Motor Control of Rapid Sequential Finger Tapping in Humans

R. Arunachalam, V. S. Weerasinghe, K. R. Mills


We investigated in 29 healthy subjects a simple model of rapid independent finger movement: the rapid sequential tapping of adjacent fingers. Inter-tap interval (ITI) was measured for adjacent pairs of fingers in each direction. ITI was shorter in the ulnar→radial direction than in the reverse direction [P < 0.001 for middle to index (M→I) compared with index to middle (I→M)]. There was a gradient across the hand such that in the ulnar→radial direction, little to ring (L→R) tapping was fastest and M→I slowest; in the radial→ulnar direction, the reverse was the case. Rectified surface electromyography (EMG) from finger extensors and flexors was averaged with respect to either the first or second tap. The interval between the flexor EMG burst and the tap was similar irrespective of the order of finger tapping, excluding a mechanical explanation of the timing difference. Transcranial magnetic stimulation (TMS) was applied at 0- to 50-ms intervals after the first tap. Interposed TMS delayed the second tap significantly more (P < 0.001) in the M→I direction than in the I→M direction. Motor-evoked potentials (MEPs) evoked by TMS interposed between taps showed a greater facilitation in the M→I than in the I→M direction (P < 0.001). Increasing intensity of TMS rendered subjects unable to produce the second tap, more frequently in the I→M direction than in the M→I direction. We have demonstrated a consistent pattern across the hand and postulate that finger-order-dependent differences in ITI and the gradient of these across the hand may reflect the mechanism of grasping and further that the cortical programming of finger tapping differs depending on finger order.


The ability to move the fingers rapidly and independently in such activities as typing or playing a musical instrument is one feature that sets humans apart from other primates. A professional violinist, for instance, can play a three octave scale in <2 s involving the accurate, sequential placement of fingers at intervals of ∼80 ms. The task is accomplished by some 27 intrinsic and extrinsic hand muscles that are activated in a complex temporal sequence while other forearm, upper arm, and shoulder muscles stabilize the hand and arm in space. Such a feat is achieved only after practice with auditory feedback, but there are many other examples from everyday life where rapid independent finger movements (RIFM) are needed. The importance of RIFM becomes even more apparent when lesions of the nervous system are studied. Small cortical lesions caused by such conditions as stroke or multiple sclerosis may manifest clinically solely as loss of dexterity due to impaired RIFM. During recovery from larger lesions causing more profound motor disturbance, RIFM is often the last function to return (Denny-Brown 1950) if indeed it does. Study of such a complex system is clearly difficult, and for that reason we have chosen here to investigate a much simpler movement: the rapid, sequential tapping of adjacent fingers.

Comparative work has shown that fast monosynaptic cortico-motoneuronal connections have evolved pari passu with the development of dexterity (Kuypers 1981; Napier 1961; Phillips 1971), and it is likely that these connections mediate RIFM in humans (Heffner and Masterton 1983; Kuypers 1981). The large-diameter pyramidal axons from large pyramidal neurons form only a small fraction of the cortico-spinal tract but are most densely distributed to lower cervical motoneurons subserving hand muscles (Bortoff and Strick 1993). It is this population of neurons that are most easily excited by transcranial magnetic stimulation (TMS), and we have used this to study the disruption and delay of finger taps (Day et al. 1989) as well as using it as a probe of cortico-motor excitability during tapping.

Execution of finger tapping movements has been the subject of a number of previous investigations. The speed and accuracy of single-finger tapping (Aoki and Kinoshita 2001; Billon et al. 1996; McManus et al. 1986), double-finger tapping (Kitamura et al. 1993), the synchrony of bimanual finger tapping, the effect of cuing (Deiber et al. 1999; Rivkin et al. 2003), and the changes in cortical organization associated with the learning of finger tapping sequences (Gerloff et al. 1998) have all been investigated using neurophysiological or imaging techniques (Harrington et al. 2000; Sadato et al. 1996). PET studies (Aoki et al. 2005) have shown that in contrast to single-finger tapping, alternating paired-finger tapping results in additional activation of bilateral dorsal premotor, contralateral primary sensorimotor cortex, and ipsilateral anterior cerebellar areas. Rapid, sequential, adjacent finger tapping has received less attention. We show here that tapping speed is finger-order dependent with tapping in the ulnar to radial direction always being faster than in the reverse direction, that tapping speed varies in a consistent way across the hand, and that the effect of TMS when applied between the two finger taps is also finger-order dependent. We hypothesize that the central control of finger tapping differs depending on finger order and that the differences may reflect a specialization related to the rapid grasping of objects.


Experiments were performed on a total of 29 healthy subjects (M: 15, F: 14) aged between 21 and 57 yr (mean age: 33.9 yr). The dominant hand (right: 27) was studied in all. Subjects were naïve to the task at the start of the experiment.


Subjects sat with the wrist and forearm prone and supported on a table with the fingers actively extended above the rigid bar of a strain gauge. Subjects were required to tap adjacent fingers so as to minimize the interval between the taps achieved by active flexion and then extension at the metacarpo-phalangeal joint of each finger in turn. There was no constraint such that the second tap could only be executed after the first finger had returned to its starting position. Synchronous tapping of the two fingers [intertap interval (ITI) < 20 ms] produced by wrist flexion and nonindependent finger tapping, which gives rise to very short ITI, were avoided. At the start of the task, the fingers were extended 2 cm above the bar, and after tapping they returned to this position. Subjects were not allowed to view monitors to see tap timings. In some experiments, single-finger taps were used, and in these, the other fingers were held extended throughout. Finger tapping was self-paced with a minimum of 4 s between trials. Naïve subjects required a short time to gain facility with the task, and in 18 subjects, 100 trials were collected during familiarization to document this effect, but all 29 subjects were familiarized in the same way. The study began when subjects were able to produce finger taps with a consistent intertap interval (see Fig. 1).

FIG. 1.

Characteristics of intertap interval (ITI). A: ITIs (ms) of index to middle (I→M) finger taps sequentially recorded. In this naïve subject, ITI progressively shortened over the 1st 20 trials during familiarization with the task but was subsequently stable. B: in another subject, ITI was stable from the start. C: distribution of ITI in the middle to index finger (M→I) and I→M directions. Both are normally distributed; in this subject, mean I→M ITI (134.9 ms) is greater than M→I ITI (88.9 ms; P < 0.001). D: mean − SD (n = 100) M→I ITI and mean + SD I→M ITI in 10 subjects ranked in order of M→I ITI. There is marked inter-individual variability, but in all M→I ITI is shorter than I→M ITI.

Data collection

Finger taps were recorded with a strain gauge (TBM 4, World Precision Instruments, Sarasota, FL) allowing the timing of the tap to be measured. The force signal was digitized at a sampling rate of 5 kHz and also passed to a level discriminator that produced a pulse at the start of the force change. This pulse was used to synchronize electromyographic (EMG) data collection and to trigger magnetic stimuli.

Surface EMG recordings were made from electrodes over the forearm flexor and extensor compartments and from the lumbrical muscles in the palm. Pregelled 1 cm square electrodes (Neuroline, Ølstykke, Denmark) were placed at one-third of the elbow-to-wrist distance up from the wrist, just medial to the tendon of flexor carpi radialis, with the reference electrode 5 cm distal, and over the muscle belly of the common finger extensor (4 cm distal to the lateral epicondyle) again with the reference electrode 5 cm distal. In the palm, electrodes were placed in the first and second web spaces with the reference electrodes on the proximal phalanges. EMG signals were amplified (Nicolet, Madison, WI), band-pass filtered between 10 Hz and 1 kHz, and digitized at 5 kHz using an analogue to digital converter (CED1401plus, Cambridge Electronic Design, Cambridge, UK). Epochs of 1 s of data were collected from 250 ms before the first tap. Off-line rectification, and averaging, either synchronized with the first tap or with the second was performed (Cambridge Electronic Design Signal Software).


With approval of the King's College Hospital Research Ethics Committee, a magnetic stimulator driving a figure-8 coil with outside diameter of each wing of 9 cm (Magstim, Dyfed, S. Wales) was used. Stimulus intensity relative to the resting corticomotor threshold for the finger extensors was used throughout. Initially, a point on the scalp 5 cm lateral to the vertex and 3 cm anterior was marked and stimuli of increasing intensity were applied until a clear motor-evoked potential (MEP) was obtained. The coil was held with its central segment at 45° to the parasagittal plane (Mills et al. 1992) and maintaining this orientation, scalp locations 1–2 cm distant from the marked position were explored to identify the most sensitive coil position. This was then marked. Threshold was determined with the muscle relaxed (confirmed by auditory and visual monitoring) by a standard bracketing procedure (Mills and Nithi 1997). This identified a lower level of intensity at which 10 consecutive stimuli gave no MEP and an upper level at which 10 consecutive stimuli always evoked an MEP. The mean of upper and lower intensity levels was used as threshold. To compare changes in MEP amplitude across subjects, MEPs were normalized to the mean MEP (n = 10) evoked at resting threshold TMS intensity during a 10% maximum force isometric finger press. In the middle to index finger (M→I) direction, MEPs were normalized to the mean MEP during isometric middle finger pressure and in the I→M direction, to the MEP from isometric index finger pressure. The duration of the silent period (stimulus to resumption of EMG activity) after each TMS pulse was measured in all trials.

Experiment 1

In 10 normal subjects, after familiarization, ITIs and the associated EMG activity was recorded during 100 tapping movements of pairs of adjacent fingers across the hand. This gave six combinations of finger pairs: I→M, M→R, R→L, L→R, R→M and M→I where I, M, R and L refer to the index, middle, ring, and little fingers. For the middle-index finger pair, ITI for M→I and I→M taps was compared with the arm in both the prone and supine positions. Data were also collected for tapping using nonadjacent fingers (I→R and R→I). In each subject, finger order and direction was randomized. Data from an additional 19 subjects were available for the I→M versus M→I comparison.

Experiment 2

In 11 familiarized subjects, changes in corticomotor excitability and ITI were assessed during I→M and M→I taps. This finger pair combination was chosen because the ITIs had the least variability (see Fig. 1). The pulse generated by the first tap was used to trigger TMS at intervals of 0, 10, 20, 30, 40, and 50 ms. The intensity of TMS was set at resting threshold for the finger extensors. Tap to stimulus intervals were randomized, and finger order was changed from I→M to M→I at each interval. Ten trials at each tap to stimulus interval were collected. In addition, MEPs evoked after a single index or middle finger tap were studied in the same way, at the same intervals and with the same TMS intensity. Trials (n = 100) without TMS were collected at the start, midway within and at the end of each experimental run and the mean ITI from these trials were used as baseline ITI.

Experiment 3

In seven familiarized subjects, the effects of increasing stimulus intensity of TMS delivered coincident with the first tap in both M→I and I→M directions was investigated. Intensities were varied from 1.0 times threshold to 1.4 times threshold in random order. Ten trials were collected at each intensity and in each finger tapping direction. The proportion of trials that caused failure of production of the second tap was measured. Failure of the second tap was defined as an absence of force change after the first tap and before the next trial began.

Data analysis

All values are cited as means ± SD unless stated otherwise. Repeated-measures ANOVA (RM-ANOVA) was used to determine if there were overall differences in MEP amplitude, ITI, and failure of the second tap with time after the stimulus in the two directions of tapping as the main factor. Duncan's post hoc range test was used to determine which range of tap to stimulus interval produced significant differences. Paired Student's t-tests were used to determine if means of ITI in the two directions of tapping and the EMG peak to tap intervals differed. A probability level of P ≤ 0.05 was accepted as significant.


ITIs of adjacent finger pairs

In 6/18 subjects, increasing familiarity with the task lead to a reduction in ITI (Fig. 1A) but in the remainder, there was no trend in ITI over the initial 100 trials (Fig. 1B). In those subjects in whom a learning effect was seen, the ITI became stable after 20–50 trials and did not differ with respect to finger order. ITI was normally distributed (Kolmogorov-Smirnoff test, P = 0.88 for M→I taps and P = 0.37 for I→M taps; Fig. 1C).

There was, however, considerable inter-subject variability in mean ITI (Fig. 1D). Regressions of ITI in both directions of tapping against age of the subject had slopes not significantly different from zero (P = 0.69 for M→I and P = 0.81 for I→M, df = 27). The range for M→I taps was 32–206 ms and for I→M taps was 43–214 ms. In all subjects, ITI in the ulnar→radial direction (L→R, R→M, and M→I) was faster than in the corresponding reverse direction (Fig. 2). For example, taking the M→I and I→M pairing, data were available from all subjects (n = 29) and mean ITI for M→I was 87.7 ± 43.3 ms and for the I→M direction was 116.9 ± 47.3 ms. Taking each finger pair and comparing the ITI (n = 10), in the ulnar→radial direction, ITI was always significantly shorter than in the radial→ulnar direction (Student's paired t-test, P < 0.0001 for I→M vs. M→I, P < 0.003 for M→R vs. R→M and P < 0.001 for R→L vs. L→R, 9 df).

FIG. 2.

ITI across the hand. Mean ± SE in the 6 combinations of finger pairings normalized to the L→R ITI for each subject. In the radial→ulnar direction of tapping, ITI becomes progressively longer across the hand (I→M < M→R < R→L) whereas in the ulnar→radial direction ITI becomes progressively shorter across the hand (M→I > R→M > L→R).

Comparing ITI for finger pairings across the hand (n = 10), the ulnar-sided fingers in general could tap faster than the radial-sided fingers in the ulnar→radial direction (Fig. 2). Thus L→R was faster than R→M (Student's paired t-test with 9 df, P < 0.0005), but R→M was not different from M→I (P = 0.16). In the radial→ulnar direction, I→M was faster than M→R (P = 0.06) and M→R was faster than R→L (P < 0.05). Additionally, L→R was faster than M→I (P = 0.004) and I→M was faster than R→L (P < 0.01).

ITI was measured in the M→I and I→M directions after changing the position of the hand from prone to supine position in 10 subjects. The ITI was again significantly faster in the M→I direction (80.8 ± 19.3 compared with 101.3 ± 18.6 ms for the I→M direction, P < 0.004, 9 df).

Nonadjacent finger tapping

For the index and ring fingers (n = 10), R→I was significantly faster than I→R tapping. The ITIs were 69.9 ± 20.6 and 102 ± 43.7 ms respectively (P < 0.02).

EMG burst to tap intervals

EMG recordings were made to measure the intervals between electrical excitation and mechanical output. This was required to decide whether biomechanical constraints were relevant in any difference with respect to finger order that were detected. The interval between the peak of the flexor EMG burst and the tap was used for analysis. This was done by averaging the same data, initially with the first tap as the time reference and then with the second tap as the time reference. The EMG bursts associated with M→I and I→M taps in a single representative subject are illustrated in Fig. 3. As expected the electrical event preceded the mechanical tap by some 30-40ms. Also as expected there was a marked reduction in extensor activity in anticipation of and during finger flexion. The mean flexor burst to tap interval associated with index finger flexion was 36.5 ± 7.5 (SD) ms (n = 18) in the M→I direction and 37.3 ± 7.1 ms in the I→M direction. These intervals are not significantly different (Student's paired t-test: P = 0.63). The mean flexor burst to tap interval associated with middle finger tapping was 40.1 ± 10.1 ms in the M→I direction and 33.1 ± 7.6 ms in the I→M direction. The burst to tap interval was significantly shorter in the I→M direction (Student's paired t-test, P = 0.02).

FIG. 3.

Electromyography (EMG) during tapping. Averaged rectified EMG from finger flexors and extensors during M→I (A and B) and I→M (C and D) finger tapping in a single representative subject. In A and C, the 1st tap has been used as the time reference for averaging. In B and D, the 2nd tap has been used as the time reference. The time point for averaging is marked with a dotted vertical line. The mean ± SD timing of the nonreference tap is marked above each trace. The peak of each flexor burst is marked with a short vertical line. Peak flexor burst to tap intervals are unrelated to the direction of tapping.

Delay of second tap by TMS

The effect of interposing a magnetic stimulus between the two taps is seen in Fig. 4. In the I→M direction, the ITI did not differ from baseline ITI, whereas in the M→I direction, there was significant delay of the second tap (ANOVA, F: 15.9, P < 0.001). Post hoc analysis of each tap to stimulus interval, however, revealed no interval at which delays were significantly different in the two directions.

FIG. 4.

Effect of transcranial magnetic stimulation (TMS) after tap 1 on ITI. ITI (mean ± SD) normalized to the mean ITI without TMS of each subject, during M→I and I→M tapping. Threshold intensity TMS was given at 0–50 ms after tap 1. Overall, tap 2 is delayed in the M→I direction but not in the I→M direction (P < 0.001), but the difference was not significant at any particular interval.

Because there is variability in ITI, stimuli applied at intervals after the first tap fall at variable intervals before the second tap. To investigate the effects on ITI of TMS applied before the second tap, ITI for each trial was normalized to the control ITI without TMS for each subject and stimulus to tap-2 intervals from all subjects (n = 11) placed in 10 ms time bins before the second tap (Fig. 5). The number of data points in each time bin was different and differed between the two directions of finger tapping; the minimum number of data points averaged in each time bin was 20, and time bins with fewer data points (all more than 200 ms before the second tap) were not included in the analysis. The effect of TMS in delaying the second tap was different in the two directions of tapping (ANOVA, F: 62.49, P < 0.0001). The range over which the index finger tap was significantly delayed in M→I tapping when compared with the middle finger in I→M tapping was when the stimulus fell 75 to 145 ms before the second tap (Duncan's Range test, P < 0.05).

FIG. 5.

Effect of TMS before tap 2 on ITI. Mean ITI, normalized to the control ITI, with respect to time before the 2nd tap for M→I taps and I→M taps. The index finger tap in M→I tapping is delayed (*P < 0.05) when the stimulus falls 75–145 ms before the tap, whereas the middle finger tap in I→M tapping is not delayed.

MEPs evoked during tapping

MEPs, normalized to the mean amplitude of the MEP obtained at resting threshold intensity with the finger exerting a 10% maximum flexion, during single and paired taps are seen in Fig. 6. With single middle finger taps (Fig. 6A), MEPs evoked 0–50 ms after the tap were on average larger by a factor of 1.11 ± 0.07 for finger flexors and 6.05 ± 1.14 for finger extensors. With single index finger taps (Fig. 6C), MEPs evoked 0–50 ms after the tap were on average larger by a factor of 1.05 ± 0.12 for flexors and 3.33 ± 0.37 for finger extensors. With sequential M→I finger taps (Fig. 6B), flexor MEPs were larger by a factor of 2.64 ± 0.33 and extensor MEPs were larger by a factor of 7.13 ± 1.36. Similarly, with sequential I→M finger taps (Fig. 6D), flexor MEPs were larger by a factor of 1.85 ± 0.17 and extensor MEPs were larger by a factor of 4.93 ± 0.59 (Student's paired t-test P < 0.001, 9 df). The facilitation of flexor MEPs in the M→I direction was significantly greater than in the I→M direction (P = 0.008) as was the facilitation of extensor MEPs (P < 0.02). MEPs of lumbrical muscles showed essentially the same findings as with finger flexors.

FIG. 6.

MEPs during finger tapping. Mean normalized flexor and extensor MEPs evoked at tap to stimulus intervals of 0–50 ms after the 1st tap. A: after single middle finger taps; C: after single index finger taps; B: during M→I tapping; D: during I→M tapping.

To investigate whether corticomotor excitability changes in advance of the second tap, MEPs from individual trials were measured and stimulus to tap-2 intervals placed in 10-ms time bins before the second tap. Again the number of data points averaged in each time bin was different and differed in the two directions of tapping; the minimum number of data points averaged was 20 and time bins with fewer data points (all >200 ms before the 2nd tap) were not included in the analysis. The pooled results from 11 subjects are seen in Fig. 7. There was a significant difference in flexor facilitation prior to the second tap in the two directions (ANOVA, F: 22.9, P < 0.0001). In M→I tapping, flexor MEPs were facilitated 85–95 ms prior to the index finger tap (Duncan's range test, P < 0.0001), but in I→M tapping, there was no modulation of facilitation of flexor MEPs prior to the middle finger tap (Duncan's range test, P = 0.06). Extensor facilitation was also different in the two directions (ANOVA, F: 21.7, P < 0.0001). In M→I tapping, extensor MEPs were facilitated 75–105 ms before the index tap (Duncan's range test, P < 0.01), but in the I→M direction, no such facilitation was seen.

FIG. 7.

Facilitation of MEPs before tap 2. Mean normalized MEP amplitude as a function of time prior to the 2nd tap in finger flexors (above) and finger extensors (below). In the M→I direction, flexor MEPs are significantly greater at 75 ms (** P < 0.001) and 95 ms (* P < 0.02) than in the I→M direction. Extensor MEPs are greater at 75–105 ms (** P < 0.001) and 65 ms (* P < 0.04) before the 2nd tap in M→I tapping than in I→M tapping.

Duration of silent period in the two directions

The mean silent period evoked by TMS at 0 ms after single index taps was used to normalize the silent period obtained during M→I and I→M taps. In all 10 subjects and in all tap to stimulus intervals (0–50 ms), there was no significant difference (P > 0.05, 9 df) between silent period obtained during M→I and I→M taps. The mean silent period varied between individuals with a range of 58.0 to 103.2 ms for M→I and 58.3 to 100.4 ms for I→M taps.

The duration of silent period in relation to tap 2 was also measured and placed in 10-ms bins before the second tap. The number of data points was different in each time bin and in the two directions of tapping. The pooled results from all 10 subjects are seen in Fig. 8. There was no significant difference in the duration of silent period between M→I and I→M taps (repeated-measures ANOVA, F: 0.1, P = 0.75).

FIG. 8.

Silent Period during M→I and I→M tapping. A: mean silent period during paired finger taps, normalized to those obtained with TMS applied at 0 ms after single index finger taps. There was no significant difference in the two directions when TMS was applied at various intervals after tap1. B: mean silent period when measured in relation to tap 2 (0 ms on the x axis coincides with the occurrence of tap 2). The data points have been binned at 10-ms intervals and again show no significant difference between the 2 directions of tapping.

Disruption of second tap by TMS

The effect of TMS stimulus intensity on the ITI in the two directions was investigated in seven subjects. As stimulus intensity was increased, finger tapping became disrupted such that the subject was unable to produce the second tap. However, the effect was different dependent on finger order (ANOVA, F: 15.9, P < 0.001; Fig. 9). At stimulus intensities of 1.3 and 1.4 times resting threshold, the second tap in the I→M direction failed more frequently than in the M→I direction (Student's paired t-test, P < 0.02 and P < 0.03, respectively). Subjects were aware that the second tap could not be executed and were often surprised that they were unable to produce the tap.

FIG. 9.

Effect of increasing intensity of TMS on failure of the 2nd tap. Mean failure rate (%) in production of the 2nd tap during M→I and I→M tapping. The stimulus was applied coincident with the 1st tap. There is a significant excess of failures in the I→M direction compared with the M→I direction (P < 0.02) when stimulus intensity is >1.3 times resting threshold.


The muscles of the forearm that flex and extend the fingers are of complex geometry. The principal finger extensor, extensor digitorum communis, has a single muscle belly and four tendons that insert into the extensor surfaces of the index, middle, ring, and little fingers. Independent finger extension must be accomplished by partitioning of function within the muscle. There are additional, separate index and little finger extensors. Flexion at the metacarpophalangeal joint involves the lumbrical muscles and the flexor digitorum superficialis. It is not possible because of volume conduction to separate the contributions of the different muscle compartments using surface EMG recordings. The approach taken here was to synchronize averaging of the rectified EMG signal to the onset of the tap (either the 1st or the 2nd), which will have the effect of maximizing the signal from those muscles contributing to the movement. However, we recognize that many forearm muscles will have contributed to the signal. The peak of the rectified signal, will, we believe, correspond with the largest EMG contribution from the particular finger flexor and extensor most active during the movement. Indeed, the lumbrical muscles which are remote from the forearm flexors, and the signal from which will be less contaminated, gave essentially the same timings. The same problem will occur with MEPs making conclusions about facilitation of individual muscle compartments impossible. However, our concern here is with differences in facilitation that are dependent on the order of fingers being tapped, there being no reason to suppose that contaminating MEPs from other forearm muscles would be different in the two directions.

It could be argued that differences in the timing of finger tapping dependent on finger order might be due to mechanical factors within these muscles. Independent extension of the middle finger, say, might cause a shearing force within the common extensor muscle that may impede the subsequent contraction of the index finger compartment. The following observation makes this an unlikely explanation. If the difference between I→M and M→I timing is solely due to mechanical factors, then this difference would not be apparent in the timing of electrical activation of the muscle. Thus the interval between the flexor burst and the tap would be longer if there was a mechanical delay, whereas the timing would be the same if the explanation was a difference in central activation. We have shown that the burst to tap intervals were the same for the index finger in both directions of tapping. For the middle finger, the burst to tap interval is significantly shorter in the I→M direction than in the reverse direction. This argues that the difference resides in central activation and not simply in the mechanical end result. We believe, therefore, that the observed differences in finger tap timing that depend on finger order are due to differences in central activation.

We postulate that the cortical mechanisms for execution of finger tapping differ depending on finger order. Evidence for differing cortical mechanisms comes from several observations. First, by using a strong magnetic stimulus to disrupt the movement, tapping in the faster direction (M→I) fails significantly less often than in the slower direction. Second, the overall facilitation of MEPs is about twice as great during M→I tapping than during I→M tapping. Third, the modulation of facilitation in MEPs that occurs prior to the second tap is different in the two directions. Furthermore, MEP facilitation prior to the second tap occurs in both flexor and extensor muscles in the M→I direction but not in the I→M direction. Given that the same peripheral apparatus is used in the two tapping movements and that the excitation of the muscle as revealed by averaged rectified EMG is essentially the same, we would argue that this reflects a difference in central command for the movement. It could be argued that because the index finger has its own independent extensor muscle (extensor indicis), that TMS applied after a middle finger tap would cause excitation of this muscle and therefore disrupt or delay the index finger tap. This would not occur in the reverse direction. However, MEP facilitation after single index taps is less than that following single middle finger taps. This would suggest that this explanation is less likely and that a difference in central activation is involved.

It is known that primate pyramidal cells begin to discharge ≤140 ms before the appearance of EMG activity (Evarts 1972). In humans, the Bereitschaftspotential begins some 1–2 s before a self-paced movement (Cui et al. 2000), and in this premovement period, fMRI studies show activity in SMA and other frontal areas as well as in M1 (Cunnington et al. 2003). Activity in SMA appears to be even more prominent during fast self-initiated finger movements as used in the current experiments (Deiber et al. 1999). Experiments in humans using TMS to probe corticomotor excitability in the premovement period have shown a rise 70 ms or so before the movement onset (Hoshiyama et al. 1996; Rossini et al. 1988; Starr et al. 1988); in addition there is evidence that intracortical inhibition as tested with the paired pulse method declines from ∼95 ms before movement onset (Reynolds and Ashby 1999). Consistent with these findings, in the current experiments, facilitation of flexor MEPs peaked some 85–95 ms before the second tap in the M→ I direction, reflecting a general increase in M1 excitability. In contrast no such facilitation was seen before the second tap in the I→ M direction (Fig. 7). Facilitation at forces <50% maximum is thought to occur mainly through spinal motoneuron excitation but at higher forces, and probably with fast ballistic movements as used here, there is a rise in cortical excitability. This is evidenced by an increase in the size and number of descending corticospinal volleys evoked by TMS (Di Lazzaro et al. 1998). This would suggest that in the M→I direction primary motor cortex (M1) has increased excitability prior to the second tap but does not prior to the second tap in the I→M direction. One possible explanation is that in the I→M direction, the motor command may be routed to the spinal motoneurons either through a pathway other than M1 or may involve elements of M1 not excited by TMS such as smaller diameter pyramidal cells. It is known for instance that a significant proportion of the corticospinal tract derives from SMA and other frontal areas (Dum and Strick 1991; Jane et al. 1967). Against this is the finding that disruption of the second tap by TMS occurs more frequently in the I→M direction. This occurs, however, only at high intensities of TMS, and in this situation, inhibition, known to be produced by TMS (Ho et al. 1998; Inghilleri et al. 1993), may predominate leading to a failure of the second tap. At lower intensities of TMS, this inhibition may manifest as a relative subexcitability of M1during tapping in the I→M direction and result in no facilitation of the MEP. However, there was no evidence from the silent period data that long-duration cortical inhibition was different in the two directions.

It is known that a stereotyped ballistic movement can be delayed by TMS (Day et al. 1989). The EMG bursts associated with such a movement emerge unchanged, merely delayed in time. It has been suggested that execution of a stored motor program is temporarily suspended by TMS. Our finding that in I→M tapping the second tap is not delayed, whereas in M→I it is, indicates that execution of the former program is not influenced in the same way by TMS. This finding would support our speculation (preceding text) that the motor commands for the two directions of movement may differ in the pathways over which they are mediated or the motor cells used in their execution.

A clear result from this study was that not only do timing differences exist between I→M and M→I taps but that similar timing differences exist between the other adjacent fingers. Perhaps surprisingly, the fastest sequential tap was achieved with the L→R pair with progressive slowing as we proceed to the M→I pair (Fig. 2). Furthermore, tapping of nonadjacent fingers follows the same pattern. One can speculate that this may reflect the mechanism of grasping of objects such as ropes or branches in which first contact with the object would be by the little finger with progressive tightening of the grip as the other fingers are engaged. The little finger when flexed forms the smallest diameter with progressive increase in the diameter across the hand. Grasping such an object with the index finger first would be less efficient and indeed feels very awkward. Although the timing sequence of individual finger flexion during grasping has not been studied, there is evidence of differential force generation of individual fingers (Li 2002) suggesting a gradient in the ulnar→radial direction. We speculate that the cortical programming in place for rapidly grasping an object may be hard-wired and that what we have examined here is sequential fragments of this program. Grasping of objects or moving fingers in the reverse, unnatural direction requires separate programmed movements for each finger and is therefore slower.

It is known from primate work that the planning of grasping movements resides in supplementary and premotor areas. Specifically, there appears to be a specialized region of anterior premotor cortex (F5) where visuomotor integration of grasping occurs (Fogassi et al. 2001). Cells in F5 have been found to be sensitive to particular sequences of movements in trained monkeys (Clower and Alexander 1998; Shima and Tanji 2000); activity in these cells ceased once the movement had been initiated. Other cells were sensitive to the rank order of planned movements. It is also known that there is a fast, potent connection between F5 and primary motor cortex (M1) (Cerri et al. 2003). It seems likely that premotor cortex in humans performs a similar task; fMRI evidence points to the intraparietal sulcus and ventral premotor cortex being active during active manipulation of objects requiring sequential finger movements (Binkofski et al. 1999). In the current study, we have shown that corticomotor excitability is increased 75–95 ms in advance of the index tap in M→I tapping but does not change in advance of the middle finger tap in I→M tapping. This suggests that the commands emanating from SMA or F5 to M1 and directly to the spinal motoneurons differ depending on the required finger order. It is possible that in the faster direction commands are routed predominately through M1 and hence are associated with marked MEP facilitation, whereas in the slower direction SMA may have contributed directly to spinal motoneuron excitation with less involvement of M1 and therefore less MEP facilitation. Further studies including functional imaging and using TMS to disrupt activity in SMA (Schluter et al. 1998) are clearly required to test this hypothesis.


V. S. Weersinghe received a Commonwealth Fellowship for which we are grateful.


Present address of V. S. Weerasinghe: Dept. of Physiology, Faculty of Medicine, University of Peradeniya, Sri Lanka.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract