Journal of Motor Behavior, Vol. 47, No. 3, 2015 Copyright © Taylor & Francis Group, LLC
RESEARCH ARTICLE
Anticipatory Adjustments to Abrupt Changes of Opposing Forces Katrin Rapp, Herbert Heuer IfADo – Leibniz Research Centre for Working Environment and Human Factors, Dortmund, Germany. ABSTRACT. Anticipatory adjustments to abrupt load changes are based on task-specific predictive information. The authors asked whether anticipatory adjustments to abrupt offsets of horizontal forces are related to expectancy. In two experiments participants held a position against an opposing force or moved against it. At force offset they had to stop rapidly. Duration of the opposing force or distance moved against it varied between blocks of trials and was constant within each block, or it varied from trial to trial. These two variations resulted in opposite changes of the expectancy of force offset with the passage of time or distance. With constant force durations or distances in each block of trials, anticipatory adjustments tended to be poorest with the longest duration or distance, but with variable force durations or distances they tended to be best with the longest duration or distance. Thus anticipatory adjustments were related to expectancy rather than time or distance per se. Anticipatory adjustments resulted in shorter peak amplitudes of the involuntary movements, accompanied by longer movement times in Experiment 1 and faster movement times in Experiment 2. Thus, for different states of the limb at abrupt dynamic changes anticipatory adjustments involve different mechanisms that modulate different mechanical characteristics.
and 0 otherwise. In both cases predictability would be perfect. From the perspective of a human agent, however, these probabilities neglect the fact that the human capabilities of time and distance estimation are limited. With these limitations the expectancy (or subjective probability) will be less than perfect (less than 1). The level of expectancy when the event occurs should critically depend on the noise in sensing or estimating the variables such as time or distance on which the offset of the horizontal force depends. Time can be explicitly represented by some kind of central timing mechanism, as posited by certain models of motor timing (cf. Vorberg & Wing, 1996). However, the timing of motor actions could also rely on task-specific timing mechanisms (cf. Merchant, Zarco, & Prado, 2008) or on time-related changes such as the decay of proprioceptive inflow of earlier portions of the movement (cf. Schmidt, 1971) or the state of an active movement (e.g., Conditt & Mussa-Ivaldi, 1999). Instead of time-related bodily changes, the timing of motor actions can also be geared to time-related environmental changes such as the approach of an object, for example a falling ball (Heuer, 1993; Lee, Young, Reddish, Lough, & Clayton, 1983). In view of the manifold of potential sources of predictive information, the question arises which sources provide effective information that is used for properly timed anticipatory adjustments to abrupt changes of external forces. According to the existing evidence, this information is task specific. A first type of task is the classic unloading task, in which an object is removed from the horizontally held hand. For this task, anticipatory adjustments are present when the subject removes the object with the other hand, but they are absent when the experimenter removes the object at some rather unpredictable time (Hugon et al., 1982; Lum et al., 1992). More importantly, they are also absent when the unloading is temporally predictable (e.g., when it occurs 500 ms after a warning tone or when the subject presses a key which triggers the unloading; Dufosse, Hugon, & Massion, 1985). Thus, in this task expectancy of the unloading is not accompanied by anticipatory adjustments. With active button presses to trigger the unloading, however, anticipatory adjustments can be learned in the course of a series of trials (Diedrichsen, Verstynen, Hon, Lehman, & Ivry, 2003). These results suggest
Keywords: anticipation, involuntary movement, limb impedance, unloading
H
umans move in a world of variable external forces, and they do so quite skillfully. Often the variations are continuous and predictable as in moving a hand-held object in different directions relative to gravity (e.g., Hermsd€orfer et al., 1999, 2000). Sometimes, however, external forces change abruptly. These challenges of the motor system result in involuntary movements (e.g., Lum, Reinkensmeyer, Lehman, Li, & Stark, 1992). Typical examples are the sudden removal of an object held in the hand or the catching of a falling ball. When the abrupt unloading or loading of the hand can be predicted, anticipatory adjustments such as deactivation or activation of load-carrying muscles or an increase of limb impedance can be observed (e.g., Biryukova, Roschin, Frolov, Massion, & Dufosse, 1999; Hugon, Massion, & Wiesendanger, 1982; Lacquaniti & Maioli, 1989a; Shiratory & Latash, 2001). The anticipatory adjustments serve to reduce or even prevent the involuntary movements. Here we examine the effects of the predictability and thus the expectancy of abrupt offsets of horizontal forces on initial conditions and kinematic characteristics of the resulting involuntary movements. Predictability of an event such as the offset of a horizontal force can be defined in terms of the probability of its occurrence. For example, if the offset occurs 1 s after a warning signal, the probability is 1 for a delay of 1 s and 0 otherwise. If the offset occurs after the hand has moved a distance of 1 cm, the probability is 1 for the 1 cm distance
Correspondence address: Herbert Heuer, IfADo - Leibniz Research Centre for Working Environment and Human Factors, Ardeystraße 67, 44139 Dortmund, Germany. e-mail: [email protected] Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/vjmb. 167
K. Rapp & H. Heuer
that in the classic unloading task anticipatory adjustments are coupled to an individual’s own active movements that produce the unloading and thus to efferent signals, whereas other predictive information is ineffective. A second type of task is catching a falling ball. In this task anticipatory adjustments are elicited by the observation of the ball (Aruin, Shiratory, & Latash, 2001; Lacquaniti & Maioli, 1989a; Shiratori & Latash, 2001). When vision of the ball is prevented, but its fall is announced by an auditory signal, anticipatory adjustments are initially absent, but they are learned in the course of a series of trials (Lacquaniti & Maioli, 1989b). Thus, in catching a falling ball, no efferent signals are required to trigger anticipatory adjustments, but afferent information on the impending impact seems to be sufficient. In addition, with sufficient practice specific time intervals become associated with anticipatory adjustments as well. The factors that contribute to the discrepant findings with different types of task are unknown. Among the candidates, as listed by Shiratori and Latash, are the different muscle groups involved in anticipatory adjustments, the different levels of predictability of the abrupt change of the load, and the possibly limited time available for preparation and generation of the appropriate adjustments. However, certain anticipatory adjustments might also be triggered by highly specific conditions only rather than by expectancy in general. For example, some adjustments are insensitive even to subjective certainty and thus to extreme expectancy of a future event (Fukui, Kimura, Kadota, Shimojo, & Gomi, 2009; Reynolds & Bronstein, 2007). A striking example is the broken-escalator phenomenon (Reynolds & Bronstein, 2003), which can be experienced in everyday life when stepping on a broken escalator belt: even when an individual knows the phenomenon and is fully aware of the standstill of the belt, walking is perturbed by the now inadequate anticipatory adjustments to the perturbation by the running belt. Whereas in both the classic unloading task and the ball-catching (or loading) task the effective external forces are gravitational, in the present study we used artificially generated horizontal forces. Anticipatory adjustments to variations of gravitational forces could be special in that the motor system seems to incorporate representations of these forces (McIntyre, Zago, Berthoz, & Lacquaniti, 2001; Papaxanthis, Pozzo, Vinter, & Grishin, 1998). These representations are likely to be resistant to change (Sangals, Heuer, Manzey, & Lorenz, 1999) because of the presence of gravity during the whole of human evolution. For tasks such as catching even dedicated processes have been hypothesized (e.g., Bootsma, 1989). Therefore the findings obtained with gravity-related tasks could be task specific and not apply to tasks in which abrupt load changes occur which are unrelated to gravity. In the latter type of task, anticipatory adjustments could be bound to expectancy independent of the source of predictive information 168
rather than to specific sources such as an individual’s own movements. Given the existing evidence of task specificity of effective predictive information, we studied two different tasks with artificial horizontal forces in two experiments. In the task of the first experiment time was predictive. It was modelled after the everyday task of removing a tip cap from the nozzle end of a syringe. Participants maintained a certain position of the hand against a horizontal force, which was switched off abruptly after a certain time interval. For the task of the second experiment distance moved was predictive. This task was modelled after the everyday-task of cutting against a resistance, which ends suddenly and thus was similar to a surface-penetration task studied by Klatzky et al. (2013). Participants had to produce a voluntary forward movement in the horizontal plane against an opposing force and to stop the movement as soon as the force was switched off abruptly. The time intervals of the first experiment, after which the horizontal forces were switched off, corresponded roughly to the mean times needed to cover the distances of the second experiment. The involuntary movements that can be observed with the tasks of the present experiment should be affected by the anisotropy of limb impedance (Darainy, Towhidkhah, & Ostry, 2007; Gomi & Kawato, 1997; Gordon, Ghilardi, Cooper, & Ghez, 1994; Lametti, Houle, & Ostry, 2007; Tsuji, Morasso, Goto, & Ito, 1995; Wong, Wilson, Malfait, & Gribble, 2009). Given that anticipatory adjustments could also serve to modulate the impedance of the arm (Biryukova et al., 1999), their effects should be stronger in those directions in which impedance is small without such modulations. Therefore, in the first experiment we varied the direction of horizontal forces in addition to the predictability and thus the expectancy of their offset. In the second experiment we varied the strength of the opposing force in addition rather than its direction. In both experiments expectancy of the force offset as a function of the waiting period (time or distance until force offset) was manipulated in two different ways. First, the waiting period was varied across blocks of trials, but was constant for trials within each block. Second, the waiting period varied from trial to trial within each block. From reaction-time experiments (cf. Niemi & N€a€at€anen, 1981) it is known that the variation of expectancy as a function of time is different for constant and variable durations within each block of trials. The reasons for these differences, as outlined below, should also apply to constant and variable distances within each block. The two experimental conditions in which expectancy varies differently as a function of waiting period allow us to separate the effects of expectancy on anticipatory adjustments from the effects of time or distance per se. In terms of the statistical analysis, effects of time or distance would be revealed by main effects of time or distance, whereas effects of expectancy would be revealed by interactions of time or distance Journal of Motor Behavior
Abrupt Force Changes
with the sequence of times or distances—constant or random. Such interactions are thus of focal interest for the present study. The reasons for the different variations of expectancy with constant and variable waiting periods, as they have been studied extensively for the duration of the foreperiod in reaction-time experiments, are the following. First, with increasing duration the variability of time estimates increases (e.g., Peters, 1989), and with increasing distance the variability of movement amplitudes becomes larger (e.g., Woodworth, 1899). Because of the increasing noise, expectancy of the offset of the horizontal force should decline with longer durations or distances, and anticipatory adjustments should become poorer. With variable waiting periods, there is a second factor that affects expectancy. This factor favors the longest waiting period and typically dominates the increasing temporal or spatial uncertainty. With variable waiting periods, the conditional probability of force offset, given that it has not yet occurred, increases. For example, when force offset can occur randomly after three different durations with t1 < t2 < t3, then the conditional probability for the offset at t1 is 0.33 for t
RESEARCH ARTICLE
Anticipatory Adjustments to Abrupt Changes of Opposing Forces Katrin Rapp, Herbert Heuer IfADo – Leibniz Research Centre for Working Environment and Human Factors, Dortmund, Germany. ABSTRACT. Anticipatory adjustments to abrupt load changes are based on task-specific predictive information. The authors asked whether anticipatory adjustments to abrupt offsets of horizontal forces are related to expectancy. In two experiments participants held a position against an opposing force or moved against it. At force offset they had to stop rapidly. Duration of the opposing force or distance moved against it varied between blocks of trials and was constant within each block, or it varied from trial to trial. These two variations resulted in opposite changes of the expectancy of force offset with the passage of time or distance. With constant force durations or distances in each block of trials, anticipatory adjustments tended to be poorest with the longest duration or distance, but with variable force durations or distances they tended to be best with the longest duration or distance. Thus anticipatory adjustments were related to expectancy rather than time or distance per se. Anticipatory adjustments resulted in shorter peak amplitudes of the involuntary movements, accompanied by longer movement times in Experiment 1 and faster movement times in Experiment 2. Thus, for different states of the limb at abrupt dynamic changes anticipatory adjustments involve different mechanisms that modulate different mechanical characteristics.
and 0 otherwise. In both cases predictability would be perfect. From the perspective of a human agent, however, these probabilities neglect the fact that the human capabilities of time and distance estimation are limited. With these limitations the expectancy (or subjective probability) will be less than perfect (less than 1). The level of expectancy when the event occurs should critically depend on the noise in sensing or estimating the variables such as time or distance on which the offset of the horizontal force depends. Time can be explicitly represented by some kind of central timing mechanism, as posited by certain models of motor timing (cf. Vorberg & Wing, 1996). However, the timing of motor actions could also rely on task-specific timing mechanisms (cf. Merchant, Zarco, & Prado, 2008) or on time-related changes such as the decay of proprioceptive inflow of earlier portions of the movement (cf. Schmidt, 1971) or the state of an active movement (e.g., Conditt & Mussa-Ivaldi, 1999). Instead of time-related bodily changes, the timing of motor actions can also be geared to time-related environmental changes such as the approach of an object, for example a falling ball (Heuer, 1993; Lee, Young, Reddish, Lough, & Clayton, 1983). In view of the manifold of potential sources of predictive information, the question arises which sources provide effective information that is used for properly timed anticipatory adjustments to abrupt changes of external forces. According to the existing evidence, this information is task specific. A first type of task is the classic unloading task, in which an object is removed from the horizontally held hand. For this task, anticipatory adjustments are present when the subject removes the object with the other hand, but they are absent when the experimenter removes the object at some rather unpredictable time (Hugon et al., 1982; Lum et al., 1992). More importantly, they are also absent when the unloading is temporally predictable (e.g., when it occurs 500 ms after a warning tone or when the subject presses a key which triggers the unloading; Dufosse, Hugon, & Massion, 1985). Thus, in this task expectancy of the unloading is not accompanied by anticipatory adjustments. With active button presses to trigger the unloading, however, anticipatory adjustments can be learned in the course of a series of trials (Diedrichsen, Verstynen, Hon, Lehman, & Ivry, 2003). These results suggest
Keywords: anticipation, involuntary movement, limb impedance, unloading
H
umans move in a world of variable external forces, and they do so quite skillfully. Often the variations are continuous and predictable as in moving a hand-held object in different directions relative to gravity (e.g., Hermsd€orfer et al., 1999, 2000). Sometimes, however, external forces change abruptly. These challenges of the motor system result in involuntary movements (e.g., Lum, Reinkensmeyer, Lehman, Li, & Stark, 1992). Typical examples are the sudden removal of an object held in the hand or the catching of a falling ball. When the abrupt unloading or loading of the hand can be predicted, anticipatory adjustments such as deactivation or activation of load-carrying muscles or an increase of limb impedance can be observed (e.g., Biryukova, Roschin, Frolov, Massion, & Dufosse, 1999; Hugon, Massion, & Wiesendanger, 1982; Lacquaniti & Maioli, 1989a; Shiratory & Latash, 2001). The anticipatory adjustments serve to reduce or even prevent the involuntary movements. Here we examine the effects of the predictability and thus the expectancy of abrupt offsets of horizontal forces on initial conditions and kinematic characteristics of the resulting involuntary movements. Predictability of an event such as the offset of a horizontal force can be defined in terms of the probability of its occurrence. For example, if the offset occurs 1 s after a warning signal, the probability is 1 for a delay of 1 s and 0 otherwise. If the offset occurs after the hand has moved a distance of 1 cm, the probability is 1 for the 1 cm distance
Correspondence address: Herbert Heuer, IfADo - Leibniz Research Centre for Working Environment and Human Factors, Ardeystraße 67, 44139 Dortmund, Germany. e-mail: [email protected] Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/vjmb. 167
K. Rapp & H. Heuer
that in the classic unloading task anticipatory adjustments are coupled to an individual’s own active movements that produce the unloading and thus to efferent signals, whereas other predictive information is ineffective. A second type of task is catching a falling ball. In this task anticipatory adjustments are elicited by the observation of the ball (Aruin, Shiratory, & Latash, 2001; Lacquaniti & Maioli, 1989a; Shiratori & Latash, 2001). When vision of the ball is prevented, but its fall is announced by an auditory signal, anticipatory adjustments are initially absent, but they are learned in the course of a series of trials (Lacquaniti & Maioli, 1989b). Thus, in catching a falling ball, no efferent signals are required to trigger anticipatory adjustments, but afferent information on the impending impact seems to be sufficient. In addition, with sufficient practice specific time intervals become associated with anticipatory adjustments as well. The factors that contribute to the discrepant findings with different types of task are unknown. Among the candidates, as listed by Shiratori and Latash, are the different muscle groups involved in anticipatory adjustments, the different levels of predictability of the abrupt change of the load, and the possibly limited time available for preparation and generation of the appropriate adjustments. However, certain anticipatory adjustments might also be triggered by highly specific conditions only rather than by expectancy in general. For example, some adjustments are insensitive even to subjective certainty and thus to extreme expectancy of a future event (Fukui, Kimura, Kadota, Shimojo, & Gomi, 2009; Reynolds & Bronstein, 2007). A striking example is the broken-escalator phenomenon (Reynolds & Bronstein, 2003), which can be experienced in everyday life when stepping on a broken escalator belt: even when an individual knows the phenomenon and is fully aware of the standstill of the belt, walking is perturbed by the now inadequate anticipatory adjustments to the perturbation by the running belt. Whereas in both the classic unloading task and the ball-catching (or loading) task the effective external forces are gravitational, in the present study we used artificially generated horizontal forces. Anticipatory adjustments to variations of gravitational forces could be special in that the motor system seems to incorporate representations of these forces (McIntyre, Zago, Berthoz, & Lacquaniti, 2001; Papaxanthis, Pozzo, Vinter, & Grishin, 1998). These representations are likely to be resistant to change (Sangals, Heuer, Manzey, & Lorenz, 1999) because of the presence of gravity during the whole of human evolution. For tasks such as catching even dedicated processes have been hypothesized (e.g., Bootsma, 1989). Therefore the findings obtained with gravity-related tasks could be task specific and not apply to tasks in which abrupt load changes occur which are unrelated to gravity. In the latter type of task, anticipatory adjustments could be bound to expectancy independent of the source of predictive information 168
rather than to specific sources such as an individual’s own movements. Given the existing evidence of task specificity of effective predictive information, we studied two different tasks with artificial horizontal forces in two experiments. In the task of the first experiment time was predictive. It was modelled after the everyday task of removing a tip cap from the nozzle end of a syringe. Participants maintained a certain position of the hand against a horizontal force, which was switched off abruptly after a certain time interval. For the task of the second experiment distance moved was predictive. This task was modelled after the everyday-task of cutting against a resistance, which ends suddenly and thus was similar to a surface-penetration task studied by Klatzky et al. (2013). Participants had to produce a voluntary forward movement in the horizontal plane against an opposing force and to stop the movement as soon as the force was switched off abruptly. The time intervals of the first experiment, after which the horizontal forces were switched off, corresponded roughly to the mean times needed to cover the distances of the second experiment. The involuntary movements that can be observed with the tasks of the present experiment should be affected by the anisotropy of limb impedance (Darainy, Towhidkhah, & Ostry, 2007; Gomi & Kawato, 1997; Gordon, Ghilardi, Cooper, & Ghez, 1994; Lametti, Houle, & Ostry, 2007; Tsuji, Morasso, Goto, & Ito, 1995; Wong, Wilson, Malfait, & Gribble, 2009). Given that anticipatory adjustments could also serve to modulate the impedance of the arm (Biryukova et al., 1999), their effects should be stronger in those directions in which impedance is small without such modulations. Therefore, in the first experiment we varied the direction of horizontal forces in addition to the predictability and thus the expectancy of their offset. In the second experiment we varied the strength of the opposing force in addition rather than its direction. In both experiments expectancy of the force offset as a function of the waiting period (time or distance until force offset) was manipulated in two different ways. First, the waiting period was varied across blocks of trials, but was constant for trials within each block. Second, the waiting period varied from trial to trial within each block. From reaction-time experiments (cf. Niemi & N€a€at€anen, 1981) it is known that the variation of expectancy as a function of time is different for constant and variable durations within each block of trials. The reasons for these differences, as outlined below, should also apply to constant and variable distances within each block. The two experimental conditions in which expectancy varies differently as a function of waiting period allow us to separate the effects of expectancy on anticipatory adjustments from the effects of time or distance per se. In terms of the statistical analysis, effects of time or distance would be revealed by main effects of time or distance, whereas effects of expectancy would be revealed by interactions of time or distance Journal of Motor Behavior
Abrupt Force Changes
with the sequence of times or distances—constant or random. Such interactions are thus of focal interest for the present study. The reasons for the different variations of expectancy with constant and variable waiting periods, as they have been studied extensively for the duration of the foreperiod in reaction-time experiments, are the following. First, with increasing duration the variability of time estimates increases (e.g., Peters, 1989), and with increasing distance the variability of movement amplitudes becomes larger (e.g., Woodworth, 1899). Because of the increasing noise, expectancy of the offset of the horizontal force should decline with longer durations or distances, and anticipatory adjustments should become poorer. With variable waiting periods, there is a second factor that affects expectancy. This factor favors the longest waiting period and typically dominates the increasing temporal or spatial uncertainty. With variable waiting periods, the conditional probability of force offset, given that it has not yet occurred, increases. For example, when force offset can occur randomly after three different durations with t1 < t2 < t3, then the conditional probability for the offset at t1 is 0.33 for t
