Exp Brain Res (1996) 110:117-130
9 Springer-Verlag 1996
M . B i l l o n 9 A . S e m j e n 9 J. C o l e . G. G a u t h i e r
The role of sensory information in the production of periodic finger-tapping sequences
Received: 3 July 1995 / Accepted: 19 December 1995
Abstract A subject lacking proprioceptive and tactile sensibility below the neck and a group of control subjects performed sequences of periodic finger taps involving a pattern of accentuation. The required intertap interval was 700 ms. In some situations, the taps were synchronized with the clicks of a metronome. Feedback conditions were manipulated by either allowing or not allowing the subjects to hear the taps and see their finger movements. We recorded the trajectory of the subjects' finger displacement in the vertical plane, and the force and moment of contact of the finger with the response key. The control subjects achieved precise timing of the finger taps by trading off downstroke onset for movement duration, e.g., they initiated shorter-duration tapping movements with a delay. This strategy did not vary depending on task demands (e.g., synchronization) or feedback conditions. The deafferented patient produced intertap intervals on average close to the required value. However, his tap timing was characterized by increased variability and severe distortion (lengthening) after the accentuated tap, regardless of feedback conditions. He did not manifest the compensatory strategy whereby, in control subjects, movement onset was adjusted to movement duration. Thus, such a strategy in controls seems to M. Billon ( ~ ) 9A. Semjen Laboratoire de Neurosciences Cognitives, CNRS, 31 Chemin Joseph Aiguier, F-13402 Marseille Cedex 20, France J. Cole Clinical Neurological Sciences, University of Southampton, Southampton, UK J. Cole Department of Clinical Neurophysiology, Poole Hospital, Poole, Dorset, UK G. Gauthier Equipe contr61es sensorimoteurs, URA 1166, CNRS, UMR Mouvement et Perception, Laboratoire de Contr61es Sensorimoteurs, Universit6 de la Mrditerranre, Facult6 des Sciences du Sport, 163 Avenue de Luminy CP 910, F-13288 Marseille Cedex 9, France
depend on intact proprioceptive and/or tactile information from the moving limb. Upon withdrawal of visual and acoustic feedback, the deafferented subject increased the force of the taps and the amplitude of tapping movements; his mean synchronization error with the metronome also increased. However, he did not lose correct phasing between the taps and the clicks of the metronome. These findings suggest that, under normal circumstances, sequential movements are timed by an internal timekeeper which paces sensory consequences relating to the occurrence of behaviorally important events (e.g., finger taps), and not the onset of the movements eliciting those events. In the synchronization task, the timekeeper may be phase locked to the periodic acoustic stimuli by direct entrainment. Feedback information may be needed, however, for keeping any synchronization error as small as possible. Key words Movement timing 9 Sequence timing Synchronization 9Internal clock
Introduction There are contrasting views on how fine motor control is modulated by movement-generated afferent signals. On the one hand, it has been claimed that afferent signals are necessary for the generation of purposeful movements (e.g., Mott and Sherrington 1895). On the other hand, it has been shown that after adequate readaptation, deafferented patients are able to perform numerous motor tasks. For instance, Rothwell et al. (1982) observed that even with eyes closed, a deafferented subject could touch his thumb with each finger in turn, produce repetitive alternating movements of the wrist and hand such as tapping and waving, and also make different shapes in the air using his wrist and fingers. However, these authors also showed that without vision of the limb, the deafferented subject could not maintain a hand posture or generate a long sequence of repetitive movements. In addition, when vision was removed and movement trajectory
Closed loop 1
Closed loop 2 MOVEMENT L PROC~_.J
'M0V!ME I PROCESS
Fig. 1 Internal timer models illustrating open-loop and closedloop hypotheses (see text)
disturbed, the deafferented subject failed to reach the intended end position. Although purposeful movements can be selected and initiated in subjects with impaired sensory feedback, movement-generated sensory information may be critical for maintaining over time the correspondence between intended and achieved motor actions (see also Bossom 1974). In the present study, we examined the role of movement-generated afferent information in the timing control of periodic finger taps by comparing the results from a subject suffering from loss of tactile and kinesthetic afferentation with those from a group of healthy subjects. Two versions of the periodic finger-tapping task were used: self-paced tapping and synchronization with an external metronome. The synchronization task consists of placing the finger taps as close as possible to the sounds emitted by the metronome. Most current models of synchronization assume that in performing a task of this kind, the subjects predict with some degree of precision the time of occurrence of the metronome sound, produce the tap at that anticipated moment, and adjust via a feedback loop the synchronization error between the taps and the metronome sounds from one response cycle to the next (Fraisse and Voillaume 1971; Voillaume 1971; Hary and Moore 1987; Schulze 1992). Without such a corrective feedback loop, the sequence of taps would progressively drift away from the sequence of metronome sounds because of the variability inherent in the timing system that anticipates the metronome sounds (Church 1984; Treisman et al. 1990, 1992). The sensory modalities used in the synchronization error correction processes have been conjectured to include acoustic, proprioceptive and tactile feedback signals (Fraisse 1980; Hary and Moore 1987; Aschersleben and Prinz 1995). Assuming that these theories are valid, a subject deprived of all this information should experience great difficulties in both keeping the synchronization error as small as possible, and stabilizing the phase relationships between the metronome sounds and the tap series.
TAPS Unlike synchronization, self-paced tapping does not involve any coordination between finger taps and external events. It simply consists of maintaining a given rhythm of tapping during a certain period of time. Models of self-paced tapping differ from each other in whether or not they include sensory feedback. Open-loop models assume that tapping at a pre-specified rate does not require any sensory information about the ongoing movement. A model of this kind is the two-stage timing model of Wing and Kristofferson (1973); Wing (1980) which assumes two functionally independent modules: a timer and a motor implementation module (Fig. 1A). The output of the timer consists of series of pulses emitted with some temporal variability. The mean interpulse interval can be set at a predetermined value. Each pulse is assumed to initiate a chain of processes internal to the motor system, which - after a certain delay - ends in an observable movement. The model makes the assumption that the interpulse intervals and the motor delays are uncorrelated, random variables. This model does not make use of movement-generated afferent signals either for initiating the successive timer periods (i.e., the timer is assumed to be "free-running") or for adjusting the intertap intervals with reference to some memorized standard. Closed-loop models, on the contrary, assume at least one of these functions. For instance, one such model would suggest that each timer interval is triggered by a feedback signal from the previous response (Fig. 1B). In this model, any extra delay in the response feedback loop would cause the next response to be delayed by the same amount. Wing (1977) tested this model by occasionally delaying the occurrence of a response-produced auditory feedback. The responses after delayed feedback were also delayed but by an amount shorter than the delay introduced in the feedback loop, contrary to the model's expectation. In a model described by Shaffer (1982), a free-running internal timer would generate clock pulses the function of which would not be to trigger movements but to provide temporal goal points for movement-produced effects. As Shaffer (1982) pointed out, a model of this kind implies that the motor system has procedural knowledge and computational power to generate movement trajecto-
119 ries with durations appropriate to the timing goals. Questions arise, however, about whether feedback information is necessary for such a system be able to work. One logical requirement seems to be that feedback information (FB r in Fig. 1C) be available about the occurrence of the expected movement effects (for instance a finger tap on a key). Any discrepancy between the expected and actual moment of occurrence of those effects could then be used for refining the motor commands in the subsequent movement cycle(s). Updating the motor commands from one cycle to the next, or during movement execution, may also require dynamic action feedback (FB a in Fig. 1C) concerning, for instance, the velocity or acceleration of the movement. 1 Two previous studies provided support for the notion that periodic finger taps were paced by movement termination rather than movement onset (Billon and Semjen 1995; Billon et al. in press). In these studies, finger movement trajectories were recorded in the vertical plane during the production of tapping sequences. In a timing model in which the onset of the tapping movements is triggered by the pulses of the internal timekeeper in an open-loop fashion, the time intervals measured between successive taps (i.e., intertap intervals) should be m o r e variable than those measured between downstroke onsets (i.e., onset intervals) because motor execution adds variability to the taps (Wing 1980). Contrary to this expectation, Billon and Semjen (1995) and Billon et al. (in press) found the intertap intervals to be less variable than the onset intervals, suggesting that the internal timekeeper constrained the occurrence of the movement endpoints (taps) rather than the occurrence of movement onsets. One implication of this is that the variations of the onset times may be partly compensated for by opposite variations of the movement execution times. For instance, later initiation of the downstroke may be correlated with faster execution of the downward movement, and earlier initiation of the downstroke, with slower execution. Trial-by-trial correlations provided evidence for such a trade-off in both of the above-cited studies. A further finding consistent with endpoint (tap) timing was related to the effect of accentuating one of the taps in the sequence. The forceful downstroke that produced the accentuated tap was executed in less time than other downstrokes, and the immediately following downstroke was executed with longer duration than other downstrokes. However, subjects adjusted the initiation of these downstrokes in a compensatory way. That is, the average onset interval preceding the fast downstroke was systematically lengthened, whereas the average onset interval preceding the slow downstroke was systematically shortened. It must be noted that despite such refinement
of their timing control strategy, the subjects never succeeded in producing perfectly even tap timing when accentuation was required. The intertap interval immediately following the accentuated tap was systematically longer than the required value, independently of the required tapping rate (between 300 ms and 900 ms), musical expertise, and the presence or absence of an external metronome (Billon and Semjen 1995). In the present study, the subjects produced the fingertapping sequences in two feedback conditions. In one condition, the subjects could hear the taps they produced on the response key and see their finger movements. In the other condition, visual and auditory feedback was removed. The subjects produced the finger-tapping sequences either with all taps being similar in intensity, or by accentuating every fifth tap. If endpoint (tap) timing requires movement-related sensory feedback, as suggested in Fig. 1C, then the timing performance of a subject deprived of all feedback information should not manifest the abovementioned characteristics, namely the inverse relationship between the moment of downstroke onset and downstroke duration and, hence, the reduced variability of the intertap intervals as compared with the onset intervals. Rather, the timing performance of such a subject (e.g., our deafferented subject with visual and auditory feedback removed) should resemble "open-loop timing" as described in Fig. 1A. A further question that could be tackled by comparing the deafferented subject's timing performance with that of controls concerned the long intertap interval that appears regularly after an accentuated tap. If this phenomenon is linked to a systematic deformation of the internal timer's output, then it should be present in the timing performance of both the control subjects and the deafferented subject. The alternative hypothesis would be that a subject's movement-related feedback afferents undergo qualitative and quantitative change when a forceful accentuated tap is produced. Such variations in feedback signals might somehow disorganize the computation of movement trajectories adapted to the downstrokes' temporal goal. If this were the case, a lengthened intertap interval after the accent would characterize only the controls' timing performance. In summary, this study addresses the question of whether movement-related feedback contributes to the timing performance in self-paced periodic tapping; whether it is necessary for keeping the tapping sequence synchronized with the sound sequence of an external metronome; and whether it plays a role in the mistiming of the sequence after an accentuated tap.
Subjects and methods
The elaboration of a process model of how timing goals may Subjects organize movement trajectories in space and time is beyond the scope of this study. The working principles of the neural me- A control group of six healthy male subjects (mean age 33 years, chanisms underlying this achievement may resemble, for in- SD 11.4 years) and a deafferentedsubject, AN (age 41 years), perstance, those responsible for the shaping and timing of the speech formed the experiment. motor system as a function of anticipated sequences of phonemic Details of the case history of AN can be found in Cole and goals. Sedgwick (1992). At the age of 19 years, AN suffered a purely
Finger tin g
Response kkey .
(3) ~-,-, (4) I A ]I ".-Velocity
i /___J _ _
Time Fig. 2 Schematic representation of the apparatus and definition of the dependent variables. (1) intertap intervals between successive contacts with the response key, (2)force (arbitrary units), (3) onset intervals between onsets of the successive downstrokes, (4) downward movement time (from the onset of the downstroke to the time of contact with the key), (5) peak amplitude sensory neuropathy with acute onset following an illness documented at the time as infectious mononucleosis. Neurophysiological tests have confirmed the loss of large myelinated fibers below the neck (Cole and Katifi 1991). There were no sensory nerve action potentials and no Hoffman reflex. Cutaneous-muscular reflexes were absent (Baker et al. 1988). Needle electromyographic investigation of the muscles was normal. Pain and temperature sensations were still present, suggesting significant sparing of small myelinated and unmyelinated fibers. All AN's everyday movements require sustained attention and visual control. For instance, AN is unable to move accurately in the dark or to carry out simultaneous motor tasks such as maintaining a precision grip while walking. Apparatus The response key and the manipulandum were mounted on a wooden support fixed to a table. The manipulandum (Fig. 2) consisted of a molded forearm rest and a metal finger bar which pivoted in the vertical plane. The subject was seated with forearm flexed at 90 ~ and resting on the molded support. The support height was 6 cm. The finger bar could be adjusted in length to be
in line with the forearm. At its distal end the finger bar supported a ring in which the index finger was inserted. The axis of rotation of the finger bar about the support was directly beneath the wrist, with the response key (a circular metal plate 2 cm in diameter) beneath the distal end of the finger bar. Rotation of the finger bar (i.e., downward and upward movements of the hand about the wrist) during tapping with the index finger was recorded with a potentiometer. Measurements All experimental events and measurements were controlled by an Olivetti M290 PC computer. The definition and measurement of the dependent variables are shown in Fig. 2. The time intervals from onset of a contact to the onset of the next contact [i.e., intertap intervals: see (1) in Fig. 2] were measured to the nearest millisecond. Tapping on the response key also activated a piezoelectric crystal. Tapping force [see (2) in Fig. 2] was measured in arbitrary units as the peak output voltage of the piezoelectric crystal. The output of the potentiometer was sampled and digitized at a frequency of 250 Hz, Instantaneous velocity of the tapping movements was derived from the displacement recording. The velocity profile was smoothed by a moving average of five points and the time occurrence of velocity zero crossing was taken to indicate the onset of the downward movement. These time points were used for calculating both the intervals between the onsets of the downstrokes [i.e., onset intervals: see (3) ~n F~g. 2]; and the duration of the downward movement from the zero crossing point of the velocity curve to the contact of the index finger with the response key [movement time: see (4) in Fig. 2]. Peak amplitude [see (5) in
Fig. 2] and the time to reach peak amplitude were also determined. The peak amplitude is the highest point on the tap trajectory. The time to peak is the delay from the tap contact offset to the occurrence of the peak amplitude. Task and procedure The subjects produced sequences of finger taps in eight different conditions which resulted from orthogonal combination between two feedback conditions (normal feedback/no feedback), two pacing conditions (without metronome/with metronome), and two accentuation conditions (without accent/with accent). 2 A trial started with the presentation via headphones of a series of metronome clicks. The clicks were generated by electric pulses of 5 ms duration paced at 700 ms. The subjects were instructed to listen to several clicks before starting to tap. In the self-paced ("no metronome") condition, the metronome was switched off upon detection of the first tap, and the subjects continued tapping according to the imagined cadence of the metronome until the completion of 35 taps. In the synchronization condition, the metronome remained on during the whole series of 35 taps. The accentuation task required the subjects to tap every fifth tap stronger. In the so-called no feedback condition, auditory and visual information of hand movements and taps was removed; visual information by covering the subject's hand and forearm by a large cardboard box, auditory information by delivering continuous white noise in the headphones and by providing the subjects with ear plugs. The intensity of the white noise (about 80 dB) and of the metronome clicks was set so that the subject could not hear the taps on the response key, but still could hear the metronome. The different experimental conditions were presented to the subjects systematically. First, each subject was tested in the no feedback condition. Within this condition, six synchonization series were performed first - three without and three with accentuation. Another block of six self-paced ("no metronome") tapping series followed - three without and three with accentuation. Then the subjects moved to the normal feedback condition and performed the synchronization and self-paced series, without or with accentuation, as in the no feedback condition. Tapping sequences with an intertap interval two times larger or smaller than the required value were discarded and replaced by a new sequence. Sequences with incorrect accent position were treated similarly.
Results The dependent variables were analyzed in most cases by taking into account the recurrence of the accentuated tap. To this end, each trial sequence was subdivided into subsequences of five taps with one accentuated tap in each. Means were calculated over sub-sequences according to the sequential index (1 to 5) of the dependent variables. These computations were carried out trial by trial for each subject.
Movement characteristics The characteristics of the tapping movements in the vertical plane, i.e., peak movement amplitude, upward movement velocity, downward movement time, and tap 2 In a previous study (Billon and Semjen 1995), sequences of 65 taps were used. During the work with the deafferented subject it soon appeared that in the absence of visual and acoustic feedback, he lost contact with the response key after about 40 tapping cycles. It was then decided to request 35 taps in each sequence.
force, are given in Fig. 3. Data were averaged over metronome conditions (presence or absence of the metronome sounds during tapping) since this variable did not have any systematic effects on the movement characteristics considered here.
Peak amplitude and average velocity of upward movements
With the control subjects, the upward movement immediately before the accentuated tap (Fig. 3A, position 1) and the one that immediately followed it had larger amplitude than the other movements. Moreover, the amplitude of the upward movements tended to become smaller when visual and acoustic feedback was suppressed. A 5x2x2 (Tap positionxFeedbackxMetronome) analysis of variance (ANOVA) performed on the subjects' individual means revealed a significant position effect, IF(4, 20)=11.04, P