UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Changes in learned motor behavior due to the effects of various forms of augmented kinematic feedback Hale, Trevor A. 1999

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1999-0122.pdf [ 3.56MB ]
Metadata
JSON: 831-1.0077247.json
JSON-LD: 831-1.0077247-ld.json
RDF/XML (Pretty): 831-1.0077247-rdf.xml
RDF/JSON: 831-1.0077247-rdf.json
Turtle: 831-1.0077247-turtle.txt
N-Triples: 831-1.0077247-rdf-ntriples.txt
Original Record: 831-1.0077247-source.json
Full Text
831-1.0077247-fulltext.txt
Citation
831-1.0077247.ris

Full Text

CHANGES IN LEARNED MOTOR BEHAVIOR DUE TO THE EFFECTS OF VARIOUS FORMS OF AUGMENTED KINEMATIC FEEDBACK by TREVOR A . H A L E B.H.K., The University of British Columbia, 1995  A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE  THE FACULTY OF G R A D U A T E STUDIES School of Human Kinetics  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH C O L U M B I A January 1999 © Trevor A . Hale, 1999  In  presenting  degree freely  at  this  the  available  copying  of  department publication  of  in  partial  fulfilment  of  University  of  British  Columbia,  I agree  for  this or  thesis  reference  thesis by  this  for  his thesis  and study. scholarly  or for  her  The University of British Vancouver, Canada  DE-6 (2/88)  Columbia  purposes  gain  shall  requirements that  agree  may  representatives.  financial  permission.  I further  the  It not  be is  that  the  Library  permission  granted  by  understood be  for  allowed  an  advanced  shall for  the that  without  make  it  extensive  head  of  my  copying  or  my  written  ABSTRACT The present study was conducted to determine the relative effectiveness of four types of augmented feedback on the acquisition of a rapid aiming movement. Static graphic feedback depicting the primary submovement and error correction phases of the just performed movement (Static group) was compared to numeric KR (KR group). In addition, concurrent kinematic feedback (CD group) was compared to delayed kinematic feedback (DD group). It was predicted that delaying kinematic information would facilitate the performance of the DD group during no-KR retention trials, due to an increased processing of kinesthetic feedback during the movement and subsequent calibration of this information with the delayed visual feedback. It was found that participants in the CD group produced less absolute error, consumed a smaller proportion of TMT in the primary submovement phase, produced less variable primary submovement, and produced longer secondary submovements than participants in the other three groups. In addition they had significantly more trials containing an error correction phase than participants in the Static and DD groups. However, there was no statistical difference between the four groups in their overall Score, TMT, or constant error. These results indicated that participants in the CD group utilized a two movement strategy during Acquisition while those in the DD group did not. As participants in the DD group did not frequently use an error correction phase, it was concluded that delayed visual kinesthetic feedback was not used to calibrate kinesthetic feedback modalities. In Retention, significant condition by group interactions, on most measures, indicated that the removal feedback immediately and significantly affected the  ii  performance of the CD group, where as the performance of the KR, Static and DD groups deteriorated after the one week retention interval. It was concluded from these data that the delayed feedback did not induce the disrupting or guiding qualities that concurrent visual feedback did. In addition, these findings supported the Guidance Hypothesis (Salmoni etal., 1984). During both Acquisition and Retention, shifts in end position were found to be concomitant to shifts in start position. That is participants overshot the target when they began their movement with their elbow in the most extended position, and progressively shortened their movement as the start position moved closer to their mid-line. As participants were aiming to a common end point and not a series of targets, 45° from the start positions, it appeared that participants coded the movement endpoint. This effect was most pronounced for participants practicing with delayed feedback; as participants in the CD group accurately hit the target in acquisition by making visually based on-line corrections to the initial end-point code. In Retention, however, the CD group overshot the target and, like the other three groups, produced movements to a common end point. That participants were coding their movements to end at a single common end point rather than traverse a particular distance is well accounted for by the Equilibrium Point Hypothesis (Feldman, 1986). Indeed an equilibrium point interpretation was used to explain these findings.  iii  T A B L E OF CONTENTS  Abstract Table of contents List of Figures Acknowledgment Dedication  1.  Introduction  2.  Methods Participants Apparatus and Task Procedure and Experimental Design Analysis of Movement Dependent Measures Statistical Analysis  3.  Results Acquisition Performance measures Submovement analysis Retention  iv  Page  4.  Performance measures  33  Submovement analysis  37  Positional analysis  40  Acquisition  40  Retention  40  Discussion  46  On the comparison of Concurrent and Delayed forms of Feedback  47  On the difference between KP and KR  51  On amplitude versus endpoint coding (what is represented?)  53  Conclusion  58  References  61  Appendix A  64  Abstract  64  Introduction  65  Method  69  Results  72  Discussion  75  v  LIST OF FIGURES  Figure 1.  Page  Sample position, velocity, and acceleration profiles showing parsing of movements into their primary submovement and error correction phases. Movements may contain (a) no error corrections, or they may contain error corrections initiated either by (b) a change in direction, (c) a zero line crossing in acceleration, or (d) a significant deviation in the acceleration profile. Markers indicate the beginning of movement, the end of the primary submovement phase, and movement endpoint (PS -primary submovement phase, EC -error correction phase). (After Khan et al., 1998)  2.  16  Mean trial score as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  3.  20  Mean total movement time (in milliseconds) as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  4.  22  Mean absolute target accuracy (AE -in degrees) as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  23  vi  Figure 5.  Page  Mean target accuracy (CE -in degrees) as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  6.  24  Mean primary submovement duration (in milliseconds) as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  7.  26  Mean proportion of total movement time consumed by the primary submovement phase (as a percentage) of total movement time as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  8.  27  Mean primary submovement variability as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  9.  29  Mean proportion of total movement distance consumed by the error correction phase (as a percentage) of total movement distance as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  31  vii  Figure 10.  Page  Mean duration of error correction phase as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  11.  32  Mean number of trials (within a block of 50 trials) that contain an error correction phase as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention).  12.  34  Index of error correction effectiveness (as a percentage) as a function of Block and Condition. (A - Acquisition, B -Block, R i -Immediate Retention, R2 -1 week Delayed Retention).  13.  35  Analysis of mean target accuracy (CE -in degrees) as a function of start position. Each graph depicts CE as a function of Block and Condition. (A -Acquisition, B -Block, RI -Immediate Retention, R2 -1 week Delayed Retention). Position 1 represents the start position at which the participant's elbow was most extended (130° of elbow extension where 180° is considered full elbow extension). Amount of elbow extension decreases from position 1 to 5 (position 5 represents 110° elbow extension).  41  viii  Figure 14.  Page  Graphic representation of the positional effect bn target accuracy (CE -in degrees). Main effect of position. Position 1 represents the start position at which the elbow was most extended (130° of elbow extension where 180° is considered full elbow extension). The amount of elbow extension decreases from position 1 to 5 (position 5 represents 110° of elbow flexion).  15.  42  (a) Main effect of start position and group for all Acquisition trials, (b) Main effect of start position and group for all Retention trials (Immediate and Delayed Retention). Position 1 represents the start position at which the elbow is most extended (130° of elbow extension where 180° is considered full elbow extension). Amount of elbow extension decreases from position 1 to 5 (position 5 represents 110° elbow flexion).  16.  43  Target accuracy (CE -in degrees) for each group as a function of condition (last block of 50 trials from Acquisition -ACQ, and the block of 50 trials in Immediate Retention -RI, and Delayed Retention -R2) and position. Position 1 represents the start position at which the elbow was most extended (130° of elbow extension where 180° is considered full elbow extension). Amount of elbow extension decreases from position 1 to 5 (position 5 represents 110° elbow flexion).  45  ix  Figure 17.  Page  Mean primary submovement time (A), proportion of total movement time spent in the primary submovement phase (B), and mean number of trials (per 20 trial block) containing a secondary submovement (C) as a function of acquisition, immediate (RI) and delayed Retention (R2) for the Static and Dynamic groups.  x  74  ACKNOWLEDGMENT  There are many people I would like to thank for their roles in helping me to complete this thesis. First my advisor -Dr. Ian Franks, for his interest and continual support in my endeavor as both an undergraduate and a graduate student. He introduced me to the field of motor control and learning and has since provided encouragement and insight into my work in addition to challenging me to pursue difficult yet achievable goals. I would also like to thank my other committee members: Dr. John Dickinson and Dr. Tim Lee for their constructive advice and participation in this project. Thanks to my fellow students Nicola Hodges, Mike Khan, Mike Garry, and Tim McGarry who's insight and advice have contributed to my knowledge, and the quality of my work; and who's friendship, and humor has made the journey much more enjoyable. In addition Paul Nagelkerke for his patience, humor, and invaluable contributions to my data collection and experimental setup. I am also grateful to all of the individuals who volunteered to participate in my research. I must also thank my immediate and extended family for their unconditional support in all my pursuits. To my mother Betty-Rae, for her calm; to my father Terry for his persistence (yes dad it is finished!); to my sister Nicole, for her kindred spirit; and all my aunts, uncles and cousins for my sense of family. To Monica for her love and support. Finally to all my close friends for everything they bring to my life.  xi  Dedicated to the memories of my Grandmothers Evelyn Hale and Margaret Rafter.  xii  INTRODUCTION  During and after the production of movements two sources of feedback are available to the performer. Intrinsic feedback can be defined as information gained from an individuals sensory modalities, while extrinsic feedback is information which augments this intrinsic feedback. Further, extrinsic feedback is provided by an external source; for example, a teacher or experimenter. As such extrinsic feedback can only be used to modify subsequent movement plans (or a subsequent portion of a movement) and presents information that is almost always supplemental to intrinsic feedback. Finally, the presentation form, content and temporal contiguity of extrinsic feedback may vary; for example, it may be verbal or graphic, static or dynamic, represent an aggregate score, or relate specifically to each response, and it may be presented immediately after a response or be delayed by some length oftimeor number of trials. The present research is concerned with the effects of providing several forms of extrinsic feedback to individuals who have been asked to learn an aiming task. A considerable body of motor learning research has focused on the benefits of one form of extrinsic feedback -knowledge of results (KR), as it pertains to both performance and learning (cf. Salmoni Schmidt & Walter, 1984). It has, however, been suggested that KR, which only provides information about the success of performance, might not be the most useful form of feedback as it is discrete and does not provide any prescriptive information to an individual (Brisson & Alain, 1996). While Gentile (1972) has been cited as one of the first investigators to examine the effects of knowledge of performance (KP), it has recently become a more frequent subject of investigation. Researchers have  1  suggested that, due to the more detailed and prescriptive qualities of KP (which often provides information about the kinematic or kinetic qualities of the movement) an individual identifies which portions of the movement were in error and therefore which portions need to be corrected on future trials (Hale & Franks, 1998; Newell, Quinn, Sparrow & Walter, 1983; Schmidt & Young, 1991; Young & Schmidt, 1992). Typically, studies which have demonstrated an advantage for KP over KR have used tasks which are isomorphic with the movement required to attain it. That is, the task and the movement required to attain it are identical. For example, using a force production task, Newell, Sparrow and Quinn (1985) demonstrated that the performance of participants learning to produce a specific isometric force more closely resembled the criterion when they received KP (a force-time trace superimposed over the criterion trace) rather than KR (peak force provided numerically). Several additional studies have provided support for presenting KP to individuals learning to perform isometric and coordinated multiple degree of freedom tasks. Newell and Carlton (1987) illustrated that learning to produce a complex force time trace was facilitated by KP; as participants who received a force time trace superimposed by the criterion trace produced a significantly different shaped (closer to the criterion shape) impulse than those receiving KR (the total impulse presented verbally). In a later study Newell, Carlton, and Antoniou (1990), investigated the effects of feedback on familiar and novel tasks. They demonstrated that when the criterion required participants to draw a circle (an action considered to be familiar to participants) KR alone provided enough information to learn the task adequately. However when participants were required to  2  produce an irregular shape their performance benefited from receiving KP which included both the criterion and their performance trace. In 1991 Schmidt and Young sought to develop a paradigm which they considered novel yet similar to a "real world" task in that it required participants to use a unique combination of movement parameters (timing, force, duration, etc.) to achieve the task successfully. Further their goal was to use a task with no isomorphic attainment constraints thus allowing both KR and KP to be manipulated independently. Participants were required to coincide a horizontal arm movement with a light moving towards them along a trackway perpendicular to their mid-line. This task is described by Schmidt and Young as being "an analog of hitting a moving ball with a bat"(p. 17; cf., Schmidt and Young for a detailed description of the task). Participants were provided with both knowledge of the optimum reversal point (165°) and their actual reversal point (KP), and a score (KR) which was a function of both speed and accuracy at the coincident point. While Schmidt and Young found that providing participants with KR plus KP benefited learning they did not systematically determine which kinematic variables best facilitated learning (e.g. peak back swing variability, mean forward swing initiation etc.) Subsequently, Young and Schmidt (1992) performed a study in which several kinematic variables served as KP. Four optimal spatial or temporal criteria were identified from previous research and used in conjunction with the coincident task as the criterion. KP reflected the participants ability to match this criterion while KR was presented as a score. As in their previous study, they demonstrated that kinematic information about the reversal position resulted in the best performance, and learning. However, Brisson and  3  Alain (1996) argued that the task and feedback in these studies were isomorphic because participants were encouraged to match the optimum aspects (e.g. reversal point) of the best performers' movement. Therefore it was still not possible to determine whether KP enhanced learning or just matched the goal of the task more directly than did KR. Using the coincident timing task of Schmidt and Young (1991), Brisson and Alain (1996) sought to determine if a single common movement pattern could be used as a criterion to facilitate learning a coincident timing task. Here the kinematic characteristics of the best participant's performance were used as the criterion for one group of participants, while the characteristics of the participant's personal best pattern (which was determined from an earlier portion of the study) was used as the criterion for a second group. While particular aspects of the best participant's performance were met, Brisson and Alain concluded that a common optimal template did not facilitate performance over a personal template. Moreover, they argued that "the movement pattern used by different subjects to attain these critical features may be unique to each subject...[and] that perhaps a personal pattern, unique to the physical attributes and morphology of each learner, may be more appropriate for learning the task" (p. 222). However, as with Schmidt and Young (1991) and Young and Schmidt (1992) the task in this study and KP were isomorphic. That is, subjects were encouraged to pay particular attention to critical features of the criterion and their own kinematic movement trace which were provided (along with KR) as feedback after each trial. The purpose of a recent study in our laboratory (Hale, Hodges, Khan, & Franks, Appendix A) sought to investigate the value of two forms of augmented feedback for  4  learning a rapid aiming task with no attainment constraints. An advantage of using aiming movements is that programming and sensory processes can be investigated by parsing the movement kinematics into primary submovement and error correction phases. (Abrams & Pratt, 1993; Abrams, Meyer & Kornblum, 1990; Khan, Franks & Goodman, 1998; Meyer, Abrams, Kornblum, Wright & Smith, 1988; Pratt & Abrams, 1996; Woodworm, 1899). The primary submovement is a ballistic phase and assumed to be preprogrammed so as to end at the target. However, if the primary submovement fails to reach the target the limb movement may enter an error correction phase consisting of one or more secondary submovements. The error correction phase is indexed by discontinuities in the kinematic profile and is said to indicate feedback based adjustments to movement. These on-line modifications are an attempt to minimize the error between the end of the primary submovement and the target position. In performing rapid aiming movements, several strategies are at the participants' disposal; according to Meyer et al. (1988), achieving maximum performance is achieved by adjusting initial impulse velocity such that the combined duration of the primary submovement and error correction phase is minimized. As aiming movements have been shown to consist of two submovement phases; which make up differing components of the total movement with increasing practice, we (Hale et al., Appendix A) investigated the effects of two forms of augmented feedback (KP) which provided information specific to these phases. Although Meyer et al,. (1988) and Pratt and Abrams (1996) have suggested that the control strategy was independent of the availability of on-line sensory information, recent work by Khan, Franks and Goodman (1998) and Khan and Franks (1998) has shown that different strategies do emerge after  5  extensive levels of practice. Participants modulated either the distance traveled (Khan, Franks and Goodman) or the time spent (Khan & Franks) in the primary submovement depending on the availability of visual feedback. It was suggested that participants who practiced with visual feedback became effective at processing on-line sensory information and therefore programmed their movements to make optimal use of vision to guide their movements to the target. In contrast, when visual feedback was not available, movements were planned to decrease reliance on sensory feedback. Using a rapid aiming paradigm, we required participants to perform a 45° elbow flexion which moved a cursor to a target on an oscilloscope (Hale et al, Appendix A). While participants were not permitted vision of their limb or the cursor their goal was to learn to move 45° as accurately as possible while minimizing movement time. A l l participants received numeric KR, presented on a computer screen, which consisted of target accuracy (constant error in degrees), total movement time (TMT), primary submovement time (MT1), and error correction duration (MT2). In addition they received either static KP, a static displacement time trace, or dynamic KP, a dynamic replay of the cursor movement in real time, (see Appendix A for more details). The results of this study indicated that static KP was adequate for learning a rapid 45° aiming task. While there was no statistical difference between participants receiving static or delayed dynamic feedback in the overall measures of performance, the strategies used by the two groups to perform the task were different. Participants receiving dynamic KP had more trials containing an error correction phase both in Acquisition and Retention as compared to those receiving static KP and spent less time in the primary submovement phase. This  6  provided more time for the error correction based secondary submovements, a process recently thought to be reserved for visually based on-line error corrections. Following this study we (Hale et al., Appendix A) proposed that participants who received dynamic KP engaged in a process by which delayed visual information was used to calibrate kinesthetic feedback. This process of calibration is one whereby augmented visual information (which provides information about how the movement was performed with respect to the task parameters) is used as a reference to which the kinesthetic consequence of the movement is compared. The result was an internal reference that may then be used to make on-line corrections to the movement. Finally this internal calibration process was likely responsible for the maintenance of performance in the no feedback Retention, where after one week the frequency of the error correction phase was significantly greater for participants who received KP in the form of a dynamic cursor presentation. Even in the absence of concurrent visual feedback, the development of a sensory motor representation seems to have been facilitated by providing a dynamic visual replay of the response outcome (Appendix A). Indeed the two movement strategy demonstrated by participants receiving delayed dynamic feedback was previously thought to be evidence for a process of on-line visual control. However, we (Hale, Hodges, Khan & Franks, Appendix A) suggested that a similar on-line kinesthetic mode of control may be adopted in the absence of concurrent visual information. In addition to contrasting several forms of KP which were equivalent in informational content, but different in presentation form; the current study sought to  7  investigate the interplay between KP, visual feedback and the kinesthetic consequence of the movement in attempt to better understand the process by which augmented feedback is used to learn an aiming task. Additionally, the two movement strategy previously employed by participants receiving delayed visual information in the form of dynamic cursor feedback suggested that kinesthesia can be calibrated by visual feedback and used on-line to control corrections effectively during rapid aiming movements (Appendix A). However, it remained unclear if this process was the same / or similar to that of the two movement strategy previously demonstrated by Khan et al. (1998) or Proteau and colleagues (1987,1990,1993). Therefore an additional goal of the present work was to expand upon the results of our previous study (Appendix A) by comparing the performance of participants receiving numeric KR, static KP, delayed dynamic visual feedback or concurrent dynamic visual KP while performing a rapid aiming task with no prescribed attainment constraints. Specifically, if the delayed visual information can be used to tune or calibrate kinesthesia then the movement characteristics (particularly the use and effectiveness of secondary submovements) of individuals practicing with concurrent and delayed visual feedback should be similar. In the present study participants were required to make a rapid, elbow flexion movement aimed to a target. The movement amplitude was 45° and the target width was 1.5°, yielding an index of difficulty of 5.9 bits (Fitts, 1954). Forty-five degrees of forearm flexion or extension was translated to a cursor moving between a home and target position on an oscilloscope screen. Similar to our previous study participants did not have vision of their arm, and for three of the four groups vision of the cursor was not available during  8  the movement. All participants received numeric feedback (KR) consisting of both accuracy (constant error -CE) and their trial SCORE (which was a composite of both total movement time -MT and accuracy -AE) immediately following each trial. This is all the information the Numeric group received. In addition to the numeric information, participants in the remaining three groups also received KP. One group received a kinematic displacement time graph (Static group). Another received a dynamic cursor display on the oscilloscope screen after they had completed their trial (Delayed Dynamic group -DD), while the fourth group received this information concurrent with their arm movement (Concurrent Dynamic group -CD). Here the Static and Delayed Dynamic groups receive essentially the same information as participants receiving static and delayed dynamic visual information in the previous study (Hale et al., Appendix A). Based on our previous results (Hale et al., Appendix A) it was hypothesized that, compared to a static graphic displacement time trace, dynamic feedback (delayed and concurrent) would encourage participants to consistently use a two movement strategy. While both modes of KP (static and dynamic) provided the same information, aspects of the dynamic information (such as acceleration, and velocity) may have been more salient, and therefore more easily used to calibrate kinesthetic feedback sources. Thus, if delayed dynamic KP provided information similar to that which was available in a full vision situation, then the Acquisition performance of participants receiving delayed dynamic information should not be different to that of participants receiving concurrent dynamic cursor information. Consistent with Proteau and colleagues (1987,1990, 1993) and Khan et al., (1998), who have used rapid aiming tasks and found that participants remained  9  reliant on visual feedback even after extensive practice, it was expected that the CD group would suffer a significant decrement in performance when this information was removed during Retention. Indeed, Proteau and Marteniuk (1993) suggested a guidance mechanism (which suppresses the processing of kinesthetic information) might be associated with concurrent visual information. In contrast, Adams (1971) proposed a process of internal error evaluation and correction (subjective reinforcement); if participants in the DD group used the delayed visual information to calibrate kinesthetic feedback, they were expected to rely on kinesthetic feedback during no KP Retention trials, and therefore maintain a proficient level of performance. Thus the findings of this study were expected to provide insight into the process by which individuals practicing under both concurrent visual, and no-vision conditions use both augmented and intrinsic feedback sources to learn and improve performance of a rapid aiming task with no prescribed attainment constraints. METHOD  Participants Twenty eight self-declared right handed volunteers were required. All participants were naive to the hypothesis being tested and randomly assigned into groups consisting of 7 participants each. The experiment was conducted according to the ethical guidelines laid down by the Behavioral Ethics Committee on Research of the University of British Columbia.  10  Apparatus and Task Participants were seated at a table with their shoulders harnessed to a chair. A computer monitor and an oscilloscope screen were mounted 35 cm. above the table 80 cm. in front of the subject. The participants' left arm was secured with Velcro straps to a manipulandum consisting of a padded aluminum bar 45 cm. long mounted, at the proximal end, to a bearing mounted vertical axial (allowing 360° of rotation). A 20 X 13 cm. board, used to secure the hand in pronation, was affixed to the distal end of the manipulandum. Vision of the participants' limb and hand was occluded beneath a shield. An optical encoder (Dynapar E20-2500-130), attached to the shaft of the manipulandum and custom made computer interface card enabled high speed sampling of angular position (sampling rate was 1000 Hz). Additionally a Kisler accelerometer (type 8638B50, + 50G), mounted at the distal end of the manipulandum (45 cm. from the axis of rotation), was used to record angular acceleration. Its signal (measured in milli-volts) was filtered with an active low pass filter (Krone-Hite, # 3750) set at 50 Hz and sampled at 1000 Hz. Kinematic and Numeric feedback were presented on the computer screen, while the home and target positions as well as the Dynamic feedback were presented on the oscilloscope screen. Angular distance between the home and target positions was 45° of elbow rotation. Five different start positions were used: 110°, 115°, 120°, 125°, 130° where 180° is defined as full elbow extension. Start positions were randomly presented in blocks of five; such that each of the five positions would be presented within a block of  11  five trials. Thus the same start position could only occur in succession if it were the last trial of one block and the first in the next. Participants were required to perform a 45° elbow flexion movement. Horizontal flexion and extension of the forearm caused a cursor on the oscilloscope screen to move left and right respectively, where each degree of forearm movement resulted in 2 mm. of cursor movement. Participants were not given any instructions about how to produce the movement, only that it needed to be performed as accurately as possible while minimizing movement time. Movements were to be to a target with a width of 1.5° which yielded an index of difficulty of 5.9 bits (Fitts, 1954). Procedure and Experimental Design Upon entering the laboratory participants read and signed an informed consent form. They were then positioned and secured to the apparatus in the above mentioned fashion. All participants were then required to perform 25 pre-trials in which they performed an elbow flexion movement to a wide target (20°) as fast as possible. Participants were then randomly assigned to one of four feedback groups. The four feedback groups were: Numeric KR, Graphic KP, Concurrent dynamic KP, and Delayed dynamic KP. Feedback was presented after each trial. All participants received the numeric score and accuracy information (Constant error) along the top of the computer screen. Constant error was displayed in the following phrase: "overshoot / undershoot xx°". In addition, one group (the Graphic group) received feedback consisting of a displacement time trace after they had completed their movement. Markers were placed along this trace indicating movement onset, end of primary submovement, and the end  12  error correction phase. Two additional groups were presented with cursor trajectory information. One of these groups (Concurrent Dynamic -CD) received this information concurrent with their arm movement, the other group (Delayed Dynamic -DD) received the same information but it was presented after they had completed their movement (i.e. it was a replay of the cursor moving across the oscilloscope screen). To begin a trial, participants aligned the cursor with the home position. At this point a (high frequency) 1000 Hz tone was presented; 1500 ms later a 2000 Hz tone sounded indicating participants should move. It was emphasized to participants that this was not a reaction time paradigm, but that they should move any time after hearing the second tone. For the Graphic, DD, and Numeric groups, the cursor disappeared once movement exceeded 8° / s, and did not reappear until it was time to prepare for the next trial. For the CD group the cursor remained on the screen throughout the movement, disappeared after the movement had come to a full stop, and reappeared in time to prepare for the next trial. All participants were instructed to try to hold their movement at the endpoint for approximately 500 ms after which time they could return to the home position. Approximately 5 seconds after the end of the movement, feedback was presented in accordance to group assignment for a period of 5 seconds. The study consisted of two Acquisition sessions on consecutive days. Prior to beginning the first Acquisition session all participants were required to perform 25 pretrials. Each Acquisition session consisted of 250 trials with participants receiving a 1 minute rest after every 50 trials. In addition all participants were required to complete two Retention sessions. Both sessions consisted of 50 trials with no-feedback and no  13  concurrent visual feedback of the cursor. The first Retention session followed the last Acquisition trial by 10 minutes, while the second was performed 1 week post Acquisition. To complete the experiment all participants were asked to fill out a short questionnaire. Analysis of movement The movement wasfirstanalyzed by separating the total movement into the primary submovement and error correction phases according to a movement parsing algorithm used previously and explained in detail by Khan et al., (1998). Briefly, this algorithm acts to separate the primary submovement from the error correction phase by identifying movement modifications indexed by (b) a positive to negative zero line crossing in velocity, (c) a negative to positive zero line crossing in acceleration or (d) a significant deviation in acceleration (see Figure 1) Dependent Measures The primary dependent measure was a score based on both total movement time and accuracy (absolute error). The score formula described the speed / accuracy trade off by identifying the increase in movement time associated with performing accurately. For each participant analysis of time to peak velocity (PV) from the 25 pre-trials was used to determine their theoretical minimum movement time (MT). In order to determine the minimum time a participant could traverse the 45° movement amplitude we used the following formula: MT  mm  = 2 (Mean PV - 2.5 * SD PV)  14  From the 25 pre-trials the mean and standard deviation (SD) of the time to PV was calculated. From the mean PV we subtracted 2.5 * SD in order to produce a theoretical minimum time to peak velocity . Finally we multiplied this theoretical minimum by 2 in 1  order to approximate the minimum MT (MT ) required to traverse 45°. Here it should min  be noted that it was assumed that the fastest possible way to traverse the 45° would be to have a symmetrical velocity profile with no zero line crossings in velocity (see Figure 1, A). The difference between the movement time for any particular trial and the theoretical minimum MT can then be thought of as a movement time effectiveness (MT ) which e  represents the degree to which a participant reduced MT in order to achieve accuracy on a particular trial. However, as improvement in performance resulted in a decrease in MT and or AE, we applied a coefficient of both MT and AE such that the product of these e  values increased with improved performance. The weighting coefficient for M T was e  defined in the following fashion: when MT was 0 a weight of 10 was awarded; as MT e  approached the average pre-trial MT the coefficient decreased linearly to 0. Similarly, when A E was 0 a weighting coefficient of 10 was awarded; as AE increased from 0° to 10° the weighting coefficient decreased linearly to 0. The SCORE was therefore the product of the coefficient of MT and the AE of a particular trial. Thus: e  SCORE = (MT ) (AE) e  MT  e  = MTtrial - MTmin  2.5 standard deviations were subtracted from the mean peak velocity as this would provide a product that was theoretically within an individuals distribution of peak velocity but was at the extreme low end of their distribution and therefore very difficult to actually attain. 1  15  o ,o g ^  "55 a € ~  I  8  s  § r3  3  . s &g  o .2  2  u  g  8 A ?J IL, M O g fl  n  •ig os IS -a o o M o  2 >  4> OD <D  rj  C  ft ^  c«  2  c  8  o  o  g  a « o pq  a  3  a ^ •S „ S3 e ^  p a  g  a  S  & 5«  2  o  -1 I I «J  "8 | '5 -S > -S § -9 o  g  0 0  " •53 >  U 2* ~ 2 5 8& o ^  16  where AE is the coefficient of absolute error incurred on a particular trial and MT is the e  coefficient of the difference between M T  mm  (a constant unique to each participant  reflecting the minimum theoretical MT) and the MT produced from a particular trial. Thus MT represents the amount of time a participant increased MT in order to meet the e  accuracy constraints. As absolute error (AE) was used in the score calculation it was included in the analysis as the primary measure of accuracy. In addition constant error (CE) was included so as to provide an indication of bias in performance (i.e. an overshoot or undershoot of the target). Total movement time (TMT) was the second factor used to calculate the score and was also included as a separate dependent measure. As well a number of kinematic measures were taken. These included distance traveled, time spent in the primary submovement phase, the proportion of total movement distance and total movement time consumed in the primary submovement phase, and the number of trials containing error corrections phase. Statistical analysis Acquisition data was analyzed by a group(4) by day (2) by block (5) randomized group ANOVA with repeated measures on the second and third factors. As performance tends to become most stable with practice a separate, group(4) by block (5) randomized group ANOVA with repeated measures on the second factor, was also performed for the second day of Acquisition.  17  Similar to other research (Proteau & Marteniuk, 1993; Proteau et al., 1987) the last block (50 trials) of Acquisition was considered to represent participants performance late in Acquisition. Therefore Immediate and Delayed Retention data were compared with the final block of Acquisition so that changes in performance (due to removal of feedback and / or absence from the task) could be discussed relative to Acquisition performance. Thus a group (4) by condition (3 -final block of Acquisition, Immediate Retention, Delayed Retention) analysis with repeated measures on the second factor was performed. Finally as there were 5 different start positions an analysis was performed on the target accuracy (CE) data to ascertain if the changing start positions affected performance. For Acquisition data a group (4) by position (5) by day (2) by block (5) randomized group ANOVA with repeated measures on the second, third and fourth factors was performed. To investigate the effect of feedback removal Retention data was analyzed in a group (4) by position (5) by condition (3 -last block of Acquisition, Immediate and Delayed Retention) randomized group ANOVA with repeated measures on the second and third factors. RESULTS Acquisition Two separate analyses were performed on Acquisition data. The primary analysis included both days of Acquisition, while the other was only on the second day of Acquisition. Results of the second analysis were only reported if significance was found here and not in the primary analysis. Finally, post hoc analysis was performed on all significant comparisons using Tukey's HSD test at the 0.05 level of significance.  18  Performance measures Score: The overall measure of performance was a score based on both movement time and accuracy. The main effect of Groups failed to reach significance even when the second day of Acquisition was analyzed separately. However, while not significant, on inspection of (Figure 2) the CD and DD groups achieved higher scores than the KR and Static groups, particularly on the second day of Acquisition. Both main effects of Day and Block were significant F(l,24) = 15.62, p < .001, and F(4,96) = 10.38, p <.001. The analysis of the day effect indicated that performance improved from the first to second day of practice. TMT: All groups decreased their total movement times throughout Acquisition as indicated by significant main effects of Day F(l,24) = 21.36, p < .001 and for Blocks F(4,96) = 21.1, p < .001. Post hoc analysis indicated TMT significantly decreased from the first to second day of Acquisition; however, the stringent Tukey test failed to reveal differences between blocks. No significant interactions were found (see Figure 3). Absolute error: The main effect of Group was significant across Acquisition F(3,24) = 41.96, p < .001 (see Figure 4) and was due to the CD group performing with greater overall accuracy than the other three groups. Significant main effects of Day F(l,24) = 20.41, p < .001 and Block F(4,96) = 7.88, p < .001 were also found. Post hoc analysis of these effects indicated that participants became more accurate (decreased their error) with practice over the two days. Additionally there was a higher order Day x Block x Group interaction F (12, 96) = 1.742, p < .05. Post hoc analysis indicated that this  19  o  < / • >  (S  o  ts  < o  —  3HODS  20  o  ~+  in  o  effect was due to a more substantial decrease in AE across blocks by the KR group than the other three groups during the first day of Acquisition. Constant error: There was no significant difference between groups with respect to CE, nor was there a significant effect of Day (see Figure 5). However there was a significant main effect of Blocks F(4,96) = 5.53, p < .01. When subjected to post hoc examination it was found that there was a constant (yet gradual) decrease in CE over blocks; specifically, that block 1 was significantly different from block 4. Submovement analysis Primary submovement time (MT1): Analysis of MT1 across both days of Acquisition revealed no significant main effect of Groups F(3,24) = 2.15. Inspection of Figure 6, however indicated that the CD group was about 40 ms faster than the KR group and about 80 ms faster than the Static and DD groups, indeed the group main effect approached the conventional level of significance when a separate ANOVA was performed on the second day of Acquisition F(3,24) = 2.58, p = .07 (this trend is consistent with the two movement strategy predicted of participants practicing under full vision conditions). Although the Day and Block and Block x Group effects failed to reach significance, there was a significant Day x Block interaction F(12,96) = 5.59, p < .01. Breaking down the Day x Block interaction indicated that this effect was due to a decrease in MT1 across blocks during the first day of Acquisition, and an increase in MT1 across blocks during the second day. The increase in MT1 is consistent with the data for the number of trials containing an error correction phase (M2), where the day by block  21  06  o v-i \0  o o  on < «1  VO  22  o o U">  o  >/i  o o 'Jf  toaa) Hoaaa axmosav  23  c  interaction indicated that there was an overall decrease in M2 on the second day of Acquisition. Proportion of TMT consumed by MT1 (MT1/TMT): The proportion of total movement time consumed in the primary submovement phase was found to be significantly lower for the CD group than the other three groups throughout Acquisition as indicated by a main effect of Group: F(3,24) = 6.75, p < .01. In addition the KR group consumed a smaller proportion of TMT in MT1 than the Static and DD groups (which were not different from one another) (see Figure 7). A main effect of Blocks was also found F(4,96) = 5.91, p < .01. Two significant interactions were found. The Day x Block interaction F(4,96) = 3.32, p < .05 indicated that there was a tendency for participants to increase the proportion of TMT consumed in MT1. However it was the large increase in MT1/TMT by the Static and DD groups which was primarily responsible for this two way interaction. Indeed, the Day x Block x Group interaction F(12,96) = 1.89, p <.05 showed that there was a sharp increase in MT1/TMT by the Static group (block 1 to block 2 in particular) and the strong steady increase by the DD group on the second day while there was little change between day 1 and 2 for the KR and CD groups. Primary submovement distance (MD1): The mean distance traveled during the primary submovement phase did not differ significantly between groups; nor was there a significant day effect. While the main effect of blocks for both Acquisition sessions narrowly missed the conventional level of significance F(4,96) = 2.72, p = .057, the main  25  «  O co  O co  O  O cn  Q c o c o  26  O  O (N  O <S  O tN  cs  27  effect of blocks for the second day of Acquisition was significant F(4,96) = 3.24, g < .05 and post hoc analysis indicated that the primary submovement distance decreased across blocks on the second day. Primary submovement variability (PSV): While there was little difference between the mean distance traveled between groups during the primary submovement phase (MD1) the main effect of Groups for the variability in MD1 was found to be significant F(3,24) = 9.63, p<.001. As is evident in Figure 8, the CD group produced their primary submovement with significantly less variability than the other three groups. In addition there were significant main effect of Day F(l,24) = 42.31, p < .001 and Block F(4,96) = 10.79, p < .001 indicating that variability decreased for all groups across blocks on both days of Acquisition. Proportion ofTMD consumed in MD1 (MD1/TMD): The proportion of total movement distance traveled during the primary submovement phase is shown in Figure 9. While the Group effect failed to reach significance F(3,24) = 2.11, the pattern of results resembled that which was seen in MT1/TMT. That is the KR, Static and DD groups tended to consume more of the total movement distance during the primary submovement phase than the CD group. Post hoc examination of the significant main effect of Blocks F(4,96) = 5.67, p < .01 revealed differences between block 1 and blocks 2, 3, 4, and 5. When the second day of Acquisition was analyzed separately a main effect of Blocks was again found F(4,96) = 7.70, p < .001. In addition, there was a significant Block x Group interaction F( 12,96) = 1.91, p <.05. Break down of this interaction indicated that the KR  28  oo  c—  \o  <o  "st  :oaa) AimavravA  29  c%  <N  group decreased MD1/TMD while the other three group continued to increase this ratio on the second day of Acquisition. Error correction duration (MT2): Significant main effects of group F(3,24) = 9.57, p <.001, day F(l,24) = 7.82, n< .05 and block F(4,96) = 13.76, p < .001 were found. Analysis of the group effect revealed that the CD group produced a longer secondary submovement than the other three groups; and the KR group produced longer secondary submovements than the Static or DD groups (which did not differ from each other). The Day and Block effects indicated that MT2 decreased with practice (see Figure 10). Number of trials containing an error correction phase (M2 count): Significant main effects of both group F(3,24) = 3.79, p < .05, and block F(4,96) = 4.58, p <.001 were found. Tukey post hoc analysis revealed that the CD group used an error correction phase on significantly more trials than the Static or DD groups (which did not differ from each other), and that the KR group did not differ form the CD, DD, or Static groups. There was also a significant Day x Block interaction F(4,96) = 3.77, p < .01, and a higher order, three way, Day x Block x Group interaction F( 12,96) = 2.08, p <.05. Breakdown of the day by block interaction showed the number of trials with an error correction phase decreased across blocks to a greater extent on the second day of Acquisition than on the first. Post hoc analysis of the three way interaction indicated that this effect was due to the decrease in number of trials containing an error correction phase on the second day of Acquisition by the Static and DD groups while the KR group slightly increased and the CD group generally maintained the number of trials containing an error correction phase on the second day of Acquisition (see Figure 11).  30  Q O  Q Q  ON  m o  CJ <  h-H  EL,  05  (%) a w i / i a w  31  TJ-  CO  CO  32  CN  CN  Index of Error Correction Effectiveness (IECE): For those trials containing an error correction phase the success of correcting the movement towards the target was analyzed. The main effect of Group was significant F(3,24) = 45.41, g .001. Figure 12 clearly indicates that the CD was the only group to make effective secondary submovements during Acquisition. Indeed the post hoc analysis of the Group effect indicated that the CD group was significantly more effective than the other three groups (all of which did not differ from one another). Retention The final block of Acquisition was compared with both immediate and delayed Retention. This analysis provided an indication of the effect of removing feedback on performance. Performance measures Score: A significant main effect of condition was found F(2,48) = 19.69, g < .001 which when subjected to a post hoc analysis revealed that the Score performance declined for all groups as a result of the feedback being removed and as a function of the one week retention interval. In addition there was a significant Condition x Group interaction F(6,48) = 4.17, g < .01. Examination of Figure 2 reveals that the CD group was most significantly and immediately effected by the removal of feedback. Tukey post hoc analysis supported this observation and also indicated that the Static group had a more subtle decline in Score from Acquisition to RI and from RI to R2 compared to the other three groups.  33  34  m  CO  CM  % 3031  35  T-  •  Total movement time (TMT): Significant main effects for both Group and Condition were found F(3,24) = 5.04, p < .01, and F(2,48) = 30.03, p < .001 respectively (see Figure 3). While the KR, Static and DD groups all increased TMT from Acquisition to R2 none of these groups experienced a significant change in TMT from Acquisition to RI suggesting that the one week delay had more of a detrimental effect on performance than removing KR. In addition there was a Condition x Group interaction F(6,48) = 4.56, p < .01. It was found that removing post response feedback resulted in all groups increasing TMT during the one week retention (between Acquisition and Delayed Retention, and from Immediate to Delayed Retention). However, withdrawing concurrent visual feedback had a significant immediate effect on the CD group (between Acquisition and Immediate Retention) Absolute error (AE): All groups suffered a decrement during Retention, particularly between RI and R2 (main effect of Condition F(2,48) = 22.33, p < .001). As with the Score it appeared that removing post response feedback had little immediate effect on accuracy, whereas removing concurrent visual feedback and the one week retention period both seriously impaired the participant's accuracy. Indeed a significant Group x Condition interaction (F(6,48) = 4.54, g < .01) was found to be the result of the CD group experiencing a significant decrease in accuracy upon immediate removal of feedback (from Acquisition to RI) while the other three groups experienced no significant change from Acquisition to RI (see Figure 4). While the DD group suffered a significant decrease in AE between RI and R2, Absolute error only significantly decreased between Acquisition and R2 for the KR and Static groups.  36  Constant error (CE): The pattern of results for CE are similar to those seen for AE. Specifically there were significant main effects of Group F(3,24) = 3.85, p < .05, and Condition F(2,64) = 25.61, p. < .001 and a significant interaction between Condition and Groups F(6,48) = 3.39, p .01. Post hoc analysis (see Figure 5) indicated that while the CD group immediately began to overshoot the target upon removal of feedback, the KR Static and DD groups first undershot (between Acquisition and Immediate Retention) and then overshot (between Immediate and Delayed Retention) the target when feedback was removed. Submovement analysis Primary submovement time (MT1): There was a significant main effect of Condition F(2,48) = 4.04, p < .05, and a significant Condition x Group interaction F(6,46) = 2.96, p < .05. Breakdown of the interaction indicated that the KR, Static, and DD groups all decreased MT1 between Acquisition and RI and then increased MT1 between Immediate and Delayed Retention (this effect was significant for the Static and CD groups). The CD group steadily increased MT1 from Acquisition through Immediate to Delayed Retention (see Figure 6). Proportion of TMT consumed by MT1 (MT1/TMT): There was an approximate 25 - 30% difference in the proportion of TMT consumed in MT1 between the Static and DD groups and that of the KR and CD groups (See Figure 7). Indeed post hoc analysis of main effect of Group F(3,24) = 8.60, p <.001 revealed that the KR and CD groups consumed a significantly smaller proportion of TMT in MT1 than the Static and DD groups, and that these latter groups did not significantly differ from one another. In  37  addition a significant main effect of Condition was found F(2,48) = 5.58, p < .01. Post hoc analysis confirmed that the proportion of TMT consumed in MT1 declined between Acquisition and R2 and from RI to R2; again suggesting that the one week retention interval had a greater effect on how participants performed than the removal of feedback. Primary submovement distance (MD1): Group and Condition main effects were found to be significant F(2,24) = 4.09, p < .05, and F(2,48) = 20.31, p < .001 respectively. In addition, there was a significant Condition x Group interaction F(6,48) = 2.90, p < .05. As previously found the removal of feedback most significantly effected the CD group. More specifically, the CD group markedly increased the distance traveled in MD1 such that they drastically overshot the target on the immediate Retention, and further overshot it one week later. In comparison the other three groups initially shortened their MD1 (although not significantly) on immediate Retention, and then significantly increased MD1. Primary submovement variability (PSV): All groups increased the variability of the primary submovement distance across conditions F(2,48) = 21.14, p < .001. Again there was a condition by group interaction F(6,48) = 5.58, p < .001, which when broken down was found to be the result of the CD group having a greater increase in variability in MD1 from Acquisition to the immediate Retention session (Figure 8). Proportion ofTMD consumed in MD1 (MD1/TMD): The main effect of Groups narrowly failed to reach conventional levels of significance F(3,24) = 2.86, p = .058. Inspection of figure 9 shows the CD and KR groups continued to travel a smaller portion of the total movement distance in the primary submovement phase than the Static and DD groups. No other effects were significant.  38  Error correction duration (MT2): Group and Condition main effects were found F(3,24) = 13.66, g < .001, and F(2,48) = 2.23, p < .001 respectively. MT2 was significantly longer for the CD group than the KR, Static or DD groups. In addition the KR group produced a longer MT2 than the static and DD groups (which did not significantly differ). The Day x Group interaction approached conventional levels of significance F(6,48) = 2.23, g = .059. Inspection of figure 10 suggests that the CD group was most affected by the withdrawal of feedback. Number of trials containing a secondary submovement phase (M2 count): Main effects of Group and Condition were found for the number of trials containing a secondary submovement phase F(3,24) = 4.869, p < .01 and F(2,48) = 4.24, g < .05 respectively (see Figure 11). Post hoc analysis of the Group effect indicated that the Static group performed using a secondary submovement phase on fewer trials per block than the CD group, but that there was no significant difference between the KR, DD and CD groups. Examination of the Condition effect indicated that there was a significant increase in the number of trials containing a secondary submovement phase from Acquisition to delayed Retention. Effectiveness of secondary submovement phase (IECE): There were significant main effects of Group F(3,24) = 6.13, g < .01 and Condition F(2,48) = 9.67, g < .001 as well as a Condition x Group interaction F(2,48), 6.50, g < .001. Inspection of the interaction indicated, once more, that the CD group suffered the greatest deleterious effects from the removal of concurrent feedback. Unlike the other three groups, even  39  though the CD group continued to use an error correction phase during Retention (see M2 count), this phase was ineffective in reducing error (see Figure 12). Positional analysis Acquisition A positional analysis was performed on constant error (CE) to determine whether changing start positions differentially effected accuracy. The ANOVA on Acquisition yielded main effects of both position F(4,96) = 268.96, p < .001 and block F(4,96) = 8.13, p_ < .001. Examination of figure 13 revealed the effect of position; that is when participants were required to initiate their movement with their elbow in the most extended position (position 1) they overshot the target. As the start position moved closer to the participants mid-line, movements became progressively shorter until they undershot the target (see figure 13 & 14). There was also a significant position x group interaction F(12,96) = 20.73, p < .001. Inspection of figure 15 indicated the concomitant shifts in endpoint accuracy, exhibited by all groups, was less dramatic for the CD group. Retention The analysis of Retention revealed significant main effect of group F(3,24) = 4.29, p < .01, position F(4,96) = 98.07, p < .001, and condition F(l,24) = 30.52, p < .001. As was the case with the above data, the CD group was most severely affected by the removal of feedback. Figure 13 and 15b clearly reveal the main effect of position. Indeed start position did have an effect on performance. As in Acquisition, when participants  40  ^  .52  o  as  0) OH  e  0)  (030) 3 0  (D3Q)33  41  (D30) 3 3  0>  S3  H  J3  ^  5  u  O  l>  as £  42  .2 o S •M  ~  IT!  in  co  3 O  .8  (Sop) 3 0  43  were asked to begin their movement with their elbow in the most extended position (position 1) they produced the longest movements. As the degree of elbow extension decreased participants progressively made shorter movements until they were undershooting the target (see Figure 14 & 15b). Performance across conditions was found to first improve (on immediate Retention) and then decrease on delayed Retention for participants in the KR, Static and DD groups. However the lower CE values for these groups likely was only consequence movements becoming shorter. Indeed careful consideration of figure 13 and 16 reveals that, as the degree of elbow extension decreased CE values decreased, however note that instead of maintaining accuracy on Immediate Retention the CE value becomes negative at about position 3 and become more negative at positions 4 and 5. While it appears initially that performance improves for these group on Immediate Retention the lower CE values are only a function of an initial decrease in the distance of the movement. Finally on Delayed Retention, all participants increased the distance of their movements. Additionally, there were position x condition F(8,192) = 4.26, g < .001 and position x condition x group F(24,192) = 2.39, g <.01 interactions found. Reduction of the three way interaction revealed that on the final block of Acquisition all groups progressively decreased the distance of their movements across positions. In Immediate and Delayed Retention, however, the CD group overshot the target at all positions while the other three groups further shortened their movements on Immediate Retention and then, on Delayed Retention, increased the distance of their" movements (see figure 16).  44  • 3LLVIS  a  ^-v o cn '-a o  i  o  o tN  aa DUV1S  as o  45  DISCUSSION  Questions investigating the pertinence of feedback when learning a skill have long been the focus of motor learning research. In particular attention has been directed at identifying optimal feedback schedules, or the most effective presentation form of feedback. However, little focus has been directed towards identifying the process by which feedback is incorporated into a movement strategy and used to enhance the performance of that movement on future trials. Vision has been credited as providing information about the early portion of a movement which is then used to make corrective secondary submovements (Proteau et al., 1987,1990,1993; Khan et al., 1998). Indeed when vision is available individuals tend to rely heavily upon it to guide their movements. The effect of concurrent visual feedback was highlighted in research which investigated the effectiveness of vision used to guide rapid movements in individuals who suffer form large fiber sensory neuropathy (Ghez, 1991). However, a second intrinsic source of information is available to the performer. Kinesthesia also provides information about the movement, and may be a source of feedback which can be used to correct movement errors. The goal of this research was to investigate the effects of several forms of feedback, which provided information specific to both phases of the movement. More specifically it was designed to explore the interplay between these feedback formats and the intrinsic sources of feedback on learning a rapid aiming task.  46  On the comparison of Concurrent and Delayed forms of Feedback The process of making on-line corrective secondary submovements has been thought to be reserved only for movements made under full vision conditions. However, previous research in our laboratory (see Appendix A) suggested that delayed visual feedback may be used to calibrate the kinesthetic consequence of a movement and that this sense could then be used on subsequent trials to direct on-line corrective secondary submovements. This process, then, is similar to the two movement strategy normally exhibited by individuals practicing under full vision conditions; and would be subject to the predictions made in Meyer's submovement optimization model (Meyer et al, 1982). Consequently two predictions were made: first that participants receiving dynamic (concurrent or delayed) visual feedback would use a two movement strategy and perform better in Acquisition than the KR and Static groups. Second, as it has been shown that when participants are presented with a form of feedback which provides an easy solution to the error correction problem they come to rely upon this form of information rather than developing intrinsic error correction processes (Salmoni, Schmidt, and Walter, 1984). Therefore it was proposed that participants in the DD group would recognize the kinesthetic consequence of the movement prior to receiving KP and then make an internal comparison between intrinsic and extrinsic feedback sources. Consequently, using this intrinsic form of error detection and correction, participants in the DD group were predicted to maintain their movement strategy and demonstrate a superior level of performance during Retention when augmented visual information was absent.  47  The score variable was a function of both total movement time and accuracy; and therefore is an overall index of performance. Analysis of Acquisition data indicated that feedback and or practice did facilitate improvement in the score variable for all participants during Acquisition. Recall, however, that both the CD and DD groups were predicted to use a two movement strategy to achieve the task. Examination of the submovement and kinematic data, however, indicated that the process by which these two groups achieved their high scores during Acquisition was different. Though significantly more accurate then the DD group, the CD group had an average TMT that was approximately 40 ms slower than the DD group (this difference was not significant p = .160). Furthermore, the CD group performed using a two movement strategy throughout Acquisition as predicted; while by the DD group significantly decreased the use of a error correction phase from the first to second day of Acquisition. It has been demonstrated on several occasions that individuals practicing under full vision conditions use a two movement strategy to achieve the goal of moving to a target as fast and as accurately as possible (Proteau and colleagues 1987, 1990, & 1993; Khan et al., 1998). In particular Khan et al., argued that there is an increased reliance on vision as a function of practice, and that vision is used to guide secondary submovements to the target. Additionally Meyer et al., (1988) proposed that participants work to minimize the total movement duration and end-point error by optimally balancing the primary submovement and error correction (secondary submovement) durations. Indeed, the performance of the CD group supports these proposals. Furthermore there is considerable agreement between the submovement data of the CD group in this study and that which  48  was exhibited by the participants practicing with full vision in the Khan et al study. That is, in addition to achieving the greatest accuracy participants with full vision of the cursor trajectory, performed the task by producing the fastest primary submovements, consumed a smaller proportion of TMT in MT1 and traveled a smaller proportion of TMD in MD1 than participants in the other three groups. In short, the CD group effectively used a strategy whereby secondary submovements were used to achieve the task during Acquisition, while the KR, Static and DD group did not. These data, then, do not provide support for our previous findings that individuals practicing with delayed dynamic feedback use a two movement strategy to achieve the target (Appendix A). In addition, support for the proposal that delayed dynamic feedback would facilitate an effective calibration of kinesthetic feedback such that this form of feedback could be used to guide corrective secondary submovements was not found. On the contrary, participants in the DD group apparently worked to eliminate the use of an error correction phase, and worked to achieve the target within the primary submovement. When the Retention data were considered, it was again evident that the CD group had been heavily reliant on vision to accurately guide their movements. Although the score performance of all groups was lower when feedback was removed, analysis of the condition by group interaction indicated, as predicted, that the CD group was most severely (and immediately) effected by the removal of feedback (see figure 2). Indeed, the condition by group interactions, seen in many of the dependent measures, were caused (primarily) by the sharp decline in performance, from Acquisition to the first Retention session, by the CD group compared to a less dramatic decline in performance by the other  49  three groups. Furthermore, the Score, TMT and AE performance for the KR, Static, and DD groups did not significantly change from Acquisition to the Immediate Retention. This suggests that these delayed forms of feedback did not have the same disrupting or guiding qualities that the concurrent visual feedback had; and is in line with the principals of the Guidance hypothesis (Salmoni, Schmidt, & Walter, 1984). In fact, the decline in performance for these three groups was most dramatic (and most often only significant) after the one week retention. That is, the immediate removal of feedback often failed to induce a significant decrement in performance for participants in the KR, Static and DD groups. Rather the decline in performance appears to have been a function of time away from the task. Proteau et al., (1993) suggested that, while kinesthesia may be calibrated from visual feedback, this process is unstable and often results in poor performance when concurrent visual feedback is removed. In the present study, the performance of the CD group immediately deteriorated when concurrent visual feedback was removed. This pattern of performance was predicted from the outset of this study. Based on the Guidance hypothesis (Salmoni, et al., 1984) it was hypothesized that participants in the CD group would come to rely upon concurrent visual feedback instead of developing other kinesthetic sources of error detection and correction. Indeed, if the CD group were using vision to calibrate the kinesthetic consequence of the movement then the effect of feedback withdrawal would not have been as immediate and or severe. Finally there was a less dramatic decay in performance exhibited by the other groups indicating that the  50  delayed feedback formats were not as disruptive to intrinsic feedback processing. These findings are well accounted for by the Guidance Hypothesis. On the difference between KP and KR It has already been argued that participants receiving CD feedback effectively used visually guided secondary submovements to achieve the target during Acquisition. Indeed concurrent visual feedback of the cursor movement enabled these participants to realize a strategy whereby producing a very fast and consistent primary submovement followed by secondary submovements would result in relatively high scores. However, based on our previous research (Appendix A) it was proposed that participants in the DD group would use a similar strategy as those participants receiving concurrent visual feedback during Acquisition. Furthermore, the DD group was predicted to use a different strategy than that of participants in the Static and KR groups. This was not the case. Unlike our previous research (Appendix A) the submovement and kinematic data indicated that participants receiving delayed dynamic feedback and those receiving the static displacement trace generally did not differ with respect to how the movement was produced. Both groups tended to use a single movement to hit the target; that is, with increasing practice, the number of trials (per block) containing an error correction phase significantly decreased. Interestingly, while participants in the KR group were as accurate as those receiving DD and Static feedback, they performed using a two movement strategy. Nevertheless the data indicated that KR did not facilitate an effective use of secondary submovements. While it could be argued that the task was slightly different for  51  the C D g r o u p , i n that they p e r f o r m e d a task w h i c h a f f o r d e d c o n c u r r e n t v i s u a l f e e d b a c k , 2  t h e r e m a i n i n g t h r e e g r o u p s p e r f o r m e d the i d e n t i c a l t a s k u n d e r d i f f e r i n g f e e d b a c k  conditions.  T h u s it c a n be a s k e d :  w h a t attributes o f the d i f f e r e n t f o r m s o f d e l a y e d  f e e d b a c k e n c o u r a g e d the D D a n d S t a t i c g r o u p s to d e c r e a s e the u s e o f the  secondary  s u b m o v e m e n t s w h i l e K R g r o u p d i d not?  It w a s d i s c u s s e d a b o v e that t h e S t a t i c a n d D y n a m i c f o r m s o f f e e d b a c k p r o v i d e d  i n f o r m a t i o n a b o u t b o t h the p r i m a r y s u b m o v e m e n t a n d e r r o r c o r r e c t i o n p h a s e s o f a r a p i d  a i m i n g m o v e m e n t , a n d that this i n f o r m a t i o n is c r i t i c a l f o r a p a r t i c i p a n t s i m p r o v e m e n t .  I n d e e d , o n e o f t h e g o a l s o f t h i s s t u d y w a s to i n v e s t i g a t e t h e e f f e c t s o f f e e d b a c k  p r o v i d e i n f o r m a t i o n s p e c i f i c to these p h a s e s o f the m o v e m e n t .  which  O n the o t h e r h a n d ,  participants receiving K R practiced with a somewhat impoverished f o r m o f feedback.  R e c a l l that K R o n l y p r o v i d e d i n f o r m a t i o n a b o u t the o v e r a l l a c c u r a c y o f the  a n d the  score  a c h i e v e d o n that t r i a l .  movement,  A l t h o u g h movement time information was  not  d i r e c t l y p r o v i d e d w i t h i n the K R d i s p l a y it c o u l d b e d e d u c e d , f o r e x a m p l e , that a t r i a l w i t h  g o o d a c c u r a c y y e t y i e l d i n g a p o o r s c o r e w a s the r e s u l t o f a m o v e m e n t t i m e that w a s  long.  too  H o w e v e r , the i n f o r m a t i o n w i t h i n the K R d i s p l a y d i d n o t differentiate, i n a n y w a y ,  the m o v e m e n t i n t o the p r i m a r y a n d s e c o n d a r y s u b m o v e m e n t p h a s e s .  Static K P w a s  the  o n l y f o r m o f f e e d b a c k c o n t a i n i n g a p h y s i c a l d e l i n e a t i o n b e t w e e n the p r i m a r y a n d  s e c o n d a r y s u b m o v e m e n t phase;  although, participants receiving d e l a y e d d y n a m i c  i n f o r m a t i o n c o u l d d e t e r m i n e these m o v e m e n t p h a s e s w h i l e the c u r s o r m o v e d a c r o s s  oscilloscope  the  screen.  1 would not consider this to be so. Rather presenting concurrent visual information was simply one manipulation of feedback. 2  52  Previously we (Hale et al., Appendix A) proposed "that dynamic visual information (i.e. a visual play-back of the cursor trajectory) provided in real-time, would be especially helpful in calibrating the kinesthetic response-produced feedback so that on subsequent trials the sensory-motor representation (Proteau, 1992), would be more developed, allowing for greater sensitivity to any errors in the movement". Accordingly, delayed visual feedback would then sharpen a participants awareness of the secondary submovements made during the error correction phase. A quote we have previously used (Hale et al Appendix A) speaks nicely to this proposal "The role of KP may be simply to augment information about the movement pattern that may otherwise go undetected, and to make it more salient so that the calibration between the movement and the outcome is facilitated" (Brisson & Alain, 1996, p. 222). Thus, while kinesthesia was available to all participants, only those in the DD and Static groups had submovement information augmented with the feedback display. As such participants in the DD and Static groups likely realized that the error correction phase was costly with respect to total movement time and worked to eliminate it. On the other hand, participants receiving KR may not have come to such a conclusion. Indeed on a post test questionnaire many participants in the DD and Static groups made comments such as "I have to eliminate the wiggling around at the end of my movement" indicating that they were aware of the time cost of these movements. On amplitude versus endpoint coding (what is represented?)  It is interesting that the Static and DD groups performed similarly in the current study when we previously found evidence suggesting that the DD group would use a two  53  movement strategy, and the Static group a single movement strategy (Appendix A). To understand why there is a discrepancy between our previous research and the current study differences between the two are briefly discussed. Several changes in the design of this study may in part or in combination have accounted for the disagreement between the data. Performance plateaus later in practice suggested that the task in our previous study may have been too easy. Thus changes were made to the design of this study to make the task more novel and or difficult. The left arm rather than the right was used. Participants were required to make a 45° flexion rather than a 45° extension movement, and instead of a consistent start and finish location, there was five start and finish positions (all requiring a 45° movement). It was felt that these changes would be sufficient to increase the difficulty of the task, but that they would not impact on the process by which participants completed the task. Variable start positions were used in an attempt to reduce the likelihood that participants would use only end point spatial locations to remember the target position particularly during Retention. However, in hindsight it was felt that this change in design may have disrupted the participants ability to use the delayed visual feedback to calibrate the kinesthetic representation of the movement. Therefore a separate analysis of start position was performed on constant error (CE). From this analysis it was apparent that the different start positions did have a significant effect on accuracy, and therefore likely disrupted the reliability of the kinesthetic feedback. Most notable were the significant main effect of position and the position by group interaction in both Acquisition and  54  Retention. These effects provided strong support for the Equilibrium point (EP) or Mass Spring model of motor control (Feldman, 1966 a,b; 1986; Polit & Bizzi, 1979). The EP hypothesis is based on the premise that muscle has spring like qualities. Thus when the flexors and extensors of a limb are producing equal but opposing forces the limb will be stationary and in a state of equilibrium. For movement to occur an imbalance or a shift in the equilibrium point is created and the limb moves until the flexors and extensors are once again in a state of equilibrium. Furthermore according to the EP hypothesis the end position (or equilibrium point) is the only information necessary for the successful completion of the movement. Therefore, regardless of where a movement begins it should reach the programmed end position (Feldman, 1986, Polit & Bizzi, 1979). Indeed the CE performance of participants in the present study was biased such that, regardless of start position, the end positions of all movements were similar. Examination of figures 13 and 14 clearly reveal the incremental shifts in accuracy with the shifts in start position. The main effect of position seen in figures 13,14 and 15 indicates that all participants overshot the target when their elbow was in the most extended position (130°), and as the start position shifted closer to the participants midline they progressively made shorter movements until they were undershooting the target. These concomitant shifts in CE indicate that participants were coding the movement endpoint rather than coding elements of the entire movement. That is, participants were aiming to a common end point and not a series of targets coinciding with the start positions (see figure 14). This finding is directly in line with Jaric, Corcos, Gottlieb, Ilic and Latash (1994) who used the equilibrium point model (Feldman 1966 a,b) to explain  55  the performance of participants in their study. Jaric et al. required participants to perform an elbow flexion aiming task. One group was instructed to aim from one of seven start positions to a single target location (the location group) while the second group was instructed to produce the same distance movement from one of seven start positions (the distance group). These authors found that on no vision transfer trials participants were not able to accurately reproduce distance. However, they were able to accurately reproduce an end position. In addition Jaric et al. found that when participants initiated a movement with their elbow in the most extended position they tended to overshoot the target, and like the results of the current study (see figures 14 and 15), as the start position shifted towards the participants mid-line the movements progressively became shorter. Indeed, as in Jaric et al., the finding that participants, in the present study, were coding their movements to end at a single common end point rather than to traverse a particular distance (45°) is well accounted for by the EP hypothesis, There were also position by group interactions in Acquisition and Retention. During Acquisition, although biased in the same direction as the KR, Static, and DD groups, the CD group was considerably more accurate at all positions (see figures 13 & 15 a & b). Recall, also, that during Acquisition the CD group used secondary submovements more often and more effectively than the other three groups. Thus suggesting that during Acquisition a visually based feedback loop may have been used to reset or possibly override a pre-set equilibrium point. In Retention, however, vision was not available and the participants in the CD grossly overshot the target (figure 15 b). Interestingly, like the other three groups, the CD group was again biased in Retention such that regardless of  56  start position they produced movements to similar end points. In sum, when vision was available there seemed to be an interplay between vision and the equilibrium point such that vision was being used to correct the equilibrium point and guide the movement towards the target. Indeed Ghez (1991), Khan et al. (1998) and Proteau and colleagues (1987,1990,1993) have all indicated that participants use vision to accurately guide their movements in the final phases of the movement execution. The CD group performed differently than the other three groups at each of the start positions in Acquisition and Retention. However, examination of the condition by group interaction revealed it was due to the CD group, who's performance differed between Acquisition and Retention. Once again the performance of the CD group was immediately and deleteriously affected by the removal of concurrent visual feedback while the affect of feedback removal on the other three groups was less severe (see figure 16). A guidance interpretation (Salmoni et al., 1984) can account for the decrement in performance experienced by the CD group in Retention. The performance of the CD group across conditions indicated that, while they practiced in concurrent visual feedback environment, they became increasingly reliant upon vision to update or modify the equilibrium point bringing their movements close to the target. In doing so they became increasingly reliant upon vision rather that improving their final endpoint representation. Therefore, when concurrent visual feedback was removed their movement error dramatically increased; and, similar to the other three groups, was more dramatically biased according to shifts in start position. In comparison, the effect of feedback removal on the performance of participants in the other three groups was much less severe; in fact  57  performance for these participants deteriorated to a greater extent with the one week absence from the task than the removal of feedback. Conclusion One of the goals of this study was to identify the process by which individuals incorporated feedback information into a movement strategy, such that their performance on subsequent trials improved. The results of the CD group provided support for the recent finding of Khan et al., (1998), that when vision is available, participants do come to rely upon it to guide corrective secondary submovements. However the data from this same group also provided evidence from KP data that supported the well established Guidance Hypothesis. That is, the dramatic and harmful effects of feedback removal strongly suggest that participants receiving concurrent visual feedback relied upon this source of information to make on-line corrections to their movements while ignoring intrinsic feedback modalities. When vision is available it is a powerful source of information and tends to override other available sources of feedback. An additional goal of this research was to determine if, when dynamic visual feedback was delayed, participants would use a two movement strategy to achieve the target; a process hypothesized to be similar to the two movement strategy used by participants receiving concurrent visual feedback. However, it was proposed that they would use the delayed visual feedback to calibrate their kinesthetic feedback modalities such that on subsequent trials they would make kinesthetically based secondary submovements. The performance of participants receiving delayed dynamic feedback, however, did not support this hypothesis. That is, they did not use a two movement  58  strategy to achieve the task. On the contrary, they reduced the frequency of secondary submovements and attempted to hit the target using only a single movement. Thus, unlike our previous findings (Appendix A), the process by which participants receiving delayed dynamic feedback and those receiving delayed static feedback produced the task did not differ. However varying the start positions may have made the kinesthetic feedback an unreliable source of information. Indeed the positional analysis indicated that all participants, particularly those practicing without concurrent visual feedback, coded their movements to end at a single common endpoint. Finally a positional analysis on CE confirmed that the performance of all participants was disrupted by the changing start locations. Incremental shifts in CE indicated that all participants (at least to some degree) were coding their movements to reach a particular end point rather than coding a particular amplitude of movement. Furthermore, this finding implied delayed visual feedback was not a superior form of error information. Moreover, the shifts in start position likely resulted participants disregarding kinesthetic representation of the movement because it would have appeared to be unreliable and, therefore, not contribute to producing and or improving the next trial. As a result participants practicing without concurrent visual feedback did not match a particular amplitude movement (one which would have reached the appropriate target location) with the start positions; instead movements appeared to be coded to a functional median end position (recall Figure 14). Conversely, when concurrent visual feedback was available, participants could make on-line corrections to the initial end-point code by producing visually based secondary submovements which accurately achieved the target. It is to be  59  stressed however, that vision was used in concert with the endpoint coding; as careful examination of figure 15a indicates that even the CD group was biased (albeit very slightly) according to the initial start positions. In Conclusion, the process by which participants produce their movements appears to depend upon the initial conditions associated with the practice setting. Here all participants coded their movements to a particular endpoint. It is likely that the variable start positions made the kinesthetic consequence of the movement an unreliable source of information. However when vision was available this powerful source of feedback was used to successfully compensate for the less accurate initial representation.  60  REFERENCES  Adams, J. A. (1971). A closed-loop theory of motor learning. Journal of Motor Behavior, 3(2), 111-149. Abrams, R.A., Meyer, D.E. & Kornblum, S. (1990). Eye-hand coordination: Oculomotor control in rapid aimed limb movements. Journal of Experimental Psychology: Human Perception and Performance, 16, 248-267. Abrams, R.A. & Pratt, J. (1993). Rapid aimed limb movements: Differential effects of practice on component submovements. Journal of Motor Behavior, 25, 288298. Brisson, T. A., & Alain, C. (1996). Should common optimal movement patterns be identified as the criterion to be achieved? Journal of Motor Behavior, 28(2), 211-223. Fitts, P. M . (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology, 47, 381391. Feldman, A. G., (1966a). Functional tuning of the nervous system with control of movement or maintenance of a steady posture-II. Controllable parameters of the muscles. Biophysics, 11, 565-578. Feldman, A. G., (1966b). Functional tuning of the nervous system with control of movement or maintenance of a steady posture-IH. Mecanographic analysis of the execution by man of the simplest motor tasks. Biophysics, 11, 667-675. Feldman, A. G. (1986). Once more on the equilibrium-point hypothesis (k) model for motor control. Journal of Motor Behavior, 18, 17-54. Gentile, A. M. (1972). A working model of skill acquisition with application of teaching. Quest, 17, 3-23. Ghez, C. (1991). The Control of movement. In E. R. Kandel, J. H. Schwartz, & T. M . Jessell (eds.), Principals of neural science (pp. 533-547). New York: Elsevfies Science Publishing Co. Hale, T & Franks, I. M . (1998). A comparison between graphic and numeric forms of summary KR in a movement reproduction task. Journal of Human Movement Studies, 35, 1-19. Hale, T., Hodges, N., Khan M . A. & Franks, I. M. (in preparation). A comparison of discrete versus continuous forms of augmented feedback when the task goal is not isomorphic with the movement pattern.  61  Henry, F. M., (1968). Specificity Vs. generality in learning motor skill. In R.C. Brown & G. S. Kenyon (Eds.), Classical studies on physical activity (pp. 331-340). Englewood Cliffs, NJ: Prentice-Hall. (Original work published in 1958). Jaric, S., Corcos, D. M., Gottlieb, G. L., Ilic, D. B., & Latash, M . L. (1994). The effects of practice on movement distance and final position reproduction: implications for the equilibrium-point control of movements. Experimental Brain Research, 100, 353359. Khan, M. A., Franks, I. M. & Goodman, D. (1998). The effect of practice on the control of rapid aiming movements: Evidence for and interdependency between programming and feedback processing. Quarterly Journal of Experimental Psychology, 51 A, 425-444. Meyer, D.E., Abrams, R.A., Kornblum, S., Wright, C E . & Smith, J.E.K. (1988). Optimality in human motor performance: Ideal control of rapid aimed movements. Psychological Review, 95, 340-370. Meyer, D.E., Smith, J.E.K., & Wright, C E . (1982). Models for the speed and accuracy of aimed movements. Psychological Review, 89,449-482. Newell, K. M., & Carlton, M. J. (1987) Augmented information and the acquisition of isometric tasks. Journal of Motor Behavior, 19(1), 4-12. Newell, K. M., & Carlton, M . J. & Antoniou, A. (1990). The interaction of criterion and feedback information in learning a drawing task. Journal of Motor Behavior, 22, 536-552. Newell, K. M., Sparrow, W. A., & Quinn, J. T., Jr. (1985). Kinetic information feedback for learning isometric tasks. Journal of Human Movement Studies, 11, 113-123. Newell, K. M., Quinn, J. T., Sparrow, W. A., & Walter, C. B. (1983). Kinematic information feedback for learning a rapid arm movement. Human Movement Science, 2 255-269. Polit, A., & Bizzi, E. (1979). Characteristics of motor programs underlying arm movements in monkeys. Journal of Neurophysiology, 42,183-194. Proteau, L. & Cournoyer, J. (1990). Vision of the stylus in a manual aiming task: The effects of practice. Quarterly Journal of Experimental Psychology, 42 A, 811-828. Proteau, L. & Marteniuk, R. G. (1993). Static visual information and the learning and control a of manual aiming movement. Human Movement Science, 12, 515-536.  62  Proteau, L., Marteniuk, R. G., Girouard, Y. & Dugas, C. (1987). On the type of information used to control and learn an aiming movement after moderate and extensive training. Human Movement Science, 6, 181-199. Pratt, J. & Abrams, R.A. (1996). Practice and component submovements: The roles of programming and feedback in rapid aimed limb movements. Journal of Motor Behavior, 28, 149-156. Salmoni, A. W., R. A. Schmidt & Walter, C. B. (1984). Knowledge of results and motor learning: A review and critical reappraisal. Psychological Bulletin, 95, 355-386. Schmidt, R. A., Young, D. E. (1991). Methodology for motor learning: A paradigm for kinematic feedback. Journal of Motor Behavior, 23, 13-24. Schmidt, R. A., Young, D. E., Swinnen, S. & Shapiro, D. C. (1989). Summary knowledge of results for skill acquisition: Support for the guidance hypothesis. Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 352-359. Young, D. E. & Schmidt, R. A. (1992). Augmented kinematic feedback for motor learning. Journal of Motor Behavior, 24,261-273.  63  APPENDIX A ABSTRACT In this study two forms of augmented feedback were compared to examine theneffects on the acquisition of a rapid aiming movement. One group received a static, graphic representation of their movement displacement as a function of time while another received a real-time, dynamic playback of their movement. The two groups did not differ significantly with respect to accuracy and movement time. However, on analysis of the kinematic data, the process of goal attainment differed between groups. Specifically, the Dynamic group tended to produce faster primary submovements and had more trials containing corrective secondary submovements than the Static group. Consequently the Dynamic group consumed a smaller proportion of the total movement time and distance in the primary submovement phase as compared to the Static group. These different strategies were evident in both immediate and delayed Retention. It is argued that the dynamic feedback provides a richer source of afferent information which enables participants to form a stronger association (calibration) with visual information and the kinesthetic consequence of the movement.  64  A comparison of static and dynamic forms of augmented feedback during the acquisition of a rapid aiming movement  Augmented feedback has long been considered a critical element for the acquisition and development of motor skills. In particular it has been proposed to enhance intrinsic error detection and correction capabilities (Adams, 1971; Schmidt, 1975). Many different types of augmented information have been explored in relation to their effects on performance and learning. For example, researchers have compared movement outcome (knowledge of results, KR) to movement characteristics (knowledge of performance, KP), quantitative versus qualitative information and visual versus verbal feedback. In this paper, we compared two forms of KP, one which provides static information about limb trajectory and another which provides this information in dynamic form. What makes this comparison especially interesting is that both forms of KP were similar in terms of the quantity of information, but differed in how this information was presented. This is an important consideration for KP research since it allows inferences to be made regarding the qualitative aspects of KP that enhance performance and learning. KP typically provides information about the kinematic or kinetic aspects of the movement, that are usually provided in a static form, post-movement execution (e.g., displacement-time graph, or force production curve). KP can also be provided in real-time as a dynamic video replay of the action. Given the compatibility of this real-time feedback with the movement trajectory, one may suspect this to be especially helpful in enhancing the calibration and development of sensory-motor representations that are postulated to  65  guide movement production (see Proteau & Marteniuk, 1993). Whether static feedback in the form of limb kinematics would be as, or more informative than this more naturalistic, dynamic feedback is questionable. The static nature of the kinematic feedback and the greater skill in interpretation may lead one to suspect not. Dynamic kinematic feedback may be more informative about specific aspects of the movement trajectory, such as the timing and force components of the initial acceleration phase. In this way performance may be enhanced by focusing the individual's attention on potentially changeable aspects of the movement trajectory. Indeed, Kernodle and Carlton (1992) found that benefits of video-feedback were more apparent when subjects were directly cued into important aspects of the movement trajectory. Studies that have examined the effects of KP have often used tasks that require the production of a particular movement pattern (criterion template). Therefore, the effects of KP are limited by the task constraints. That is, it is unclear how KP impacts on learning beyond it's function as comparative information (cf., Brisson & Alain, 1997). In the present experiment we used a rapid aiming task that had no task attainment constraints, where participants were instructed to move from a home position to a target as quickly and as accurately as possible, but were not given any instruction as how this goal was to be accomplished An advantage of using aiming movements is that programming and sensory processes can be investigated by parsing kinematics into their primary and secondary submovements. The primary phase is believed to be of central origin, and planned in advance of the movement. This primary submovement traverses most of the distance from  66  the start position to the target. In contrast, the secondary phase has been termed an errorcorrection phase, containing one or more secondary submovements (e.g., Meyer et al., 1988; Woodworth, 1899). Error corrections are indexed by discontinuities in the kinematic profiles and are said to reflect the presence of on-line adjustments to the movement (e.g., van Donkelaar & Franks, 1991). It is here, based on sensory information, that an attempt is made to reduce any discrepancy between the endpoint of the primary submovement and the position of the target. In performing rapid aiming movements, several strategies are at the participants' disposal. For example, they can produce very rapid primary submovements and then rely on sensory feedback to home in on the target. On the other hand, they can produce slower, less variable primary submovements, thereby reducing the need for secondary submovements. According to Meyer et al. (1988), achieving maximum performance entails adjusting initial impulse velocity such that the combined duration of the primary submovement and error correction phase is minimized. Although past research has shown that the control strategy was independent of the availability of on-line sensory information (Meyer et al., 1988; Pratt & Abrams, 1996), recent work in our lab. has shown that different strategies do emerge after extensive levels of practice. Participants modulated either the distance traveled (Khan, Franks & Goodman, 1998) or the time spent (Khan & Franks, 1998) in the primary submovement depending on the availability of visual feedback. It was suggested that participants who practiced with visual feedback became effective at processing on-line sensory information and therefore programmed their movements to make optimal use of vision to guide their  67  movements to the target. In contrast, when visual feedback was not available, movements were planned to decrease reliance on sensory feedback. Given the proposed importance of on-line sensory information we were interested in the role augmented feedback would play in the control of rapid aiming movements, where on-line visual information is not available. Augmented feedback has been said to be important for developing sensory motor representations that can be used during the online control of a movement. If augmented feedback serves to strengthen error-detection and correction capabilities then one would predict that post-response augmented feedback may enhance subjects ability to use the sensory information provided on-line (i.e., kinesthetic feedback). Particularly, it was predicted that dynamic visual feedback (i.e., a visual play-back of the cursor trajectory) provided in real-time, would be especially helpful in calibrating the kinesthetic response-produced feedback, so that on subsequent trials the sensorimotor representation (Proteau, 1992) would be more developed, allowing for greater sensitivity to any errors in the movement. Subsequently, participants who receive dynamic visual feedback will show increased performance relative to a group who does not receive such a rich source of information. For participants who received this information in the form of a displacement-time graph (the Static group), it was expected that due to the static nature of the feedback, the calibration of this KP with the kinesthetic information would not be as strong as the dynamic visual information. If this dynamic source of information does serve to enhance error correction capabilities it would be interesting to investigate whether movements are indeed planned to take advantage of this enriched sensory information. Analysis of movement kinematics was expected to illustrate  68  differences between the two groups with respect to programming and feedback processing. METHOD  Participants Twenty (8 male, 12 female) right handed individuals volunteered to take part in the experiment. All participants were paid $20 CDN. for participation. In addition, a $25 CDN. prize was offered to the top performer in each group. The experiment was conducted according to the ethical guidelines of the University of British Columbia. Apparatus Participants were seated at a table with their shoulders harnessed to a chair. Both a computer monitor and an oscilloscope screen were mounted 35 cm. above the table 80 cm. in front of the subject. The right forearm of each participant was positioned on a horizontal manipulandum such that the elbow was co-axial with the vertical axis of rotation. The right hand was placed palm down on a platform and secured with Velcro straps. This allowed the elbow to rotate freely in the horizontal plane. The arm and hand were hidden from the subjects' view by an opaque shield. Visual displays of the home position, target region and a cursor representing limb position appeared on the oscilloscope screen that was positioned in front of the participants. Extension and flexion movements of the elbow caused the cursor to move to the right and left, respectively. The target was defined by a single point (approximately  69  one pixel) with its center 45° of angular distance from the home position (this translated to 10 cm of movement of the cursor on the oscilloscope screen). An optical encoder (Dynapar E20-2500-130), attached to the shaft of the manipulandum and custom made computer interface card enabled high speed sampling of angular position (sampling rate was 1000 Hz). A Kistler accelerometer (type 8638B50, + 50G), mounted at the distal end of the manipulandum (45 cm from the axis of rotation), was used to record angular acceleration. Its signal (measured in milli-volts) was filtered with an active low pass filter (Krone-Hite, # 3750) set at 50 Hz and sampled at 1000 Hz. Procedure At the beginning of each trial, participants positioned the manipulandum at a fixed stop (such that the elbow angle was 75°, where 180° is defined as full elbow extension). At this point a high frequency (1000 Hz) tone was presented. Participants were free to initiate their movements anytime within 1500 ms following the onset of the tone and were informed that it was not necessary to minimize reaction time. The cursor disappeared once movement velocity exceeded 8° / s, and did not reappear until the following trial (or until feedback presentation for half of the participants). Participants were instructed to move from the home position to the target as accurately as possible while minimizing movement time. They were also required to hold their movement at the endpoint for at least 500 ms, after which they could return to the home position. Approximately 5 s after the end of the movement, feedback was presented in accordance to group assignment and was available for about the same amount of time.  70  Participants were randomly assigned to one of two feedback groups; Static or Dynamic. Participants in both groups received numeric KR consisting of target accuracy (constant error in degrees), total movement time (TMT), primary submovement time (MT1), and secondary submovement time (MT2). In addition to numeric feedback, participants in the Static group also received a displacement-time graph. Markers were placed along this trace indicating movement onset, end of primary submovement, and the end of secondary submovement(s). The placement of markers was determined by a movement parsing algorithm (see below). Participants in the Dynamic group were presented with a replay of the cursor moving across the oscilloscope screen in real time. After receiving task instructions, all participants were given one practice trial after which all possible forms of feedback were presented and explained to them. The study involved three acquisition sessions (consisting of 120 trials each) separated by 24 hours. There were also two no KR Retention sessions consisting of 60 trials each. The first was performed 5 minutes after the last acquisition trial, and the second one week later. Data reduction Movements were parsed into their component submovements using a movement parsing algorithm that was originally developed by Meyer et al. (1988) and subsequently modified by Khan et al. (1998). Movements were judged to contain an error correction phase if one of the following movement modifications were observed; (a) a positive to negative zero line crossing in velocity, (b) a negative to positive zero line crossing in the acceleration trace, or (c) a significant deviation in the acceleration trace, i.e., a relative minimum in the absolute value of the acceleration while the acceleration is negative. In  71  order to qualify as a significant deviation, neither a preceding nor succeeding absolute maximum could lie within 30 ms of the relative minimum (see also van Donkelaar & Franks, 1991). Afso, the difference in the absolute values of acceleration between the minimum and maximums has to be at least 1007 s. On those trials in which none of the above modifications had occurred, the movement was recorded as containing only an initial impulse phase. Dependent measures included total movement time (TMT), primary and secondary submovement time (MT1 and MT2) and the proportion of TMT in these respective phases. Constant and variable error (CE and VE) provided measures of accuracy. A l l dependent measures were analyzed in acquisition by a 2 group (Static, Dynamic) x 3 day x 6 block mixed ANOVA, with repeated measures on the last two factors. Immediate and delayed Retention data were compared to the final block of acquisition and analyzed in a 2 group (Static, Dynamic) x 3 experimental phase (acquisition, immediate (RI) and delayed (R2) Retention) x 3 block, mixed ANOVA, with repeated measures on the final two factors.  RESULTS Acquisition Performance measures CE, VE and TMT failed to yield significant differences between the two groups F > 1. However, all groups showed improvements in performance as a function of practice as evidenced by significant day and block effects for VE (F(2,36) = 35.29, p<.001; F(5,  72  90) = 42.35, p<.001) and TMT (F(2,36) = 9.95, p<.001; F(5,90) = 10.65, p<.001). CE yielded a main effect for block only, F(5,90) = 11.45, p<.001. Submovement Analysis The absolute time spent in the primary submovement (MT1) and the proportion of the total movement time spent in the primary submovement (MT1/TMT) tended to be greater for the Static group as compared to the Dynamic group, F(l, 18) = 2.42 p = .13 and F(l, 18) = 4.13, p = .069, respectively. For both groups this ratio increased as a function of day, F(2, 36) = 12.48, p < .001 and block, F(5, 90) = 9.54, g < .001. When a separate analysis was conducted on the last day of acquisition for MT1/TMT, the Static group was found to spend a significantly greater portion of their total movement time in the primary submovement phase as compared to the Dynamic group, F(l, 18) = 4.77 p<.05. (see Figure 17 A and B) Consistent with the above data, the number of trials with an error correction phase was observed to be significantly different between the groups F(l, 18) = 5.16, p <.05. The Dynamic group had significantly more trials with an error correction phase than the Static group. In addition, there were significant day F(2,36) = 11.52, p < .001 and block F(5, 90) = 6.38, p < .001 effects indicating that both groups decreased the number of secondary submovements with increasing practice (see Figure 17 C).  73  74  Retention Performance measures Similar to the acquisition data, there were no significant differences between groups with respect to the performance measures, i.e., CE, VE and TMT. Submovement analysis The Dynamic group continued to produce a primary submovement that was on average 50 ms faster than the Static group, F(l, 18) = 3.37, p = .08 (see Figure 17 A). Consistent with these findings the proportion of total movement time (MT1/TMT) and total movement distance spent in the primary submovement phase was greater for the Static group than the Dynamic group, F(l, 18) = 6.714, p. < .05 and F(l, 18) = 4.515, p < .05, respectively (see Figure 17 B). The Dynamic group also produced more trials containing an error correction phase than the Static group F(l, 18) = 7.15, p < .05. There was no group x experimental phase interaction F < 1, indicating that the difference between groups did not change between acquisition and Retention (see Figure 17 C). DISCUSSION  The goal of this study was to contrast two forms of augmented kinematic feedback (KP) and evaluate their effect on learning to hit a target under instructions to move as fast and as accurately as possible. More specifically, it was our intent to have participants perform a task with no prescribed attainment constraints in order to examine the effects of static and dynamic modes of KP on the performance and strategies used to acquire the task goal. We predicted that the more natural and dynamic form of KP would facilitate a  75  calibration process between KP and proprioceptive feedback, resulting in superior performance and learning when compared to static feedback. The performance of both groups during acquisition resembled that which is commonly seen in learning studies (e.g., Schmidt, Young, Swinnen & Shapiro, 1989). That is, a rapid improvement in performance early in acquisition after which there was little improvement or change in accuracy measures. These performance plateaus may be due to both the simplicity of the task and the fact that all participants received detailed, numeric feedback, relating to goal attainment after every trial. While our measures of performance indicated that both feedback groups attained similar levels, a more detailed examination of the submovement data indicated that the process by which these groups attained the task goal differed. For the Dynamic group, the primary submovements were on average 50 ms faster than the Static group. In addition, the Static group had significantly fewer trials containing an error correction phase. Indeed, by the end of acquisition, the Static group was performing the task using an error correction phase on only 40 % of their trials as compared to the Dynamic group who produced submovements on 75% of their trials. As a consequence, the Static group spent a greater proportion of total movement time and traveled a larger proportion of the total distance in the primary submovement phase, as compared to the Dynamic group. These differences in movement strategy were maintained in Retention. Interestingly, these differences between the Static and Dynamic groups are similar to those reported in past research which has investigated the importance of on-line sources of sensory information. Specifically, participants produced more error corrections when  76  on-line visual feedback is available compared to when it is not. It was reasoned that vision, which provides a richer source of afferent information than proprioceptive feedback, enabled participants to make effective error corrections to attain the target (e.g., Khan et al., 1998). In the present study, although both groups practiced without on-line visual information, the Dynamic group produced significantly more error corrections than the Static group. We believe that the enriched dynamic feedback (which was basically a visual replay of the movement) was such that participants could use it to calibrate an internal reference of the movement. This interpretation is in agreement with Brisson and Alain (1996) who proposed that "The role of KP may be simply to augment information about the movement pattern that may otherwise go undetected, and to make it more salient so that the calibration between the movement and the outcome is facilitated" (p. 222). The difference in strategies suggest that participants receiving dynamic feedback underwent similar processes of on-line error detection and correction. That is, they may have calibrated their proprioceptive feedback mechanisms from the visual information provided in the dynamic display and then used this internal reference to make on-line corrections. Proteau and Marteniuk (1993) have suggested that the calibration of proprioceptive feedback from visual information is volatile. In their work, one group of participants was provided with on-line visual feedback. It is possible that since visual feedback was presented concurrent with the movement in both cases, attention may not have been directed to developing an internal reference. Another group received visual feedback only at the endpoint of the movement (visual KR group). Because visual  77  feedback was only available about the outcome of the movement, there was probably not sufficient information to properly calibrate proprioceptive feedback. In contrast, the Dynamic group in the present study received delayed visual information of the entire movement trajectory. This dynamic form of KP might have enabled participants to form a stronger association (calibration) between visual information and the kinesthetic consequence of the movement. Indeed further evidence that this calibration process remains resistant to the decay is that the Dynamic group continued to use a two movement strategy during the no feedback transfer, even after a one week Retention interval. In conclusion, both forms of KP had differential effects on how participants performed the task. In particular it appears that participants can and do benefit differentially based on the kinematic information (KP) that they receive. Although no differences were observed with respect to goal attainment it may be that with a more complex task performance differences will be detected as a function of the type of kinematic feedback provided. It is expected that the more descriptive / continuous feedback formats provided after the completion of the movement, such as a video replay of the just-completed movement, will prove to be more useful for the acquisition of motor skills, facilitating calibration with intrinsic sources available during the movement itself.  78  R E F E R E N C E S  Adams, J.A. (1971). A closed-loop theory of motor learning. Journal of Motor Behavior. 3.111-149. Brisson, T. A., & Alain, C. (1996). Should common optimal movement patterns be identified as the criterion to be achieved? Journal of Motor Behavior. 28, 211-223. Brisson, T.A., & Alain, C. (1997). A comparison of two references for using knowledge of performance in learning a motor task. Journal of Motor Behavior. 29, 339350. Kernodle, M.W. & Carlton, L.G. (1992). Information feedback and the learning of multiple-degree-of-freedom activities. Journal of Motor Behavior. 24, 187-196. Khan, M.A., & Franks, L M . (1998). The effect of practice on component submovements is dependent on the availability of visual feedback. Manuscript submitted for publication. Khan, M . A., Franks, I. M., and Goodman, D. (1998). The effect of practice on the control of rapid aiming movements: Evidence for an interdependency between programming and feedback processing. The Quarterly Journal of Experimental Psychology. 51A. 425-444. Meyer, D.E., Abrams, R.A., Kornblum, S., Wright, C E . & Smith, J.E.K. (1988). Optimality in human motor performance: Ideal control of rapid aimed movements. Psychological Review. 95, 340-370. Pratt, J., & Abrams, R.A. (1996). Practice and component submovements: The roles of programming and feedback in rapid aimed limb movements. Journal of Motor Behavior. 28,149-156. Proteau, L. (1992). On the specificity of learning and the role of visual information for movement control. In L. Proteau & D. Elliott (Eds.), Vision and motor control (pp. 67-103). Amsterdam: North Holland. Proteau, L., & Marteniuk, R.G., (1993). Static visual information and the learning and control of a manual aiming movement. Human Movement Science. 12, 515-536. Schmidt, R. A. (1975). A schema theory of discrete motor skill learning. Psychological Review. 82, 225-260.  79  Schmidt, R. A., D. E. Young, S. Swinnen & Shapiro, D. C. (1989). Summary knowledge of results for skill acquisition: Support for the guidance hypothesis. Journal of Experimental Psychology: Learning. Memory and Cognition. 15, 352-359. Woodworth, R. S. (1899). The accuracy of voluntary movement. Psychological Review. 3, (Monograph Supplement), 1-119. van Donkelaar, P. & Franks, I. M. (1991). The effects of changing movement velocity and complexity on response preparation: Evidence from latency, kinematic, and EMG measures. Experimental Brain Research. 83,618-632.  80  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0077247/manifest

Comment

Related Items