C. Differential transfer of training in a rotary activity. J Exp Psychol 1939;24:640-6. 6. Halsband U. Left hemisphere preponderance in trajectorial learning. Neuroreport 1992;3:397-400. 7. Teixeira LH. Timing and force components in bilateral transfer of learning. Brain Cogn 2000;44:455-69. 8. Pinzur MS, Angelats J, Light TR, Izuierdo R, Pluth T. Functional outcome following traumatic upper limb amputation and prosthetic limb fitting. J Hand Surg [Am] 1994;19:836-9. 9. Malone JM, Fleming LL, Roberson J, et al. Immediate, early, and late postsurgical management of upper-limb amputation. J Rehabil Res Dev 1984;21:33-41. 10. Ammons RB, Ammons CH. Bilateral transfer of rotary pursuit skill. Am Psychol 1951;6:294. 11. Parker-Taillon D, Kerr R. Manual asymmetries within the performance of a complex motor task. Hum Mov Sci 1989;8:33-44. 12. Schmidt RA, Lee TD. Motor control and learning: a behavioral emphasis. Champaign (IL): Human Kinetics; 1999. 13. Brashers-Krug T, Shadmehr R, Bizzi E. Consolidation in human motor memory. Nature 1996;382:252-5.