Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Preprogramming vs. on-line preparation in simple movement sequences Van Donkelaar, Paul 1990

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

Item Metadata

Download

Media
831-UBC_1990_A7_5 V36.pdf [ 5.3MB ]
Metadata
JSON: 831-1.0077255.json
JSON-LD: 831-1.0077255-ld.json
RDF/XML (Pretty): 831-1.0077255-rdf.xml
RDF/JSON: 831-1.0077255-rdf.json
Turtle: 831-1.0077255-turtle.txt
N-Triples: 831-1.0077255-rdf-ntriples.txt
Original Record: 831-1.0077255-source.json
Full Text
831-1.0077255-fulltext.txt
Citation
831-1.0077255.ris

Full Text

PREPROGRAMMING VS. ON-LINE PREPARATION IN SIMPLE MOVEMENT SEQUENCES by PAUL VAN DONKELAAR B.P.E, The University of British Columbia, 1987 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF PHYSICAL EDUCATION in THE FACULTY OF GRADUATE STUDIES School of Physical Education and Recreation We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA February 1990 © Paul van Donkelaar, 1990  In presenting degree  this thesis  in partial fulfilment of  at the University of  of this thesis for scholarly  department  or  by  his  or  requirements  for  her  I further agree that permission for  purposes  may be granted  representatives.  permission.  Physical  Education  The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  F e b . 19,  1990  advanced  It  is  extensive  by the head of my  understood  that  publication of this thesis for financial gain shall not be allowed without  Department of  an  British Columbia, I agree that the Library shall make it  freely available for reference and study. copying  the  copying  or  my written  ABSTRACT If movement control is afforded through the advance planning, or preprogramming, of upcoming actions, then one of the behavioral outcomes should be an increase in reaction time (RT) as the movement becomes more complex. In some situations, however, RT does not increase across levels of complexity, rather it remains invariant. In these cases, on-line preparation is typically inferred. That is, the sequence is said to be prepared in parts throughout the movement, as opposed to entirely beforehand. Given that there is some planning occurring during the sequence, then evidence of this process should be apparent within the movement itself. Three such dependent variables appear to provide such evidence. Specifically, the number of times the underlying accelerations cross the zero line within the movement, the number of "significant deviations" within the acceleration trace, and the length of time for which the muscles are active (as measured by EMG) in relation to the duration of the movement. In the present experiment, then, these variables were measured in addition to the time required to prepare and initiate a movement performed under conditions conducive to either preprogramming or on-line preparation. Specifically, the movements were either completed as fast as possible, or at a considerably slower, more controlled speed. Each of the dependent variables displayed evidence of preprogramming in the movements completed at the fast velocity, and on-line preparation in the slower paced movements. Thus, in the slow condition, subjects appeared to rely more heavily on on-line prepared adjustments to produce an accurate outcome. The convergence attained between the various dependent measures lends power to the conclusions regarding hypothesized modes of control within the different speeds of movement. ii  TABLE OF CONTENTS Abstract  .  .  .  .  .  .  .  ii  List of Tables.  .  .  .  .  .  .  v  List of Figures  .  .  .  .  .  .  vi  Acknowledgment 1. Introduction  ix  .  .  .  .  .  .  .  1  2. Methods  9  Subjects  .  .  .  .  Task and Apparatus.  .  .  .  .  .  . .  9 .  9  Independent Variables  12  Experimental Procedure and Design  14  Dependent Variables and Data Reduction  15  Statistical Analysis  17  3. Results and Discussion  .  18  Latency Analysis  18  Kinematic Analysis .  .  .  .  .  .  25  EMG Analysis  37  4. General Discussion References .  .  . .  .  .  .  .  .  .  .  .  .  46  .  56  Appendix A: Review of Literature  .  .  .  .  .  .  61  RT Studies  61  Preprogramming and On-Line Preparation  67  Kinematic Studies  69  .  .  .  .  EMG Studies Representationalism  72 .  .  . iii  .  .  74  Appendix B: Optimization of Displacement Cutoff Frequencies  .  .  .  .  .  .  .  .  77  Appendix C: EMG Sampling and Analysis  .  .  .  82  Appendix D: Pilot Study Introduction  87 .  .  .  .  .  .  .  Method Results and Discussion  87 91  .  .  .  General Discussion .  .  .  95 108  iv  LIST OF TABLES Table 1. Mean Number of Acceleration Zero Line Crossings at each Level of Complexity in each Movement Speed C o n d i t i o n . . . . . .  V  LIST OF FIGURES Figure  Page  1. Experimental Apparatus  10  2. Reaction Time as a function of Number of Cycles  19  3. PremotorTime as a function of Number of Cycles  .  .  .  .  .  .  20  .  .  .  .  .  22  4. Motor Time as a function of Number of Cycles  .  5. Average Velocity (degrees/sec) as a function of Number of Cycles  .  .  .  .  24  .  .  .  .  26  6. Time to the 1st Peak in Acceleration as a function of Number of Cycles 7. Kinematics of Fast Movement .  .  .  .  .  8. Kinematics of Slow M o v e m e n t . . . . .  27 28  9. Number of Zero Line Crossings as a function of each Segment .  .  .  .  .  30  .  .  .  .  33  10. Number of Deviations as a function of each Segment  .  vi  Figure 11. Relative Temporal Location of the Deviations - SLW Condition. 12. Time Between each Deviation SLW Condition  .  .  .  .  13. Typical EMG Data - SLW Condition . 14. Typical EMG Data - FAP Condition . 15. Ratio of EMG Duration to Segment Duration as a function of each Segment 16. RMS Error Between Raw Displacement Data and Filtered Data as a function of Filter Cutoff Frequency 17. Comparison of Accelerometer and Differentiated Data across a range of Cut-off Frequencies - SLW Condition 18. EMG Start and Stop Times as a Function of Sampling Rate for 4 Different Algorithms 19. EMG Start and Stop Times as a Function of Sampling Rate for 2 Different Algorithms  vii  Figure  Page  20. Reaction Time as a function of Number of Cycles  96  21. Reaction Time as a function of Number of Cycles  98  22. Kinematics of Control Movement  102  23. Kinematics of Fast Movement.  103  24. Distribution of the Deviations Collapsed Across all Conditions - SLW vs. CNT  105  25. Time Between each Deviation -SLWvs. CNT  106  viii  ACKNOWLEDGEMENTS This thesis would not have been realized without the combined help of many people: Dr. Ian Franks - for his stringent, yet commonsensical approach to research, and driving curiosity of motor control; Dr. Sanderson - for his biomechanical expertise; Dr. Schutz - for keeping me on the right statistical track; Dr. Ward - for his knowledge in the areas of perception and cognition; Paul Nagelkerke - for his invaluable programming and hardware skills; my parents and family - for their nutritional, residential and, most important, financial support; and, last, but not least, Cat - for showing me what hard work and patience can bring.  ix  CHAPTER ONE INTRODUCTION One of the underlying questions in motor behavior research is: "How do we control our movements?". Those adopting a more cognitive approach have attempted to answer this question by suggesting that we plan (preprogram) upcoming actions in advance. If it is assumed that this planning process occurs in a serial manner, then one of the predictions from this potential answer is that a simple movement will take a shorter amount of time to plan than a more complex movement. This prediction has been tested empirically by measuring the reaction time (RT) required to prepare and initiate movements which vary in complexity. Typically, it has been found that the more complex movements require a greater amount of time to prepare and initiate than the simpler movements. If it is assumed that each unit of complexity causes a constant increase in RT, then, taken to a theoretical-logical extreme, a very complex movement would require a relatively long time to initiate. However, this does not typically occur. In some situations, increasing movement complexity does not lead to increased latencies prior to the movement. Instead, RT remains relatively invariant across levels of complexity. Indeed, the relationship between RT and movement complexity appears to break down when the movement sequences under consideration are of long duration or composed of a large number of elements. How can this lack of increase in RT with increased movement complexity be explained while still maintaining the hypothesis that movements are planned in advance? One possibility is that in such situations some aspect of the interaction between the actor and the action allows the movement to be prepared in parts during the movement as opposed to entirely beforehand. Such a form of control is typically referred to as "on-line" preparation. The purpose of the present investigation was to compare movements l  performed under conditions conducive to either preprogramming or on-line preparation. In previous research within the RT/movement complexity paradigm, on-line preparation has typically only been inferred from a lack of increase in RT with increased movement complexity. However, there appear to be other dependent  variables which  are  sensitive to the  differences  between  preprogrammed and on-line prepared movements. Two such measures, the angular acceleration traces and EMG profiles, were used in the present study in an effort to uncover the process by which on-line preparation may occur. Henry and Rogers (1960) were among the first to show that the time required to initiate a movement sequence is related to the complexity inherent in that sequence. Specifically, they demonstrated that a simple key lift response was initiated significantly more quickly than a response composed of a key lift and additional movements to specified targets. Since their paper, the RT/movement complexity paradigm has been extensively studied. Various researchers have shown that RT increases with increased movement complexity using such tasks as keystrokes and tapping (Fischman, 1984; Klapp & Rodriguez, 1982; Rosenbaum & Patashnik, 1980), speech (Erikson, Pollack, & Montague, 1970; Sternberg, Monsell, Knoll, & Wright, 1978), and handwriting (Hulstijn & van Galen, 1983; Teulings, Mullins, & Stelmach, 1986). The parameter which appears to be most strongly related to changes in RT in these studies is the number of response elements that comprise the movement. Thus, it is the number of taps or keystrokes, number of stress groups in a sequence of speech, and number of strokes taken in writing a letter which have the greatest impact on the time required to prepare and initiate these responses. Studies which have confirmed the relationship between RT and movement complexity have typically used maximal speeds of response. Specifically, 2  subjects were not only required to initiate their movements as quickly as possible but also complete them in as short a time as possible. More recently, a number of investigators, using variable rates of response, have uncovered some interesting results (Canic & Franks, 1989; Franks & van Donkelaar, in press; Garcia-Colera & Semjen, 1987; 1988; van Donkelaar & Franks, 1989a; 1989b). They found RT increased linearly in relation to the number of movement elements at the quickest rates, but either failed to do so, or did so in a nonlinear fashion at slower rates. Thus, when a movement sequence is completed at a less than maximal speed, changing the complexity of it by adding extra elements does not reliably cause an increase in the RT required to prepare and initiate it. This finding has been explained by suggesting that some form of "on-line" or parallel system of preparation is taking place (Rosenbaum, Hindorff, Munro, 1986). The logic is as follows: If subjects preprogram a movement sequence, then increasing the number of elements within that sequence should lead to an increased latency during response initiation. However, if increasing the number of elements does not cause an increase in RT, then the subjects must not have to plan the entire sequence beforehand; rather, some aspect of the this process can carry on into the period of movement execution. If the planning process does not occur during the period prior to actual movement, then some evidence of it should be present within the movement itself. Since, by definition, the planning process takes time, several researchers have looked to the interresponse intervals (IRI's) for such evidence. Ostry (1980, 1983) has been perhaps the most successful in this regard. He has found that in typing a series of keystrokes, subjects tend to lengthen the IRI's in the middle of the sequence [Sternberg, Monsell, Knoll, & Wright (1978) found a similar result but did not discuss it at any length]. Ostry suggested that this midsequence slowing was used by the subjects to prepare on-line the terminal elements in the 3  sequence. Such a form of preparation is required because, according to Ostry, we have a limited "motor output span". In a similar vein, another line of IRI evidence has been used to demonstrate the differences between preprogrammed movements and those prepared on-line. Specifically, Canic (1988) has found that the RT required to prepare and initiate isochronous sequences of tapping movements increased as the number of taps increased at fast rates (200 and 300 msec/tap), but failed to do so at slow rates (400, 600 and 800 msec/tap). In addition, the IRI's produced in such sequences displayed lengthened beginning and terminating elements at the fast rates but not the slow rates. Povel and Essens (1985) have suggested that such temporal accenting (termed "agogic" accenting) reflects a tendency to perceive the sequence under question as a unitary whole. Thus, in Canic's study it appears that the individual taps in the sequences completed at the faster rates were both perceived and prepared (as evidenced by the IRI and RT data, respectively) as a group. On the other hand, the taps produced in the sequences completed at the slower rates appeared to have been perceived and prepared individually. In such a situation, it is likely that each tap was prepared during the execution of those preceding it. Evidence from serial pattern learning studies has also demonstrated the viability of analyzing the variations within IRI's. In studies by Povel and Collard (1982) and Rosenbaum, Kenny, and Derr (1983) the IRI's between movement sequence subunits in a discrete movement pattern were lengthened in relation to those within the subunits. This was used as evidence to suggest that the patterns were prepared in parts as the movement was being executed. Similarly, in a study by Franks, Wilberg, and Fishburne (1985) probe RT to a secondary task was measured at points within and between subunits of the pattern. If the patterns were planned entirely before their execution, then probe 4  RT should remain similar at each point in time. However, it was found that probe RT was greater during the intervals between subunits, suggesting that more processing was required during this part of the pattern. This increase in probe RT also indicated that parts of the pattern were prepared on-line as opposed to entirely beforehand. Unfortunately, other than the studies by Canic (1988) and Ostry (1980, 1986), little else has been done within the RT/movement complexity paradigm to uncover how and when the on-line preparation process occurs during the movement sequence. However, other studies which do not use RT as a dependent variable have been completed which have revealed some interesting aspects regarding this process. Two such types of studies - those examining the kinematics of a movement (specifically, the acceleration traces), and those concerned with the EMG activity underlying the movement - appear to allow considerable insight into the differences between preprogrammed movements and those prepared on-line. There are two aspects of an acceleration trace which reflect the presence of on-line prepared adjustments in a movement. The first concerns the number of times the acceleration trace crosses the zero line during a particular movement segment.  Brooks, Cooke, and Thomas  (1973)  have  suggested  that  preprogrammed movements result in acceleration traces which cross the zero line only once during each movement segment. On the other hand, movements in which on-line prepared adjustments are made as the result of feedback and/or feedforward lead to acceleration traces with several zero line crossings within each segment. The second way that on-line prepared adjustments can be inferred from an acceleration trace is through the analysis of "significant deviations". Typically, a preprogrammed movement results in a relatively smooth symmetrical 5  acceleration trace (Crossman & Goodeve, 1963/1983). However, when adjustments are made during the movement, these show up in the acceleration trace as deviations from the normally smooth curve (Young, Allard, & Marteniuk, 1988). Carlton (1981) found that such adjustments occur as a target is being approached in a single aiming movement task. Thus, looking at the topology of the acceleration trace may provide a clue as to the underlying mode of control. Analysis of the EMG activity associated with a particular movement can provide pertinent information regarding the duration and intensity of firing within a specific set of muscles (Hof, 1984). Like the acceleration traces, differences have also been found in the EMG activity underlying preprogrammed movements and movements prepared on-line. Desmedt and Godaux (1979) have shown that during a loaded ballistic (fast as possible) index finger abduction movement to a target the integrated EMG of the interosseous muscle appeared as a brief burst 50 to 100 msec in duration. Furthermore, muscle firing started before the movement was initiated and was completed by the time angular displacement had reached mid-amplitude. In contrast, ramped (slow, controlled) movements were characterized by integrated EMG's extended over the length of the abduction movement. This evidence in itself does not specifically prove that the fast movements were preprogrammed and the slow ones, prepared on-line. In order to show that this was, indeed, the case, Desmedt and Godaux included a second condition in which the load was unexpectedly removed from the finger. Subjects were still required to move the finger as accurately as possible to the target. It was assumed that in the ballistic preprogrammed movement no compensations would be possible for the unloading, and, as a result, the target would be overshot. However, in the ramped movement such adjustments could be made to effect an accurate outcome. This was, indeed, what occurred: not only were adjustments 6  observed in the displacement of the limb, but also in the underlying EMG activity. Similar results have been found by Cooke (1980) using a larger scale arm movement. Another advantage of measuring EMG activity is that it allows RT to be fractionated into its premotor and motor time components (Botwinick & Thompson, 1966). Premotor time is the time from the imperative stimulus to the beginning of EMG activity. It has typically been associated with delays in programming activity. Motor time is the time from the beginning of EMG activity to the start of external limb movement. It is believed to reflect the duration of nonprogramming events [e.g. electromechanical delay, and development of sufficient torque to initiate movement (Anson, 1982)]. Since motor time, in theory, does not reflect delays associated with central planning, it is important to separate this time out of the RT. This process allows a more confident interpretation of any differences in programming time (Anson, 1989; Christina & Rose, 1985) From the preceding discussion it appears that analysis of both acceleration traces and EMG profiles can contribute to the understanding of the differences between preprogrammed movements and those prepared on-line within the RT/movement complexity paradigm. Unfortunately, these two dependent variables have rarely been combined with RT in a single study; despite the fact that they each appear to be dealing with the same phenomenon. Indeed, it is interesting to note how these three different methodological perspectives within motor behavior use identical terms (preprogramming and online preparation), yet never communicate as to the possibilities of what should be theoretically convergent data. The present investigation was an attempt to amalgamate such data in a meaningful manner. Towards this end, the angular acceleration traces as well as the 7  underlying EMG activity from a horizontal repetitive arm extension/flexion movement were measured in addition to the RT required to prepare and initiate such movements. Since the speed at which a movement is completed appears to have the greatest impact on whether it is preprogrammed or prepared on-line, this variable was manipulated in the present study. Norman and Komi (1979) have shown that different velocities of movement about the same joint can lead to differences in motor time. This, in turn, can make it difficult to interpret any RT differences between conditions (Falkenberg & Newell, 1980; Sidaway, 1988). In order to avoid such a situation, subjects initially moved as quickly as possible, then either remained at this maximum velocity, or slowed down to a more controlled pace.  In addition, the movement complexity was varied by  manipulating the number of extension/flexion segments the subjects completed during each trial. It was hypothesized that in the faster movements RT would increase with increases in movement complexity. In addition, the angular acceleration traces associated with such movements would contain only one zero line crossing and no significant deviations within each movement segment. Also, the EMG activity would be short in relation to the duration of the movement itself. On the other hand, in the slow movements the direct relationship between RT and movement complexity would not be observed. Similarly, the angular acceleration traces would have several zero line crossings as well as significant deviations from a smooth curve. Finally, the EMG activity would be extended over the length of each movement segment duration. In addition to testing the convergence of these three dependent variables, this study also allowed for a more accurate determination of where in time on-line adjustments were made during a movement sequence.  8  CHAPTER TWO METHOD Subjects: Twelve university students served as subjects in the present study. All were naive as to the hypotheses being tested and inexperienced at the experimental task. To motivate the subjects to perform well, a 50 dollar reward was offered to that subject who best combined the dual task of reacting quickly and reproducing the movement sequence accurately. Task and Apparatus: The subjects performed a repetitive arm extension/flexion movement in the horizontal plane through a range of 45° (from 50° to 95° - where 180° was defined as full extension). The right forearm was positioned on a padded horizontal lever attached to a bearing-mounted vertical shaft such that the elbow was coaxial with the point of rotation. The hand was supinated to grasp a vertical handle at the end of the lever. The position of the handle was adjusted to accommodate for varying forearm lengths (see Figure 1). The shoulders were restrained in a harness in order to keep the contribution from these muscles constant within each condition. In addition, the height at which the subjects were seated was adjusted so that the shoulder angle remained constant in the frontal plane across subjects. The subjects viewed a stimulus cursor that described a cosine wave cycle back and forth across an oscilloscope screen at a predetermined rate and amplitude. One cycle of the stimulus cursor started at the left most edge of the screen (corresponding to 50° extension) proceeded to the right most edge ( 95° extension) then reversed direction and ended up at the original starting point. The subject's task was to initiate as quickly as possible, after an imperative signal, and reproduce as accurately as possible, from memory, the stimulus cursor's 9  FIGURE 1 EXPERIMENTAL APPARATUS  path by flexing and extending about the elbow joint. The subject's movements were sampled by a linear potentiometer (stated linearity - 0.1%) which converted the angular displacements into a voltage signal ranging from +10 to -10 volts. This signal was then fed into a 12 bit A/D board which converted the voltage values into integer values ranging from +2047 to -2047. These values were subsequently read at a rate of 500 Hz. by a "Turbo Pascal" data acquisition program resident in an Olivetti 286 "AT' computer. The program sent the values back through the D/A board to the oscilloscope screen, where the subjects' movements were represented by a response cursor. The range of possible integer values from +2047 to -2047 represented an angular displacement of approximately 225° at the elbow joint (obviously, this full range of values was not used in the present movement). The resolution, then, of the potentiometer/A/D interface was approximately 18 values per degree of movement. Thus, if a subject moved one degree at the elbow, this would be represented by a 18 point difference in the data points recorded by the computer. The linearity of the potentiometer was confirmed through the use of a calibration program. First, a protractor was positioned coaxial with the point of rotation of the potentiometer. The potentiometer was then calibrated at +90° and -90° as determined by the protractor. These values were then set as the endpoints with the calibration program, through the potentiometer/A/D board interface. The program displayed on the computer screen the angular values obtained from the potentiometer; thus, allowing comparison to those read off the protractor. With this procedure it was determined that the values obtained from the potentiometer were the same as those which were actually produced across the entire range of movement. The angular acceleration traces were obtained by differentiating the displacement data twice. To avoid excessive noise in these traces, the 11  displacement data were first low-pass digitally filtered. The upper frequency cutoff of this filter was set at 12 Hz. for the slow movements and 18 Hz. for the fast movements (see Appendix B for a more complete description of the justification of these cut-off frequencies). Movements which cycle at frequencies above these values do not appear to be under volitional control (Brooks, Cooke, & Thomas, 1973; Teulings & Maarse, 1984). The shift in phase produced by the filtering procedure was eliminated by filtering the data forwards and then backwards in time (Pezzack, Norman, & Winter, 1977). EMG activity was recorded with Ag/AgCI surface electrodes positioned 3 cm. apart over the bellies of the long head of the biceps brachii and the lateral head of the triceps brachii. The raw signal was amplified, sampled at 500 Hz., and stored on-line by the Olivetti 286 "AT' computer. It was subsequently fullwave rectified and filtered with lower and upper frequency cut-offs of 10 and 1000 Hz., respectively. The onset and offset of activity was determined as those points which either rose above or fell below baseline levels for at least 40 msec as determined by an algorithm (see Appendix C for a more complete description of this process). In cases where the algorithm failed to correctly locate the start and/or stop of activity (due to noise in the data), the points were manually determined using a computer program which displayed the EMG signal and allowed positioning of a moveable cursor to the appropriate locations. In order to check the reliability of the experimenter in positioning the cursor correctly, 20 randomly assigned trials were processed 10 times. The correlation between these trials was approximately .95, indicating that the experimenter was highly consistent in positioning the cursor. Independent Variables: Two variables were manipulated in the present investigation. The first was the frequency of the stimulus cursor. As described in the introduction the stimulus 12  cursor's initial segment (1st .5 cycle) was held constant at a frequency corresponding to each subject's maximum. At the first reversal the frequency of the stimulus cursor either remained at the subject's maximum (Fast As Possible or FAP condition) or decreased (Slow or SLW condition). From the pilot study (see Appendix D) it was demonstrated that frequencies in the FAP condition ranged from 3.6 Hz. to 6.3 Hz. (corresponding to an average angular velocity about the elbow joint of 314° to 550° sec" ) depending upon the individual. In 1  the SLW condition the stimulus started to cycle at a frequency of 1.1 Hz. (100° sec" .) after the first reversal. In each case the amplitude of the stimulus cursor 1  remained constant at a value corresponding to 45° about the elbow joint. This range of speeds,  from fastest to slowest,  has been shown to elicit  preprogramming and on-line control, respectively, in previous research (Desmedt & Godaux, 1979; Engelhorn, 1983; Hallett, Shahani, & Young, 1975). The second variable to be manipulated in the present investigation was the complexity of the movement. As mentioned in the introduction, movement complexity is reliably varied (as evidenced by changes in RT) by manipulating the number of response elements within the movement. From the pilot study it was found that a response element in this task was composed of a reversal in direction. Thus, complexity was varied by changing the number of segments which were completed in each trial. Specifically, subjects completed movements consisting of 1, 2, 3, or 4 movement segments (analogous to .5, 1, 1.5, or 2 cycles, respectively). Previous research has shown that RT increases up to six (Sternberg, Monsell, Knoll, & Wright, 1978) or eight (Canic & Franks, 1989)  response  elements; however, the tasks used in those experiments (speech, keystrokes and finger tapping) could be completed at a faster rate than the arm extension/flexion task used in the present study. If that many elements were used in this task, the 13  duration of the longer movement sequences would reach approximately 3 to 4 seconds in the SLW condition. Traditionally, variations in RT are found for movement sequences lasting less than 1200 to 1600 msec in duration (Franks & van Donkelaar, in press; Klapp & Wyatt, 1976). Therefore, in order to stay within these temporal limits, and to maximize the chances of getting differences in RT, the number of elements was limited to 4. Experimental Procedure and Design: The experiment consisted of three sessions, the first lasting approximately 15 minutes and the remaining two, 45 minutes. Prior to the first session the experiment and task was described to the subjects and informed consent was obtained. During the first session all of the subjects completed 10 trials in which they flexed and extended in a repetitive fashion as fast as possible for 3 seconds; the only stipulation being that their movements remain as close as possible to the criterion amplitude of 45° degrees. The average speed at which each subject moved was calculated from this data and used as the individual stimulus cursor speeds in the FAP condition and the first .5 cycle in the SLW condition. The remaining two sessions were devoted to the SLW and FAP conditions. Each session began with 10 trials of tracking the stimulus cursor for 20 cycles. During these trials the stimulus cursor cycled back and forth at either the SLW or FAP speed and the subjects were required to keep their response cursor as close as possible to it by flexing and extending about the elbow joint. Next, the RT/reproduction task for each level of complexity was completed in two parts. First, a series of practice trials were completed, followed by 10 performance trials for each level of complexity. The order of presentation of each level of complexity condition was counterbalanced across subjects to control for any order effects. Similarly, in order to control for this confound between conditions, half of the subjects completed the FAP condition first, followed by the 14  SLW condition, while the other half of the subjects completed the sessions in the reverse order. Each trial consisted of the following sequence: 1. a variable warning tone (500-1500 msec) to signal the start of the trial; 2. the presentation of the stimulus (starting at the left edge of the screen and moving to the right); 3. a 500 msec delay; and 4. a second variable warning tone (500-1500 msec), the termination of which acted as the imperative stimulus. Catch trials, in which the second warning tone lasted 5 sec in duration, occurred 20% of the time during both the practice and performance trials. After each trial the RT and integrated spatial error (root mean squared error) were given as feedback . Any trials with an error score of greater than 30 mm or an RT of less than 100 msec (indicating anticipation) or greater than 500 msec (indicating a lack of attention) were discarded from further analysis. In addition, trials in which the subject failed to complete the entire movement sequence during the sampling period were also discarded. These criteria were not difficult for the subjects to achieve and were merely meant to insure that the subjects were performing in the required manner. Dependent Variables and Data Reduction: Three main types of measures (the latency, angular accelerations, and EMG durations) were used to detect changes in performance as a result of the manipulations to the stimulus frequency and complexity. RT was used to determine the time required to prepare and initiate the movement sequence. It was measured as the time from the imperative stimulus to the start of angular displacement about the elbow joint. For the purposes of this study, angular displacement was said to have occurred at that point after which at least three consecutive data points continually increased above a "twitch range" (the range within which the subject's arm oscillated prior to the imperative stimulus). 15  In addition, through the use of EMG recordings, the premotor and motor components of the RT period were determined. The premotor time was defined as the time from the imperative stimulus until the beginning of muscle activity. Muscle activity was said to have occurred at the point at which EMG activity rose above baseline levels (see Appendix C). Motor time was measured as the time from the beginning of EMG activity to the beginning of actual limb displacement. Fractionating reaction time into its premotor and motor time components allowed for a more confident interpretation of any differences in RT between the conditions. Within the angular acceleration traces two measures were taken. The first was the number of zero line crossings within each movement segment. A zero line crossing was defined as a series of at least two points - one of which was below the zero line and the other which was above it. The second measure taken from the angular acceleration traces was the presence of "significant deviations". A significant deviation was said to have occurred at a "peak" or "valley" in the data immediately followed by 10 points below or above it, respectively. Pilot work indicated that the 10 point cut-off most effectively accounted for the majority of the deviations - including those of any substance, but leaving out any of lesser importance. The number of deviations within each segment, where they occurred in time, as well as the duration between each one were the measures of interest. In terms of the EMG, the duration of agonistic muscular activity in relation to the duration of the corresponding movement segment was the measure of interest. It was expressed as a ratio. Thus, a value equal to or exceeding 1.0 indicated that the activity was present for the entire duration of that particular movement segment.  16  Statistical Analysis: The design was a 2 X 4 repeated measures in which each subject completed all 8 conditions. The results from each dependent variable were subjected to a repeated measures ANOVA. Post-hoc tests (Tukey's for the levels of complexity effect and Scheffe's for the interaction between the two independent variables) were used to analyze any differences (at the .05 level) found between the conditions after the ANOVA.  17  CHAPTER THREE RESULTS AND DISCUSSION Latency Analysis: The mean and standard deviation for reaction time (RT), premotor time (PMT), and motor time (MT) was calculated for each subject. From these individual data, group means were computed for each of the eight conditions. As can be seen in Figure 2, RT appeared to be greater overall in the SLW condition than in the FAP condition. In addition, the increasing trend across levels of complexity was similar in both conditions. These observations were confirmed statistically with a 2 X 4 repeated measures (RM) ANOVA. Specifically, significant effects of speed of movement (SLW = 239 msec, FAP = 221 msec) [F(1,11)=12.27, p=.0049], and number of cycles (.5 = 207 msec, 1 = 234 msec, 1.5 = 235 msec, 2 = 243 msec) [F(3,33)=46.28, p<.0001; Huynh-Feldt Epsilon Factor = 1.0] were found. A post-hoc Tukey's test revealed that the significant number of cycles effect was due to an increase in RT from the .5 cycle condition to the 1, 1.5 and 2 cycles conditions. Thus, in both the SLW and FAP conditions, RT increased with the addition of a reversal in direction to the movement task. Subsequent reversals did not reliably cause any further increases in the RT values. If it is assumed that RT accurately reflects the planning requirements prior to movement, then the difference between the SLW and FAP conditions for this variable suggests that it was a more complex endeavor to prepare a movement which was initially fast, and subsequently slowed down, than it was to prepare a continuously fast movement. The roots of these RT differences were examined by analyzing the premotor and motor time components of the latency period. The premotor time (PMT) results closely followed those of RT (see Figure 3). That is, for both the SLW and FAP conditions, PMT initially increased then 18  FIGURE 2 RT AS A FUNCTION OF NUMBER OF CYCLES  FIGURE 3 PMT AS A FUNCTION OF NUMBER OF CYCLES  leveled off as complexity was increased. A 2 X 4 RM ANOVA revealed a significant main effect for number of cycles only (.5 = 136 msec, 1 = 158 msec, 1.5 = 159 msec, 2 = 166 msec) [F(3,33)=31.62, p<.0001; Huynh-Feldt Epsilon Factor = 1.0]. The effect of speed of movement and the interaction both failed to reach significance. Subsequent Tukey's tests performed on the data revealed that the significant number of cycles effect was due to an increase in PMT from the .5 cycle condition to the 1, 1.5, and 2 cycles conditions. Since PMT reflects the delays associated with central programming activity, it can be concluded from the present results that both the SLW and FAP conditions demanded the same degree of planning prior to movement initiation. The increase and subsequent plateau in PMT from the .5 cycle condition to the 1, 1.5, and 2 cycles conditions mirrors the result found in the RT data. It suggests that the addition of reversals to the movement sequence caused the greatest increase in programming requirements. After the initial reversal had been planned, however, subsequent reversals did not reliably cause any further increases in the latency period. In terms of the motor time (MT) results, it was assumed that because subjects were required to minimize the duration of the initial extension segment there would be no differences in this measure, either within or between the conditions. However, as can be seen in Figure 4, this was not the case. A 2 X 4 RM ANOVA performed on the MT data revealed significant effects for speed of movement (SLW = 82 msec, FAP = 69 msec) [F(1,11)=66.34, p<.0001], number of cycles (.5 = 72 msec, 1 = 76 msec, 1.5 = 77 msec, 2 = 77 msec) [F(3,33)=13.37, p<.0001; Huynh-Feldt Epsilon Factor = 1.0], and the interaction between these two variables [F(3,33)=7.56, Huynh-Feldt p=.0048; Huynh-Feldt Epsilon Factor = .8647]. A subsequent post-hoc Scheffe's test revealed that the differences between the SLW and FAP conditions at 1, 1.5, and 2 cycles (12, 16, 21  FIGURE 4 MT AS A FUNCTION OF NUMBER OF CYCLES SLOW 130  •  FAST  r  120 110 -  50 40 1/2  1  11/2  # OF CYCLES  2  and 20 msec, respectively) were significantly larger than the difference at .5 cycles (7 msec). The question to be answered is: why was MT increased in the 1, 1.5 and 2 cycles conditions during the SLW movements? One possible explanation may be related to the difficulty associated with slowing down at a relatively precise location when attempting to complete a movement in a minimal amount of time (as was required in the 1, 1.5, and 2 cycle movements in the SLW condition). Specifically, some subjects were unable to perform this movement very accurately. In order to overcome this difficulty, they may have adopted a strategy in which they moved more slowly than was required. This possibility was assessed by measuring the average angular velocity in the initial extension segment. Figure 5 displays this variable for the SLW and FAP conditions across the 4 levels of complexity. A 2 X 4 RM ANOVA revealed significant effects of speed of movement (SLW = 237° sec , FAP = 360° sec" ) [F(1,11)=53.82, -1  1  p<.0001], number of cycles (.5 =333° sec" , 1 =291° sec" , 1.5 =285° sec" , 2 1  1  1  =285° sec" ) [F(3,33)=19.79, Huynh-Feldt p<.0001; Huynh-Feldt Epsilon Factor 1  = .7466], and the interaction between these two variables [F(3,33)=18.14, HuynhFeldt p<.0001; Huynh-Feldt Epsilon Factor = .8449]. A subsequent post-hoc Scheffe's test demonstrated that the differences between the SLW and FAP conditions at 1, 1.5, and 2 cycles (115, 156, and 154° sec" , respectively) were 1  greater than the difference at .5 cycles (65° sec" ). Thus, subjects appeared to 1  move more slowly, on the average, during the initial extension in the SLW condition; despite the fact that they were asked to complete this segment in a minimal amount of time in all trials. One drawback of measuring average angular velocity is that it takes into account the entire movement segment. Perhaps subjects started the first extension in a similar manner across the two speeds of movement, but then 23  FIGURE 5 AVERAGE VELOCITY (DEGREES/SEC) AS A FUNCTION OF NUMBER OF CYCLES SLOW  FAST  A V E V E L O C I T Y  CM  1/2  1  1 1/2  # OF CYCLES  changed before the first reversal. In order to assess this possibility the time to the first positive peak in angular acceleration was also measured. This provided a clue as to any similarities or differences between the SLW and FAP conditions during the very beginning of the movement. Figure 6 displays this variable for both speed of movement conditions across the 4 levels of complexity. A 2 X 4 RM ANOVA revealed a significant effect of speed of movement (SLW = 45 msec, FAP = 40 msec) [F(1,11)=17.95, p=.0014]. In addition, a significant interaction between speed of movement and number of cycles was also attained [F(3,33)=7.47, p=.0006; Huynh-Feldt Epsilon Factor = 1.0]. A post-hoc Scheffe's test revealed that the differences between the SLW and FAP conditions at 1, 1.5, and 2 cycles (6, 10, and 7 msec, respectively) were significantly greater than the difference at the .5 cycle condition (0 msec). Therefore, even at this early stage in the movement, the task requirement of slowing down at the first reversal had an impact on the manner in which the task was performed. This, in turn, appeared to influence the MT results. Kinematic Analysis: Figures 7 and 8 display typical angular displacement, velocity, and acceleration traces for the 2 cycle condition done under both SLW and FAP constraints. The angular acceleration traces were analysed in several ways. First among these was simply the number of times the acceleration curve crossed the zero line (changing from a positive to a negative angular acceleration) within each segment (extension or flexion) of the movement. It was hypothesized that the zero line would be crossed less often (indeed, usually only once - as is required by the mechanics of movement) within the movement segments that were completed as fast as possible than in the segments completed at a slower speed. A 2 X 4 RM ANOVA was first performed on the zero line crossings data 25  FIGURE 6 TIME TO 1ST PEAK IN ACCELERATION AS A FUNCTION OF NUMBER OF CYCLES  FIGURE 7 - KINEMATICS OF FAST MOVEMENT F i l t e r 18 Hertz  Angular Displacenent (deg)  Uelocity (deg/s)  Acceleration Cdeg/s/s) I  2QQ ns  |  /  i  - -5000  FIGURE 8 - KINEMATICS OF SLOW MOVEMENT F i l t e r 12 Hertz  Angular Displacement Cdeg)  Uelocity Cdeg/s)  Acceleration Cdeg/s/s) 200 ns  oo CM  from the initial extension movement in each condition. Since the subjects were required to complete the first extension segment in a minimal amount of time, it was expected that there would be no significant effects in the analysis. Indeed, this was what was found (see Figure 9). Specifically, the effect of speed of movement was not significant (SLW = 1.02, FAP = 1.01) [F(1,11 )=0.31, p=.5863]; indicating that there were the same number of zero line crossings in the initial extension during both the SLW and FAP conditions. The effect of number of cycles and the interaction between speed of movement and number of cycles also failed to reach significance. A second 2 X 2 X 3 [speed (SLW vs. FAP) X movement segment (1st extension vs. 1st flexion) X number of cycles (1 vs. 1.5 vs. 2)] RM ANOVA was performed on the data from the first extension and flexion segments for the 1,1.5 and 2 cycles conditions. This time a difference between the SLW and FAP conditions was expected; and, in fact, was found (SLW = 1.3, FAP = 1.0) [F(1,11)=34.78, p=.0001]. In addition, a significant movement segment effect was found (E1 = 1.0, F1 = 1.3) [F(1,11)=34.78, p=.0001]; as well as an interaction between movement segment and speed of movement [F(1,11)=34.78, p=.0001]. This last result is of the most theoretical importance because it points to the fact that the number of zero line crossings increased significantly from the first extension to the first flexion in the SLW condition, but not in the FAP condition. Such an increase indicates that, after the initial extension, subjects began to rely more heavily on on-line prepared adjustments to successfully meet the goals of the task in the SLW condition. A final 2 X 2 X 2 RM ANOVA was completed on the data from the first flexion and the second extension in the 1.5 and 2 cycles conditions in order to assess whether any further changes occurred in the number of zero line crossings data. As expected, a significant speed of movement effect was found 29  FIGURE 9 # OF ZERO LINE CROSSINGS AS A FUNCTION OF EACH SEGMENT 1  SLOW  —•  FAST  1 1/2 CYCLES 1 CYCLE  -  2 CYCLES  1/2  CYCLE  +  J  i  l  l  i  i  i  l  I  l  E1  E1  F1  E1  F1  E2  E1  F1  E2  SEGMENT  l  F2  (SLW = 1.7, FAP = 1.0) [F(1,11)=42.13, p<.0001]. In addition, the effect of number of cycles (1.5 = 1.5, 2 = 1.2) [F(1,11)=20.02, p=.0009], and the interaction between this variable and speed of movement [F(1,11)=17.12, p=.0016] were also significant. The interaction indicated that the decrease from 1.5 to 2 cycles was due mostly to the SLW condition. In terms of the movement segments, it was found that there were more zero line crossings during the second extension than during the first flexion (F1 = 1.1, E2 = 1.6) [F(1,11 )=28.28, p=.0002]. In addition, the interactions between this variable and speed of movement [F(1,11)=26.03, p=.0003], and number of cycles [F(1,11)=10.59, p=.0077] also reached significance. These indicate that the increase in the number of zero line crossings from the first flexion to the second extension was larger in the SLW condition than in the FAP condition, and, larger in the 1.5 cycle condition than in the 2 cycle condition. Finally, a three way interaction was found between speed of movement, number of cycles and movement segment. This effect demonstrated that the interaction between speed of movement and movement segment was more pronounced in the 1.5 cycle condition than in the 2 cycle condition. From these analyses it can be concluded that when subjects moved as fast as possible, the resulting angular acceleration trace contained only one zero line crossing. When they slowed down to the required speed in the SLW condition, the number of zero line crossings increased significantly. Brooks, Cooke, and Thomas (1973) felt that the number of zero line crossings accurately reflected the underlying mode of control: one zero line crossing represented open loop, preprogrammed behavior, and more than one, closed loop movements containing feedback based corrections. Using this definition, it appears that the movements completed in the FAP condition were entirely preprogrammed, whereas, those in the SLW condition were subject to feedback based corrections 31  prepared as the movement was being completed. A second indicant of the type of programming which occurred prior to or within a movement sequence is the presence of "significant deviations" within the acceleration trace (Carlton, 1981). In the present experiment the number of deviations within each segment, their temporal location, as well as the duration between each was analysed. As in the zero line crossings data, it was hypothesized that there would be many deviations within the slow movements, and very few, if any, in the movements completed as fast as possible. It follows, then, that the requirement to minimize movement time in the first extension would result in no differences between the SLW and FAP conditions for this segment. This hypothesis was, again, confirmed (see Figure 10). Specifically, a 2 X 4 RM ANOVA revealed a nonsignificant effect of movement speed (SLW = 0.2, FAP = 0.1) [F(1,11)=3.88, p=.0745]. Thus, subjects produced basically the same number of deviations during the initial extension in both the SLW and FAP conditions. In addition, the effect of number of cycles and the interaction between this variable and speed of movement both failed to reach significance. In order to test whether there were more deviations in the first flexion than in the first extension, a 2 X 2 X 3 (speed of movement X movement segments X number of cycles) RM ANOVA was conducted on the data from the 1, 1.5, and 2 cycles conditions. It revealed a significant speed of movement effect (SLW = 1.5, FAP = 0.1) [F(1,11)=359.93, p<.0001]. Thus, more adjustments were made in the SLW condition to control the required movement outcome than in the FAP condition. Although the effect of number of cycles was not significant, the interaction between this variable and speed of movement was [F(2,22)=5.48, Huynh-Feldt p=.0192, Huynh-Feldt Epsilon Factor = .7945]; indicating that there was a greater decrease in the number of significant deviations across movement 32  FIGURE 10 # OF DEVIATIONS AS A FUNCTION OF EACH SEGMENT — I — SLOW  -  1 1/2 CYCLES  1 CYCLE  1/2 CYCLE  --•  FAST  2 CYCLES  ^  /  /  /  / _  + i  i  i  i  i  i  i  i  i  E1  E1  F1  E1  F1  E2  E1  F1  E2  SEGMENT  i  F2  segments in the SLW condition from 1 to 2 cycles than in the FAP condition. In addition, a significant effect of movement segment (E1 = 0.1, F1 = 1.5) [F(1,11)=199.29, p<.0001], as well as interactions between this variable and number of cycles [F(2,22)=17.64, Huynh-Feldt p=.0001; Huynh-Feldt Epsilon Factor = .8921], and speed of movement [F(1,11)=348.61, p<.0001] were found. The interaction between movement segment and speed of movement sheds the most light on possible modes of control given the task constraints. Specifically, it demonstrates that the number of deviations increased from the first extension to the first flexion in the SLW condition, but not in the FAP condition. A final 2 X 2 X 2 RM ANOVA was conducted in order to confirm the presence of these differences in the second extension, and to assess whether changes occurred between this segment and the first flexion. The effect of speed of movement was, once again, significant (SLW = 3.1, FAP = 0.1) [F(1,11)=113.75, p=.0087]; demonstrating that the movements in the SLW condition contained more significant deviations within each segment than those in the FAP condition. The effect of movement segment (F1 = 1.2, E2 = 2.0) [F(1,11)=20.56, p=.0454], and the interaction between this variable and speed of movement [F(1,11 )=19.95, p=.0466] also reached significance. Thus, more deviations were present in the second extension than in the first flexion; furthermore, this increase was more pronounced in the SLW condition than in the FAP condition. The results from the number of deviations analyses closely follow those from the zero line crossings data. Specifically, if deviations within the acceleration trace of a movement suggest the presence of feedback based corrections prepared on line, then subjects appeared to rely more heavily on this form of control in the SLW condition than in the FAP condition. A single characteristic of the SLW condition, that being the decreased speed of movement, appeared to 34  account for this on-line control. However, whether moving slowly simply "allows" one to prepare movements on-line, or whether it "forces" one to do so is the question. That is, are the subjects taking advantage of the conditions under which they are being asked to operate, or are they being taken advantage of by the conditions, and forced to pursue the only line of control which remains functional? This issue will be addressed more fully in the General Discussion. In an attempt to gain a further insight into the nature of on-line control, the relative temporal locations of the deviations within each segment were plotted as a frequency distribution. This provided an indicant of where the majority of deviations were located; and, thus, could be used to suggest possible characteristics of the control mechanisms. Figure 11 displays the distribution of the deviations collapsed over conditions and separated into flexion and extension segments for the SLW speed of movement. The general trend in both directions of movement was an increase in the number of deviations in the second half of each segment. Such a result concurs with that found in most single aiming studies (Carlton, 1981; Crossman & Goodeve, 1963/1983). Specifically, a movement from a home position to a target is typically composed of one distance-covering submovement, followed by a series of adjustments until the target is attained. Such adjustments tend to show up in the acceleration trace as deviations from the smooth curve. In addition, the grouping of the deviations at particular points in time, as found in the pilot study (see Appendix D), appears to be only partially reproduced in the present results. Specifically, during the flexion segments there appeared to be a tendency to make adjustments at specific temporal locations; this is especially noticeable 60% into the flexion segments, where there is a large increase in the frequency distribution. A final measure which sheds some light on the nature of the control 35  FIGURE 11 RELATIVE TEMPORAL LOCATION OF THE DEVIATIONS - SLW CONDITION EXTENSIONS  FLEXIONS  140 ,  0  10  20  30  40  50  60  RELATIVE TIME  70  80  90  100  process is the time between each deviation. Previous research has produced equivocal results regarding the frequency with which adjustments can be made. If a strict serial approach to the perception-action loop is assumed, the time required to make an adjustment includes latencies associated with the processing and interpretation of the stimulus array, as well as response selection and execution. With this in mind, original research in this area posited a value in the neighborhood of 200 msec (Keele & Posner, 1968). However, subsequent papers have indicated that this value may have been an overestimation (Carlton, 1981). Indeed, more recent unpublished research using single aiming tasks has shown that if the target is moved as soon as 10 msec before the subject has attained it, adjustments can still be made which effect an accurate outcome. In Figure 12, a frequency distribution of the number of deviations in the SLW condition separated by durations from 1 through to 240 msec displays a sharp peak at approximately 60 to 70 msec. Thus, on the average, subjects were making adjustments to their course of action about every 65 msec. Although this value is quite fast, it is not below that which is theoretically possible given previous results. Such a low value does, however, suggest that the visual perception feedback loop was bypassed in some manner. It may be that once the subjects undertook a specific sequence of movements, they did not require a complete or thorough run through all of the stages associated with visual movement control in order to make the appropriate adjustments. In some way they may have already been primed to make such adjustments; thus, the latency associated with them may be reduced substantially. EMG Analysis: The length of time for which EMG activity occurred in relation to the total length of time for each movement segment was the measure of most interest in this analysis. It was calculated as a ratio, where, for example, a value of .5 37  FIGURE 12 TIME BETWEEN EACH DEVIATION SLW CONDITION 140  0  25  50  75  100  125  150  TIME (MSEC)  175  200  225  250  indicated the presence of EMG activity for half of the movement segment duration.  Desmedt  and Godaux  (1979)  found that  during  ballistic,  preprogrammed movements, EMG activity was usually close to or below this .5 value; however, in slower, ramped movements the ratio of EMG activity duration to movement segment duration was considerably larger (i.e. in the range of .9 to 1.3). Figures 13 and 14 provide an illustration of the EMG activity of the triceps (middle trace) and biceps (lower trace) in relation to the displacement (upper trace) in the 2 cycles condition for the SLW and FAP constraints, respectively. The results from each segment in each condition of the present experiment are displayed in Figure 15. Again, 3 separate ANOVA's were completed to check for meaningful significant differences. A 2 X 4 RM ANOVA done on the data from the first extension in each condition revealed a nonsignificant effect for speed of movement (SLW = .476, FAP = .512) [F(1,11)=2.19, p=.167]. Again, this result demonstrates the similarity between the SLW and FAP conditions during this initial movement. Specifically, the underlying muscular activity is essentially the same in terms of relative temporal duration in all conditions during the first extension. A second 2 X 2 X 3 (speed of movement X movement segment X number of cycles) RM ANOVA was completed in order to compare any changes in the EMG ratio value across the first extension and flexion in the 1, 1.5, and 2 cycles conditions. As expected a significant speed of movement effect was found (SLW = .743, FAP = .630) [F(1,11 )=17.19, p=.0016]. Thus, within the SLW condition, muscular activity was present for a greater relative duration during each segment than within the FAP condition. In addition to this result, a significant effect of movement segment was also revealed (E1 = .514, F1 = .857) [F(1,11)=115.28, p<.0001]; as well as an interaction between this variable and speed of movement [F(1,11)=38.37, p=.0001]. Once again, the interaction is the most important result 39  FIGURE 13 - TYPICAL EMG DATA - SLW CONDITION - 95 - 50 Angular Displacement (deg)  TRICEPS Cnv)  -  0.5  ^-  0  - -0.5  0.5 0 BICEPS |2QQ  HS[  -0.5  FIGURE 14 - TYPICAL EMG DATA - FAP CONDITION - 95 •v  -  Angular Displacement (deg>  50  -  2.5  -  0  TRICEPS - -2.5 -  2.5  -  0  BICEPS - -2.5  FIGURE 15 RATIO OF EMG DURATION TO SEGMENT DURATION AS A FUNCTION OF EACH SEGMENT — I — SLOW  E1  E1  F1  E1  •  F1  E2  SEGMENT  E1  FAST  F1  E2  F2  from a theoretical perspective. It indicates that the EMG ratio value increased more substantially from the first extension to the first flexion in the SLW condition than in the FAP condition. This trend suggests that even at the level of muscular activity the constraints imposed on the subjects in the SLW condition were more conducive to on-line control than in the FAP condition. A final 2 X 2 X 2 RM ANOVA was completed on the data from the first flexion and second extension in the 1.5 and 2 cycles conditions. Significant effects of speed of movement (SLW = 1.07, FAP = .802) [F(1,11)=104.35, p<.0001]; number of cycles (1.5 = .986, 2 = .889) [F(1,11)=2.07, p=.0009]; and an interaction between these two variables [F(1,11)=4.64, p=.05] were found. These indicate, in turn, that the EMG ratio value was greater in the SLW condition than in the FAP condition, greater in the 2 cycles condition than in the 1.5 cycles condition; and that this decrease from 1.5 to 2 cycles was more pronounced in the SLW condition than in the FAP condition. There was also a significant effect of movement segment (F1 = .882, E2 = 1.03) [F(1,11)=66.89, p<.0001]; as well as an interaction between this variable and speed of movement [F(1,11)=24.02, p=.0005]. The interaction demonstrates that the increase from the first flexion to the second extension was greater in the SLW condition than in the FAP condition. From these analyses, then, it can be concluded that in the SLW condition, the EMG ratio remained close to or at .5 for the first extension, then increased towards values of 1 or greater as movement velocity decreased. This trend closely mirrors the results found in the acceleration traces; thus, providing converging evidence that movements in this condition were controlled in an online fashion. In the FAP condition, the EMG ratio also increased from the .5 level after the initial extension segment. Obviously, if these movements had been entirely preprogrammed, one of the behavioral outcomes would have been EMG 43  ratios which remained at this level throughout the movement. Although the increase in this value was not as large as in the SLW condition, it does suggest that at least some on-line preparation was occurring. Recall, however, that the acceleration data pointed to the use of preprogramming in this condition (and, as mentioned above, to on-line control in the SLW condition). How can these dependent variables be reconciled given the present results? One possibility is that the increase in the EMG ratio value after the initial extension in the FAP condition reflected the braking and accelerating functions of the muscle activity. For example, during the first extension, the biceps activity must be initiated as the reversal is being approached in order to brake the initial impulse from the triceps. After it has stopped the arm from extending, the biceps must continue to fire in order to initiate the movement in the other direction. This dual activity of the muscle resulted in the increased EMG ratio values. Unfortunately, separating out the braking and accelerating components of the EMG activity is a difficult task; thus, it is impossible to know how much of the ratio was accounted for by each component. Given the previous argument, it may be possible to conclude that the increased EMG ratio in the SLW condition was also due to the combined braking and accelerating function of the muscles. While it can be assumed that the braking activity did occur in this condition, it seems unlikely that such activity accounted for the large increase in the EMG ratio value. This hypothesis can be supported by the following argument. Since the movements were being completed at a slower speed in this condition, it seems likely that the braking function prior to each reversal would take less time than in the FAP condition (or, at the most, the same amount of time; but, certainly not a longer time). If this was the case, and the braking function was the only factor which caused increases in the EMG ratio value, then the SLW condition should have led to values below 44  those which occurred in the FAP condition. Since this was not the outcome of the present experiment, it can be concluded that some other factor (most likely the decreased speed of movement, and consequent on-line control) can be held accountable.  45  CHAPTER FOUR GENERAL DISCUSSION The main goal of the present experiment was to investigate the differences between  movements  performed under conditions conducive to  either  preprogramming or on-line preparation. Using latencies, angular acceleration traces, and EMG durations, it was possible to "zero in" on these differences as they occurred within the movement sequence. In previous RT/movement complexity research, this ability to pinpoint where on-line control occurs has been lacking. In terms of the latency data, a difference between the SLW and FAP conditions was found for reaction time (RT) and motor time (MT), but not for premotor time (PMT). The results found in RT were a direct outcome of the changes which occurred in the PMT and MT data. Furthermore, it may be more meaningful from a theoretical standpoint to discuss these latter two variables. Therefore, the RT results will not be expanded upon, other than to recall, as mentioned above, that there was a difference between the SLW and FAP conditions as well as an increase from the .5 cycle condition to the 1, 1.5, and 2 cycles conditions. Because PMT, in theory, represents the delays associated with central programming activity, the present results suggest that both the SLW and FAP conditions demanded the same degree of planning prior to movement initiation. This result is, on the surface, incongruent with both the pilot study (see Appendix D) and previous research regarding the time required to prepare and initiate movements which vary in velocity (Baba & Marteniuk, 1983; Newell, Carlton, Carlton, & Halbert, 1980). However, having subjects in the present experiment initiate their movements at the same speed across all conditions may have led to the nonsignificant differences between the SLW and FAP conditions. Indeed, 46  despite the fact that changes were required in the speed of movement at the first reversal in the SLW condition, this did not require any additional planning time during the PMT period prior to movement initiation. The largest increase in PMT was between the .5 and 1 cycle conditions. This suggests that the planning required to make the first reversal was the most time consuming. After this initial investment of time, the subsequent reversals in the 1.5 and 2 cycles condition did not cause as great an increase in the latency period. The initial increase and subsequent plateauing that was found in the PMT scores was not entirely without precedent. Specifically, previous research has demonstrated that many different types of movements, even when completed as quickly as possible, cause their greatest increases in latency when complexity is varied from 1 to 2 elements (Fischman, 1984; Franks & van Donkelaar, in press; Garcia-Colera & Semjen, 1987; 1988; Henry & Rogers, 1960). Sternberg, Monsell, Knoll, and Wright (1978) suggested that this effect was due to the more precise control overtiming and coordination that was necessary in the sequences of two or more elements in comparison to the single element response. Presumably, this more precise control must be programmed during the reaction time period and thus causes the increases observed in this variable. Subsequent increases in complexity do not appear to require any greater degree of control, over and above that found in the two element condition. How is it that an initial increase in movement complexity leads to a large increase in the latency period, but subsequent increases of similar magnitude do not. MacKay (1983) has offered a rule-based explanation for such results, similar to that used by Restle and his colleagues to explain the processes involved in serial pattern learning (Restle & Brown, 1970a, 1970b). Specifically, MacKay has suggested that a series of sequencing rules govern the execution of a movement sequence. As the complexity of the movement increases in some manner, the 47  rules which guide such movements either become more complex, or, simply, greater in number. The latency required to prepare and initiate such sequences is increased in comparison to simpler sequences because more time is taken in retrieving, discriminating, and applying the more complex rules. As an example of how latency increases as the rules become more complex, MacKay cites evidence found in Klapp's "dit-dah" paradigm (Klapp & Wyatt, 1976). In this experiment, subjects were asked to initiate as quickly as possible tapping sequences composed of the morse code keypresses "dit" or "dah". Klapp and Wyatt found that more time was required prior to initiating the "dah" keypresses in comparison to the "dit" keypresses. MacKay suggests that the increase in latency for the sequences beginning with the "dah" result from the additional rule specifying the duration for which the response key must be held down; such a rule is not required in the "dit" response. The plateauing observed in the RT data can be accounted for if it is assumed that one rule representing a particular characteristic of the sequence is simply iterated the number of times that characteristic occurs within the sequence. For example, Garcia-Colera and Semjen (1987) found that tapping sequences composed of three or more elements did not vary in terms of initial latency. Furthermore, they suggested that the initial and terminal elements in the sequence were fundamentally different from the intermediate elements, but, that these intermediate elements were not different from each other. Thus, in sequences with three or more taps, the only factor which varied was the number of intermediate elements. Since these elements were basically the same, increasing their number did not add any further processing demands prior to movement initiation, thus, the initial latency remained invariant. It is suggested that the processing demands did not increase because the rule representing the intermediate elements was simply repeated the required number of times. All that 48  needed specification, then, was the number of repetitions of the rule. If a rule-based approach was adopted by the subjects in the present experiment, the next question is: which characteristic of the movements was represented in the rule. Since the largest increase in PMT occurred when a reversal in direction was added to the movement sequence (i.e. the .5 cycle condition contained no reversal, whereas, the 1 cycle condition did), it is suggested that this aspect of the movement was the factor which the rule specified. Adding subsequent reversals in the 1.5 and 2 cycles conditions did not cause any further increases in PMT; thus, the rule appears to have been simply iterated the number of times required to successfully complete the sequence. As mentioned above, although the PMT results were similar in both conditions, the MT results did not follow the same trend. From an analysis of the kinematics of the initial extension, it appeared that MT varied directly with the average angular velocity and the time to the first positive peak in angular acceleration. Specifically, in the .5 cycle condition of the SLW movements and in all four conditions in the FAP movements, these three variables remained essentially unchanged. This indicated that in these conditions the task requirement of minimizing the duration of the initial extension segment was maintained. However, when the subjects were required to slow down at the first reversal (i.e. in the 1, 1.5, and 2 cycles conditions of the SLW movements), the values for these three variables changed. Specifically, both MT and time to the first positive peak in angular acceleration increased, and average angular velocity decreased. Thus, some aspect of the task demands forced the subjects to adopt different strategies in an effort to control their own movements. It is suggested that this was the first sign of on-line controlled adjustments which subsequently differentiated the movements in the SLW condition from those in the FAP condition. 49  Traditionally, MT has been viewed as something which is not associated with planning activity. Instead it is a variable which is related to peripheral, physical delays. However, this distinction seems rather arbitrary. On the one hand, if the notion of on-line preparation is accepted, then it appears plausible for planning to occur during the MT period. On the other hand, if all of the planning takes place prior to movement initiation, then it seems faulty to assume that as soon as the muscles become active, the planning can no longer occur. Indeed, if this assumption is made, then the planning activity must stop before the muscles become active in order to accommodate for the delays associated with neural transmission; thus, greatly reducing the usefulness of the MT measure. It is suggested that the present results point to the fact that some form of planning occurred during the MT period. Specifically, the increased MT values in the 1, 1.5, and 2 cycles conditions in the SLW movements (in conjunction with the lack of difference in these conditions between the SLW and FAP movements in the PMT data) indicate that future characteristics of the sequence can influence the MT period. If this assumption is correct, the next question is: what caused this increase in the MT values? The most likely possibility is that the act of slowing down to the required speed and maintaining this speed during subsequent movements required some sort of timing mechanism in order to be accurate. Such a timing mechanism was not required in the FAP condition: the subjects simply moved as fast as they could. The lack of increase in MT after the implementation of this component (i.e. from the 1 cycle to the 2 cycles conditions) suggests that it can be applied in an on-line fashion. Thus, the subject knows that a certain amount of time must pass between each reversal in direction; however, because this duration is so lengthy, any adjustments required to meet the criterion time interval can be accomplished while the movement is underway. 50  The consequences  of preparing a change in movement velocity,  specifically, a decrease in velocity, have been previously demonstrated. In 1967, Vince and Welford, using a psychological refractory period paradigm, found that the delays in responding to a signal requiring a decrease in movement velocity were significantly greater than the delays occurring prior to an increase in velocity. Indeed, the latency associated with a signal to increase movement velocity was found to be less than the usual refractory period delays. Glencross (1980) suggested that this discrepancy may be due to the fact that in speeding up a slow movement, one simply has to increase the gain on an already existing parameter of the controlling process; whereas, in effecting a decrease in movement velocity, one has to restructure such a process. There is also evidence to suggest that the maintenance of a less than maximal speed requires additional planning prior to movement initiation. Ivry (1986) found increases in RT (he did not fractionate RT) when he added a timing requirement to an isometric force task. Similarly, Klapp and Wyatt (1976) found that the time required to prepare and initiate a pattern of morse code keypresses was influenced by the length of each tap. Specifically, the production of the "dah" keypress appears to require greater programming demands (as reflected in increased RT) than the simpler "dit" keypress. Thus, from the latency analysis two hypotheses can be formed regarding possible modes of control. First, since planning for a reversal in direction caused an increase in PMT for both the SLW and FAP conditions, but subsequent reversals did not cause any further increases in this variable; it is suggested that the control of reversals was accomplished through the application of a rule which was simply iterated the required number of times. Second, since the additional task demand of slowing down at the first reversal caused an increase in MT in the SLW condition, but no additional increases in the multiple cycle conditions, it 51  appears that the maintenance of proper speed was accomplished by a timing mechanism which was applied as the movements were being completed (i.e. in on-line fashion). Evidence for this on-line control was gained from the remaining dependent variables. Specifically, movements in the SLW condition displayed more acceleration zero line crossings, more deviations within the acceleration trace, and greater EMG ratio values than movements in the FAP condition. The presence of on-line control in the SLW condition appeared to be the result of the decreased speed of movement used in this condition. This can be concluded from the changes which occurred in each dependent variable from the first extension to the first flexion. These changes indicated that the subjects began to rely more heavily upon on-line controlled adjustments to produce the required speed of movement in the segments after the first extension. Such evidence was not apparent to the same degree within any of the movement segments completed as fast as possible (i.e. all of the FAP conditions, and the initial extension in each of the SLW conditions). Given that the evidence points to the use of on-line control in the SLW condition, one can question whether the subjects were "allowed" to use this type of behavior, or whether they were "forced" to do so. Specifically, was on-line control something that the subjects brought with them to the SLW condition and used to their advantage as they completed these slow movements; or, did they have to perform in this manner because it was their only behavioral recourse? In one sense, the increased duration from one segment to the next certainly allowed for feedback and subsequent adjustments. Thus, it may be that on-line control simply "fell out" of the system due to the context. Alternatively, the subjects, being willful creatures, may have found that through trial and error only movements controlled in an on-line fashion led to a successful outcome. In either case, these suggestions point to the possibility that the presence of on-line 52  control may have resulted from a choice to use it. However, one wonders whether such temporally lengthy movements could be accomplished without adjustments made during the course of movement. Specifically, is it possible to plan in advance a movement which is completed at a relatively slow pace? Evidence from learning studies indicates that it takes a great deal of practice before a movement can be completed without any feedback control (Moore & Marteniuk, 1986). Indeed, the movements made during the early stages of learning are typically jerky and discontinuous; two characteristics which result from the inefficient use of the perception-action feedback loop. In the present experiment, remnants of this type of behavior were apparent in the SLW condition but not in the FAP condition. Thus, given the relatively minimal levels of practice in this study, it may be that the subjects were unable to perform the movements in the SLW condition without the use of on-line controlled adjustments. Perhaps, if allowed to practice the task to a sufficient degree, the subjects may have learned to perform both the slow and the fast movements in a similar manner. If these arguments are accepted, then it appears as if on-line control was the only option for the subjects during the SLW condition. Further to this point, on the one hand, on-line control can be seen as something that simplifies the movement control process, because it allows the performer to "put off" certain "parameterizations" until they are relevant to the task at hand. However, on the other, the fact that the controlling process must allow for feedback and subsequent alterations makes it seem as something more complex. Thus, although on-line control may appear as a simplification of preprogrammed behavior from the outside, the internal workings of this process are much more malleable; and, as a result, may be more complex. Ostry (1980) has suggested that preprogrammed control can be likened to a self-contained computer program, and on-line control to an interactive program. In the former 53  case, all of the information required to successfully execute the program is specified within the program. When it comes time to use the program, it is simply compiled and run off. Because it has all of the information it needs, the program can complete its task without being updated or adjusted to accommodate new information. Such a program can be viewed as relatively simple. On the other hand, an interactive program has to have the ability to take in and use information in order to function successfully. Indeed, it cannot complete its task without this capability. As such, an interactive program must contain subroutines which allow for the updating and accommodation of new information. In this sense, it can be viewed as a more complex piece of programming. Given the above statements, the assumption is that subjects used on-line control in the SLW condition because it was the only way in which they could successfully accomplish the task. Since they were relatively unskilled at the task, if they had resorted to the easier, preprogrammed mode of control, the outcome would have been far less acceptable. A second goal of the present experiment was to test for convergence between several dependent measures which each had previously been used to differentiate  preprogrammed movements  from those  prepared on-line.  Specifically, latency data, acceleration traces, and EMG signals each are characterized by specific trends when the mode of control varies. Whether these trends would all be apparent within a single experiment was the question of interest. From the results it appears that a relatively high degree of convergence was attained, especially between the acceleration and EMG data. The similarity between the measures from these data are best observed in Figures 9, 10, and 15. These graphs illustrate the group means for each individual segment across all eight conditions (2 speeds of movement X 4 number of cycles). All three 54  display a distinct increase in the relevant dependent variable from the first extension to the first flexion in the SLW condition. This indicates that the decrease in movement velocity between these two segments caused the subjects to rely more heavily on on-line prepared adjustments to move successfully. Similar increases were not in evidence in the FAP condition (except, of course, in the EMG ratio values - due to the braking and accelerating functions of the muscles as explained in the Results and Discussion section). The latency data did not follow in this trend as closely as anticipated. Specifically, it was hypothesized that the fast movements would be entirely preprogrammed, and, thus, lead to increased RT values as complexity was increased; whereas, the SLW movements would allow on-line preparation, and, thus, not lead to such increases. However, as outlined above, the results for both speed of movement conditions were similar and in between these two extremes. It appears that subjects may have adopted a rule-based approach in executing their movements. Such an approach can efficiently account for the initial increase and subsequent plateau in the latency values. In conclusion, by combining kinematic and EMG measures with those from the initial latency period, it was possible in the present experiment to accurately assess the mode of control as it occurred within a movement sequence. Using the RT/movement complexity paradigm as a basis, it was found that subjects relied more heavily on preprogramming when asked to move as fast as possible, and, on on-line preparation when asked to move at a slower pace. It appears that this on-line preparation took the form of adjustments which served to effect an accurate temporal outcome. Such adjustments were not required in the FAP condition because the subjects simply moved as fast as they could.  55  REFERENCES Anson, J.G. (1982). Memory drum theory: Alternative tests and explanations for the complexity effects on simple reaction time. Journal of Motor Behavior. 14, 228-246. Anson, J.G. (1989). Effects of moment of inertia on simple reaction time. Journal of Motor Behavior. 21, 60-71. Botwinick, J., & Thompson, L.W. (1966). Premotor and motor components of reaction time. Journal of Experimental Psychology, 71, 9-15. Brooks, V.B., Cooke, J.D., & Thomas, J.S. (1973). The continuity of movements. In R.B. Stein, K.G. Pearson, & J.B. Redford (Eds.), Control of posture and locomotion. New York: Plenum Press. Canic, M.J. (1988). Perceptual and response organization of rhythmic patterns. Unpublished doctoral dissertation, University of British Columbia. Canic, M.J., & Franks, I.M. (1989). Response preparation and latency in patterns of tapping movements. Human Movement Science. 8,123-139. Carlton, L.G. (1981). Processing visual feedback information for movement control. Journal of Experimental Psychology: Human Perception and Performance. 7, 1019-1030. Christina, R.W., & Rose, D.J. (1985). Premotor and motor reaction time as a function of response complexity. Research Quarterly for Exercise and Sport, 56, 306-315. Cooke, J.D. (1980). The organization of simple skilled movements. In G.E. Stelmach & J. Requin (Eds.), Tutorials in motor behavior. Amsterdam: NorthHolland. Crossman, E.R.F.W., & Goodeve, P.J. (1983). Feedback control of handmovements and Fitts' Law. Quarterly Journal of Experimental Psychology. 35A, 251-278. Original work presented at the meeting of the Experimental Psychology Society, Oxford, England, July 1963. Desmedt, J., & Godaux, E. (1979). Voluntary motor commands in human ballistic movements. Annals of Neurology. 5, 415-421. Ells, J.G. (1973). Analysis of temporal and attentional aspects of movement control. Journal of Experimental Psychology. 99, 10-21. Engelhorn, R. (1983). Agonist and antogonist muscle EMG activity pattern changes with skill acquisition. Research Quarterly for Exercise and Sport. 54, 315-323. Erikson, C.W., Pollack, M.D., & Montague, W.E. (1970). Implicit speech: Mechanism in perceptual encoding? Journal of Experimental Psychology, 84, 502-507. 56  Falkenberg, L.E., & Newell, K.M. (1979). Relative contribution of movement time, amplitude, and velocity to response initiation. Journal of Experimental Psychology: Human Perception and Performance, 6, 760-768. Fischman, M.G. (1984). Programming time as a function of number of movement parts and changes in movement direction. Journal of Motor Behavior. 16, 405-423. Fitts, P.M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology. 47, 381-391. Franks, I.M. & van Donkelaar, P. (in press). The effects of demanding temporal accuracy on the programming of simple tapping sequences. Acta Psvchologica. Franks, I.M., Wilberg, R.B., & Fishburne, G.J. (1985). The planning, organization and execution of serially ordered movement patterns: A coding perspective. In D. Goodman, R.B. Wilberg, & I.M. Franks (Eds.), Differing perspectives in motor learning, memory, and control. Amsterdam: North-Holland. Freeman, F.N. (1907). Preliminary experiments Psychological Monographs. 8, 301-333.  on writing  reactions.  Garcia-Colera, A., & Semjen, A. (1987). The organization of rapid movement sequences as a function of sequence length. Acta Psvchologica. 66, 237250. Garcia-Colera, A., & Semjen, A. (1988). Distributed planning of movement sequences. Journal of Motor Behavior. 20, 341 -367. Glencross, D.J. (1980). Levels and strategies of response organization. In G.E. Stelmach & J. Requin (Eds.), Tutorials in motor behavior. Amsterdam: NorthHolland. Hallett, M., Shahani, B.T., & Young, R.R. (1975). EMG analysis of stereotyped voluntary movements in man. Journal of Neurology. Neurosurgery, and Psychiatry. 38, 1154-1162. Henry, F.M., & Rogers, D.E. (1960). Increased response latency for complicated movements and a "memory drum" theory of neuromotor control. Research Quarterly for Exercise and Sport. 31, 448-458. Higgins, J.R., & Angel, R.W. (1970). Correction of tracking errors without sensory feedback. Journal of Experimental Psychology. 84, 412-416. Hof, A.L. (1984). EMG and muscle force: An introduction. Human Movement Science. 3,119-153. Hulstijn, W., & van Galen, G.P. (1983). Programming in handwriting: Reaction time and movement time as a function of sequence length. Acta Psvchologica. 54, 23-49. 57  Keele, S.W. (1968). Movement control in skilled motor Psychological Bulletin. 70, 387-403.  performance.  Keele, S.W. (1981). Behavioral analysis of motor control. In V. Brooks (Ed.), Handbook of physiology: Motor control. New York: Academic Press. Keele, S.W., & Posner, M.I. (1968). Processing of visual feedback in rapid movements. Journal of Experimental Psychology, 77,155-158. Kelso, J.A.S., Holt, K.G., Kugler, P.N., & Turvey, M.T. (1980). On the concept of coordinative structures as dissipative structure: II. Empirical lines of convergence. In G.E. Stelmach & J. Requin (Eds.), Tutorials in motor behavior. Amsterdam: North-Holland. Kerr, B. (1975). Processing demands during movement. Journal of Motor Behavior. 7, 15-27. Kerr, B. (1978). Task factors that influence selection and preparation for voluntary movements. In G.E. Stelmach (Ed.), Information processing in motor control and learning. New York: Academic Press. Klapp, S.T. (1977). Reaction time analysis of programmed control. In R.S. Hutton (Ed.), Exercise and sport science reviews (Vol. 5). Santa Barbara, CA.: Journal Publishing Affiliates. Klapp, S.T. (1980). The memory drum theory after twenty years: Comments on Henry's note. Journal of Motor Behavior. 12,169-171. Klapp, S.T. (1981). Motor programming is not the only process which can influence RT: Some thoughts on the Marteniuk and MacKenzie analysis. Journal of Motor Behavior. 13, 320-328. Klapp, ST., & Rodriguez, G. (1982). Programming time as a function of response duration: A replication of "dit-dah" without possible guessing artifacts. Journal of Motor Behavior. 14, 46-56. Klapp, ST., & Wyatt, E.P. (1976). Motor programming within a sequence of responses. Journal of Motor Behavior. 8,19-26. Kugler, P.N., & Turvey, M.T. (1987). Information, natural law and the self assembly of rhythmic movement. Hillsdale, N.J.: Erlbaum. McLeod, P. (1980). What can RT tell us about the attentional demands of movement? In G.E. Stelmach & J. Requin (Eds.), Tutorials in motor behavior. Amsterdam: North-Holland. Moore, S.P., & Marteniuk, R.G. (1986). Kinematic and electromyographic changes that occur as a function of learning a time-constrained aiming task. Journal of Motor Behavior. 18, 397-426. Newell, K.M., Carlton, M.J., Carlton, L , & Halbert, J.A. (1980). Movement velocity as a factor in movement timing accuracy. Journal of Motor Behavior, 12, 4756. 58  Norman, R.W., & Komi, P.V. (1979). Electromechanical delays in skeletal muscle under normal movement conditions. Acta Phvsiologica Scandinavica. 106, 241-248. Ostry, D.J. (1980).Execution time control. In G.E. Stelmach & J.Requin (Eds.), Tutorials in motor behavior. Amsterdam: North-Holland. Ostry, D.J. (1983). Determinants of interkey times in typing. In W.E. Cooper (Ed.), Cognitive aspects of skilled typewriting. New York: Springer-Verlag. Pezzack, J.C., Norman, R.W., & Winter, D.A. (1977). An assessment of derivative determining techniques used for motion analysis. Journal of Biomechanics, 10, 377-382. Povel, D.J., & Collard, R. (1982). Structural factors in patterned finger tapping. Acta Psvchologica, 52, 107-123. Povel, D.J., & Essens, P. (1985). Perception of temporal patterns. Music Perception. 2, 411-440. Rosenbaum, D.A., Hindorff, V., & Munro, E.M. (1986). Programming of rapid finger sequences. In H. Heuer & C. Fromm (Eds.), Generation and modulation of action patterns. Berlin: Springer-Verlag. Rosenbaum, D.A., Kenny, S., & Derr, M.A. (1983). Heirarchical control of rapid movement sequences. Journal of Experimental Psychology: Human Perception an Performance. 9, 86-102 Rosenbaum, D.A., & Patashnik, O. (1980). A mental clock setting process as revealed by reaction times. In G.E. Stelmach & J.Requin (Eds.), Tutorials in motor behavior. Amsterdam: North-Holland. Sidaway, B. (1988). Fractionated reaction time in lower leg responses: A note on response programming time. Research Quarterly for Exercise and Sport. 59, 248-251. Sternberg, S., Monsell, S., Knoll, R.L, & Wright, C.E. (1978). The latency and duration of rapid movement sequences: Comparisons of speech and typewriting. In G.E. Stelmach (Ed.), Information processing in motor control and learning. New York: Academic Press. Stetson, R.H., & McDill, J.A. 1923. Mechanisms of the types of movements. Psychological Monographs. 32,18-40. Teulings, H.L., & Maarse, F.J. (1984). Digital recording and processing of handwriting movements. Human Movement Science. 3,193-217. Teulings, H.L, Mullins, P.A., & Stelmach, G.E. (1986). The elementary units of programming in handwriting. In H. Kao, G.P. van Galen, & R. Hoosain (Eds.), Graphonomics: Contemporary research in handwriting. Amsterdam: NorthHolland. van Donkelaar, P., & Franks, I.M. (1989a). The preparation and initiation of simple rhythmical patterns. Manuscript under review. 59  van Donkelaar, P., & Franks, I.M. (1989b). RT and reproduction accuracy in tapping movements with varying temporal structure. Manuscript under review. Vince, M.A, & Welford, A T . (1967). Time taken to change the speed of a response. Nature. 213, 532-533. Woodworth, R.S. (1899). The accuracy of voluntary movement. Psychological Monographs. 3, 1-114. Wells, R.P., & Winter, D.A. (1980). Assessment of signal and noise in the kinematics of normal, pathological and sporting gaits. In Human Locomotion I: "Pathological Gait to the Elite Athlete". Proceedings of the Special Conference of the Canadian Society for Biomechanics. London, Ontario, October 27-29. Young, R.P., Allard, F., & Marteniuk, R. (1988). The kinematics of visuallyfeedback-based corrections. Paper presented at the 19th annual meeting of the Canadian Society for Psychomotor Learning and Sports Psychology. Collingwood, Ontario, November 10-13.  60  APPENDIX A REVIEW OF LITERATURE RT Studies: The use of reaction time (RT) has a long history in motor behavior research. The underlying assumption behind its use is that differential latencies can help uncover the processes involved in movement planning. It is suggested that to plan a movement a number of processes must be completed. When the movement is a relatively simple one, these planning processes take a relatively short amount of time. However, as the movement becomes more complex, the time required to run through each process lengthens; with the result being an increased RT. One prerequisite to the use of RT as an inferential tool is the acceptance of the hypothesis that we somehow represent movements centrally. Those adopting the motor systems approach readily accept such a scenario. However, action systems theorists reject this idea, opting instead to explain movement control from a physical, noncognitive perspective. This issue will be dealt with in greater detail in the section concerning the plausibility of representationalism later in this review. Freeman (1907) was likely the first to measure the relationship between RT and movement complexity. He found that when subjects drew geometric figures such as a straight line, a circle, or a pentagon, the RT became longer as the figure increased in complexity. These variations in latency were said to have occurred because of antagonistic muscular tensions originating from anticipation of the necessary movement reversals which occurred in the more complex sequences. Perhaps the most influential researcher to work within the RT/movement complexity paradigm was Franklin Henry. In a series of papers in the 50's and 61  60's, Henry and his co-workers laid the groundwork in terms of ideas and methodology which still guide investigators in the area today. Henry's work was underscored by his "Memory Drum Theory" (Henry & Rogers, 1960). In this theory, the basis of movement control was likened to the memory drum hardware in the computers of the day. It was suggested that the more complex the movement, the more complicated the memory drum program representing it. This more complicated program takes longer to initiate because: " a larger amount of stored information will be needed, and thus the  neural impulses will require more time for  coordination and direction into the eventual motor neurons and muscles." (Henry & Rogers, 1960, p. 450) Henry and Rogers tested this theory by having subjects produce either a single key lift response or a key lift response concatenated with movements to targets. They found that the simple movement required less time to prepare and initiate than the more complicated sequence of movements. Since that time a variety of researchers have investigated the relationship between RT and movement complexity. Basically, what has been found is that the relationship does not always hold. Although some have viewed the equivocality of the results as evidence that RT may not be a reliable dependent variable, it seems more viable that such a state of affairs is due more to the "maturation" of the paradigm. Specifically, each new finding appears to more clearly define the conditions and limits within which the relationship holds. Several issues have arisen concerning these conditions and limits. Each of these will be considered below. The first concerns the use of simple RT as opposed to choice RT. In a series of articles, Klapp (1977, 1980, 1981) put forth the idea that simple RT cannot reflect the processes involved in movement planning because the subject, 62  knowing in advance exactly the required movement, can prepare a response prior to the imperative stimulus. Thus, in Klapp's view, simple RT would not change with variations in movement complexity because it simply measures the latency associated with peripheral factors. Choice RT, on the other hand, was seen as a more accurate indicant of response planning because subjects in this situation had to wait until the imperative stimulus to initiate the processes associated with each specific movement. The idea that simple RT reflects only peripheral delays has been addressed by fractionating RT into its premotor and motor time components. Several researchers,  most notably Christina and his colleagues, have  demonstrated that the increase in simple RT as movement complexity is increased is almost exclusively due to an increase in premotor time (Anson, 1982; Christina & Rose, 1985; Fischman, 1984). Since premotor time is thought to reflect delays associated with central planning activity, these findings appear to refute Klapp's explanation of simple RT effects as a function of peripheral factors. Although Klapp's ideas concerning simple RT appeared to be unfounded, this, in itself, did not make choice RT any less efficacious as an inferential tool. However, in an influential paper Sternberg, Monsell, Knoll, and Wright (1978) pointed out several possible confounds in the choice RT methodology. Specifically, any differences in compatibility between a stimulus and its response, one response and another, and entire stimulus-response ensembles can make the interpretation of choice RT a difficult task. For this reason, some have chosen to avoid the use of choice RT (Sternberg, Monsell, Knoll, and Wright, 1978). Others, however, have found it useful to use both simple and choice RT in an effort to compile a more complete picture of the processes involved in movement planning (Garcia-Colera & Semjen, 1987; 1988). In the present study only simple RT was used. This choice was justified by 63  the recent findings of Canic and Franks (1989). In this study subjects responded in a simple RT paradigm with no uncertainty as to what movement was required or when it was to be initiated. That is, the trials were blocked across conditions and the foreperiod and imperative stimulus were temporally invariant. In addition, there were no catch trials to keep the subjects from anticipating. In spite of the certainty of these conditions, subjects still required more time to prepare and initiate the more complex movements. This finding confirms that movements cannot be entirely preprogrammed in a simple RT paradigm. Rather, some time must be spent after the imperative stimulus planning the movement sequence prior to its execution. With this in mind, the present investigation was undertaken using only a simple RT task. Another consideration which has arisen from the research conducted within the RT/movement complexity paradigm over the past 30 years is the selection, and appropriate manipulation of a suitable task. It is important in this decision to choose a task which is not confounded with other variables (Fischman, 1984), can easily be broken down into invariant elements (Sternberg, Monsell, Knoll, & Wright, 1978), and is meaningful in terms of movement control [i.e. not contrived to suit the needs of the experiment (Kerr, 1978)]. Each of these will be considered in turn. The task must not be confounded with other variables which also have an effect on RT. If this was to occur it would make the interpretation of any RT differences very arduous, indeed. One of the most difficult confounds to control for is the correspondence between the number of elements within a response and the duration of the response. That is, if the rate of movement is held constant, it is impossible to increase the number of elements in a sequence without increasing the total duration of the sequence. There is evidence to suggest that both of these variables have an effect on RT (Klapp & Wyatt, 1976). 64  However, it appears that the number of movement elements is the more important variable (Sternberg, Monsell, Knoll, & Wright, 1978) Another difficult confound is that between movement time, velocity, and distance moved. Again, it is impossible to manipulate one without having some effect on at least one of the others. For example, if movement time is held constant, and the distance to be moved is experimentally varied, subjects must produce their movements at different velocities to achieve the required outcome. Thus, it becomes impossible to determine if the resulting differences in RT were due to changes in the distance moved or the resultant velocity. In the present experiment movement time was confounded with the main experimental manipulation of movement velocity. However, several studies have shown that the rate at which a movement is completed has more of an influence on the nature of the underlying control than the actual duration of the movement (Canic, 1988; Garcia-Colera & Semjen, 1987; 1988). It is also important that each element within the task be similar. This is essentially Sternberg, Monsell, Knoll, and Wright's (1978) element invariance requirement. Basically what Sternberg and his colleagues suggested was that if any of the elements within the response were unique, then the task of interpreting any variance in the RT scores in a theoretically meaningful manner becomes difficult. That is, one can never be sure in this case if an increase in RT is due to the fact that an extra element has been added to the sequence or simply that the element was different than the others. Kerr (1978) has warned against the possibility of using a particular type of task simply because it suits the need of the experiment or reliably causes increases in RT. This can lead to filling the subject's head with movement parameters which are unrealistic, at best, and useless, at worst. This problem arose because, originally, research in the RT/movement complexity paradigm 65  was concerned with defining the parameters which needed specification prior to movement execution. That is, a movement parameter was said to be important in response planning if increasing the complexity of it in some manner led to an increase in RT. More recently, these parameters have been more clearly defined. In the present investigation a reversal in direction was the defined movement parameter. This parameter has been explicitly tested before by Glencross (1975). He found that RT did not increase significantly as the number of reversals increased. However, he did not have accurately defined points of reversal in his experiment. Subjects merely had to break a photoelectric cell then change direction. As a result, it was likely that the reversals were not as accurately controlled in space or time as in the present study. In more recent experiments reversals in direction have been implicit in movements which are linearly related to RT. For instance, tapping motions contain a reversal in direction at the apex of the movement (Canic, 1988; Franks & van Donkelaar, in press; van Donkelaar & Franks, 1989a, 1989b). Similarly, writing motions are produced with definite points of reversals and have been shown to be related to RT (Hulstijn & van Galen, 1983; Teulings, Mullins, & Stelmach, 1986). Also, the task originally used by Henry and Rogers (1960) and subsequently by Christina and his coworkers (Anson, 1982; Christina & Rose, 1985; Fischman, 1984) contains abrupt changes in direction at each target. Thus, it is likely that a movement reversal or change in direction is an important part of movement control. Intuitively, this point seems viable. That is, points of reversal appear to provide a natural transition from one aspect of a movement sequence to another. It may be that in producing particular movements we plan each segment from one reversal (or change in direction) to the next. This point is theoretically supported by the equilibrium-point hypothesis [i.e. setting new weights to the agonist-antagonist pair to affect a different point of 66  equilibrium (Feldman, 1968)], motor systems theory [i.e. planning a movement to a particular point in extracorporeal space and parameterizing a motor program to achieve that point (Schmidt, 1988)], and action systems theory [i.e. phase transitions from one type of movement to another (Kelso, Holt, Kugler, & Turvey, 1980)]. Preprogramming and On-Line Preparation: RT studies have been used to gain insights into the process of preprogramming. In this process the entire movement is planned in advance of its execution. In this sense, it can be viewed as open-loop behavior. Specifically, it is assumed that preprogrammed movements run their course uninfluenced by feedback. Typically, such movements are completed in such a short period of time or at such a fast rate that error corrections are impossible to make. For these latter types of movements (i.e. those completed at a fast rate) the overall duration may appear to provide ample opportunity for feedback and appropriate adjustments. However, because the movement is being done at such a fast rate, by the time a correction can be made, it would be nonfunctional in the updated conditions in which the limb is now situated. Under such conditions the only recourse for the subject is to plan the movement in advance. As a consequence, the RT required to prepare and initiate such movements increases as the movement becomes more complex. However, for longer more slowly completed movements feedback can be used in a functional manner. The challenge has been how to detect and quantify the presence of the feedback/adjustment loop. RT can not be used in this analysis because, by definition, it measures only the activity occurring prior to movement execution. As discussed in the introduction, several researchers have looked to the interresponse intervals (i.e. the time from the beginning of one movement element to the next) for evidence of feedback and the subsequent 67  adjustments (Canic, 1988; Ostry, 1983). They justify the use of this measure by suggesting that variations in the IRI's reflect delays associated with planning activity as the movement is being executed. Although this type of data can tell us when in time any on-line prepared adjustments were produced, it can not tell us the nature of such adjustments (other than to demonstrate that the adjustment increased the duration of the movement segment under consideration). Another method which has been employed to investigate the presence of feedback and adjustments has been probe RT (e.g. Ells, 1973; Kerr, 1975). In this paradigm a secondary task is completed during the main movement sequence. Usually this involves reacting as quickly as possible to a stimulus unrelated to the main task. The point in time at which this stimulus occurs in the movement is manipulated in a systematic manner. Assuming that attention has a limited capacity, it is suggested that elements or segments of the main task which require greater attentional demand will lead to increased probe RT scores for the secondary task. If the main task is entirely preprogrammed, attentional demands should remain invariant throughout the movement; however, if the movement is prepared on-line the attentional demands should increase during the parts of the movement at which such processing occur; and this would be reflected in increased probe RT scores. Franks, Wilberg, and Fishburne (1985) used this methodology in a serial pattern learning study. Their purpose was to demonstrate that more processing activity occurred between subunits of a pattern than during such subunits (termed "runs"). They did, indeed, find that probe RT increased during the intervals in between the runs in comparison to the intervals during the runs. This can be used as evidence supporting the assumption that on-line control was occurring during these sequences. Unfortunately, the use of probe RT as an inferential tool was questioned 68  by McLeod (1980). He demonstrated that several confounds within the probe RT methodology could make subsequent interpretations difficult. The most obvious was the interference caused by using the same response modality (e.g. motor, vocal, auditory, etc.) for both the main and secondary task. For example, using an arm extension/flexion movement as the main task and a push-button response for the secondary task. When McLeod (1980) changed the secondary task from a manual to vocal one, he found a different pattern of probe RT results. From this he concluded that previous probe RT studies had found properties of movement which were only relative to the competing manual tasks and not absolute in terms of control. Kinematic Studies: Because of the difficulties associated with the inferential analysis of temporal data, some have suggested that examining the movement per se for evidence of on-line control (i.e. via the kinematics) may be a more fruitful endeavor. Analysis of the accelerations produced in a movement sequence has a long history in motor behavior research. Stetson produced several studies in the early part of this century which demonstrated the usefulness of this measure (Stetson & McDill, 1927). His basic thesis was that ballistic, preprogrammed movements resulted in acceleration traces composed of smooth symmetrical acceleration and deceleration portions with a period of minimal activity in the middle. During this time the limb under consideration was said to be moving simply via its own momentum. On the other hand, slow controlled movements resulted in acceleration traces with several peaks and valleys, suggesting the presence of several submovements. This evidence was elaborated upon by Brooks, Cooke, and Thomas (1973). They found that single aiming movements produced by monkeys could be divided into two categories: "continuous" and "discontinuous". In continuous 69  movements the angular acceleration traces were relatively smooth and crossed the zero line only once. During discontinuous movements, however, these traces contained large deviations which led to several zero line crossings within each movement segment. Brooks and his coworkers suggested that the continuous movements were composed of one preprogrammed movement segment; whereas, the discontinuous movements contained several submovements which were, of necessity, prepared on-line. Thus, using this criteria, the degree to which a movement is preprogrammed is reflected in the number of times the underlying acceleration trace crosses the zero line. Another measure of preprogramming that can be gleaned from the acceleration trace is the presence or absence of "significant deviations" (where a significant deviation is said to have occurred after a criterion number of data points have been produced in the direction opposite to that which is normally found in a particular movement segment). Such deviations are not normally found in preprogrammed movements. Indeed, the acceleration trace in a movement completed as quickly as possible is usually smooth and symmetrical (Crossman & Goodeve, 1963/1983). However, movements which have the potential to be prepared on-line usually result in acceleration traces which contain deviations. One of the more extensively studied tasks employing the kinematic analysis of feedback based corrections has been that of single aiming movements. The impetus for this paradigm grew out of the research concerning speed-accuracy trade-offs in motor skills (Fitts, 1954; Woodworth, 1899). The question this research has attempted to answer is: what is the mechanism of control which leads to such a trade-off? The most often cited suggestion has been that feedback is used to correct for any errors in the initial part of the movement. A response which requires a high degree of terminal accuracy will need more corrections to achieve such accuracy than one which occurs under 70  more lenient conditions. Because corrections take time, a movement performed under stringent accuracy requirements will take longer to complete than one performed under lenient conditions. This hypothesis, known as the deterministic iterative-corrections model, was put forth by Crossman and Goodeve (1963/1983) and subsequently elaborated upon by Keele (1981). Basically, Crossman and Goodeve suggested that in moving from a home position to a target region a subject makes a series of submovements until the target is attained. These submovements are made on the basis of sensory feedback such that the error from one movement is used to make corrections during the subsequent movements. Each of the submovements supposedly has a well-defined beginning and end, requires a constant amount of time to complete, and travels a constant proportion of the remaining distance to the target region. The model is "deterministic" in that it does not incorporate any random noise due to neuromotor variability into the movement outcomes. If such a mode of control were used by subjects in the present experiment, then the distribution function of the acceleration deviations should have been of a different form. Specifically, the deviations should have occurred progressively closer together in time as the point of reversal was approached. However, it appears that they occurred, instead, at equidistant specific locations within each segment. Thus, the perception-action loop in this particular task results in a discontinuous, but, nevertheless, consistent behavioral outcome. One of the reasons that the results from this experiment were different from those of previous papers may be due to differences in the task itself. Specifically, the majority of studies requiring arm movements as a task have typically used only a discrete flexion or extension. Rarely have repetitive flexion and extension cycles similar to the movement used in the present study been examined [see Benecke, Rothwell, Day, Dick, and Marsden (1986), and Schmidt, 71  Sherwood, and Walter (1988) for exceptions]. This type of movement, however, is important if an understanding of the mechanisms involved in on-line control is to be gained. In addition, it has also been very rare for a less than maximal speed of movement to be used. By their very nature, fast, discrete movements minimize the potential for control in an on-line fashion; and, thus, have been preferred by researchers  concerned  with  elucidating  the  processes  involved in  preprogramming. However, it is naive to think that preprogrammed movements represent the majority of actions produced in everyday life. Indeed, the equivocal relationship between RT and movement complexity, especially as the movement becomes more realistic [e.g. completed at a less than maximal rate (Canic, 1988; Garcia-Colera & Semjen, 1987; 1988), or with accuracy constraints (Franks & van Donkelaar, in press)], suggests that preprogramming occurs only within limited conditions. When operating outside of such conditions, some form of online control must be used to successfully produce a functional action. EMG Studies: Electromyography (EMG) is the graphical representation of the electrical activity of muscles. Branches of a motor neuron activate the motor endplate of a muscle fibre. This results in two depolarization waves which travel to either end of the muscle fibre. Because the tissue in and around the active fibre is electrically conductive, the fibre depolarization can be recorded at the surface (Hof, 1984). The EMG signal has a variety of uses and interpretations. At the most basic level, the EMG recording can tell the investigator when the muscle activity occurs in time. If movement is recorded in conjunction with the EMG then the correlation between when the muscle is active and when the movement occurs can be addressed. 72  The amount of muscular activity can be analyzed by full-wave rectifying and integrating the EMG signal. The relationship between this signal and movement can provide more information about how muscle activity affects movement outcomes than the temporal analysis described above. Finally, by measuring the force produced in a movement as well as the EMG signal underlying the movement, one can gain the greatest insight into the relationship between these two variables. This process is the most complicated of the three due to the complexities of the muscle contraction mechanics. In the present study, the most basic level of EMG analysis, that of the temporal relation between the signal and the consequent movement, was used. It was felt that this was sufficient for the purposes of the study, and that analysis of the amount of activity or force of movement were beyond the scope of the investigation. Given the constraints chosen above, what has been shown in studies using EMG recordings regarding the differences between preprogrammed movements and those prepared on-line? Desmedt and Godaux (1979) demonstrated that the duration of activity in relation to that of the movement is a viable indicant of the degree of preprogramming. As described in the introduction, these researchers had subjects produce index finger abduction movements either as fast as possible (ballistically) or at a more controlled pace (ramped). They found that the relative duration of the EMG activity was greater in the slower movements (i.e. the activity was extended over a greater duration of the movement) than in the fast movements. A similar result was found by Hallett, Shahani, and Young (1975) using a larger scale arm movement similar to the one used in the present investigation. Such results indicate that ballistic movements are initiated by a large impulse of muscle activity after which the limb continues moving under its own 73  momentum (Stetson & McDill,  1928). Such movements appear to be  preprogrammed and, thus, uninfluenced by feedback. On the other hand, ramped movements are characterized by continual muscle activity, indicating some form of on-line feedback/ adjustment control. Such hypotheses can be confirmed by perturbing the movement in some manner to see whether adjustments can be made. The rationale being that movements under on-line control will exhibit such adjustments, whereas, preprogrammed movements will not because of the absence of a feedback loop. Desmedt and Godaux (1979) included such conditions in their experiment. Specifically, they unexpectedly removed a load on the index finger while the movement was underway. They found that adjustments were made in both the limb trajectory and the underlying EMG activity in the slow movements but not in the fast movements. Again, similar results have been found in larger scale arm movements (Cooke, 1980). In the present study, load perturbations were not included as an experimental manipulation. Although this would increase the confidence with which conclusions could be made regarding the presence of preprogramming or on-line control, it was felt that the combination of the RT, acceleration, and EMG temporal analysis was sufficient for this purpose. Representationalism: One of the main assumptions of the motor systems approach is that movements are in some way represented centrally. Traditionally, this representation has been characterized as a motor program. In this context a motor program is an abstract generalization of the movement under consideration. During each iteration of the movement, parameters of the motor program are calculated which allow the movement to satisfy the goals of the task under the conditions in which it occurred. According to this view, then, the motor 74  program controls the eventual outcome of the actor/environment interaction. When the motor program concept was first defined (Keele, 1968) it was at the expense of a large loan on intelligence. That is, information from the environment had to be taken in and integrated with the goals of the performer. This process allowed the parameterization of the motor program; which, in turn, led to movement production. Such a process required a very smart central controller. Some suggested that the controller could be modelled after a computer. However, since a computer can only respond when given commands, this hypothesis begs the question: Who gives the commands? The answer, of course, is a computer programmer in the mind. But, then, who controls the programmer's actions. This line of reasoning obviously leads to an infinite regress with very little resolution regarding the question of movement control. These problems are greatly reduced in the action systems approach (Kelso, Holt, Kugler, & Turvey, 1980; Kugler & Turvey, 1988). Researchers from this perspective suggest that there is no need to represent movements centrally. Instead, control is gained via the inherent dynamical properties present in the limbs themselves. In effect, these researchers are attempting to see how much control can be gleaned from a purely physical system without resorting to a higher authority (i.e. a motor program). Within the action systems approach, then, control is seen as a potentiality that "falls out" of the system; and not as something that we govern. However, this theory of movement control still requires some sort of "go" signal to start the limbs in motion. That is, a decision regarding choice amongst several movement alternatives must still be made. Only after such a decision has been reached can the limbs be left to their own "intelligent" devices. The role of the periphery, then, may be to set limits on the possible types of movements that can be accomplished given a particular environmental context. For example, 75  when confronted with an obstacle which blocks our normal progression, we can proceed using a number of different strategies. For instance, we could simply step, hop, or climb over it, take another route, walk around it, etc. Despite the large number of alternatives, this list is not endless. There are certain forms of movement which are not possible given the context in which the obstacle occurs. However, it does not seem possible that the limbs would know this. Rather, some mental decision must be made which eliminates the impossible movements (given the context) and leads to a choice among those that are possible. This decision-making process may be analogous to a simplified motor program. That is, one which just gives a go signal for a particular movement but does not control the movement in any way. The methodology and underlying theory of the present experiment is obviously motor systems in nature. That is, the RT/movement complexity paradigm assumes that movements are represented and controlled centrally. However, as mentioned in the previous paragraph, this representation and control may take the form of a simple yes/no decision if one assumes that the periphery is capable of controlling itself by taking advantage of its own inherent dynamical properties. Thus, one can still postulate the existence of "weak" motor programs (and the resulting outcomes given certain experimental manipulations) without taking out a large loan on intelligence.  76  APPENDIX B OPTIMIZATION OF DISPLACEMENT CUT-OFF FREQUENCIES In order to obtain the acceleration traces produced during a movement, the displacement data must be differentiated two times. During this process any small perturbations in the data become progressively larger. This results in an acceleration trace composed mostly of high frequency noise. In order to avoid such a situation, the displacement data must first be sent through a filter which removes any of the small perturbations but allows the larger more meaningful ones through. This is accomplished by setting an upper frequency cut-off within the filter algorithm. Thus, movements which cycle at a frequency above the cutoff will be removed; whereas, those below the cut-off will be kept, and subsequently differentiated. The question is: at what value should the cut-off frequency be set? For normal arm and hand movements volitional control appears to extend to frequencies of approximately 10 Hz (Brooks, Cooke, & Thomas, 1973; Teulings & Maarse, 1984). That is, anything below this frequency is under the control of the subject; whereas, anything above it is due to noise within the subject (i.e. physiological tremor) or noise within the data collection system. Thus, this cut-off frequency was chosen for the pilot study (see Appendix D). However, because three different speeds of movement were used in the pilot study, it may have been more appropriate to use a different cut-off frequency in each condition. Biomechanical studies show that a cut-off frequency of 5 times the base frequency of the actual movement may be the most optimal. Thus, in the pilot study the cut-off frequency for the FAP movement (3.6 to 6.3 Hz.) should have been anywhere from 18 to 31.5 Hz. By including these higher frequency components in the differentiation process, the resulting acceleration trace may have been less smooth than that which was originally obtained. 77  A third possibility was presented by Wells and Winter (1980). They suggested that the cut-off frequency should be individualized for each separate trial. In order to accomplish this task, they first measured the goodness of fit between raw and filtered displacement data over a range of cut-off frequencies. The optimum cut-off frequency for any particular trial was set at that point at which the signal component accounted for at least 20% of the total system noise. A graphical presentation of such a situation is provided in Figure 16. It plots the RMS deviation of the residual (the difference between the raw and filtered data) at each cut-off frequency for separate trials in both the SLW and FAP conditions. The straight line which slopes downwards represents the noise inherent in the system. The curved line which rises above the noise level represents the increase in RMS deviation as the cut-off frequency is lowered. The cut-off frequency is reached when this error is 20% greater than the noise level of the system. Thus, for the SLW condition trial the cut-off frequency was set at 23 Hz.; whereas, in the FAP condition trial it was set at 32 Hz. When the data were processed using this procedure, there were some noticeable differences between the results and what was found in the pilot study; especially in the SLW condition. Most notably, the deviations were not grouped in distinct locations, and the time between each deviation was greatly reduced. This second result is of the most importance from a theoretical point of view because it provides a temporal duration for the process by which control is afforded. Specifically, in the pilot study it was found that the majority of the deviations were separated by approximately 85 msec; however, using the Wells and Winter process, that duration was shortened to 30 msec. Since the movements were left basically unchanged, this difference can be directly attributed to the increased filter cutoff frequencies. In order to test whether this procedure produced an accurate 78  FIGURE 16 RMS ERROR BETWEEN RAW DISPLACEMENT DATA AND FILTERED DATA AS A FUNCTION OF FILTER CUTOFF FREQUENCY SLOW CONDITION  10  15  20  25  30  35  40  FILTER CUTOFF FREQUENCY (Hz.) RESIDUAL  SYSTEM NOISE  FAST CONDITION  45  18  23  28  33  38  FILTER CUTOFF FREQUENCY (Hz.) RESIDUAL  S Y S T E M NOISE  43  representation of the angular acceleration profile, data was also collected for several trails in both the SLW and FAP conditions with both an accelerometer and the potentiometer. The displacement data was filtered at cut-off frequencies from 5 to 30 Hz, and was subsequently differentiated. The resulting angular acceleration traces were analysed for the presence of zero line crossings and deviations. By simultaneously collecting data with an accelerometer, a reference was provided for comparing the values obtained for each dependent variable. In the  FAP condition no differences were found between the  accelerometer data and the differentiated data for either the zero line crossings or deviations data across all cut-off frequencies. From a topological perspective however, the accelerometer and differentiated data were most similar when frequencies above 18 Hz. (not 32 Hz. as obtained from the Wells and Winter algorithm) were removed from the displacement curve. Thus, this frequency was chosen for the FAP condition. In the SLW condition, the number of zero line crossings and the number of deviations both increased as the filter cut-off frequency increased. As can be seen in Figure 17, the values obtained for these dependent variables in the differentiated data equalled the values from the accelerometer data at 10 Hz. for the number of deviations and 12 Hz. for the zero line crossings. This is well below the value of 17 Hz. obtained from the Wells and Winter algorithm. Because the zero line crossings measure more discretely represents any changes which may occur in terms of control, it is less sensitive than the deviations measure to variations in cut-off frequency. As such, the cut-off frequency at which the differentiated data equals the accelerometer data in terms of zero line crossings appears to provide the most accurate account of the angular acceleration trace. Thus, the value obtained in the zero line crossings measure of 12 Hz. was chosen as the cut-off frequency for the SLW condition. 80  FIGURE 17 COMPARISON OF ACCELEROMTER AND DIFFERENTIATED DATA ACROSS A RANGE OF CUT-OFF FREQUENCIES - SLW CONDITION # OF DEVIATIONS  # OF ZERO LINE CROSSINGS 7  5  6  7  8  9  10  11  12  13  14  CUT-OFF FREQUENCY (Hz.)  - — A C C E L E R O M E T E R DATA  -+-  15  DIFFERENTIATED  16  DATA  5  —  6  7  8  9  10  11  12  13  14  CUT-OFF FREQUENCY (Hz.)  A C C E L E R O M E T E R DATA  -+-  15  DIFFERENTIATED  16  DATA  APPENDIX C EMG SAMPLING AND ANALYSIS In order to find the most suitable sampling frequency for the EMG data collection, as well as the most suitable algorithm for determining the start and end of EMG activity, the following procedure was carried out. First, the activity produced in the triceps muscle during a ballistic extension movement was sampled at a frequency of 5006 Hz (which is faster than the recommended rate of 500 to 1000 Hz). The resulting data was then "sampled" at various other rates by the computer and subsequently analysed using several different algorithms. This "sampling" was carried out as follows: taking every second data point as the data to be analysed yielded a sampling rate of 2503 Hz; taking every third point yielded a sampling rate of 1686.6 Hz, etc. The sampling rates ranged from the original of 5006 Hz to 98 Hz (in which every fifty-first point of the original data was used). The modified data was then rectified so that the normally bipolar EMG data was made unipolar. Next, the rectified data was subjected to a digital filter with a cut-off frequency of 10 Hertz. The cut off frequency sets the limit of the increase in amplitude from point to point and so eliminates all the large changes in amplitude of the original data. By including these two steps, the originally "spiked" appearance of the EMG data was replaced with a smooth curve. In the next step, a series of algorithms were applied to the smoothed and rectified EMG data. The purpose of these algorithms was to locate the starting and ending points of the EMG activity. For each algorithm, the data was sent through at the different "sampling" frequencies determined using the procedure described above. This resulted in separate EMG starting and ending points for each "sampling" frequency. Graphs were then made displaying these starting 82  and ending points as a function of the frequency at which the data had been "sampled". Typically, at the higher frequencies, the starting and ending points remained  relatively unchanged.  However, as the  "sampling"  frequency  decreased, these points became more variable. The lowest frequency at which both the starting and ending points remained unchanged was defined as the minimally acceptable sampling frequency. The first algorithm used a criterion based on the mean and standard deviation of the baseline to determine the start and end of EMG activity. Specifically, EMG activity was said to have started when a point was found which was more than two standard deviations above the mean of the baseline activity. Similarly, EMG activity was said to be completed at the last point which satisfied this requirement. Unfortunately, this algorithm did not work because the mean and standard deviation of the baseline activity were very close to zero. Thus, any small deviations which occurred during the period prior to or after the actual EMG activity were incorrectly identified as the main impulse of activity. The second algorithm used the largest EMG value in the baseline activity as a criterion. Specifically, the first of five successive points which were greater than twice the amplitude of this maximum was considered to be the starting point of EMG activity. Similarly, the last of five successive points to be greater than this value was considered to be the ending point of EMG activity. This algorithm worked better than the first one, but it still captured some early bursts of EMG activity at the lower "sampling" frequencies. As can be seen in Figure 18, the actual EMG starts at around 820 msec, but there are bursts at 520 and 720 msec. At the lower "sampling" frequencies the algorithm would detect these short bursts as the beginning of the main EMG activity. Increasing the duration stipulation to ten points did not produce any different results than the five point criterion (see Figure 18). 83  FIGURE 18 - EMG START AND STOP TIMES AS A FUNCTION OF SAMPLING RATE FOR 4 DIFFERENT ALGORITHMS ALGORITHM 3 10 HZ FILTER, 10 INCREASING POINTS  ALGORITHM 2 10 HZ FILTER, 5 INCREASING POINTS 2000  2000  • STOP TIME  1760 "J"  M E M S E C  1750 "j"  1600  M E  1250 1000 750  S  600 -  c  250  1500 1250 1000  ' « 600 250  o — 1  400  4000  40  SAMPLING RATE (HZ)  400  4000  SAMPLING RATE (HZ)  ALGORITHM 4 15 HZ FILTER, 5 INCREASING POINTS  ALGORITHM 5 20 HZ FILTER, 5 INCREASING POINTS  2000 1760 "j"  M E M S E C  - 8TART TIME  — STOP TIME  1600 1250 1000 760 500 260 _i—i—*  * <' •  i  i  i  i  • • ' •'  400  SAMPLING RATE (HZ)  4000  i _  400  SAMPLING RATE ( Hz )  4000  The second algorithm (with the five point duration stipulation) was subsequently used with data which was filtered at upper frequency cut-offs of 15 and 20 Hertz. Figure 18 also shows the results of this procedure. From these graphs it appears that increasing the cut-off frequency led directly to an increase in the minimally acceptable sampling frequency. A third algorithm used an absolute duration criterion. Specifically, EMG activity was said to have occurred if the data points were more than two times the maximum of baseline activity, and lasted a minimum length of time. Two different durations were chosen: 40 and 100 msec (see Figure 19). Using the 100 msec duration as a criterion, the algorithm successfully located the start of EMG activity; disregarding the small bursts occurring at 520 and 720 msec. The algorithm also constantly gave a smaller value for the stop time of the EMG in comparison to the previous algorithms (i.e. 1280 msec vs 1480 msec) This is attributed to the bypassing of short duration EMG bursts at the trailing end of the physical movement. Unfortunately, there were times within the movement trials during which EMG activity lasted less than 100 msec, especially during the first burst of the triceps. For the 40 msec duration criterion, the algorithm was still able to locate the starting and ending points consistently. In addition, it was felt that this duration was short enough to capture any meaningful activity at the start of movement. Thus, for the purposes of this study, the algorithm requiring the signal to last 40 msec in duration and be at least two times the maximum of the baseline was used to detect the presence of muscular activity.  85  FIGURE 19 - EMG START AND STOP TIMES AS A FUNCTION OF SAMPLING RATE FOR 2 DIFFERENT ALGORITHMS  ALGORITHM 6 10 HZ FILTER, 100 MSEC OF INCREASE  ALGORITHM 7 10 HZ FILTER, 40 MSEC OF INCREASE 2000 -  2000 r 1750 -  T I M E M S E C  - START TIME  • STOP TIME  1750 -  T I M E  1500 1250 1000 -  M S E C  750 500 250 -  40  400  SAMPLING RATE (HZ)  4000  - START TIME  • STOP TIME  1500 1250 -  /hrvvv-  1000 750 500 250 0 40  400  SAMPLING RATE ( Hz )  4000  APPENDIX D PILOT STUDY INTRODUCTION It has been suggested that the time taken to initiate a movement sequence is related to the complexity inherent in that sequence (Henry & Rogers, 1960). Specifically, reaction time (RT) increases as the movement becomes more complex. This increase in RT has been used as evidence to suggest that we control our movements through a series ofinternal processes; each of which takes a finite amount of time to complete. The assumption being that the more complex the movement, the longer it takes to run through one or several of these processes - with the result being a longer RT. There are, however, some conditions in which increases in movement complexity do not lead to increases in RT. In such situations some aspect of the interaction between the actor and the action allows the sequence to be prepared in parts during the movement as opposed to entirely before the movement. The present investigation was an attempt to detect and quantify these preparatory processes as they occurred within various movement sequences. Henry and Rogers (1960) were among the first to show the relationship between RT and movement complexity. They found that RT was greater when subjects had to move to a series of specified targets than when they merely had to make a simple key lift response. More recently, this effect has been investigated using a variety of tasks, including keystrokes and tapping (Fischman, 1984; Klapp & Rodriguez, 1982; Rosenbaum & Patashnik, 1980a) speech (Erikson, Pollack & Montague, 1970; Sternberg, Monsell, Knoll, & Wright, 1978) and handwriting (Hulstijn & van Galen, 1983; Teulings, Mullins, & Stelmach, 1986). The parameter which appears to be most strongly related to changes in 87  RT in these studies is the number of response elements that comprise the movement. Thus, it is the number of taps or keystrokes, number of stress groups in a sequence of speech, and number of strokes taken in writing a letter which have the most influence on the time required to prepare and initiate these responses. Studies that have found increases in RT with increases in movement complexity have commonly used maximal speeds of response. Specifically, subjects were required to not only initiate their movements as quickly as possible, but also complete them in as short a time as possible. More recently, a number of investigators, using variable rates of response, have uncovered some interesting results (Canic & Franks, 1989; Franks & van Donkelaar, in press; Garcia-Colera & Semjen, 1987; 1988; van Donkelaar & Franks, 1989a; 1989b). Specifically, they have found that RT increased linearly in relation to the number of movement elements at the quickest rates, but either failed to do so, or did so nonlinearly at slower rates. Thus, when a movement sequence is done at a less than maximal speed, changing the complexity of it by adding extra elements does not reliably cause an increase in the RT required to prepare and initiate it. This finding has been explained by suggesting that some form of "on-line" or parallel system of preparation is taking place (Rosenbaum, Hindorff, & Munro, 1986). The logic here is as follows: If subjects produce a movement sequence by programming or preparing it prior to its actual execution (from here on termed preprogramming), then increasing the number of elements within the sequence should also cause an increase in this preparation time; with the result being a longer RT. However, if increasing the number of elements does not cause an increase in RT, then the subjects must not have to plan the entire sequence beforehand; rather, some aspect of this process can carry on into the period of movement execution. 88  Unfortunately, a lack of increase in RT when movement complexity is increased appears to be the most popular explanation of on-line preparation in the literature. Although variations in the inter-response intervals within a movement sequence have been used by several researchers as evidence for the on-line preparation process (Canic, 1988; Garcia-Colera & Semjen, 1988; Ostry, 1983; Rosenbaum, Hindorff, & Munro, 1986), other potential measures which appear to be sensitive to the differences between preprogramming and on-line preparation have been relatively absent within the RT/response complexity paradigm. One way in which preprogrammed movements have been differentiated from those prepared on-line in studies which have not used RT as a dependent variable is through the analysis of the acceleration traces from the movement. For example, Brooks, Cooke, and Thomas (1973) found that the angular acceleration traces produced by monkeys during "continuous" step tracking movements crossed the zero line only once from flexion to extension; whereas, those produced during "discontinuous" movements had several zero line crossings. The former traces have historically been associated with fast ballistic movements assumed to be under preprogrammed control (Stetson & McDill, 1923); the latter, with slower movements that display small adjustments related to feedback-based corrections (Woodworth, 1899). Similarly, when a discrete aiming movement to a target is the required task, the resulting acceleration trace can provide information relating to the time course of adjustments made as a result of on-line feedback (Young, Allard, & Marteniuk, 1988). Carlton (1981) found that such adjustments were easily detected as an abrupt change in the deceleration pattern or an increase in acceleration near the target. In movements in which feedback is not used, the acceleration traces are typically smooth and symmetrical (Crossman & Goodeve, 89  1963/1983). Thus, by analyzing the acceleration traces produced during movements, it appears that the extent and timing of on-line preparation can be determined. Specifically, the number of zero line crossings within the acceleration trace as well as any significant deviations from a smooth curve may be a potential indicant of adjustments made during the movement. In the present investigation the angular acceleration traces from a horizontal repetitive arm extension/flexion movement were measured in addition to the RT required to initiate such movements. The number of extension/flexion movement segments required in each trial as well as the speed with which these movements were completed was manipulated in such a way so as to lead to either preprogramming or on-line preparation. It was hypothesized that the movements which were preprogrammed would lead to increases in RT with increases in complexity. In addition, the angular acceleration traces associated with such movements would display only one zero line crossing and no significant deviations from a smooth curve within each extension or flexion segment. By contrast, the movements prepared on-line would not lead to increases in RT with increases in complexity. In addition, the resulting angular acceleration traces would have several zero line crossings and display significant deviations from a smooth curve during each movement segment.  90  METHOD Subjects: Twenty undergraduate physical education students took part in the study as part of a course requirement. All were inexperienced at the task and naive as to the experimental hypothesis. A $50 reward was offered to that subject who best combined the task of reacting quickly and completing the movement accurately. Task and Apparatus: The subjects performed a repetitive arm extension/ flexion movement through a range of 45° degrees (from 50° to 95° degrees - where 180° degrees was defined as full extension). The right forearm was positioned on a padded horizontal lever attached to a bearing-mounted vertical shaft such that the elbow was coaxial with the point of rotation. The hand was supinated to grasp a vertical handle at the end of the lever. The position of the handle was adjusted to accommodate for varying forearm lengths. The height at which the subject was seated was adjusted so that the angle at the shoulder remained constant across subjects. Subjects viewed a stimulus cursor that cycled back and forth (describing a cosine wave) across an oscilloscope screen at a predetermined rate and amplitude. One cycle of the stimulus cursor started in the middle of the screen, proceeded to the right side, reversed direction and went to the left side, then reversed direction again to end up back at the center. The subjects were instructed to react as quickly as possible, after an imperative stimulus, and reproduce as accurately as possible from memory the stimulus cursor's path by flexing and extending about the elbow joint. The angular displacement of the lever was transmitted by means of a belt to a linear potentiometer, and represented on the oscilloscope screen by a 91  response cursor (5° of arm movement equalled 10 mm of cursor movement). The displacement data was sampled at a rate of 250 Hz and stored on-line by an IBM "XT" microcomputer for subsequent analysis. The angular acceleration traces were obtained by differentiating the displacement data twice. To avoid excessive noise in these traces, the displacement data was first low-pass digitally filtered at an upper frequency cutoff of 10 Hz. Movements which cycle at frequencies above this value do not appear to be under volitional control (Brooks, Cooke, & Thomas, 1973; Teulings & Maarse, 1984). The shift in phase produced by the filtering procedure was eliminated by filtering the data forwards and then backwards in time (Pezzack, Norman, & Winter, 1977). The stimulus cursor cycled at three different frequencies. These were .45 Hz in the slow (SLW) condition (corresponding to an average angular velocity about the elbow joint of 40° sec" ); 1.1 Hz in the control (CNT) condition (100° 1  sec" ); and in the fast as possible (FAP) condition, 3.6 to 6.3 Hz (314° to 550° 1  sec-1) depending upon the individual (the frequency of the stimulus cursor corresponded to each individual's maximum speed of movement). The frequency and amplitude (80 mm -> 45° of elbow movement) of the stimulus cursor remained constant within each block of trials. The complexity of the movements were varied by having subjects complete from 1 to 4 cycles. Thus, there were a total of 12 conditions: 3 different stimulus frequency conditions with 4 levels of complexity within each. Two dependent variables were used to assess any changes which occurred as a result of the manipulations to the speed and complexity of the movements. The first was RT. It was measured as the time from the imperative stimulus to the start of angular displacement about the elbow joint. The second involved the angular acceleration traces. Within these traces the number of zero line crossings per movement segment was determined. In addition, the number 92  and timing of deviations within the angular acceleration traces was also examined. A significant deviation was said to have occurred at that point after which at least 5 consecutive data points decreased (or increased) in the direction opposite to that normally found within the movement segment. Design: The experiment was comprised of two sessions, each lasting about 45 minutes. In the first session, all of the subjects completed the CNT condition as well as 10 trials in which they moved as fast as possible for 3 seconds. The only stipulation during these maximum velocity movements was that the amplitude of the movement not deviate substantially from that required in the other conditions (45°). For the second session the subjects were split equally (in terms of their RT from the CNT condition) into two groups. Half of the subjects were assigned to the FAP condition and half to the SLW condition. The 10 trials done as fast as possible in the CNT condition were used to calculate the individual speeds of movement for the subjects assigned to the FAP condition. Both sessions began with practice at tracking the stimulus cursor (10 trials X 20 cycles). Next, the RT/reproduction task was completed in two parts. First, 5 practice trials were completed, followed by 10 performance trials for each level of complexity. The 4 different levels of complexity within each speed condition were blocked and counterbalanced to control for any order effects. Each trial consisted of the following sequence: 1. a variable warning tone (500-1500 msec) to signal the start of the trial; 2. the presentation of the stimulus (starting at the center of the screen and moving to the right); 3. a 500 msec delay; and 4. a second variable warning tone (500-2500 msec), the termination of which acted as the imperative stimulus. Catch trials, in which the second warning tone lasted 5 sec. in duration, occurred 20% of the time during both the practice and performance trials. 93  After each trial the reaction time and integrated spatial error (RMSE) was given to the subjects as feedback. Any trials with an error score of greater than 40 mm or an RT less than 100 msec (indicating anticipation) or greater than 500 msec (indicating a lack of attention) were discarded from further analysis. Subjects were able to attain these criteria on 98% of the trials.  94  RESULTS AND DISCUSSION Reaction Time: The mean and standard deviation for RT was calculated for the 4 levels of complexity within each of the 3 speed conditions (see Figure 20). Although the effect of variations in movement complexity appeared to be similar at each speed of response, this was not confirmed statistically. Analysis within each condition demonstrated that no significant increases in RT occurred as a result of increases in movement complexity in either the CNT condition (F[3,57]=2.12, p=.1082) or the SLW condition (F[3,27]=2.17, p=.1149). However, in the FAP condition there was a significant increase in RT (F[3,27]=4.78, p=.0085). Preplanned comparisons revealed that this difference was due to the 1 cycle condition being initiated significantly quicker than the 2, 3, and 4 cycle conditions (F[1,9]=14.32, p=.0043). Low between subject variability appeared to be the reason for the significant difference in the FAP condition. That is, everyone demonstrated the same basic trend in their data in this condition. In the other two conditions, however, this variability was much higher; with the result being no significant differences between the levels of complexity. According to the RT data, then, the SLW and CNT conditions appear to allow for on-line preparation, while in the FAP condition subjects appear to preprogram the execution of at least 2 cycles of movement. The differences between each stimulus frequency condition (CNT vs SLW - F[1,9]=57.2, p<.0001; CNT vs FAP - F[1,9]=20.85, p=.0014) cannot be confidently interpreted because RT was not fractionated into its premotor and motor time components (Botwinick & Thompson, 1966). Norman and Komi (1979) have shown that different velocities about the same joint can lead to differences in motor time. This, in turn, can make it difficult to interpret any RT differences between conditions which vary in velocity (Falkenberg & Newell, 95  FIGURE 20 RT AS A FUNCTION OF NUMBER OF CYCLES 295  1951  1  1  1  2  1  h  3  4  # OF CYCLES  1979; Sidaway, 1988). In addition, as Newell, Carlton, Carlton, and Halbert (1980) have noted, it is a biomechanically difficult task to initiate a slow movement quickly. One strategy that subjects may use to overcome this difficulty is to slow down the entire process by which they prepare slow movements. This may be one reason why RT became greater as the speed of movement decreased. The plateaus in RT which occurred after 2 cycles in the FAP condition suggests that the limits of preprogramming had been reached in this situation. Such an outcome leads to the conclusion that the unit of complexity in this movement was most likely smaller than a complete cycle. In order to test this hypothesis, 6 more subjects completed the RT/reproduction task for the FAP condition with movements of .25, .5, 1.0 and 1.5 cycles. In effect, these conditions tested whether a reversal in direction was the basic unit of complexity. In addition, the 1 cycle condition provided a replication of this condition from the main experiment. As can be seen in Figure 21, RT increased in a linear fashion (F[1,5]=49.7, p=.0009 - no higher order trends were significant) as the number of cycles increased from .25 to 1.5. This difference was confirmed with a one way repeated measures ANOVA (F[3,15]=7.69, p=.0024). Preplanned comparisons revealed that .25 cycles required less time to initiate than .5, 1, and 1.5 cycles (F[1,5]=35.64, p=.0019); similarly, .5 cycles was initiated more quickly than 1 and 1.5 cycles (F[1,5]=382.69, p<.0001). In addition, the data closely followed the trend from the FAP condition in the main investigation. Specifically.the 1 cycle condition had a similar value in both experiments and the 1.5 cycle condition fell in line with the increase in RT from 1 to 2 cycles. From these results it is suggested that the unit of complexity in this task was composed of a reversal in direction. While not explicitly examining the role of 97  FIGURE 21 RT AS A FUNCTION OF NUMBER OF CYCLES  movement reversals in response planning, previous research has demonstrated that the above conclusion is a viable one. Specifically, reversals in direction have been components of several movement tasks which cause increases in RT with increased complexity [e.g. tapping (Canic & Franks, 1989; Garcia-Colera & Semjen, 1988); writing (Teulings, Mullins, & Stelmach, 1986); and hand movements to a series of targets (Anson, 1982; Henry & Rogers, 1960)]. Angular Acceleration Traces: The number of zero line crossings per movement segment was calculated and comparisons were made across conditions (see Table 1). Separate ANOVA's were completed at each level of complexity within each group of subjects (i.e. those that did the CNT-SLW conditions and those that did the CNTFAP conditions). Subjects in the CNT-SLW group had significantly more zero line crossings within each movement segment in the SLW condition compared to the CNT condition. The average number of zero crossings per movement segment in the SLW condition was approximately 9.1 in comparison to 1.1 in the CNT condition. The differences between the CNT and FAP condition were not nearly as large (1.1 for the CNT vs. 1.0 for the FAP) yet they were still significant due mainly to the fact that there was no variability in the FAP condition. One interesting aspect of this data was that there were consistently more zero line crossings during extension than during flexion in the CNT and SLW conditions [F(1,3)=209.08, p=.0047]. It was thought that this may be due to the subjects simply moving more slowly during extension; and, as a result, increasing the probability of crossing the zero line. However, when the average angular velocities of the extension and flexion movements were compared in the CNT and SLW conditions, no statistically significant differences were found [F(1,3)=2.22, p=.233]. An alternative explanation was that significant deviations in  99  TABLE 1: MEAN NUMBER OF ACCELERATION ZERO LINE CROSSINGS AT EACH LEVEL OF COMPLEXITY IN EACH MOVEMENT SPEED CONDITION CNT-SLW GROUP NUMBER OF CYCLES 1  2  3  4  CNT  1.05  1.09  1.15  1.17  SLW  7.09  9.29  9.90  10.19  CNT-FAP GROUP NUMBER OF CYCLES 1  2  3  4  CNT  1.04  1.11  1.23  1.14  FAP  1.00  1.00  1.00  1.00  100  the angular acceleration trace occurred near the zero line more often during extension. This would make it more likely that the zero line would be crossed in these situations. This hypothesis was tested during the analysis of deviations reported below. The results from the analysis of the zero line crossings in the angular acceleration traces suggest that movements in the CNT and FAP conditions were preprogrammed (Brooks, Cooke, & Thomas, 1973). Although there were significantly more zero line crossings in the CNT condition, this difference appeared to be due more to the lack of variance in the FAP condition. Indeed, in the CNT condition the majority of the movement segments contained only one zero line crossing. It was those segments which had more than one crossing which caused the increase in the mean and variance and the subsequent difference between the conditions. In the SLW condition the number of zero line crossings was increased substantially. Thus, in this condition it appeared that the movements were prepared on-line. The results from the zero line crossings analysis do not completely correspond to those found from the RT data. From the former it was concluded that movements in the FAP and CNT conditions were preprogrammed and those in the SLW condition prepared on-line; while in the latter only movements in the FAP condition were preprogrammed. This discrepancy may stem from the fact that the zero line crossings measure is relatively discrete in nature. In using such a measure it is possible to conclude that two movements were performed similarly, when, in fact, they may have been quite different. This appeared to be the case here. That is, the CNT and FAP conditions were similar at a quantitative level (i.e. they contained basically the same number of zero line crossings), but differed qualitatively in that the traces themselves were more variable in the CNT condition (see Figures 22 and 23). These qualitative differences were assessed 101  FIGURE 22 - KINEMATICS OF CONTROL CONDITION  FIGURE 23 - KINEMATICS OF FAST CONDITION ANGULAR  /K  DISPLACEMENT  I.  UELOCITY —A,—  \*  4  V  ACCELERATION .  o  statistically by measuring the timing and number of significant deviations from a smooth curve within each movement segment. No significant deviations were found in the FAP condition, reconfirming the finding from the previous measures that movements in this condition were entirely preprogrammed. However, in both the CNT and SLW conditions there were significant deviations from a smooth curve within each movement segment. In the SLW condition these deviations were spread throughout each movement segment, as can be observed in the frequency distribution of the relative temporal location of the deviations (see Figure 24). In the CNT condition, however, the majority of the deviations appeared to be grouped at different points in the movement depending upon whether the arm was being extended or flexed. During the extension segments, the deviations appeared at 3 distinct locations: 44% (194 msec), 64% (282 msec), and 90% (396 msec) into the movement. On the other hand, during the flexion segments, there were 4 distinct locations: 8% (35 msec), 30% (132 msec), 63% (277 msec), and 90% (396 msec) into the movement (see Figure 24). In both conditions the time between each deviation was calculated and subsequently displayed in a frequency distribution. As can be seen in Figure 25, the largest majority of the deviations were separated by a duration of approximately 85 msec. This value was similar in both conditions and during both extension and flexion segments. It also falls in line with that found in previous research concerned with the time course of adjustments made during movement sequences (Brooks, Cooke, & Thomas, 1973; Carlton, 1981; Higgins & Angel, 1970). As mentioned above, one of the interesting aspects of the zero line crossing data was that more crossings were found in the extension segments than in the flexion segments. It was suggested that this may be due to a greater 104  FIGURE 24 - RELATIVE TEMPORAL LOCATIONS OF DEVIATIONS  20  30 40 60 60 70 60 TIME (% O F SEGMENT)  20  30 40 50 60 70 BO 80 TIME (% O F SEGMENT)  100  CONTROL - EXTENSION  CONTROL - FLEXION  SLOW - EXTENSION  SLOW - FLEXION  W  20  30 40 60 60 70 60 TIME (% O F SEGMENT)  00  100  20  30 40 60 60 70 60 TIME (% OF SEGMENT)  00  o  100  FIGURE 25 TIME BETWEEN EACH DEVIATION  SLOW CONDITION 200  1600 i  F R E Q U E N C Y 0  SO  100  150  200  250  TIME (MSEC)  300  350  400  CONTROL CONDITION  150 O  100  50  0  50  100  150  200  250  TIME (MSEC)  300  350  400  number of deviations occurring near the zero line during extensions. This hypothesis was tested by calculating the number of deviations which subsequently crossed the zero line during either flexion or extension. It was found that significantly more such deviations occurred during extension [F(1,4)=99.54, p=.0006]. This fact is also reflected in the frequency distributions of the deviations within the extension and flexion segments (see Figure 24). Specifically, a larger proportion of deviations occurred 50% into the extension segments (which is approximately where the zero line was crossed) than in the flexion segments. Carlton (1981) has suggested that deviations within the acceleration trace, when they occur in single aiming movements, reflect the process of on-line adjustments as the hand approaches the target. Thus, in the CNT and SLW conditions corrective adjustments in the movement were prepared on-line to meet the goals of the task. In the FAP condition no such adjustments were observed. These results are similar to those found in the RT data and support the notion of on-line preparation  in the CNT and SLW conditions and  preprogramming in the FAP condition.  107  GENERAL DISCUSSION The aim of the present investigation was twofold: First, to uncover differences between preprogrammed movements and those prepared on-line; and, second, to attempt to pinpoint the location in time at which the on-line preparation process occurred within the movement sequences. In previous research the first goal has been attained using several different dependent variables, including RT (Franks & van Donkelaar, in press; Garcia-Colera & Semjen, 1987), acceleration traces (Brooks, Cooke, & Thomas, 1973; Carlton, 1981), EMG recordings (Desmedt & Godaux, 1979; Hallet, Shahani, & Young, 1975) and interval timing data (Ostry, 1983; Rosenbaum, Hindorff, & Munro, 1986). In this study both RT and angular acceleration traces were used to meet this goal. Differences were found in each of these dependent variables suggesting that either preprogramming or on-line preparation had occurred. It appeared that when the movement sequence was done as quickly as possible preprogramming took place. However, as the speed of response was decreased, on-line preparation became the more dominant form of control. Thus, as the movement was slowed down (CNT and SLW conditions) the subjects were able to prepare upcoming responses as they executed previous ones. As a result, they were not required to prepare the entire response beforehand; rather, they only had to prepare the initial elements. Thus, the processing demands prior to movement remained constant across levels of complexity; and, as a consequence, RT did not increase as the number of movement cycles increased. Evidence for on-line preparation was also found in the angular acceleration traces. Specifically, these traces contained more than one zero line crossing and also more "significant deviations" within each movement segment; although the zero line crossing data did not clearly reflect 108  the presence of on-line preparation in the CNT condition. On-line preparation was not possible in the FAP condition because there was not enough time within each cycle to make the appropriate adjustments. In this context, then, the subjects were required to prepare the entire response prior to its execution. This resulted in RT increasing as the movement became more complex.  Again,  convergent  evidence  suggesting  the  presence  of  preprogramming was gained from the angular acceleration traces in that they contained only one zero line crossing and no significant deviations from a smooth curve within each flexion or extension segment. The relationship between speed of movement and type of preparation has been suggested only recently in the RT/movement complexity paradigm (Canic, 1988; Garcia-Colera & Semjen, 1987; 1988). This is so because traditionally the speed of movement in this paradigm has been maximal. The present experiment extends this evidence by providing convergent data from the angular accelerations produced during the fast and slow movements. Through the analysis of the angular acceleration traces it also became possible to achieve the second goal of the experiment: namely, to pinpoint where in the movement adjustments were made on-line. RT cannot be expected to reveal any evidence regarding where in time on-line preparation occurred because it measures only those processes occurring prior to movement. However, several researchers have looked to the interval timing data (i.e. the length of time from one element within a sequence to the next) for evidence of on-line preparation. Since, by definition, movement planning takes time, it is assumed that differential durations within the interval timing data may reflect delays associated with programming activity as the movement is being executed. The evidence put forward by Ostry (1983) appears to be the most successful in this regard. He found that the intermediate interresponse intervals (IRI's) 109  produced in a sequence of typing keystrokes were consistently longer in duration than the beginning and terminating intervals (similar evidence was also found by Sternberg, Monsell, Knoll, & Wright, 1978). He suggested that this increased amount of time was used by the subjects to prepare on-line the terminal elements of the response. According to Ostry, this on-line preparation is required because we have a limited "motor output span". Similar inferences have been made about the IRI's produced in studies of serial pattern learning (Franks, Wilberg, & Fishburne, 1985; Povel & Collard, 1982; Rosenbaum, Kenny, & Derr, 1983). In these studies the IRI's between subunits of a pattern are greater in duration than those within each subunit. Again, these differential durations appear to reflect the extra time required to prepare the upcoming series of movements. Although the interval timing data described above can tell us where in time on-line preparation occurred, it cannot tell us how this process occurred. The question still remains: What was the subject doing differently in terms of movement outcomes during this increased length of time? By its nature, the interval timing data cannot directly answer this question (other than to say that one aspect of the movement took longer than another). Instead, some measure of the movement itself must be used. This is where the kinematics of the movement can play an important role. As mentioned in the introduction, analysis of the acceleration traces can reveal both the presence and location of on-line prepared adjustments. Brooks, Cooke, and Thomas (1973) suggested that the number of zero line crossings within the angular acceleration trace provides an accurate index of the planning process used in that movement. Others have suggested that deviations within the normally smooth acceleration curve can also be used in this regard (Carlton, 1981; Crossman & Goodeve, 1963/1983; Young, Allard, & Marteniuk, 1988).  no  Surprisingly, in the present experiment these two lines of evidence (i.e the zero line crossings and the deviations) did not completely agree. Specifically, in some situations (most notably the CNT condition) significant deviations occurred concurrently with only one zero line crossing; despite the fact that the former reflects on-line preparation and the latter, preprogramming. This discrepancy in the present investigation may be due to the different conditions used in the originally cited articles (Brooks, Cooke, & Thomas, 1973; Carlton, 1981). Specifically, Brooks and his colleagues used an arm manipulandum similar to the one used in the present study and had their subjects move through 50 degrees in a time ranging from 900  msec (continuous movement) to 1300  msec  (discontinuous movement). The angular acceleration traces produced in the discontinuous movements were very similar to those from the SLW condition in the present experiment (i.e. several large deviations from a smooth curve, and, as a result, more than one zero line crossing per movement segment). The continuous movements, on the other hand, had angular acceleration traces which resembled those produced in the CNT condition (i.e. containing significant deviations, but generally only one zero line crossing in each movement segment). In contrast, Carlton (1981) used a Fitts task in which subjects were required to move a hand held stylus to a target placed 32 cm away as quickly and accurately as possible (350-400 msec). Again, the acceleration traces produced in this task appeared most similar to those found in the CNT condition in the present experiment. The only difference between these three sets of data (Brooks, Cooke, & Thomas's continuous movement, Carlton's movement, and those in the CNT condition) was the extent of the deviations. In Brooks' the deviations were relatively large and spread throughout the movement segment; in contrast, the  in  deviations in Carlton's data were quite small and generally occurred as the target was being approached, with the rest of the trace being smooth. The deviations in the CNT condition from the present experiment were somewhere between these two extremes. It may be, then, that the smoothness of the angular acceleration trace lies on a continuum, such that, as the movement becomes slower the extent of the deviations increase. Eventually, these deviations begin to cross the zero line, as happened in Brooks, Cooke, and Thomas' discontinuous movement and the SLW condition from the present experiment. Where these deviations occurred within the movement was also of interest. Carlton (1981) found that the deviations started to appear approximately 290-300 msec into the movement he used. Specifically, as the hand approached the target. However, since he also varied the amount of time for which vision was available, this duration overestimated the time required for the visual feedback loop. As aresult, Carlton suggested that the length of time from when the lights came on to when the adjustments in the acceleration trace were displayed more accurately reflected the latencies involved in visual feedback. This duration was calculated to be approximately 135 msec, and, thus, was considerably less than the traditional estimate (e.g. Keele & Posner, 1968). In the present investigation the distribution of the deviations was different depending upon the condition. In the CNT condition the largest majority of the deviations occurred as the point of reversal was approached. However, in the SLW condition the deviations were spread throughout the movement segment. This may be the result of these two conditions being at different points on the acceleration continuum described above. At one end of the continuum the deviations, if they occurred, were concentrated near the endpoints of the movement. As the movement was slowed down, the deviations began to spread out more into the movement. 112  From this analysis it can be concluded that, in the CNT condition, the flexion or extension movements were partially planned and executed. This preparation fulfilled part of the goal of the task, but on-line adjustments (both temporal and spatial) were required to accurately approach and execute the reversal, and begin the movement in the opposite direction. Such a process appears to account for the distinct groups of deviations which occurred in the latter portions of the flexion and extension segments. However, those which occurred earlier in each segment (i.e. 8% and 30% into flexion), are more difficult to explain. It seems unlikely that adjustments should be required at such an early stage in the movement if their function is related to the achievement of a temporally accurate reversal in direction at the end of the movement. However, this is exactly what occurred in the SLW condition. Specifically, adjustments were made more or less continuously in order to ensure that the proper temporal criterion was met. Indeed, subjects appeared to have difficulty moving slowly enough in the SLW condition. This difficulty was less noticeable in the CNT condition and nonexistent in the FAP condition. Such a state of affairs appears to be analogous to that described by Klapp and Wyatt (1976) in their well-known "dit-dah" experiment. Specifically, the RT required to prepare and initiate the "dit" keystroke of the Morse code was less than that required for the "dah" keystroke. Klapp and Wyatt suggested that the increase in RT for the "dah" response was due to the extra processing required to specify the length of time for which the key was to be held down. In the present experiment, this extra processing occurred on-line and was reflected in the number of adjustments in the angular acceleration trace. Specifically, in the SLW condition more adjustments were required to accurately meet the temporal demands of the movement than in either the CNT or FAP conditions. Although evidence for preprogramming or on-line preparation has been 113  found previously in studies which have used either acceleration traces or RT alone, these two dependent measures have rarely been combined in a single study. Since both appear to be sensitive to the same phenomenon, they should converge theoretically in such a study. This was, indeed, what occurred in the present experiment. Both the RT and angular acceleration data confirmed that preprogramming occurred in the FAP condition and on-line preparation in the CNT and SLW conditions. In addition, through the analysis of the deviations within the angular acceleration traces, a more accurate description of where in time the on-line adjustments took place was possible. Previous research in the RT/movement complexity paradigm has had difficulty in this regard because of the measures which have been used. In the future, other dependent variables which have demonstrated a sensitivity to the differences between preprogramming and online preparation should be included in an RT study to further the understanding of these processes.  114  

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.831.1-0077255/manifest

Comment

Related Items