1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Pre-programmed mechanisms of motor control are known to influence the gain of artificially evoked stretch reflexes. However, their interaction with stretch reflexes evoked in the context of unimpeded natural movement is not understood. We used a landing movement, for which a stretch reflex is an integral part of the natural action, to test the hypothesis that unpredicted motor events increase stretch reflex gain. The unpredicted event occurred when a false floor, perceived to be solid, collapsed easily on impact, allowing the subjects to descend for a further 85 ms to a solid floor below. Spinal stretch reflexes were measured following solid floor contact. When subjects passed through the false floor en route to the solid floor, the amplitude of the EMG reflex activity was double that found in direct falls. This was not due to differences in joint rotations between these conditions. Descending pathways can modify H- and stretch-reflex gain in man. We therefore manipulated the time between the false and real floor contacts and hence the time available for transmission along these pathways. With 30 ms between floors, the enhancement of the reflex was extinguished, whereas with 50 ms between floors it reappeared. This excluded several mechanisms from being responsible for the doubling of the reflex EMG amplitude. It is argued that the enhanced response is due to the modulation of reflex gain at the spinal level by signals in descending pathways triggered by the false platform. The results suggest the future hypothesis that this trigger could be the absence of afferent signals expected at the time of false floor impact and that salient error signals produced from a comparison of expected and actual sensory events may be used to reset reflex gains.

Movements are controlled by motor programmes stored in the brain and by spinal reflex mechanisms. These two means of control, once thought of as separate, are now known to interact (Prochazka, 1989; Hultborn, 2001). This paper concerns their interaction in a human landing movement.

In man, the gain of spinal stretch reflexes and H reflexes is modulated during locomotion (see review by Zehr & Stein, 1999). Furthermore, there is evidence that descending pathways play a significant part in this control. For example: (1) gain increases with the intention of contracting a muscle and before the EMG activity starts (Burke et al. 1992); (2) transcranial magnetic stimulation of the contralateral motor cortex decreases presynaptic inhibition of the soleus muscle Ia afferent fibres (Iles, 1996); (3) when peripheral sensory input is removed by blocking transmission in the common peroneal nerve, fictive dorsiflexion still depresses the soleus H reflex (Nielsen et al. 1992); (4) imagination of movements augments H reflexes without actual skeletomotor activity (Gandevia et al. 1997).

Although this evidence is substantial, it is not clear what role these modulations in gain fulfil in a natural movement, because the electrical stimulation and imposed joint rotations are not a normal part of the natural movement. However, in landing from a jump, the action investigated herein, strong reflexes are a functional part of the normal action (Duncan & McDonagh, 2000). Thus, this paradigm allows the interaction of predictive control and reflex gain to be investigated in a natural setting.

One example of this interaction is when unexpected mechanical events interrupt the predicted course of a movement. There are hints in previous work across a range of organisms that these events modify reflex amplitudes. For example, in crustacea, stretch-reflex gains are increased following unexpected slips (Barnes, 1977). In mammals, reflexes are released when events do not follow the pre-set movement programme (Grillner, 1975). In man, Dietz & Noth (1981) found that during running the natural stretch-reflex response of the lateral gastrocnemius muscle was larger if, in a single stride, the height of the contact surface was unexpectedly lowered.

The unexpected mechanical event we used to investigate the interaction of predictive and reactive control was the absence of solid floor impact at the predicted time. Subjects jumped down onto a floor that they perceived to be solid, but which easily gave way, allowing a continuous fall to a solid floor below.

We asked two main questions: is the gain of a naturally occurring spinal stretch reflex altered by this sudden and unexpected mechanical event, and is there any evidence that this alteration is brought about via descending pathways?


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References


Four main protocols were carried out, all of which involved jumping down from a platform onto a landing surface, absorbing the impact, and then regaining the standing position. In this paper, this whole movement will henceforth be referred to as ‘landing’. For each protocol the subjects came to the laboratory twice. The first visit was a habituation session. The second visit was for the main experimental session. The sequence described below was common to all of the protocols, apart from differences in fall distances.

The first (habituation) session was used to familiarise the subjects with the requirements of the experiment and to habituate them to the landing movement. During this session, each subject performed four sets of 10 downward jumps to distance ‘a’ and four sets of 10 downward jumps to a greater distance ‘b‘. Five minutes of rest was allowed between each set. At least two of the sets at each height were performed with the electrogoniometers and EMG electrodes attached to the subject.

During the second (main experimental) session, the subjects experienced three conditions (types of jumps): (1) a jump down to a false floor at distance ‘a’ from the take-off platform. They fell through this false floor to land on a solid floor at distance ‘b’ from the take-off platform; (2) a jump to a solid floor seen at distance ‘a’; (3) a jump to a solid floor seen at distance ‘b’. In the second session, the subject was first asked to perform the landing to a floor at distance ‘a’ that was, unbeknown to them, a false floor. Only one trial with this deception was performed for each subject so that no adaptation to the task could occur. The subject was then informed that every other situation would be as practised in the habituation session. A box in the landing stage was then replaced to produce a solid floor. The subject then performed five practice landings to distance ‘a‘, to become familiar with the height, and then 15 recorded landings. The first landing of the 15 was used as the short-fall control to compare with the surprise landing. The box was then removed and the subject performed five practice and 15 recorded landings to distance ‘b’ with no false floor en route. The first landing of the 15 was used as the long-fall control to compare with the surprise landing. With respect to the EMG, taking data for each subject in turn, the landing used as the control was compared to the average of the 15 landings. If it had been found to lie outside the mean amplitude of the 15 ± 1 s.d. it would have been discarded. All of the controls lay within this range. The normal trials (post-surprise) were compared to the habituation trials of the previous day (no prior surprise) and were found to be indistinguishable. All falls were self-initiated, and were performed bare-footed with the hands on the hips and the elbows turned outwards. In each landing, the subjects were asked to focus their eyes on a stylised ‘face’ drawn on the landing surface. They were instructed to ‘roll off’ rather than jump off the platform, whilst keeping both feet together throughout the movement. They were also asked to break the fall smoothly and naturally.

Four different protocols were performed using this combination of surprise and normal landings. The fall distances and flight times between floors are summarised in Table 1 and Fig. 1. However, in brief, for each protocol the values of fall distances ‘a’ and ‘b’ were as follows: protocol 1 a= 0.45c m, b= 0.70 m; protocol 2 a= 0.64 m, b= 0.70 m; protocol 3 a= 0.45 m, b= 0.5 m; protocol 4 a= 0.45 m, b= 0.6 m. The rationale for this range of fall distances is explained in the results. Prior to participation in the experiments, written informed consent was obtained from the subjects, all of whom were healthy, physically active students from the department and were not experienced in landing manoeuvres of the kind studied. A separate group of similar subjects was used for each protocol and there were 10 subjects in each group. This was done to preserve the naivety of the subjects with respect to the surprise element that was common to each protocol. The subjects were 22 females and 18 males; mean ±s.d. age 20.7 ± 2.7 years; body mass 67.0 ± 9.6 kg; height 1.71 ± 0.09 m.

Table 1.  Comparison of reflex EMG amplitudes for surprise and control landings
 EMG amplitude (%)
 Control landingSurprise landingP
  1. Protocol 2: time between floors ≈30 ms; false floor at 0.64 m; real floor at 0.7 0m.

  2. Protocol 3: time between floors ≈30 ms; false floor at 0.45 m; real floor at 0.50 m.

  3. Protocol 4: time between floors ≈50 ms; false floor at 0.45 m; real floor at 0.60 m.

  4. Summary data from protocols 2, 3 and 4. Values are means ±s.d. of determinations from 10 subjects. Each protocol was carried out on a different set of 10 subjects to retain subject naivety to the false floor condition. *Statistically significant difference (P≤ 0.05, two-tailed t test for paired values) between the raw results from the same muscle obtained under different conditions. Prior to calculation of the mean across subjects, all of the absolute EMG amplitudes from each subject were normalised with respect to the amplitude in the 80 ms prior to touchdown in the 0.45 m landing; all data are expressed as a percentage of this value.

Protocol 2
  Gastrocnemius433.8 ± 264.2312.7 ± 206.50.214
  Rectus femoris1202.6 ± 1143.91104.8 ± 674.10.949
  Biceps femoris613.5 ± 951.9598.3 ± 1009.60.519
Protocol 3
  Gastrocnemius705.0 ± 575.0900.8 ± 749.80.423
  Rectus femoris1174.6 ± 877.11301.9 ± 774.50.633
  Biceps femoris348.8 ± 240.8498.2 ± 776.80.464
Protocol 4
  Gastrocnemius197.3 ± 127.6571.5 ± 495.50.050*
  Rectus femoris1067.8 ± 546.71661.7 ± 1214.70.054
  Biceps femoris482.1 ± 689.31168.5 ± 1464.30.035*

Figure 1. Schematic representation of the four surprise landing protocols

The top, continuous line represents the take-off platform, the dashed line represents the false floor, and the bottom, continuous line represents the solid landing floor. Fall distances are shown together with the fall time between false and real floors. Also shown are the reflex EMG amplitudes (means for 10 subjects) following hard floor impact under surprise conditions expressed as a ratio of those found in control falls to hard floors at the same distance.

Download figure to PowerPoint

All experimental procedures were approved by the University Ethical Committee and were in accordance with the Declaration of Helsinki (2000).

Apparatus and technical procedures

The apparatus and technical procedures are described fully elsewhere (Duncan & McDonagh, 2000). In brief, EMGs were taken from the soleus, gastrocnemius, biceps femoris and rectus femoris muscles. Joint angles were measured at the knee and ankle using electrogoniometers. Pressure-pad switch mats recorded take-off and landing on the solid floor and an infrared detector was triggered by the subjects’ feet as they passed through the false floor. A false floor that looked identical to a solid floor at the same height was set up. Subjects fell through the false floor to a solid floor below. Joint velocities were calculated as the mean value during the whole unidirectional joint rotation following impact and as the mean velocity over the first 10 ms of rotation. All of the experimental recordings were made in one session so that the electrode placement for each subject remained the same throughout.

Signal processing.

Electrogoniometer and EMG signals were fed via a variable amplifier to an analogue-to-digital converter (1401+; Cambridge Electronic Design (CED), Cambridge, UK), which sampled the signals at 2 kHz for storage and analysis on a personal computer. Raw traces were saved on the computer for off-line analysis. DC offset was removed prior to full-wave rectification and subsequent averaging. This analysis and averaging of the electrogoniometer signals was performed using programs written in the Spike2, version 2 beta 6, script (CED).

EMG measurements and analysis.

To establish objectively the onset and offset of responses in the EMGs, we used a method suggested by Mr D. McIntyre. This is described fully elsewhere with a critique of other methods (Santello & McDonagh, 1998). The method enables the continuous increase or decrease of signals to be distinguished objectively from intermittent bursts and baseline noise. To compare results among subjects, all of the data were normalised with respect to the average EMG amplitude of the 80 ms preceding touchdown in the 0.45 m landing. This was calculated for each muscle in the 0.45 m landing by summing the 160 data points in the period 80 ms before touchdown and dividing by the number of data points. This average EMG amplitude preceding touchdown in the 0.45 m landing was the baseline used for normalisation of all EMG data for all landings both before and after impact. This was repeated for each muscle and for each subject. This method of normalisation of the EMG produced less variability than normalisations against a maximal isometric contraction. Once the data were rendered as a percentage for individual trials, the trials from 10 subjects could then be combined to form means and standard deviations for the group. The amplitude of EMGs averaged across subjects (e.g. Fig. 4, lower panel) is expressed in these percentage units.


Figure 4. Protocol 1: 85 ms between floors. Gastrocnemius muscle EMG traces for 0.70 m and surprise landings

The upper panel shows data from a single subject; the lower panel shows the average of the data from 10 subjects. In each panel, the upper trace is a 0.70 m landing and the lower, inverted trace is the surprise landing. The traces are aligned at the point of true contact with the ground at 0.70 m, as indicated by the continuous cursor. In the upper panel, the dashed cursor indicates the time of contact with the false floor. In this and subsequent figures, the group data have been normalised as described in Methods and false floor contact time is only shown for single-subject, single-trial data, not for data averaged across subjects.

Download figure to PowerPoint

The amplitude of the post-landing EMG activity was considered over two time frames: 0-35 ms post-landing and 35-80 ms post-landing. Data for the tibialis anterior muscle were analysed over a 0-50 ms time frame, as this was when most activity occurred for this muscle. Data were analysed using Student's two-tailed t test for paired values. Each value of the pair was recorded from a single subject, whilst EMG electrodes and electrogoniometers remained in situ. Amplitude was computed on the raw data as the average of the highest 10 data points within these time windows. For display in tables, the maximum post-landing EMGs were normalised in the manner described above.

Unless indicated otherwise, the data are presented as means ±s.d.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

Protocol 1: falls with 85 ms fall time between the false and solid floor

Overall view of recordings from a single subject.

Figure 2 shows the data from all the measured variables for a single normal landing from a single subject. Almost identical results were seen when single trials from the 10 subjects were averaged. Following take-off, the ankle joint underwent a gradual plantar flexion of ≈15 deg and the knee joint slowly extended by ≈30 deg. Just before impact, the knee joint reached maximum extension and the ankle joint reached a maximum of plantar flexion.


Figure 2. Protocol 1: 85 ms between floors. Recordings from one subject performing a single, control landing to a solid floor at 0.70 m

A pressure-pad switch on the take-off platform indicated the moment of take-off (first dashed cursor). In this and subsequent figures, cursors are shown as dashed vertical lines and are numbered sequentially from left to right. An infrared signal (second dashed cursor) indicated when the subject's feet passed a point 0.45 m below the take-off point. Another pressure pad switch (trace not shown) on the floor indicated the final ground contact as shown by the third dashed cursor. An increasing positive value in the electrogoniometer signal represents extension in the knee and dorsiflexion in the ankle. EMG signals have been full-wave rectified.

Download figure to PowerPoint

The EMG records (Fig. 2) show that after take-off there was a period of low-level activity, then by about 100 ms prior to landing, all of the muscles were active. The point of initiation of this activity, as assessed using the McIntyre algorithm (see Methods), was earlier for the gastrocnemius and soleus muscles than for the rectus femoris. In the case of the biceps femoris, gastrocnemius and soleus muscles, the activity started to decline before impact. In all of the muscles there was a rise in activity post-impact.

Figure 3 shows the data from all of the measured variables for a single surprise landing from the same subject as in Fig. 2. Cursor 1 indicates take-off, cursor 2 now registers passage through the false floor at 0.45 m and cursor 3 indicates impact with the solid floor. The electrogoniometer records show that no instantaneous change in joint rotation was caused by passage through the false floor. The time between floors in the landing shown for this subject was 84.6 ms. The mean time for all the subjects in this first protocol was 87.8 ± 4.7 ms.


Figure 3. Protocol 1: 85 ms between floors. Recordings from one subject performing a single, surprise landing

The pressure-pad switch on the take-off platform registered the moment of take-off, as indicated by the first dashed cursor. The infrared detection system registered the moment of expected ground contact (false floor) at 0.45 m, as indicated by the second dashed cursor. Another pressure-pad switch on the floor provided the moment of ground contact at 0.70 m (not shown), as indicated by the third cursor. EMG signals have been full-wave rectified. The directionality of electrogoniometer signals is the same as in Fig. 2.

Download figure to PowerPoint

In the ≈85 ms of flight time between floors (Fig. 3), the pre-landing, pre-programmed activity can be seen to diminish in the biceps femoris, soleus and gastrocnemius muscles. This pause of EMG activity is not present in the control condition (compare Fig. 2 and Fig. 3). During this pause it can be seen that there is a period of nearly constant ankle angle, whereas in the knee, some re-extension occurs, perhaps due to the force of gravity not being counteracted by sufficient muscle action. This longer pause in the EMG allows the fresh responses following impact to be clearly distinguished from ongoing activity. Following impact on the solid floor (Fig. 2 and Fig. 3), there are several bursts of strong EMG activity over the next 100 ms. The pre-impact EMG activity and its timing in surprise and normal landings have been discussed fully in our previous paper (Duncan & McDonagh, 2000). However, it should be noted again in comparing Fig. 2 and Fig. 3 and in Fig. 4, that EMG activity and pre-landing joint trajectory changes are initiated earlier in the surprise landing. For example, the peak knee-joint extension found prior to landing was shifted 63 ± 16 ms earlier in the surprise as compared to the control landing (10 subjects, P < 0.00001, Student's two-tailed t test for paired values). These points provide further evidence that the subjects believed the false floor at 0.45 m to be solid (see Duncan & McDonagh, 2000 for further discussion of this point).

Muscle activation following solid floor impact is much larger if a false floor has been traversed en route.

Figure 4 shows a comparison of rectified gastrocnemius EMG records from two experimental conditions. The records in the upper panel were taken from a single subject and show two rectified EMG signals recorded when the subject landed on a solid floor 0.70 m below the take-off point. The upper trace is a record from a normal landing to this solid floor. The inverted trace is the signal recorded in the surprise condition when the subject had passed through a false floor at 0.45 m en route to the solid floor at 0.70 m. The records in the lower panel are arranged in a similar manner, but show the mean responses from 10 subjects.

The most striking result of the present experiments was that the post-landing EMG peaks and the total EMG activity for three of the five muscles studied were very much larger in the surprise compared to the normal landing (Table 2). For all except the tibialis anterior muscle, the amplitude of the EMG activity was considered over two time frames 0-35 ms post-landing and 35-80 ms post-landing. Over the first time frame, there were no significant differences in overall EMG amplitude between surprise and normal conditions. Over the second time frame, large absolute and statistical differences were found (Table 2) and subsequent descriptions focus on this 35-80 ms period. In Table 2, the data for the tibialis anterior is shown for a different time window from the above (0-50 ms) because the strongest activity was found to be present in this time window. For the individual subject data shown in Fig. 4 (upper panel), the peak EMG activity of the gastrocnemius muscle during this phase was 5.7 times greater in the surprise as compared to the normal landing. When the total EMG activity between 35 and 80 ms post-landing was measured from single trials from all subjects, the results from the gastrocnemius showed that in the surprise trial the mean activity was 2.2 times larger than that in the control trial (Fig. 4, lower panel and Table 2). This large difference in post-landing EMG amplitude between normal and surprise conditions was true for all of the subjects.

Table 2.  Amplitude of reflex EMG and joint velocity in surprise and control landings to 0.45 m and 0.70 m
 Amplitude of reflex EMG (%)Joint velocity (deg s−1)
  1. Values are mean ± S.D. of determinations from 10 subjects for the gastrocnemius (G), soleus (S), tibialis anterior (TA), rectus femoris (RF) and biceps femoris muscles (BF). The mean of the highest 10 points in the period 35–80 ms (G, S, RF, BF) and 0–50 ms (TA) after contact with the ground was used to determine EMG amplitude for individual subjects. Prior to calculation of the mean across subjects, all of the absolute EMG amplitudes from each subject were normalised with respect to the amplitude in the 80 ms prior to touchdown in the 0.45 m landing; all data are expressed as a percentage of this value. Velocities are the mean velocities for the whole unidirectional joint rotation following impact. All statistical tests were carried out on the raw, not the normalised data. *Statistically significant difference (P < 0.05, two-tailed t test for paired values) between the results from control landings to 0.70 m and surprise landings to 0.70 m. Note that ankle rotation velocity was less in the surprise condition.

 0.70 m landing421.9 ± 243.01427.8 ± 1431.9973.7 ± 1144.71367.8 ± 806.4341.1 ± 214.1899.5 ± 361.9610.8 ± 102.9
 0.70 m landing914.9 ± 533.21431.9 ± 1417.81063.9 ± 1128.22638.3 ± 1616.2966.3 ± 974.8796.4 ± 281.7645.8 ± 374.1
 0.45 m landing186.1 ± 157.5483.3 ± 245.7647.7 ± 606.91063.6 ± 692.6286.2 ± 282.7571.4 ± 204.6490.6 ± 123.0

It can be seen from Fig. 5 that the rectus femoris also exhibited a larger post-landing EMG response in the surprise condition. Table 2 shows the data averaged across subjects. The EMG amplitude of the rectus femoris muscle post-impact was 1.9 times larger than that recorded during a normal landing; for the biceps femoris it was 2.8 times larger. The post-landing EMG signals taken from the soleus and tibialis anterior muscles did not show a difference in amplitude between the conditions (Table 2).


Figure 5. Protocol 1: 85 ms between floors. Rectus femoris muscle EMG traces for 0.70 m and surprise landings

Average of the data from 10 subjects. The upper trace is the 0.70 m landing and the lower, inverted trace is the surprise landing. All traces are aligned to the point of ground contact at 0.70 m, as indicated by the cursor.

Download figure to PowerPoint

Joint velocities following impact.

Our initial hypothesis to explain the enhancement of muscle responses in the surprise landing was to assume that there was a stronger stimulus to the muscle spindles in the surprise landing. However, despite the large differences in post-landing EMG amplitude between the surprise and the normal landings, joint excursions post-landing were not statistically different. There was also no significant difference between average velocities between the two conditions (Table 2). Joint velocity, however, is not invariant under other conditions. When the total fall distance is less, joint rotation velocity is reduced, as the data for falls to a solid floor at 0.45 m show (Table 2). To summarise, EMG recordings showed large and statistically significant differences between surprise and normal conditions, whereas joint velocities showed no such differences (Table 2).

We subjected this result to further scrutiny. The velocity value taken from each subject in order to compile Table 2 was the average velocity over the whole period of rapid joint rotation following impact. It could reasonably be argued that the first part of the joint rotation is the most crucial in initiating early spinal reflex responses. We therefore also analysed joint velocity over the first 10 ms of joint rotation. This analysis gave the same result: no significant difference in joint rotations between surprise and normal conditions, but a significant difference (P < 0.001 two-tailed t test for paired values) between control falls to solid floors at 0.45 m and 0.70 m.

These results are illustrated graphically in Fig. 6, in which the amplitude of the EMG following solid floor impact is plotted against joint velocity. Data from solid floor landings to 0.45 m and 0.70 m are shown together with data for landings to 0.70 m with a false floor en route. It is clear that in the surprise landing, the EMG amplitude for a given velocity was much larger than expected for the gastrocnemius and rectus femoris muscles, but not for the soleus.


Figure 6. Reflex amplitude plotted against velocity of joint rotation

Data from the soleus (A), gastrocnemius (B) and rectus femoris muscles (C). On each graph the group data for each of the three conditions of 0.45 m landing (▪), 0.70 m landing (•) and surprise landing (▴) are plotted. Joint velocity was measured as the average velocity in the first 10 ms of joint rotation after ground contact. Reflex amplitude was normalised before averaging. Data are the means ±s.e.m. for determinations from 10 subjects.

Download figure to PowerPoint

In summary, in a fall to a solid floor at 0.70 m, passage through a false floor en route substantially enhanced the amplitude of the post-landing EMG response in the gastrocnemius, rectus femoris and biceps femoris muscles.

It is possible that passage through the false floor might provide a signal that alters the amplitude of the pre-programmed and/or reflex activity. If this were so, then the flight time between the false and real floor would provide the time interval within which the afferent and efferent pathways of this signal could be traversed. Therefore, in protocols 2 and 3, the time interval between floors was reduced to see if the amplitude enhancement was extinguished. This was accomplished by reducing the distance between the false and solid floors, as shown in Fig. 1.

Protocols 2 and 3: falls with 30 ms between floors

In order to preserve naivety about the false floor, protocol 2 was carried out with a new set of 10 subjects. The false floor was 0.64 m below the take-off stage (Fig. 1). The solid floor was 0.70 m below in both normal and surprise landings. With the false floor at 0.64 m, the flight time between floors was reduced from the 87.8 ms recorded in protocol 1 to 27.4 ± 2.3 ms in protocol 2. Under these conditions there was no significant difference between the post-landing EMG amplitudes measured in the normal and surprise falls (protocol 2, Table 1). It is important to note that the mechanical conditions of the solid-floor landing were identical to those of protocol 1. In particular, note that the total fall distance in the surprise condition (0.70 m) was the same in protocols 1 and 2. The mechanical conditions at impact were identical, yet protocol 1 produced the enhancement and protocol 2 did not.

In both protocols 1 and 2, the solid floor was located 0.70 m below the take-off stage. However, in protocol 2 the false floor was located at 0.64 m from the take-off point, whereas in protocol 1, the false floor was at 0.45 m. It could be suggested that the difference in outcome between protocol 1 and protocol 2 was due to the fact that in the surprise condition the subjects viewed false floors at different depths. To control for this, a further group of 10 subjects was recruited for protocol 3 (Fig. 1). These subjects experienced a false floor at the same depth as in protocol 1 (0.45 m) with a flight time of 30.3 ± 2.9 ms between floors, which was similar to that in protocol 2. The solid floor was located at 0.50 m below the take-off point. Protocol 3 gave the same outcome as that of protocol 2, as shown by the results for the gastrocnemius and rectus femoris muscles (Table 1 and Fig. 7). The post-impact EMG amplitude was the same under both conditions.


Figure 7. Protocol 3: 30 ms between floors. Rectus femoris muscle EMG traces for 0.50 m and surprise landings

The upper panel shows data from a single subject. The lower panel shows the average of the data from 10 subjects. In both records, the upper trace is the 0.50 m landing and the lower, inverted trace is the surprise landing (false floor at 0.45 m, solid floor at 0.50 m). All traces are aligned at the moment of true contact with the ground at 0.50 m, which is indicated by the continuous cursor.

Download figure to PowerPoint

In both protocols 2 and 3, with ≈30 ms between floors, the post-landing EMG was not potentiated by passing through the false floor. Thus, it is the reduced time between floors in protocols 2 and 3 that extinguished the enhancement of the EMG amplitudes seen in protocol 1.

Protocol 4: falls with 50 ms between floors

For the fourth protocol (Fig. 1), yet another group of 10 subjects was recruited. The question asked was: if the time between floors is critical in influencing the presence and absence of EMG enhancement, then what is the minimum time required for the processes involved to become activated? Having reviewed the theoretical latencies of various possible pathways (see Discussion) we set the time between floors at ≈50 ms (actual timing across subjects 51.8 ± 3.7 ms). With this delay between floors, the reflex enhancement reappeared. Reflex responses of 2.9, 1.6 and 2.4 times greater amplitude compared to those in a normal landing were recorded under surprise conditions in the gastrocnemius, rectus femoris and biceps femoris muscles, respectively (Table 1).


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References

The experiment had four protocols, which are summarised in Fig. 1. Each protocol had the following three main conditions: the subject fell (1) to a solid floor, passing through a false floor en route, (2) directly to the solid floor and (3) to a solid floor at the previous false floor distance. For the gastrocnemius, rectus femoris and biceps femoris muscles, passing through a false floor en route doubled the amplitude of the post-impact muscle activity compared to that found in a direct fall, despite the fact that joint rotations and mechanical conditions were the same.

No gain enhancement in soleus and tibialis muscle activity was observed in any of the conditions. We have no definitive explanation for this. Perhaps the reflex responses of these muscles are saturated at a fall height of 0.70 m and cannot be further increased. In relation to this point, it should be noted that the soleus consistently showed larger reflexes than the gastrocnemius under all of the experimental conditions. Duysens et al. (1991) described separate control of the gastrocnemius and soleus muscles in the gait cycle, and Moritani et al. (1990) showed that modulation of the gastrocnemius H reflex over the hopping cycle was much greater than that for the soleus. Note that the soleus and tibialis muscles work only across the ankle joint, while the other three muscles, which exhibited an enhanced response, work across the knee. The rotations of the knee in both normal and surprise landings to 0.7 m were virtually identical (see Table 2).

The enhancement of the response found in the gastrocnemius, rectus femoris and biceps femoris muscles disappeared if the time between the false and real floor was reduced from 85 ms to 30 ms, but reappeared when 50 ms was allowed between floors.

How is this enhancement of the response brought about?

The enhancement of the post-landing muscle activity triggered by false floor contact might directly modify an ongoing motor programme. If this were true, the responses should be time-locked to the false floor, not the solid floor contact. However, all of our data showed that these responses were time-locked to the moment of solid floor impact, irrespective of the time between floors.

It is possible that the motor programme is constructed so that the gain of the reflex always increases (in both surprise and control) after the moment of expected touchdown. Hence, the longer the time between floors the greater the gain enhancement. There is a suggestion of this in the trend of the surprise/control EMG ratios for the rectus femoris muscle shown in Fig. 1, but this is not true for the other muscles. However, other evidence suggests that an unexpected event may change the reflex gain. Data from unexpected falls (Reschke et al. 1986, their Fig. 4) show the H-reflex amplitude increasing with flight time. Reschke et al. (1986) show that an effect on reflex gain was only observed after ≈40 ms of the unexpected initiation of a fall. Our falls do not become unexpected until after the false floor is reached. This fits rather well with our results, as we found no change in the reflex amplitude until > 30 ms post-false-floor contact.

Startle responses occur in leg muscles ≈80 ms after a sudden head drop (Bisdorff et al. 1995) or release into a fall (Greenwood & Hopkins, 1976). We could find no evidence for responses at these latencies following false-floor contact (see Fig. 3). A possible mechanism might be an increase in stretch-reflex gain by increased gamma-motorneurone activity. An afferent signal from the feet or legs triggered by false floor contact will take a minimum of 30 ms to reach the brain. If, as a result of this, an efferent signal is sent down to the gamma-motorneurones, it will reach the muscle spindles after a further 30 ms. This gives a total minimum time for the loop of 60 ms for subjects of average height. If the flight time between floors is 85 ms (protocol 1), then there is plenty of time for this loop to be traversed before the subject impacts with the solid floor. Thus, the impact occurs with the muscle spindles already set to a higher sensitivity. If, however, the flight time is 50 ms (protocol 4), then there is not enough time for this mechanism to act. However, in protocol 4, the enhanced reflex response was still present. This suggests that increased gamma activation is not the mechanism. This view is strengthened by the fact that the false floor did not cause joint rotation.

In a more plausible mechanism, the descending control signal might alter reflex gain at the level of spinal neurones rather than increasing gamma firing and spindle gain. As discussed above, the descending signal would be triggered by ascending input due to the encounter with the false platform. Reflex gain will only be enhanced if the muscle spindle signal arising from the solid floor contact (signal SF) reaches the spinal cord after the descending signal. The afferent signal evoked by false-floor contact will take a minimum of 30 ms to arrive at the brain. Following this, the gain-modifying signal passes down from the brain to the spinal cord. The minimum time for this pathway would be 15 ms, making a total time of 45 ms. However, we must subtract the 15 ms taken for the muscle spindle response (SF) to reach the spinal cord, as this will travel up as the descending signal travels down. This gives a total theoretical minimum of 30 ms required between floors to produce an enhanced reflex response. Thus, 85 ms between floors (protocol 1) will allow the enhanced response, whereas 30 ms is unlikely to permit it. However, 50 ms between floors will also allow it. Our experimental evidence fits the constraints imposed by this mechanism very well.

The above explanation requires a positive trigger signal initiated by false floor contact. Such a positive signal could not arise from stretch receptors, as the joints did not rotate with false floor contact. However, stimulation might arise from cutaneous receptors in the feet. Activation of cutaneous afferents is known to reduce important gain-setting mechanisms such as Ia presynaptic inhibition in both the legs (Iles, 1996) and the hand (Aimonetti et al. 2000).

Control of stretch-reflex gain by the fastest peripheral (e.g. cutaneous) afferents activated by false floor contact and acting directly at the spinal level would only require 20 ms (the time taken for an afferent signal to reach the spinal cord from the foot) between floors to change the gain. If this were the mechanism, all of our conditions would show the enhanced stretch reflex, which they do not. Slower cutaneous pathways with afferent conduction times between 30 and 50 ms would fit our data. Also, a long-latency signal from cutaneous receptors travelling via supraspinal centres could be allowed by the timing constraints of our protocol. However, if no expectation of a change in landing circumstances is involved, then one would expect any slight cutaneous stimulation from false floor contact to be evaluated as very minor compared to normal stimulation from hard floor impact. It is difficult to see why it would be functional, as cutaneous sensory input would be expected at this time and would not constitute an abnormal event likely to trigger a gain change. In contrast, if expectation of impact at 0.45 m has pre-set the CNS, then non-occurrence of a strong, hard-floor-like cutaneous stimulation might be a very potent signal.

This suggests the possibility that the trigger is not an event, but the absence of an expected event (Johansson, 1998). Workers from Von Helmholtz (1867) through Van Holst (1954) and Johansson & Cole (1992), to Wolpert (1997) have shown that the CNS does not just simply produce pre-programmed motor activity or respond to sensory input. Rather, prior to the performance of any action, an expectation of the sensory results of that action is prepared by means of a forward model (Lemon, 1997; Wolpert, 1997). This expectation is then compared with the actual sensory outcome of the action (Coulter, 1974).

We have shown that the brain can predict very accurately the moment of impact in landing movements (Santello & McDonagh, 1998). Thus, the brain will expect strong proprioceptive signals at the predicted time of impact with the seen floor. This expectation is unfulfilled as the subject passes through the false floor without impact. We suggest that what constitutes the signal is the absence of this burst from proprioceptors (particularly spindles) at this expected time.

There are a number of examples in the literature of positive responses to the nonfulfilment of an expected event. Hiebert et al. (1994) found that when cats stepped unexpectedly into a hole, the foot was lifted out faster than could be achieved by conscious control, and also that the speed of flexion depended on prior experience and expectancy. Response to a negative event has been noted in work on grip adjustment by Johansson & Cole (1992, their Fig. 2b). They found that if an object was heavier than expected, there was an increase in the grip force just after the load force reached the level of the predicted lift-off. This occurred despite the fact that no sensory event confirming lift-off was evoked. The unexpected event in our paradigm is the absence of solid-floor impact at the predicted time.

In conclusion, when a subject passes through a false floor, the stretch reflexes evoked from impact with a lower, solid floor are twice as large as normal. This increase in stretch reflex gain is caused by a signal related to false floor contact that most probably acts via supraspinal centres. Future work should test the hypothesis that this ‘signal’ is the absence of a predicted sensory burst at this time.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. References
  • Aimonetti, J., Vedel, J., Schmied, A. & Pagni, S. (2000). Mechanical cutaneous stimulation alters Ia presynaptic inhibition in human wrist extensor muscles: a single motor unit study. Journal of Physiology 522, 137145.
  • Barnes, W. J. P. (1977). Proprioceptive influences on motor output during walking in the crayfish. Journal of Physiology (Paris) 73, 543564.
  • Bisdorff, A. R., Bronstein, A. M., Gresty, M. A., Wolsley, C., Davies, A. & Young, A. (1995). EMG-responses to sudden onset free fall. Acta Otolaryngologica: Supplementum 520, 347349.
  • Burke, D, Gracies, J. M., Meunier, S. & Pierrot-Deseilligny, E. (1992). Changes in presynaptic inhibition of afferents to propriospinal-like neurons in man during voluntary contractions. Journal of Physiology 449, 673687.
  • Coulter, J. D. (1974). Sensory transmission through lemniscal pathway during voluntary movement in the cat. Journal of Neurophysiology 37, 831845.
  • Dietz, V. & Noth, J. (1981). Significance of spinal stretch reflexes in human locomotion. In Advances in Physiological Science, Proceedings of the 28th International Congress of Physiological Sciences, Budapest 1980, vol. 1, Regulatory Functions of the CNS. Principles of Motion and Organization, ed. Szentágothai, J., Palkovits, M. & Hámori, J., pp 169177. Pergamon Press, Budapest
  • Duncan, A. & McDonagh, M. J. N. (2000). Stretch reflex distinguished from pre-programmed muscle activations following landing impacts in man. Journal of Physiology 526, 457468.
  • Duysens, J., Tax, A. A., Van Der Doelen, B., Trippel, M. & Dietz, V. (1991). Selective activation of human soleus or gastrocnemius during walking and running. Experimental Brain Research 87, 193204.
  • Gandevia, S. C., Wilson, L. R., Inglis, J. T. & Burke, D. (1997). Mental rehearsal of motor tasks recruits alpha-motorneurons but fails to recruit human fusimotor neurons selectively. Journal of Physiology 505, 259266.
  • Greenwood, R. & Hopkins, A. (1976). Landing from an unexpected step fall and a voluntary step. Brain 99, 375386.
  • Grillner, S. (1975). Locomotion in vertebrates, central mechanisms and reflex interaction. Physiological Reviews 55, 247304.
  • Hiebert, G. W., Gorassini, M. A., Jiang, W., Prochazka, A. & Pearson, K. G. (1994). Corrective responses to loss of ground support during walking II. Comparison of intact and chronic spinal cats. Journal of Neurophysiology 71, 611622.
  • Hultborn, H. (2001). State-dependent modulation of sensory feedback. Journal of Physiology 533, 513.
  • Iles, J. F. (1996). Evidence for cutaneous and corticospinal modulation of presynaptic inhibition of Ia afferents from the human lower limb. Journal of Physiology 491, 197207.
  • Johansson, R. S. (1998). Sensory input and control of grip. Novartis Foundation Symposium 218, 4563.
  • Johansson, R. S. & Cole, K. J. (1992). Sensory motor co-ordination during grasping and manipulative actions. Current Opinion in Neurobiology 2, 815823.
  • Lemon, R. N. (1997). Sensory gating during movement, how and why? Journal of Physiology 504.P, 1S
  • Moritani. T., Oddsson, L. & Thorstensson, A. (1990). Differences in modulation of gastrocnemius and soleus H-reflexes during hopping in man. Acta Physiologica Scandinavica 138, 575576.
  • Nielsen, J., Kagamihara, Y., Crone, C. & Hultborn, H. (1992). Central facilitation of Ia inhibition during tonic ankle dorsiflexion revealed after blockade of peripheral feedback. Experimental Brain Research 88, 651656.
  • Prochazka, A. (1989). Sensorimotor gain control, a basic strategy of motor systems? Progress in Neurobiology 33, 281307.
  • Santello, M. & McDonagh, M. J. N. (1998). The control of timing and amplitude of EMG activity in landing movements in humans. Experimental Physiology 83, 857874.
  • Santello, M., McDonagh, M. J. N. & Challis, J. H. (2001). Visual and non-visual control of landing movements in humans. Journal of Physiology 537, 313327.
  • Von Helmholtz, H. (1867). Handbuch der Physiologishen Optik. 1st edn HamburgVoss.
  • Wolpert, D. M. (1997). Computational approaches to motor control. Trends in Cognitive Sciences 1, 209216.
  • Zehr, E. P & Stein, R. B. (1999). What functions do reflexes serve during human locomotion? Progress in Neurobiology 58, 185205.