Timing of anticipatory muscle tensing control: responses before and after expected impact

Transcription

1 Exp Brain Res (2010) 202: DOI /s z RESEARCH ARTICLE Timing of anticipatory muscle tensing control: responses before and after expected impact Peter M. Vishton Kristin M. Reardon Jennifer A. Stevens Received: 8 January 2010 / Accepted: 13 January 2010 / Published online: 5 February 2010 Ó Springer-Verlag 2010 Abstract It is widely accepted that human motor control is anticipatory in nature. Previous studies have used electromyography (EMG) to examine muscle responses to falling objects and identified anticipatory muscle tensing (AMT) as a spike in activation that occurs prior to object impact. Some studies have suggested that humans use an internal model of gravity to mediate precisely timed AMT responses. The present study further examines predictive motor control through the analysis of AMT during an object catching task. For some trials, participants watched an object falling toward the hand; for other trials, their eyes were closed. For some trials, the object fell downward and impacted the hand; for other randomly selected trials, the object abruptly stopped 12 cm above the hand, enabling an assessment of the effect of impact anticipation independent of the reflexive tactile response associated with an actual impact. In Experiment 1, AMT did not shift for approximately 113 ms after the abrupt stop of the ball. In Experiment 2, we randomly varied the start height of the object and found well-timed AMT with a 129-ms lag time. A control system based on simple memory for fall time duration cannot explain these findings. We argue that an AMT control system with a lag time of approximately 121 ms could not perform with human levels of accuracy without accounting for the acceleration of downward moving objects. Keywords Anticipatory muscle tensing Electromyography Interceptive actions Motor control P. M. Vishton (&) K. M. Reardon J. A. Stevens Psychology Department, The College of William and Mary, Box 8795, Williamsburg, VA , USA vishton@wm.edu Introduction There is broad agreement that human motor control functions in a predictive, future-oriented fashion (Lee and Young 1985; Morice et al. 2010; von Hofsten 1980, 1983; von Hofsten et al. 1998; for review see Zago et al. 2009). Our actions are not based on the current state of the surrounding environment but instead on its anticipated future state. This anticipation takes into account the time lags involved in visual processing and motor actions (Nijhawan and Kirschfeld 2003). How this anticipation process functions is still an unresolved issue. What optical processes mediate effective actions? Across what time scale are anticipations computed? Does anticipation make use of known physical properties such as gravity? This last question has been debated in the context of anticipatory muscle tensing (AMT). Lacquaniti and Maioli (1989a) explored AMT with electromyography (EMG). They dropped a ball onto participants waiting hands while EMG sensors recorded patterns of muscle activations. Various muscles produced bursts of activation ms before ball impact. Later studies characterized the information processing inherent in AMT. When drop height was varied, AMT timing was modulated accordingly (Lacquaniti and Maioli 1989a). When a larger, heavier ball was dropped, AMT increased in amplitude, suggesting that target mass modulates AMT. When participants eyes were closed, there was initially no AMT. If a tone signaled the ball drop, however, AMT response emerged after several trials (Lacquaniti and Maioli 1989b). An exciting variant of the procedure involved sending the apparatus into orbit with the space shuttle Columbia (McIntyre et al. 2001). The precise timing of AMT in previous studies suggested that it accounts for gravitational

2 662 Exp Brain Res (2010) 202: acceleration (Lacquaniti and Maioli 1989a, b). If so, then in the microgravity environment of earth orbit, AMT should be altered. When a spring-loaded launcher pushed the ball downward (relative to the space shuttle cabin s canonical orientation), AMT occurred earlier than in the 1G environment. AMT timing was consistent with the claim that the astronauts anticipated gravitational acceleration, but because there was none, they produced AMT too early. This study suggested that gravity is an inherent part of future-oriented motor control, at least in the domain of catching falling objects. It could be that the human visuomotor system assumes the presence of downward acceleration at a rate of 9.81 m/s 2. Alternative models have been suggested, however, which do not rely on an internal model of gravity (e.g., Baurès et al. 2007). One could achieve the same timing precision observed in prior studies without an internal model of gravity by instead relying upon very frequent velocity assessments. For instance, after 10 ms of falling time, the amount of gravitational acceleration is miniscule, such that the difference between predictions based on constant velocity and gravitational acceleration will be negligible. If humans reassess the velocity of a falling object every 10 ms, then there is no need to include an acceleration component in the predictive control system. If, however, participants control actions with a longer lag time e.g., hundreds of milliseconds then the errors of a nonacceleration system would be much larger. In the current experiments, we measured the delay between a change in the motion of a falling ball and a change in the associated AMT response. We thus determined the lag time across which the visuomotor control of AMT is processed. After determining this lag time value, we considered how it might impact a control system that does/does not include an anticipation of downward acceleration. Others have used similar techniques to assess the time course of grasping control and suggested lag times of hundreds of milliseconds. In studies by Paulignan et al. (1991), an object suddenly seemed to change size or position when participants began a reach-to-grasp action. They found that grip formation changed approximately 330 ms after an abrupt size shift; arm motion changed 100 ms after an abrupt position shift. Day and Lyon (2000) found that participants made an initial, fast correction response within 160 ms of a perturbation, but a slower process of correction took place over several hundred milliseconds. Brenner et al. (1998), in their studies of interceptive pointing, found that changing the velocity of a target influenced movements after approximately 200 ms (also see Teixeira et al. 2006). Other studies have identified similar time lags in motor performance control. Human reaching involves repeated accelerations and decelerations of the hand. Pairs of acceleration and deceleration components are commonly referred to as movement units and often described as the basis of voluntary reaching control (e.g., von Hofsten 1980, 1983). Meyer et al. (1988) developed a model that predicted the duration of an initial movement unit. Duration varied as a function of the distance to and size of the target between 81 and 426 ms under typical reaching conditions. All of these studies suggest that participants plan and control actions over the course of hundreds of milliseconds. If so, then a lack of an acceleration component in the object interpolation process that accompanies catching would result in large timing errors, an argument made forcefully by Zago et al. (2008). Experiment 1 compared the time course of AMT response when the object stopped just prior to impact with the hand during an object catching task. Experiment 1: AMT during anticipated and actual object impact In the current study, a ball was attached to a string that passed over a support attached near the ceiling, and then down below the study table, where it was attached out of the participants view (Fig. 1). After it was released, the ball dragged the string behind it as it fell, pulling excess string from under the table. If the string was long enough, the ball fell until it made contact with the participant s hand. When the experimenter reduced the string length, unknown to the participant, the ball stopped 12 cm and 110 ms before striking the participant s hand. A third condition was also used in which the participant closed his or her eyes during the trial. Prior studies have focused on the onset of AMT prior to ball impact. It has been presumed that the post-impact response is due to the reflexive tactile response and not to the anticipatory response itself. The current experiment is the first to separate the anticipatory component from the post-impact tactile response by perturbing the ball drop on selected trials. It could be that AMT produced in the absence of ball impact matches the response produced when the ball impacts the hand, suggesting that the observed response is fully anticipatory. Alternatively, the abrupt stop of the ball may cause a nearly immediate change in AMT, supporting the view that AMT is supported by a rapid visual feedback process. Regardless of the outcome, observing the onset, duration, and magnitude of AMT in the stop condition will enable an assessment of its time scale and shed light on the extent to which gravitational acceleration is embodied in human visuomotor control.

3 Exp Brain Res (2010) 202: Fig. 1 We dropped a ball from a height of 83 cm. As it fell, it pulled a string behind it. For impact trials, the ball fell until it hit the hand. For stop trials, the hidden part of the string was shortened to stop the ball 12 cm above the hand. The string did not measurably slow the fall, such that it fell as if not connected to anything until the maximum string length was reached Method Participants Fourteen (6 females, 8 males) right-handed College of William and Mary undergraduates volunteered for the study in return for course credit. Three additional participants began the procedure, but were not included in analyses due to hardware malfunction. Materials and methods Participants sat in a chair and suspended their right arm over a foam pad on a table. We instructed them to keep the arm suspended throughout each trial without resting it on the table, maintaining a stationary position as much as possible. A small box (11 cm tall) was placed adjacent to the foam pad to use as a reference. We dropped a steel ball (diameter 3.2 cm, weight 177 g) from 83 cm above the hand. A string and wire were attached to the ball. The string and wire extended up from the ball to a rod mounted on the wall above the participant s head, over the rod, and then back down underneath the table, where it was attached, out of the participants view. As the ball fell downward, it pulled the string behind it, pulling excess string from underneath the table. The wire and string were long enough to allow the ball to fall onto the participant s hand. If, however, the experimenter hooked the string to a secondary attachment, the length of hidden string was shortened such that the ball would stop 12 cm above the participant s hand. We characterized the timing of the fall by placing a stop switch at heights from 0 to 59 cm above hand height and found that the string and wire exerted no measurable effect on the rate of fall, i.e., the ball fell as if it were not attached to anything until the moment when the string was maximally extended. Muscle activation was monitored using a Delsys Bagnoli-8 EMG system. Single differential surface electrodes (Delsys DE-2.1) were attached to the biceps and flexor carpi radialis. We also recorded the ball drop time using the EMG system. We attached a third EMG sensor to a wooden board that marked the ball start position. We pressed the ball against this third EMG sensor and then released it, producing a clear signal. To record ball impact time, the differential nature of a fourth EMG sensor was defeated by covering one of its recording surfaces with insulating medical tape. Participants held this modified sensor between the palm of the non-catching hand and the thigh. When the ball made contact with the catching hand a 4.5-V electrical charge carried by the ball provided a clear signal (note that the differential EMG electrodes filter out DC noise of the type produced by the constant charge carried by the ball). EMG data were sampled at 1 khz for 5 s per trial. Design and analysis Each block of three trials consisted of one eyes open impact trial, in which the participant viewed the experimenter dropping the ball, and the ball fell until it impacted the hand; one eyes open stop trial, in which the ball stopped prior to impact with the hand; and one eyes closed impact trial, in which the participant s eyes were closed during ball drop, and the ball impacted the hand. The order of trials within each block was randomized. Participants completed approximately 28 trials. We converted EMG data to the frequency domain via a Fourier transform, removed electrical source noise (60 Hz),

4 664 Exp Brain Res (2010) 202: and then reverted to the temporal domain via an inverse Fourier transform. Data were rectified and smoothed (50 Hz cutoff). For biceps and wrist recordings, we calculated the mean and standard deviation of the signal for 200 ms prior to the ball drop. We identified AMT onset as the first time when the activation increased more than 4 SD above this mean and remained so for 10 ms. For the stop trials, we calculated the moment of expected impact by adding the average fall time for the impact trials to the recorded ball drop time. For biceps and wrist sensors, AMT onset time was compared to expected impact time using one-sample t tests. We also assessed the influence of experimental condition on AMT onset time using repeated measures ANOVA. We determined the times at which the various normalized functions significantly differed from one another using paired samples t tests, using data from each participant at particular times relative to anticipated object impact (see text for details). Procedure After obtaining informed consent and attaching EMG electrodes, we instructed participants to sit at the table, facing the apparatus, and to extend their arm over the foam pad. The experimenter reached under the table and touched the string attachment before every trial, regardless of trial type, so that participants were naïve as to when impact and stop trials would occur. Participants wore headphones to block out string drag noise. After indicating the start of each trial to the participant, the experimenter pressed the ball against the start sensor and dropped it. Results and discussion Anticipatory muscle tensing was apparent for the eyes open impact and stop conditions (Fig. 2a); a small amount of AMT was also apparent for the eyes closed impact condition. Ball impact occurred 411 ms after release; for the eyes open impact, eyes open stop, and eyes closed impact conditions, the onset of biceps activation preceded this impact by 152 (SE = 22), 147 (SE = 32), and 80 (SE = 27) ms. All three values are significantly greater than 0 [t(13) = 6.9, p \ 0.001, g 2 = 0.79; t(13) = 4.63, p \ 0.001, g 2 = 0.62; t(13) = 2.99, p = 0.005, g 2 = 0.41]. The difference between conditions was highly significant, F(2,13) = 5.71, p = 0.009, g p 2 = Posthoc comparisons also revealed significantly later AMT response times for the eyes closed condition than for either of the eyes open conditions [eyes closed impact vs. eyes open impact, F(1,13) = 17.58, p = 0.001, g p 2 = 0.57; eyes closed impact vs. eyes open stop, F(1,13) = 7.62, p = 0.016, g p 2 = 0.37]. These same AMT trends were apparent for the wrist flexor muscles. For the eyes open impact, eyes open stop, and eyes closed impact conditions, the onset of wrist activation preceded impact by 57 (SE = 15), 72 (SE = 18), and 1 (SE = 12) ms. These values are significantly greater than 0 for eyes open impact and eyes open stop conditions [t(13) = 3.79, p = 0.001, g 2 = 0.52; t(13) = 3.95, p = 0.001, g 2 = 0.55]. For eyes closed impact, the anticipation was not significantly greater than 0. The effect of experimental condition on activation time was highly significant, F(2,13) = 14.23, p \ 0.001, g p 2 = Post-hoc comparisons again revealed significantly later AMT response times for the eyes closed condition than for either of the eyes open conditions [eyes closed impact vs. eyes open impact, F(1,13) = 18.49, p \ 0.001, g p 2 = 0.59; eyes closed impact vs. eyes open stop, F(1,13) = 20.33, p \ 0.001, g p 2 = 0.61]. Figure 2a presents the average wrist flexor response profiles for all three experimental conditions. We first calculated averages for each of the participants for each condition. We then normalized these functions to eliminate general response magnitude differences so that all participants contributed equally to the group average (i.e., for each participant, we determined the peak response value from all conditions and then divided all values by it). The figure presents the average of these normalized functions. A central goal of this study was to assess the lag time of AMT control. We determined the first moment after the abrupt object stop when there was a difference between the eyes open impact and the eyes open stop conditions. After the ball stopped falling, for the duration of time over which AMT is based on inferred future motion of the ball, the muscle response should not vary from that of the impact condition. The first change in AMT would mark the end of this time window. Using this method, we identified the first significant divergence between the eyes open impact and stop conditions at 3 ms after the moment of expected impact 113 ms after abrupt stop of the fall occurred. We were surprised to find AMT in the eyes closed impact condition. The consistent periodicity of the experimental trials may have been sufficient to produce this anticipation. We tested this in Experiment 2 by changing the height of the ball drop for each trial. Experiment 2: does consistent drop height play a role in AMT timing? The results of Experiment 1 suggest that AMT response is well timed and implemented at a time scale of approximately 113 ms. The combination of these two results suggests that some acceleration component is used to infer the timing of an object impact with the hand. There is an

5 Exp Brain Res (2010) 202: alternative strategy, however, which would accomplish the same level of accuracy without any internal model of object acceleration, velocity, or even position. In Experiment 1, the object was dropped from the same 83 cm height for every trial. A simple control system could just memorize the 411 ms drop time during the first trial or two and then expect it for all later trials. This method would not work outside of our laboratory, of course, but it might explain the results obtained in Experiment 1 without a need to infer the presence of some information about object acceleration. Experiment 2 replicated Experiment 1 with one key change. On every trial, the object was dropped from a different height than the one used in the previous trial. This manipulation might have resulted in several possible outcomes. Perhaps the AMT response would vanish, indicating that it does function by capitalizing on consistent object fall times in order to function with the precision observed in Experiment 1. The theory that we have espoused predicts the opposite. If AMT is based on an inference that an object will fall downward while accelerating, the AMT response should appear nearly identical to that of Experiment 1. Method All methods were identical to those used in Experiment 1 except as noted here. Twelve (6 females, 6 males) righthanded College of William and Mary undergraduates volunteered for the study in return for course credit. At the start of each trial, the experimenter moved the start sensor to a pseudo-randomly selected height between 60 and 100 cm above the hand. Results and discussion The results of Experiment 2 were generally similar to those of Experiment 1 (Fig. 2b). AMT was apparent in both the eyes open impact and eyes open stop conditions. Ball impact occurred between 350 and 452 ms after release. For the eyes open impact and eyes open stop conditions the onset of biceps activation preceded this impact by 158 (SE = 15) and 194 (SE = 25) ms, both significantly greater than 0 [t(11) = 10.62, p \ 0.001, g 2 = 0.91; t(11) = 7.60, p \ 0.001, g 2 = 0.84]. Unlike Experiment 1, however, the eyes closed impact condition showed a significantly later AMT, with a mean onset 35 (SE = 11.6) ms prior to impact [t(11) = 3.00, p = 0.005, g 2 = 0.45]. The eyes closed impact anticipation obtained in Experiment 2 was significantly later than that of Experiment 1 [F(1,23) = 7.79, p = 0.01, g p 2 = 0.25]. The difference between experimental conditions in this experiment was again highly significant [F(2,11) = 24.59, p \ 0.001, g p 2 = 0.69]. Post-hoc comparisons again revealed significantly later AMT response times for the eyes closed condition than for either of the eyes open conditions [eyes closed impact vs. eyes open impact, F(1,11) = 33.94, p \ 0.001, g p 2 = 0.76; eyes closed impact vs. eyes open stop, F(1,11) = 25.45, p \ 0.001, g p 2 = 0.70]. Similar trends emerged for the wrist flexors. For the eyes open impact and eyes open stop conditions the onset of activation preceded impact by 59 (SE = 11) and 58 (SE = 13) ms, both significantly greater than 0 [t(11) = 5.52, p \ 0.001, g 2 = 0.73; t(11) = 4.33, p \ 0.001, g 2 = 0.63]. The eyes closed impact condition did not show anticipation, with a mean onset 18 (SE = 11) ms after impact [approached significance in comparison with 0, t(11) = 1.66, p = 0.06, g 2 = 0.20]. Post-hoc comparisons again revealed significantly later AMT response times for the eyes closed condition than for either of the eyes open conditions [eyes closed impact vs. eyes open impact, F(1,11) = 28.57, p \ 0.001, g p 2 = 0.72; eyes closed impact vs. eyes open stop, F(1,11) = 26.28, p \ 0.001, g p 2 = 0.70]. Fig. 2 a Average EMG response profiles for the wrist flexors in Experiment 1 reveal a clear AMT response prior to ball impact (0 ms). Response was significantly earlier for the eyes open conditions. For the stop condition, the activation remained above baseline for several hundred milliseconds after the expected impact did not occur. b The results of Experiment 2 were very similar, even though the height and duration of the ball drop varied on each trial

6 666 Exp Brain Res (2010) 202: In Experiment 1, the duration of the object fall time was always the same. It seems that the consistent pacing of the trials enabled some anticipation of impact, even when the eyes were closed. When this aspect of the procedure was changed in Experiment 2, anticipation was significantly reduced or eliminated from the eyes closed trials. Figure 2b presents the average wrist response profiles for all three experimental conditions. Based on this, we can again estimate the time scale of AMT control. As in Experiment 1, we identified the first time when there was a significant difference between eyes open impact and eyes open stop trials. In Experiment 2, for the wrist, this first difference occurred 129 ms after the moment when the ball stopped falling. Anticipatory muscle tensing response time did not significantly vary as a function of drop height. Indeed, there was not even a meaningful trend across the range of heights that we used. Figure 3 shows the average EMG response function for four out of the ten heights used. The AMT onset for these, and all other heights, precisely preceded the moment of ball impact regardless of drop height. It seems that participants calculated the ball impact time on a trialby-trials basis, even as the drop height differed for every trial. General discussion The current studies replicated the findings of Lacquaniti and Maioli (1989a, b). AMT was apparent approximately 100 ms prior to impact in the eyes open impact conditions. In the eyes closed impact conditions, there was also a small amount of anticipatory response when the height and duration of the target drop was constant in Experiment 1; this was significantly reduced or eliminated when the height of the object drop was varied for each trial in Experiment 2. The present results also define the lag time of AMT control with the addition of the eyes open stop condition. In the stop condition, the arm muscles became active approximately 100 ms prior to expected impact, and remained active for several hundred milliseconds after the impact did not occur. A central goal of this study was to assess the time scale of AMT control. We presented two assessments based on identifying the moment when the sudden stop of the falling ball changed the AMT response. In Experiments 1 and 2, these times were 113 and 129 ms, an average of 121 ms. This time scale roughly agrees with that obtained for interceptive actions based on a new visual occlusion paradigm, around 150 ms (Marinovic et al. 2009), suggesting that this value may apply outside the context of our particular experiments. If the time scale of AMT control were sufficiently short, then there would be little need to include Fig. 3 EMG responses preceded impact time consistently, regardless of drop height, suggesting that AMT response is coordinated on a trial-by-trial basis, even when the drop height changed for each trial. A control system that functions by memorizing a particular drop duration would not produce this data acceleration in the anticipations of falling object motion. If actions are planned over the course of 121 ms, however, an acceleration component will greatly increase accuracy. Figure 4 illustrates the height of a ball falling 60 cm over 350 ms, as in Experiment 2. The dashed line indicates what would be expected based on an assessment of falling velocity 121 ms prior to impact. If one stopped inferring the acceleration of the ball and thus assumed constant velocity starting at this moment, the resulting anticipation would be as depicted by the dashed line. The resulting timing error would be approximately 31 ms. The precision of AMT suggests that human predictive control does not conform to this description. As such, the results of our experiments support the notion that AMT is governed by a control process that does account for downward acceleration of a falling object. It should be noted that our evidence does not enable us to claim that the AMT control system embodies a principle of gravity as producing downward acceleration at a rate of exactly 9.81 m/s 2. Some aspect of the object motion interpolation does seem to use information about object acceleration. Further studies may capitalize on the methods described here in order to test if the object interpolation

7 Exp Brain Res (2010) 202: Fig. 4 The timing of the fall can be characterized by a standard physical equation. Our study suggests that participants control their AMT with approximately 121 ms of lag time. For a ball dropped from 60 cm, if a participant sampled its velocity 121 ms prior to impact and assumed no acceleration thereafter, the anticipated impact would be at 381 ms rather than 350 ms 31 ms after the ball s impact. The well-timed AMT seen for various heights in Experiment 2 suggests that this is not how humans function process also considers other aspects of the physics of object motion. References Baurès R, Benguigui N, Amorim MA, Siegler IA (2007) Intercepting free falling objects: better use Occam s razor than internalize Newton s law. Vis Res 47: Brenner E, Smeets JB, de Lussanet MH (1998) Hitting moving targets. Continuous control of the acceleration of the hand on the basis of the target s velocity. Exp Brain Res 122: Day BL, Lyon IN (2000) Voluntary modification of automatic arm movements evoked by motion of a visual target. Exp Brain Res 130: Lacquaniti F, Maioli C (1989a) The role of preparation in tuning anticipatory and reflex responses during catching. J Neurosci 9: Lacquaniti F, Maioli C (1989b) Adaptation to suppression of visual information during catching. J Neurosci 9: Lee DN, Young DS (1985) Visual timing of interceptive action. In: Ingle D, Jeannerod M, Lee DN (eds) Brain mechanisms and spatial vision. Martinus Nijhoff, Dordrecht, pp 1 30 Marinovic W, Plooy AM, Tresilian JR (2009) The utilisation of visual information in the control of rapid interceptive actions. Exp Psychol 56: McIntyre J, Zago M, Berthoz A, Lacquaniti F (2001) Does the brain model Newton s laws? Nat Neurosci 4: Meyer DE, Abrams RA, Kornblum S, Wright CE, Keith Smith JE (1988) Optimality in human motor performance: ideal control of rapid aimed movements. Psych Rev 95: Morice AH, François M, Jacobs DM, Montagne G (2010) Environmental constraints modify the way an interceptive action is controlled. Exp Brain Res (in press) Nijhawan R, Kirschfeld K (2003) Analogous mechanisms compensate for neural delays in the sensory and motor pathways: evidence from motor flash-lag. Curr Biol 13: Paulignan Y, MacKenzie CL, Marteniuk RG, Jeannerod M (1991) Selective perturbation of visual input during prehension movements. 1. The effects of changing object position. Exp Brain Res 83: Teixeira LA, Chua R, Nagelkerke P, Franks IM (2006) Reprogramming of interceptive actions: time course of temporal corrections for unexpected target velocity change. J Motor Behav 38: von Hofsten C (1980) Predictive reaching for moving objects by human infants. J Exp Child Psychol 30: von Hofsten C (1983) Catching skills in infancy. J Exp Psychol Hum Percept Perform 9:75 85 von Hofsten C, Vishton P, Spelke ES, Feng Q, Rosander K (1998) Predictive action in infancy: tracking and reaching for moving objects. Cognition 67: Zago M, McIntyre J, Senot P, Lacquaniti F (2008) Internal models and prediction of visual gravitational motion. Vis Res 48: Zago M, McIntyre J, Senot P, Lacquaniti F (2009) Visuo-motor coordination, internal models for object interception. Exp Brain Res 192: