Hand Path Priming in Manual Obstacle Avoidance: Evidence for Abstract
Spatiotemporal Forms in Human Motor Control
Robrecht P.R.D. van der Wel
Pennsylvania State University
Robin M. Fleckenstein
Duke UniversitySteven A. Jax
Moss Rehabilitation Research Institute
David A. Rosenbaum
Pennsylvania State UniversityPrevious research suggests that motor equivalence is achieved through reliance on effector-independent spatiotemporal forms. Here the authors report a series of experiments investigating the role of such forms in the production of movement sequences. Participants were asked to complete series of arm movements in time with a metronome and, on some trials, with an obstacle between 1 or more of the target pairs. In moves following an obstacle, participants only gradually reduced the peak heights of their manual jumping movements. This hand path priming effect, scaled with obstacle height, was preserved when participants cleared the obstacle with 1 hand and continued with the other, and it was modulated by future task demands. The results are consistent with the hypothesis that the control of movement sequences relies on abstract spatiotemporal forms. The data also support the view that motor programming is largely achieved by changing just those features that distinguish the next movement to be made from the movement that was just made.
Keywords:sequence production, motor equivalence, movement planning, obstacle avoidance, reaching
A key phenomenon of human perception and performance is
motor equivalence, the capacity to achieve the same output through different means (Lashley, 1930). One of the best known examples of motor equivalence comes from handwriting, where an individual’s graphic output is recognized to be his or hers regard-less of the means by which it is generated—whether scrawled across a blackboard or inscribed on a check, whether produced with the preferred or nonpreferred hand, with the feet, or even with the pen clenched between the teeth. Researchers concerned with motor equivalence have shown that written script is indeed ap-proximately invariant with the effector used, the size of the gen-erated script, and the orientation of the writing surface (Castiello & Stelmach, 1992; Lashley, 1942; Merton, 1972; Raibert, 1977; Swinnen, 1991; Wright, 1993).
How is motor equivalence achieved? One attempt at answering this question came from Meulenbroek, Rosenbaum, Thomassen,
Loukopoulos, and Vaughan (1996), who extended a general model of motion planning (Rosenbaum, Engelbrecht, Bushe, & Louko-poulos, 1993; Rosenbaum, LoukoLouko-poulos, Meulenbroek, Vaughan, & Engelbrecht, 1995; Rosenbaum, Meulenbroek, Vaughan, & Jansen, 2001) to writing and drawing. Following Berkenblit and Feldman (1988) and Keele, Cohen, and Ivry (1990), Meulenbroek et al. suggested that when people write or draw, they access abstract spatiotemporal forms. According to Meulenbroek et al., writing and drawing are achieved by generating a series of move-ments, each of which goes through a via point and then to a goal point. The via points are points of maximal speed, whereas the goal points are points of minimal speed. Using computer simula-tions rendered as stick-figure animasimula-tions, Meulenbroek et al. showed that it was possible, with these assumptions, to generate the same written output with different effectors on different planes and with different sizes. Their model therefore instantiated a possible solution to the problem of motor equivalence.
On what basis did Meulenbroek et al. (1996) argue that abstract spatiotemporal movement forms are used to guide writing and drawing, aside from the fact that their simulations worked reason-ably well? These authors noted that people can learn the order in which spatial targets are presented (Kagan, 1971; Keele et al., 1990), and they appealed to the fact that psychophysical experi-ments have shown that people can mentally project images onto different distal surfaces, even adjusting the size of the projected images if they wish (Kosslyn, 1980). These arguments supported the hypothesis that actors have access to effector-independent spatiotemporal forms. However, the arguments do not necessarily prove that those forms play a role in online movement production. Evidence for the latter proposition has recently come from Jax and Rosenbaum (in press). In their experiments, participants sat at a table and held a vertically oriented dowel that stood on a wide Robrecht P.R.D. van der Wel and David A. Rosenbaum, Department of
Psychology, Pennsylvania State University; Robin M. Fleckenstein, De-partment of Physical Therapy, Duke University; Steven A. Jax, Moss Rehabilitation Research Institute, Philadelphia, Pennsylvania.
This work was supported by National Science Foundation Grant SBR-94-96290, National Institute of Mental Health Grants KO2-MH0097701A1 and R15 NS41887-01, and grants from the Social Science Research Insti-tute and the Office of Research and Graduate Studies, College of Liberal Arts, Pennsylvania State University. We thank Peter Strick for suggesting the second experiment. We also thank Meesha Ahuja, Joshua Albert, Mike Iskoe, Christine Schiller, Allie Schubert, Mycheel Stubbs, Dana Voelker, and Matthew Walsh for help with data collection.
Correspondence concerning this article should be addressed to Robrecht P.R.D. van der Wel or David A. Rosenbaum, Department of Psychology, Pennsylvania State University, University Park, PA 16802. E-mail: [email protected] or [email protected]
circular disk with felt on its bottom, allowing the manipulandum to slide smoothly from one position to another on the table. An OPTOTRAK motion tracking device registered participants’ movements. Participants saw a computer-generated stick-figure image of their right arm on a TV screen. This stick-figure image moved as a participant’s right arm did, with no noticeable delay after the participant moved. Besides seeing an image of his or her own right arm, the participant also saw targets for movement and, in some conditions, obstacles. The targets were displayed in the context of a center-out movement task. A circle appeared in the middle of the screen, and the participant brought his or her hand marker into the circle, whereupon a circle appeared at some point along the rim of an imaginary circle around the center circle. The participant’s task was to move the hand marker to the target and then to return it to the center circle as quickly as possible.
In one control condition of Jax and Rosenbaum’s (in press) experiments, no obstacle ever appeared between the center circle and target circle. In another control condition, an obstacle always appeared between the center circle and a target circle. In the latter condition, the participant was expected to make circuitous move-ments around the obstacle. Of greatest interest were the experi-mental conditions. Here, an obstacle sometimes appeared between the center circle and target, but the obstacle’s appearance was unpredictable. If an obstacle appeared, it always appeared at the moment the target came on and always stood midway between the center circle and the target (as was the case in the control condi-tion, in which an obstacle always appeared).
The question of primary interest was what would happen on trials in which an obstacle was possible but did not appear. The answer, as Jax and Rosenbaum (in press) discovered, was that on those trials, participants made movements whose curvature ex-ceeded the curvature of movements made when obstacles never appeared. Jax and Rosenbaum called this phenomenon thehand path primingeffect.
Of special importance to the claim that there are abstract spa-tiotemporal forms for movement, the hand path priming effect generalized over the workspace. It was not necessary to repeat the same target on successive trials to get the effect. Instead, greater than normal curvature was observed for movements to obstacle-free targets that were removed from the last target tested. On the basis of this observation, Jax and Rosenbaum (in press) argued that their participants relied on abstract spatiotemporal forms (i.e., hand paths that were not tied to particular spatial positions in the workspace nor to specific muscles). Jax and Rosenbaum also suggested that an advantage of relying on these abstract spatio-temporal forms is that they help eliminate the need for planning of movements from scratch. They proposed that the spatiotemporal form of one movement could be applied to, or retained for, the plan of the next movement to come.
A question that can be raised about Jax and Rosenbaum’s (in press) study concerned the generalizability of its conclusions. Jax and Rosenbaum obtained the hand path priming effect when target positions and obstacle positions were uncertain. (Recall that these appeared suddenly on a computer screen, with the participant not knowing from trial to trial where a target would appear or whether it would be accompanied by an obstacle.) Thus, it is possible that the effect depended critically on such uncertainty. If that were the case, one might not expect the effect when target positions and obstacle positions are known in advance. If the hand path priming effect failed to materialize when target positions and obstacle
positions were certain, that outcome would curtail the generaliz-ability of Jax and Rosenbaum’s conclusions.
To address this concern, in the present experiments we used a procedure in which participants had full knowledge of targets and obstacles before interacting with them. We asked participants to hold a dowel, using a power grip, and to tap the base of the dowel on each of a series of targets in time with a metronome (see Figure 1). The targets were all fully visible before each trial began and remained fully visible while the trial was underway. The targets were arranged in an arc on a table. In the control conditions, there was no obstacle between any targets, but in the experimental trials, an obstacle (a vertical piece of cardboard) stood between a pair of targets, again in full view of the participant before and during the trial. When an obstacle was present, participants were asked to carry the dowel over the obstacle in time with the metronome, tapping the targets on either side of it in time with the metronome. The question was what would happen to the jumps from target to target after, and also before, the obstacle was cleared.
If the hand path priming effect generalizes to situations in which there is no uncertainty about targets and obstacles, one would expect to see the effect in the present experiment. In particular, jump heights between targets after obstacles are cleared should be higher than jump heights between those same targets when no obstacles are cleared. The latter outcome would be expected if hand path priming involves abstract spatiotemporal forms that are carried over in successive movements. The abstractness of the forms would be supported by the fact that the successively en-countered targets occupied different locations in the workspace and were reached with different limb positions and different mus-cle groups.
Regarding what would happen to the jumps between targets
beforeobstacles were cleared, if participants anticipated
ing jumps, one would expect to see the jump heights change as the obstacles were approached. Finding such an anticipatory effect, which was not possible in the method used by Jax and Rosenbaum (in press), would add to the list of anticipatory phenomena in perceptual–motor skills.
Experiment 1
Method
Participants. Thirty-six Pennsylvania State University stu-dents (12 male, 24 female) from an introductory psychology class participated for class credit. They ranged in age from 18 to 22 years. The Penn State Institutional Review Board approved this and all of the other experiments reported here. The rights of all the participants were protected.
Apparatus, procedure, and design. Participants sat at a table (122 cm wide, 61 cm deep, 78 cm high) with six target positions evenly spaced around a semicircle whose radius was 41 cm (see Figure 1). The targets were red foam dots (7 cm in diameter and 0.2 cm thick) that lay flat on the table 28.5 cm apart (center of one target to center of the adjacent target). Participants held a wooden dowel (20.2 cm high and 3 cm in diameter, weighing 99 g) with the right hand, using a power grip and keeping the little finger as close as possible to the base of the dowel. Participants transported the dowel from target to target using a “jumping” movement such that they lifted the dowel off the target and made an arcing movement that led to the dowel’s impact with the next target.
In the experimental conditions, an obstacle was placed between any given pair of targets, equidistant between them. The height of the obstacle was varied between participants (12 participants per obstacle height). The short obstacle was 7.5 cm high, the medium obstacle was 15.0 cm high, and the tall obstacle was 22.5 cm high. The three obstacles were made of sturdy pieces of cardboard attached to the vertical edge of a metal bookend that was secured to the table in the position being tested.
The experimenter asked participants to carry out the movements in time with a metronome, which clicked every 0.60 s (1.67 Hz). The instructions emphasized timing, so the main concern, as expressed to the participants, was that they tap the base of the dowel on the series of targets in time with the metronome. The experimenter also asked the participants to move over the obstacle rather than around it. To accommodate differences in participants’ arm lengths, the experimenter asked each participant to adjust his or her sitting position so his or her right arm was fully extended in the forward direction while holding the dowel between the two central targets.
To record the dowel position, we attached three infrared-emitting diodes (IREDs) around the top portion of the dowel so the dowel would always be in view of the OPTOTRAK 3020 motion tracking system (Northern Digital, Inc., Waterloo, Ontario, Can-ada) used to track participants’ movements. We also attached a fourth IRED to the top of the obstacle to record whether partici-pants collided with the obstacle at any time during a trial, in which case the trial was rerun. The experimenter taped the wires of the IREDs to the participant’s right arm with athletic tape. The wires were affixed to the participant’s arm in a way that allowed the participant to move the arm freely within the confines of the workspace. The OPTOTRAK sampled the positions of the IREDs at 100 Hz.
Before the start of the experiment, participants practiced moving in time with the metronome but with no obstacle present. Data collection began when the participant reported feeling comfortable with the task and when the experimenter judged the participant’s performance to comply with the instructions. This usually occurred within 1–2 min.
In each trial, the experimenter asked the participant to start either on the leftmost or rightmost target. The starting target was counterbalanced across trial blocks. The participant was invited to start moving when he or she felt that he or she had internalized the beat. If the participant started on the left, he or she moved from target to target in the rightward direction, all the way to the farthest target on the right, and then back to the left, tapping all the targets in between, whereupon he or she returned toward the right again, and so on. If the participant started on the right, the sequence was reversed. Participants performed this back-and-forth sequence five times on each trial without interruption. The experimenter told participants to keep moving until they heard the verbal instruction “Stop.” The experimenter also told the participants not to worry about making a few extra jumps after hearing the “Stop” com-mand. The experimenter counted the back-and-forth movements and issued the “Stop” command after the participant hit the first target in the sixth cycle.
Each participant completed 20 trials, with 10 obstacle-absent control trials and 10 obstacle-present trials starting on the right and on the left, randomized across participants. The experimenter told the participant to take a break whenever he or she wanted, but preferably not in the midst of a trial. After every 10th trial, the experimenter mandated a break.
To evaluate the influence of moving over an obstacle, we subtracted the peak movement heights in the control (obstacle-absent) trials from the peak movement heights in the experimental (obstacle-present) trials. The subtraction was done on a participant-by-participant, target-pair-by-target-pair, obstacle-position-by-obstacle-position, and movement-direction-by-movement-direction basis. Thus, for each participant, we estab-lished a difference score in peak height for the movement from one targeta to the adjoining targetb by subtracting the mean peak height for the jump between targetsaandbin the control condition from the mean peak movement height for the jump between targets
a and bin the obstacle-present condition. The resulting values indicated a difference in peak movement height between the ex-perimental and control conditions, with positive values indicating higher movements, negative values indicating lower movements (which were theoretically possible), and 0 indicating no difference. We defined the peak height value for each intertarget movement as the highest position of the dowel during a movement between a target pair. We excluded trials if the OPTOTRAK failed to record the position of any of the three dowel IREDs, if the participant hit the obstacle, or if the participant failed to hit a target circle.
Results
Peak movement heights. As shown in Figure 2, the peak
movement heights after clearing the obstacle only gradually de-creased back to baseline in the postobstacle movements. In addi-tion, the gradual decrease scaled with obstacle height, such that clearing a higher obstacle led to higher successive movements.
repeated measures analysis of variance (ANOVA). Obstacle height was treated as a between-subjects factor, and obstacle location and movement number were treated as within-subject factors. We applied a log10 transformation to the data prior to the analysis to correct for skew. We also excluded movements over the obstacle so as to avoid spuriously significant main effects or interactions due to those movements, which were by necessity much higher than movements for which no obstacle had to be cleared.
The ANOVA, with Greenhouse–Geisser correction to the de-grees of freedom where appropriate, revealed an Obstacle Height ! Obstacle Location ! Movement Number interaction,
F(18.342, 302.643)"4.720,p# .01. The effect of a preceding obstacle generalized over the workspace, as reflected in the Ob-stacle Location ! Movement Number interaction, F(9.171, 302.643)"60.772,p#.01. The gradual decrease in peak move-ment heights back to baseline scaled with obstacle height, such that clearing a higher obstacle led to higher successive movements. This result was reflected in an Obstacle Height ! Movement Number interaction, F(8.885, 146.603) " 2.255, p # .05. The same qualitative pattern of results was found for every obstacle location.
Regarding preobstacle jump heights, the results gave little hint of a gradual increase before clearing an obstacle, although the results did indicate a constant, overall increase in peak movement height when an obstacle was present as opposed to when no obstacle was present. The magnitude of the increase depended on
the height of the obstacle, such that a higher obstacle led to a larger constant increase in peak movement height.
The statistical tests of this effect took into account the possibil-ity that the overall increase in jump height prior to obstacles was not, in fact, only anticipatory but may have also reflected a gradual decrease in peak movement height after clearing a previous obsta-cle (i.e., a long-term perseveration effect). To test for anticipation without the possible contamination of long-term perseveration, we conducted an analysis of movements made before any obstacle was cleared within a trial, comparing those jump heights to the analogous intertarget jumps in the control condition. We included only the conditions in which the obstacle was between Targets 1 and 2 or between Targets 5 and 6 (see Figure 1), because in those conditions participants made the most movements (four of them) before confronting an obstacle for the first time on a trial. We omitted the very first movement from the analysis to avoid start-up effects.
The results of the 2 (obstacle presence) ! 2 (obstacle loca-tion)!3 (movement number) within-subject ANOVA with ob-stacle height as a between-subjects factor revealed a main effect of obstacle presence,F(1, 33)"5.169,p#.05, such that participants made higher jumps in the obstacle-present conditions than in the obstacle-absent conditions (mean difference"5.817 mm,SE"
2.559). The ANOVA revealed no other significant main effects or interactions (allps$.05). Thus, participants did in fact anticipate the obstacle from the start of the trials, as revealed by their higher jumps before an obstacle was encountered than in the comparable control conditions in which no obstacle would be encountered.
Movement times. To check that participants performed the task in time with the metronome, we calculated the mean times between successive target landings, defining the moment of landing on the target as the time when the velocity first fell below 15 mm/s in the vertical direction and the dowel was in the target area. The mean movement time for each obstacle height turned out to be 0.60 s (all
SEs!0.01), which was the same as the prescribed value. To determine the influence of the independent variables on movement, we conducted a 3 (obstacle height) ! 5 (obstacle location)!10 (movement number) repeated measures ANOVA with obstacle location and movement number as within-subject factors and obstacle height as a between-subjects factor. The ANOVA yielded an Obstacle Location!Movement Number!
Obstacle Height interaction,F(7.706, 127.145)"4.384,p#.01, and an Obstacle Location ! Movement Number interaction,
F(3.853, 7.706)"89.622,p#.01. Whereas movements over the obstacle took longer than the prescribed duration, movements after clearing the obstacle took less time than the prescribed duration. Movements in the obstacle-absent conditions did not depart from the prescribed duration.
A follow-up analysis focused on the correlation between move-ment times and peak movemove-ment heights for each obstacle height separately but with movements over the obstacles removed. The analysis revealed a significant correlation (r" %.160,p#.01) for the medium obstacle height but not for the low or high obstacle heights (ps$.10).
Discussion
The results of Experiment 1 suggest that abstract spatiotemporal forms are used during the production of movement sequences and that such forms carry over from one movement to the next. Figure 2. Peak movement heights above baseline for each obstacle
Evidence for this interpretation comes from the strong hand path priming effect for movements made after clearing an obstacle. Participants made higher movements after clearing an obstacle than when they performed with no obstacle in the workspace. The postobstacle jump heights decreased with successive postobstacle movements and scaled with obstacle height. Because this effect extended over the workspace, it apparently reflected carryover of an abstract spatiotemporal form rather than a muscle-specific form. A possible alternative explanation of the results can be based on the finding that movements over the obstacle took longer than the prescribed time interval, whereas the first and second movement after an obstacle sometimes took shorter than the prescribed time interval. In view of these observations about timing, one might venture the hypothesis that the observed scaling effects of obstacle height resulted from participants trying to minimize speed changes in successive jumps. To do so, they may have made higher jumps in postobstacle trials to lengthen the movement path so they would generate speeds similar to those needed for obstacle clearance.
This explanation cannot be the sole cause of the observed scaling effect, however, because the effect also emerged when participants closely followed the prescribed metronome rate for postobstacle movements. Furthermore, a significant correlation was found between movement times and peak movement heights only for the medium obstacle height, although the same pattern of peak heights emerged in the low- and high-obstacle conditions.
The results of Experiment 1 effectively replicate the earlier findings of Jax and Rosenbaum (in press), indicating that the hand path priming effect that they observed was not just an artifact of uncertainty about targets and obstacles in their experiments. Be-cause the hand path priming effect extended over the workspace in Jax and Rosenbaum’s study (in press) as well as in the present experiment, it seems to be a general phenomenon.
Experiment 2
Having considered one alternative explanation of the results of Experiment 1 in the previous section, we now consider another alternative account. The alternative to be considered goes strongly against the abstract spatiotemporal view. According to this alter-native explanation, the results of Experiment 1 are attributable to biomechanical properties of muscles. This alternative account builds on the fact that the contraction time of muscles is much shorter than the relaxation time of muscles (Enoka, 1988). Thus, it could be that the muscles that contracted to help clear the obstacle could not return to their initial state quickly enough to permit normal, control-level jump heights. This hypothesis would not only explain why participants moved higher after clearing obsta-cles than when no obstaobsta-cles had to be cleared. It would also explain why there was only a tiny jump-height increase before
obstacles were cleared. The latter result would have stemmed from the time needed for muscles to contract, which is much shorter than the time needed for them to relax. Thus, there may have been no need to slowly prepare the muscles for the rapid contraction that would be needed to achieve a forthcoming leap over an obstacle. We designed Experiment 2 to test whether the sequential effects in Experiment 1 could have been caused by slow muscle relax-ation. We asked participants to move over an obstacle with one hand and then continue moving between subsequent targets with the other hand. This manipulation eliminated the possibility that the muscles in the arm just used to clear the obstacle could have
contributed to the sequential effect. If slow relaxation of the muscles of the active arm caused the sequential effect in Experi-ment 1, then the effect would be eliminated when participants continued their movements with the other arm after clearing the obstacle. By contrast, if the sequential effect reflected the use of abstract spatiotemporal forms during sequence production, the effect would be preserved when participants continued their move-ments with the other arm.
Method
Twenty Penn State undergraduates (13 male, 7 female) from an introductory psychology class participated for class credit. They ranged in age from 18 to 23 years. In this experiment, we used only one obstacle (21.5 cm in height). During the hand-switch trials, participants held a dowel (20.2 cm in height and 3.4 cm in diameter, 140 g) in each hand. We asked participants to move over the obstacle with one hand and then continue with the other hand. We only used the most leftward and most rightward obstacle positions from Experiment 1, because those obstacle positions provided the most informative setup for the switch trials. In the switch trials, participants moved over the obstacle with one hand and continued moving with the other hand, starting at the next target. Thus, if the left hand jumped from Target 1 to 2, the next thing that happened was that the right hand jumped from Target 3 to 4. We used this design to prevent possible collisions between the two hands. We told participants that during the task, one of the two dowels had to be in contact with one of the targets at all times. The control trials were identical to the experimental switch trials except that no obstacle was present.
Each participant completed 20 trials, with 8 obstacle-absent trials and 12 obstacle-present trials. The order of these trials was randomized across participants. The rest of the procedure was identical to that of Experiment 1 except that participants completed three back-and-forth movement sequences instead of five. In all other respects, the method was the same as that in Experiment 1.
Results
Peak movement heights. Figure 3 shows the data for each
obstacle location. To test whether the hand path priming effect of Experiment 1 still occurred when participants switched hands after clearing an obstacle, we conducted a 2 (start location)! 2 (ob-stacle location)!6 (movement number) within-subject repeated measures ANOVA. We excluded the movements over the obstacle to avoid spurious main effects or interactions resulting from those movements. We again applied a log10 transformation to the data prior to the analysis to correct for skew. Note that the obstacle location factor incorporated hand switches because participants moved over the obstacle with the left hand and continued with the right hand when the obstacle was on the left, whereas participants moved over the obstacle with the right hand and continued with the left hand when the obstacle was on the right.
signifi-cantly higher jump heights than subsequent movements. The re-sults revealed no other significant main effects or interactions. However, the data revealed a trend for movements to be higher when the obstacle was positioned on the left side of the workspace than when the obstacle was positioned on the right side of the workspace,F(1, 17)"3.724,p".07. Thus, participants tended to move higher after clearing an obstacle with the left hand (on the left side of the workspace), continuing with the right hand, than after clearing an obstacle with the right hand (on the right side of the workspace). The clearance over the obstacle differed between the hands, such that the movement over the obstacle was higher for the left hand than for the right,F(1, 17)"7.555,p#.05.
Movement times. Movements were again timed well with the metronome. The mean movement time for each obstacle position was 0.60 s (allSEs!.06), compared with the prescribed value of 0.60 s.
We conducted a 2 (obstacle presence)!2 (obstacle location)!
8 (movement number) within-subject repeated measures ANOVA to evaluate the influence of the independent variables on move-ment time for the switch trials. The obstacle location factor incor-porated the hand combination for movements in the sequence because participants moved over the obstacle with the left hand and continued with the right hand when the obstacle was on the left, whereas participants moved over the obstacle with the right hand and continued with the left hand when the obstacle was on the right.
The ANOVA yielded an Obstacle Presence!Obstacle Loca-tion!Movement Number interaction,F(7, 98)"3.368,p#.01, and an Obstacle Presence ! Movement Number interaction,
F(4.104, 57.450)"6.023,p#.01. For the obstacle-absent con-ditions, the prescribed 0.60 s movement duration fell within the 95% confidence interval for each movement. For the obstacle-present conditions, movements over the obstacle generally took longer than the prescribed 0.60 s.
Because some of the movement times departed from the pre-scribed time, we again tested the correlation between movement times and peak movement heights, removing the jumps over the obstacle to avoid spurious results. The correlation between move-ment times and peak movemove-ment heights was not significant (r" %.019, p$ .10). Thus, the obtained results in peak movement heights for movements after clearing the obstacle did not depend on fluctuations in movement times.
Discussion
The results of Experiment 2 suggest that the sequential effect observed in Experiment 1 was not caused by residual activity in the muscles of the arm used to clear the obstacle. Instead, the results suggest that the effect originated from a more abstract source of information, as suggested by Lashley (1930) and the other authors concerned with motor equivalence in handwriting mentioned earlier (see also Harrington et al., 2000).
Special note should be made of the fact that participants moved over the obstacle with more clearance when using the left hand than when using the right hand and also, especially, that the postobstacle jumps were then higher with the right hand than with the left hand. The fact that larger obstacle clearances were ob-served for the left hand than for the right can be explained by saying that participants allowed for more variability with the nondominant hand than with the dominant hand, an idea that accords with earlier suggestions by Worringham (1991) and Sabes and Jordan (1997). Interestingly, the greater clearance over the obstacle with the nondominant hand led to higher subsequent jumps with thedominanthand. This surprising outcome suggests a true carryover of the spatial form of the movement from one hand to the next.
Experiment 3
Experiment 3 was designed to provide further information about the role of anticipation in the current task. Recall that the analysis of the preobstacle moves of Experiment 1 showed that there was anticipation of forthcoming obstacles. The relevant finding was that jump heights were higher before the first obstacle was cleared than in the control condition, in which no obstacle was cleared. This result indicated that anticipation affected jump heights. It is impossible to tell from the results of Experiment 1 or Experiment 2 whether the jump heights madeafterobstacles were cleared were only affected by their having been cleared some number of moves ago or whether there was also an effect of forthcoming obstacle-clearance demands.
In Experiment 3, we sought a more direct test of the role of anticipation. We introduced a second obstacle, reasoning that if anticipation plays a role in the generation of movements before an obstacle, jump heights after clearance of one obstacle preceding another would be different from jump heights made after clearance Figure 3. Peak movement heights above baseline when clearing an
of the same obstacle when there was no other obstacle yet to be cleared.
Within this possibility, we considered two further, subordinate hypotheses. According to one, participants would try to minimize changes of jump heights between obstacles, in which case jump heights would behigherafter an obstacle was cleared and another obstacle had to be cleared than after an obstacle was cleared and another obstacle did not have to be cleared. According to the other subordinate hypothesis, participants would try to minimize the energy expended over the series of jumps they made, in which case jump heights would belowerafter an obstacle was cleared and another obstacle had to be cleared than after an obstacle was cleared and another obstacle did not have to be cleared. The third, null, hypothesis was that there would be no effect of a second obstacle on jump heights.
Method
Thirty-two Penn State undergraduates (18 male, 14 female) from an introductory psychology class participated for class credit. They ranged in age from 17 to 25 years. The method of Experi-ment 3 was the same as that in ExperiExperi-ment 1 except where noted below. In the one-obstacle trials, the obstacle (21 cm high) stood between two of the targets. In the two-obstacle trials, a second obstacle (also 21 cm high) stood between another pair of targets. We also included a no-obstacle control condition to obtain a baseline measure.
Each participant completed 32 trials made up of 2 obstacle-absent trials, 10 one-obstacle trials, and 20 two-obstacle trials. We randomized the order of the 32 trials across participants. Partici-pants completed three back-and-forth movement sequences, as in Experiment 2.
Results
Peak movement heights. To compare the peak movement
heights after clearing the obstacle in the one-obstacle condition with the peak movement heights after clearing the obstacle in the two-obstacle condition, we first subtracted the peak movement heights of the baseline trials from each of the experimental con-ditions. Then we subtracted the baseline-relative heights in the one-obstacle condition from the corresponding heights in the two-obstacle condition. We focused on the conditions in which one of the obstacles was between Targets 1 and 2 and the other obstacle was between Targets 5 and 6, because those conditions had the most intermediate steps between the two obstacles. However, we also looked at jumps between obstacles when there were fewer intermediate steps between them.
Figure 4 shows the data for the relevant experimental condi-tions. We conducted a 2 (obstacle number)!2 (movement direc-tion)!3 (movement number) repeated measures ANOVA to test for differences in peak movement height. Movement number had three levels corresponding to the three moves between the obsta-cles. Thus, we compared the peak movement height of the nth movement after clearing a first obstacle in a one-obstacle condition to the peak movement height of thenth movement after clearing a first obstacle in a two-obstacle condition. Again, we applied a log10 transformation to the data prior to the analysis to correct for skew.
The ANOVA revealed an Obstacle Number!Movement Di-rection!Movement Number interaction effect on the peak
move-ment height, F(1.689, 52.366)" 31.455,p # .01; an Obstacle Number ! Movement Number interaction, F(1.443, 44.724) "
59.777, p # .01; an Obstacle Number ! Movement Direction interaction,F(1, 31)"20.562,p#.01; and a Movement Direc-tion!Movement Number interaction,F(1.509, 46.771)"11.202,
p#.01. The peak height of the first movement after clearing an obstacle was similar for the one- and two-obstacle conditions. However, the results indicated significantly higher peak movement heights for the second and third movement after clearing a first obstacle when a second obstacle had to be cleared than when no second obstacle had to be cleared. This tendency was stronger for leftward movements than for rightward movements.
target pair and second-to-rightmost target pair. For the second ANOVA, with Greenhouse–Geisser correction to the degrees of freedom where appropriate, we analyzed the one- and two-obstacle conditions with the obstacles placed between the second-to-leftmost target pair and the rightmost target pair.
When the obstacles were located at the leftmost target pair and at the second-to-rightmost target pair, the ANOVA revealed a significant Obstacle Number ! Movement Number interaction effect on peak movement height,F(1, 31)"6.171,p#.05, such that the first movement after clearing the obstacle was similar in the one-obstacle and two-obstacle conditions, but the second movement after clearing the first obstacle was higher in the two-obstacle condition. The results also revealed a Movement Direc-tion!Movement Number interaction,F(1, 31)"8.707,p#.01, such that the first movement after clearing the obstacle was higher for movements in the rightward direction than for movements in the leftward direction, but the second movement after the obstacle was higher for movements in the leftward direction than for movements in the rightward direction.
When the obstacles were located at the second-to-leftmost target pair and at the rightmost target pair, the ANOVA revealed a significant Obstacle Number!Movement Direction!Movement Number interaction effect on peak movement height,F(1, 31)"
50.313, p # .01; an Obstacle Number ! Movement Number interaction,F(1, 31)"68.821,p#.01; an Obstacle Number!
Movement Direction interaction,F(1, 31)"12.156,p#.01; and a Movement Direction ! Movement Number interaction, F(1, 31)"7.096,p#.05. Qualitatively, the data indicated a sequential effect in both the one-obstacle condition and two-obstacle condi-tions such that the peak movement heights only gradually de-creased back to baseline after the clearing of an obstacle. Impor-tantly, participants tended to keep moving higher during the second movement after clearing an obstacle when a second obsta-cle needed to be obsta-cleared than when no second obstaobsta-cle needed to be cleared. The results also indicated that this tendency to antici-pate the upcoming obstacle was stronger when moving to the left than when moving to the right.
Movement times. As in the previous experiments, participants timed their movements well. The mean movement time was 0.60 s (SE"0.01) for the obstacle-absent condition, 0.60 s (SE"0.01) for the one-obstacle condition, and 0.60 s (SE " 0.01) for the two-obstacle condition. The ideal movement time, as specified by the metronome, was 0.60 s.
We conducted separate ANOVAs to compare the movement times between the same one-obstacle and two-obstacle conditions that we used to evaluate the peak movement heights (described above). In a first ANOVA, we compared movement times for the one- and two-obstacle conditions when the obstacles stood at the leftmost and/or rightmost positions. The 2 (obstacle number)!10 (movement number) repeated measures ANOVA revealed a sig-nificant Obstacle Number!Movement Number interaction effect on movement times,F(2.784, 69.596)"6.611,p#.01. Move-ments over the obstacle always took longer than the prescribed duration in the two-obstacle condition, whereas movements over the obstacle only took longer than the prescribed duration in the one-obstacle condition when the obstacle was positioned at the rightmost location. Additionally, the first two movements took slightly shorter than the prescribed duration after clearing an obstacle at the rightmost location in the two-obstacle condition, and the first movement after clearing an obstacle took slightly
shorter after clearing an obstacle at the leftmost location in the one-obstacle condition. The other movement times did not depart from the prescribed duration.
In a second ANOVA, we compared movement times for the one- and two-obstacle conditions when the obstacles stood at the leftmost and/or at the second-to-rightmost positions. The 2 (obsta-cle number) ! 10 (movement number) repeated measures ANOVA revealed a significant Obstacle Number ! Movement Number interaction effect on movement times,F(2.549, 68.827)"
12.912,p#.01. Movements over the obstacle always took longer than the prescribed duration in the two-obstacle condition, whereas movements over the obstacle only took longer than the prescribed duration in the one-obstacle condition when the obstacle was positioned at the leftmost location. Additionally, the first move-ment after clearing an obstacle took slightly less time than the prescribed duration in the two-obstacle condition, and the first movement after clearing an obstacle took slightly less time after clearing an obstacle at the leftmost location in the one-obstacle condition. The other movement times did not depart from the prescribed duration.
In a third ANOVA, we compared movement times for the one-and two-obstacle conditions when the obstacles stood at the second-to-leftmost and/or at the rightmost positions. The 2 (obsta-cle number) ! 10 (movement number) repeated measures ANOVA revealed a significant Obstacle Number ! Movement Number interaction effect on movement times,F(2.766, 69.162)"
11.275,p#.01. Departures from the prescribed duration did not follow a clear pattern and occurred only for movements between two of the target pairs in the two-obstacle condition and only for movements between three of the target pairs in the one-obstacle condition.
Because some of the movement times departed from the pre-scribed duration, we tested the correlation between movement times and peak movement heights (see Experiment 1 for the rationale behind this test). When the movements over the obstacle were removed from the analysis, the correlation analyses yielded no significant results for any of the obstacle combinations (allps$
.10), suggesting that the obtained results in peak movement heights for movements after clearing the obstacle did not depend on the observed fluctuations in movement times.
Discussion
The results of Experiment 3 are consistent with the hypothesis that participants anticipated forthcoming obstacles. In accord with this hypothesis, jump heights were different following an obstacle when another obstacle had to be cleared than they were when no other obstacle had to be cleared.
General Discussion
In each of the three experiments reported here, we observed a hand path priming effect after clearance of an obstacle. Peak heights for movements after obstacle clearance only gradually decreased with more moves away from the obstacle.
The results of Experiment 1 indicated that the strength of the effect scaled with the height of the obstacle. The results also indicated that the effect extended across the workspace in much the same way as Jax and Rosenbaum (in press) showed. An implication of this finding is that the results of Jax and Rosenbaum were not due to uncertainty about targets and obstacles or other aspects of their experimental task such as the use of a feedback screen that was physically removed from, and at right angles to, the response surface. The hand path priming effect is a robust, general phenomenon.
The results of Experiment 2 allowed us to rule out the hypoth-esis that the need to relax the muscles of the arm used to clear the obstacle caused the hand path priming effect in the present setup. The effect still appeared when participants switched hands after moving over the obstacle. This finding suggested that the source of the sequential effect was not tied to particular muscles of the arm used to clear the obstacle but instead was more abstract. An intriguing secondary observation in Experiment 2 substantiated this claim further. The right hand made higher postobstacle jumps after the left hand cleared the obstacle than did the left hand after the right hand cleared the obstacle. The left hand also jumped higher over the obstacle than did the right hand, which we attrib-uted to compensation for greater variability in left-hand position-ing. The fact that the right hand made higher postobstacle jumps than did the left hand argues for reliance on an abstract spatiotem-poral form for movement, although we cannot rule out the hypoth-esis that some less specific form of muscle relaxation accounted for the results (e.g., gradual relaxation of axial or postural muscles used to enable movements with either hand).
The results of Experiment 3 indicated that the jump heights observed here reflected anticipation as well as perseveration. The relevant outcome was that the presence of a second obstacle affected jump heights. Participants maintained higher postobstacle jump heights when another obstacle was upcoming then when another obstacle was not upcoming. This outcome complements the finding of Experiment 1 that jump heights prior to the first encountered obstacle were higher than jump heights when no obstacle appeared. The fact that postobstacle jump heights were
higherrather thanlowerif another obstacle was imminent suggests that participants sought to minimize changes in jump heights (or some concomitant variable) in successive moves but did not try to minimize energy. Had they tried to minimize energy, they would have reduced their jump heightsmorewhen another obstacle was upcoming than when another obstacle was not upcoming.
The latter suggestion points to a broader conclusion that can be reached on the basis of these findings—namely, that biomechani-cal costs were not minimized here. As in the study of Jax and Rosenbaum (in press), in which participants made more circuitous movements than needed when obstacles were possible but did not appear, participants in the present experiments caused their hands to jump higher than would have been expected if their primary concern was energy minimization. The fact that our participants were willing to continue to make higher than necessary jumps
between obstacles shows that conserving energy was not their main concern.
What participants instead seemed to care about was minimizing changes in the movements they made when jumping over obstacles and carrying out subsequent movements. There may be a compu-tational advantage to this approach. When forthcoming movements differ minimally from just-completed movements, the plans for the forthcoming movements can differ by as little as possible from the plans for the movements that were just completed. This makes for an efficient means of controlling successive movements because, instead of creating new plans for each movement, plans for recent movements can be slightly altered to permit the next movements to be performed. This “editing” view of motor programming also finds empirical support in the fact that repeated movement se-quences are performed more quickly when the movements in the same serial positions have the same features than when they do not (Rosenbaum, Weber, Hazelett, & Hindorff, 1986). The latter ob-servations were made in the context of violin bowing, speaking, and buttonpressing, all performed as quickly as possible, and with the dependent measure being the likelihood of error. The present study provides a complementary measure to error likelihood. Here we have shown that movement forms are adapted to minimize changes relative to recently performed movements. The addition of the form measure, embodied here in the peak jump heights of sequential aiming movements, lends further support to the editing view. Our results suggest that the editable contents of motor programs are spatiotemporal forms. As indicated above, this sug-gestion helps explain how motor equivalence is possible in tasks such as writing and drawing.
On the basis of the present results, we cannot say exactly which aspects of the spatiotemporal forms carry over from one movement to the next. The measure reflecting hand path priming was the peak heights that participants reached on successive movements. Thus, per movement, we only considered the participants’ position at a single time point. Future research is needed to determine which parameter(s) participants carried over from one movement to the next. Possible candidates include the shape of the trajectory as a whole, the point of maximal excursion, the amount of curvature, the underlying velocity profile (or a derivative thereof), or some other parameter.
A final question is why our participants did not do more to minimize changes in successive movements by showing more dramatic anticipation effects. There is little doubt that our partic-ipants were aware that large obstacle-clearing jumps would be needed after a few obstacle-free jumps, for the obstacle could hardly be missed on the table where participants performed. It is well known that latencies of sequence-initiating movements de-pend on the lengths and other features of the forthcoming se-quences (Rosenbaum, 1987; Sternberg, Monsell, Knoll, & Wright, 1978). Furthermore, transcranial magnetic stimulation studies sup-port the hypothesis that aspects of entire forthcoming movement sequence are prepared in advance (e.g., Ashe, Lungu, Basford, & Lu, 2006; Averbeck, Sohn, & Lee, 2005; Kennerley, Sakai, & Rushworth, 2004). If our participants were aware of large forth-coming obstacle jumps, why didn’t they make much larger preob-stacle jumps?
to be made from the last movement made. Prior to jumping over obstacles, our participants had their hands resting on the table and made low, easy movements to bring the hand to the first target from which the experimental sequence would be launched. The first movements were not so different from those earlier, getting-ready movements. There also appeared to be no critical reason to make much larger jumps before the obstacles had to be circum-vented. If there had been, the obstacles could not have been circumvented as well as they were after the fairly low jumps that preceded them. The only time we saw exceptionally high jumps before an obstacle was cleared was after another obstacle was cleared, but these jumps were high to begin with.
Regardless of the exact reasons for the magnitude of the antic-ipatory and perseveratory effects obtained here, our research shows that biomechanical costs are balanced against computa-tional costs in motor control, a notion that fits well with the proposition that both kinematic and dynamic internal models can be used during trajectory planning (Kawato, 1999). Somehow, the brain manages to negotiate the very different kinds of costs asso-ciated with biomechanics on the one hand and computation on the other as it prepares for action. How it does so is an exciting challenge for future research.
References
Ashe, J., Lungu, O. V., Basford, A. T., & Lu, X. (2006). Cortical control of motor sequences.Current Opinion in Neurobiology, 16,213–221. Averbeck, B. B., Sohn, J., & Lee, D. (2005). Activity in prefrontal cortex
during dynamic selection of action sequences.Nature Neuroscience, 9, 276 –282.
Berkenblit, M. B., & Feldman, A. G. (1988). Some problems of motor control.Journal of Motor Behavior, 20,369 –373.
Castiello, U., & Stelmach, G. E. (1993). Generalized representation of handwriting: Evidence of effector independence.Acta Psychologica, 82, 53– 68.
Enoka, R. M. (1988).Neuromechanical basis of kinesiology.Champaign, IL: Human Kinetics.
Harrington, D. L., Rao, S. M., Haaland, K. Y., Bobholz, J. A., Mayer, A. R., Binder, J. R., & Cox, R. W. (2000). Specialized neural systems underlying representations of sequential movements.Journal of Cogni-tive Neuroscience, 12,56 –77.
Jax, S., & Rosenbaum, D. A. (in press). Hand path priming in manual obstacle avoidance: Evidence that the dorsal stream does not only control visually guided actions in real time.Journal of Experimental Psychology: Human Perception and Performance.
Kagan, J. (1971).Change and continuity in infancy.New York: Wiley. Kawato, M. (1999). Internal models for motor control and trajectory
planning.Current Opinion in Neurobiology, 9,718 –727.
Keele, S. W., Cohen, A., & Ivry, R. (1990). Motor programs: Concepts and issues. In M. Jeannerod (Ed.),Attention and performance XIII: Motor representation and control(pp. 77–110). Hillsdale, NJ: Erlbaum.
Kennerley, S. W., Sakai, K., & Rushworth, M. F. S. (2004). Organization of action sequences and the role of the pre-SMA.Journal of Neurophys-iology, 91,978 –993.
Kosslyn, S. M. (1980).Image and mind.Cambridge, MA: Harvard Uni-versity Press.
Lashley, K. S. (1930). Basic neural mechanisms in behavior.Psychological Review, 37,1–24.
Lashley, K. S. (1942). The problem of cerebral organization in vision. Biological Symposia, 7,301–322.
Merton, P. A. (1972). How we control the contraction of our muscles. Scientific American, 226,30 –37.
Meulenbroek, R. G. J., Rosenbaum, D. A., Thomassen, A. J. W. M., Loukopoulos, L. D., & Vaughan, J. (1996). Adaptation of a reaching model to handwriting: How different effectors can produce the same written output, and other results.Psychological Research, 59,64 –74. Raibert, M. H. (1977).Motor control and learning by the state space model
(Tech. Rep. AITR-439). Cambridge: Massachusetts Institute of Tech-nology.
Rosenbaum, D. A. (1987). Successive approximations to a model of human motor programming. In G. H. Bower (Ed.),Psychology of learning and motivation(Vol. 21, pp. 153–182). Orlando, FL: Academic Press. Rosenbaum, D. A., Engelbrecht, S. E., Bushe, M. M., & Loukopoulos,
L. D. (1993). Knowledge model for selecting and producing reaching movements.Journal of Motor Behavior, 25,217–227.
Rosenbaum, D. A., Loukopoulos, L. D., Meulenbroek, R. G. J., Vaughan, J., & Engelbrecht, S. E. (1995). Planning reaches by evaluating stored postures.Psychological Review, 102,28 – 67.
Rosenbaum, D. A., Meulenbroek, R. G., Vaughan, J., & Jansen, C. (2001). Posture-based motion planning: Applications to grasping.Psychological Review, 108,709 –734.
Rosenbaum, D. A., Weber, R. J., Hazelett, W. M., & Hindorff, V. (1986). The parameter remapping effect in human performance: Evidence from tongue twisters and finger fumblers.Journal of Memory and Language, 25,710 –725.
Sabes, P. N., & Jordan, M. I. (1997). Obstacle avoidance and a perturbation sensitivity model for motor planning. Journal of Neuroscience, 15, 7119 –7128.
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 (pp. 117–152). New York: Academic Press.
Swinnen, S. (1991). Motor control. In R. Dubelco (Ed.),Encyclopedia of human biology(Vol. 5, pp. 105–129). San Diego, CA: Academic Press. Worringham, C. J. (1991). Variability effects on the internal structure of
rapid aiming movements.Journal of Motor Behavior, 23,5– 8. Wright, C. E. (1993). Evaluating the special role of timing in the control of
handwriting.Acta Psychologica, 82,5–52.
Received October 27, 2006 Revision received February 20, 2007
