Load Rather Than Length Sensitive Feedback Contributes to Soleus Muscle Activity During Human Treadmill Walking

Richard af Klint, Nazarena Mazzaro, Jens Bo Nielsen, Thomas Sinkjaer, Michael J. Grey


Walking requires a constant adaptation of locomotor output from sensory afferent feedback mechanisms to ensure efficient and stable gait. We investigated the nature of the sensory afferent feedback contribution to the soleus motoneuronal drive and to the corrective stretch reflex by manipulating body load and ankle joint angle. The volunteers walked on a treadmill (∼3.6 km/h) connected to a body weight support (BWS) system. To manipulate the load sensitive afferents the level of BWS was switched between 5 and 30% of body weight. The effect of transient changes in BWS on the soleus stretch reflex was measured by presenting dorsiflexion perturbations (∼5°, 360–400°/s) in mid and late stances. Short (SLRs) and medium latency reflexes (MLRs) were quantified in a 15 ms analysis window. The MLR decreased with decreased loading (P = 0.045), but no significant difference was observed for the SLR (P = 0.13). Similarly, the effect of the BWS was measured on the unload response, i.e., the depression in soleus activity following a plantar-flexion perturbation (∼5.6°, 203–247°/s), quantified over a 50 ms analysis window. The unload response decreased with decreased load (P > 0.001), but was not significantly affected (P = 0.45) by tizanidine induced depression of the MLR (P = 0.039, n = 6). Since tizanidine is believed to depress the group II afferent pathway, these results are consistent with the idea that force-related afferent feedback contributes both to the background locomotor activity and to the medium latency stretch reflex. In contrast, length-related afferent feedback may contribute to only the medium latency stretch reflex.


Information about current limb mechanics is relayed to the spinal cord and is used not only to adapt locomotor muscle activity throughout the stance phase of the step cycle (e.g., af Klint et al. 2008) but also to generate corrective responses following unexpected deviations in limb trajectory. The contribution of positive load and/or length feedback, sensed by Golgi tendon organs (GTOs), muscle spindles, and both joint and cutaneous receptors, is thought to be an important feedback signal for the motor control of walking (Duysens et al. 2000).

In the cat, load feedback influences phase transitions and reinforces locomotor muscle activity (Donelan and Pearson 2004; Donelan et al. 2009; Duysens and Pearson 1980; Pearson 2004). Also in humans, the magnitude of the ground reaction forces influence the locomotor activity in infants (Yang et al. 1998) and in healthy adults (Ivanenko et al. 2002). It also modulates cutaneous reflex responses (e.g., Bastiaanse et al. 2000; Nakajima et al. 2008) and stumbling reactions (Fouad et al. 2001). Load feedback has also been suggested to contribute both to phase transitions in human walking (Stephens and Yang 1999; Yang et al. 1998) and to the generation of locomotor activity (e.g., Bachmann et al. 2008; Dietz et al. 1992; Faist et al. 2006; Grey et al. 2004, 2007; Sinkjaer et al. 2000). Thus load feedback may play an intricate part in the control of human locomotion.

Reduced weight bearing with a body weight support (BWS) system effectively reduces the ground reaction forces on the supporting limbs during walking (Ivanenko et al. 2002) and thus enables the investigation of load feedback. Sustained BWS (i.e., a constant BWS over a longer period in time) decreases the magnitude of the locomotor muscle activity (Harkema et al. 1997; Stephens and Yang 1999). However, it should be expected that the CNS adapts to changes in proprioceptive input during sustained BWS walking, as is the case for stumbling reactions during body loading (Fouad et al. 2001). Therefore the implications of the muscle activation changes need to be interpreted carefully. Decreased locomotor activity has also been observed with transient increases in BWS and these findings support the idea that load feedback contributes to the background locomotor muscle activity (Bachmann et al. 2008; Stephens and Yang 1999). However, the design of these studies was such that potential feedback from spindle and GTO related afferents could not be differentiated from one another because changes in BWS produce changes not only in ankle kinematics but also in loading of the supporting leg. This limitation can be overcome by using techniques that allow the components of the afferent mediated muscle activity to be differentiated as the transient changes in BWS are presented. One such estimate of the afferent generated muscle activity is achieved by imposing a rapid plantar flexion to unload the muscle–tendon complex of the soleus and thus removing the afferent feedback (Grey et al. 2004, 2007; Sinkjaer et al. 2000).

The response to rapid plantar-flexion perturbations, termed “the unload response” by Sinkjaer et al. (2000), was not influenced by an anesthetic block of muscle and cutaneous afferents from the foot and ankle (Grey et al. 2004), ischemic depression of the largest group I afferents, or a common peroneal nerve block (Sinkjaer et al. 2000). In summary, this implies that neither autogenic group Ia activity, cutaneous and muscle afferents from the foot and ankle, nor reciprocal inhibition from the tibialis anterior (TA) contributes to the reduced soleus electromyograph (EMG) following the plantar-flexion perturbations. However, the extent to which length and force sensitive afferents contribute to this unload response, and thus the background EMG, is still debatable. If the changes in locomotor activity induced with changes in transient BWS (Bachmann et al. 2008; Grey et al. 2002; Stephens and Yang 1999) are indeed generated by proprioceptive afferent feedback, the unload response should also change because it reflects amount of muscle activity generated by afferent feedback. Furthermore, determining which components of the afferent feedback contribute to the unload response may show the afferent origin of the proprioceptive feedback to the background locomotor activity.

In contrast to the afferent-mediated contribution to the background locomotor activity, muscle activity due to corrective responses are stereotypical short-lasting reactions to a specific perturbation. These responses may be important in the signaling of potential balance disturbances and, to enable a fully controlled reaction, it would be beneficial to signal the system's entire mechanical state. One such corrective response is the soleus stretch reflex, which may be investigated during walking by imposing a rapid dorsiflexion perturbation (Sinkjaer et al. 1996; Yang et al. 1991). The spinally mediated components of the soleus stretch reflex response are known to be composed of a velocity sensitive (group Ia) mediated short latency response (Matthews 1991; Taylor et al. 1985) and mostly length sensitive (group II) mediated medium latency response (Grey et al. 2001; Schieppati and Nardone 1997). Surprisingly, a force sensitive (group Ib) contribution to the stretch reflex has not been observed. Grey et al. (2002) hypothesized that load sensitive afferents should contribute to the reflex response during walking. They investigated stretch reflex responses during normal treadmill walking and when subjects walked with sustained 30% BWS and 30% body weight loading and, perhaps surprisingly, found no evidence for a load feedback contribution to the short and medium latency components of the stretch reflex response. However, a flaw in this design may have been the use of sustained body weight support/loading. It is most likely that the CNS adapted to the modified afferent feedback. Their results might have been different had they used transient changes in body weight support/loading.

The present study was designed to differentiate the contribution of spindle and GTO feedback to the background locomotor muscle activity and to the stretch reflex response. Mid and late-stance phase ankle perturbations were elicited shortly after a transient change in BWS. Soleus stretch reflex responses were evoked with rapid dorsiflexion perturbations. The afferent feedback contribution to background locomotor muscle activity was investigated by removing the feedback with a plantar-flexion perturbation that rapidly unloaded the soleus muscle. We hypothesized that if load feedback contributes to the stretch reflex response then walking with transiently increased body weight support (i.e., decreased load on the supporting limb) would decrease the magnitude of the stretch reflex. Similarly, we hypothesized that during a transient increase in body weight support the unload response would decrease as the load on the limb decreases. Furthermore, we hypothesized that if the length sensitive afferents contribute to the background locomotor activity, and thus the unload response, then the magnitude of the unload response would be significantly decreased as the transmission in the polysynaptic group II afferent pathway was depressed using tizanidine hydrochloride, an α2-receptor agonist shown to specifically suppress this pathway in the cat (Bras et al. 1990; Hammar and Jankowska 2003; Jankowska et al. 1998, 2000).


Twenty-four able-bodied volunteers (20 male, 4 female; mean age: 27 yr, range: 21–41 yr; mean weight: 74 kg, range: 58–104 kg) with no known history of neuromuscular disorders participated in this study. Seventeen of these subjects participated in an experiment involving transient body weight changes. Eight subjects participated in an experiment involving the pharmaceutical depression of the group II afferent pathway using the α2-adrenergic agonist tizanidine hydrochloride. The study conformed to the Declaration of Helsinki and was approved by the local ethics committee (VN 99/100). All subjects provided informed written consent.

Apparatus and instrumentation

Subjects were instrumented with bipolar surface EMG electrodes (interelectrode distance 2 cm; NeuroLine 720, AMBU A/S, Denmark) on the soleus (SOL) and tibialis anterior (TA) muscles of the left leg. The EMG signals were amplified and band-pass filtered 10–1 kHz. The subjects walked on a treadmill (Woodway) while supported by a custom-built body weight support (BWS) system (Dr. J. Fung, Jewish Rehabilitation Hospital Research Centre, Laval, Quebec, Canada) running on a low friction rail centered over the treadmill (see Fig. 1A). Briefly, the BWS system uses a pneumatic cylinder with a mechanical lever arm that creates a constant vertical lifting force over its 1 m range of motion. The level of BWS (i.e., the lifting force imposed on the subject) was computer controlled by electronic pneumatic valves that enabled rapid changes of pressure in the system. The degree of BWS is given as the percentage of the subject's body weight lifted by the system, i.e., subject's weight: 70 kg; 30% BWS implies a lifting force of 21 kg. A strain gauge placed in series with the BWS system measured the lifting force so that the controller was able to reliably switch between two force levels at any desired time. Subjects were connected to the BWS by a harness (Biodex unweighting support vest 945-476; Biodex Medical Systems, Shirley, NY), with the leg straps removed to eliminate pulling on the legs. The lack of leg straps implied that the harness needed to be tightly secured around the chest and thus only male subjects were recruited for these experiments involving BWS.

Fig. 1.

A: the subjects walked on a treadmill (I) at a self-selected speed, while they were supported by a constant force body weight support (BWS) system (II). A strain gauge (III) in series with the BWS system measured the lifting force as it was changed between 5 and 30% of body weight. The subjects were also fitted with a functional joint (IV) connected by Bowden wires to a computer controlled servo motor (V) that enabled fast rotation perturbations of the ankle. B: BWS level, ankle excursion and soleus EMG activity for a typical subject. The ensemble average records (n > 30) are shown from the onset of the transition from 5 to 30% BWS (dark thick trace) and from 30 to 5% BWS (light thick trace) as measured by the percentage of body weight being lifted by the system. Vertical lines mark the initiation of the BWS change and the left heel strike of the first and second steps. The unload response as well as the stretch reflex were investigated during the second step for both BWS conditions. An example of a late stance plantar-flexion perturbation is shown in thin traces (continuous and dotted for 5 and 30%, respectively). For a more careful examination of the perturbations see Fig. 2.

Dorsiflexion and plantar-flexion perturbations were imposed by a custom-built robotic actuator (Andersen and Sinkjaer 2003). Essentially, the actuator consists of a functional joint connected with flexible Bowden cables to a powerful servo motor. The functional joint was mounted to the subject's left leg with the rotational axis aligned with the ankle pivot. Individualized foot and a calf braces transferred the joint rotation to the subject ankle and were secured with high strength duct tape and Velcro straps for a tight fit. The rotation of the joint was controlled through a personal computer and followed the ankle movement during unperturbed trials or imposed a dorsiflexion or a plantar flexion at any desired time in the step cycle. For detailed information on the functional joint see Andersen and Sinkjaer (2003). Ankle rotation was recorded through an embedded optical encoder in the functional joint, whereas knee excursion was recorded through a surface mounted goniometer (SG150; Biometrics, Gwent, UK). Data acquisition triggering and foot contact information were obtained by force sensitive resistors (Interlink FSR, LuSense) placed under the heels of both feet. Data were sampled at 2 kHz and saved for later processing.


Seventeen subjects walked at a comfortable self-selected speed of about 3.6 m/s for an accommodation period of close to 5 min prior to data collection. During this accommodation period the subjects were supported with a constant BWS of 5%. To check for the effect of transiently changing the BWS on the kinematics and electromyographic activity, the BWS system was programmed to switch the BWS between 5 and 30% every seven to nine steps so that at least seven consecutive steps with the left leg in each BWS condition were acquired. Data were acquired until ≥20 records with seven consecutive steps in each BWS condition were stored.

Next, changes in BWS and rapid plantar flexions were presented pseudorandomly for 16 subjects. The BWS system was triggered by the right heel switch and programmed to change BWS between 5 and 30% every 3 to 5 steps (see Fig. 1B). The functional joint was programmed to plantar flex (5.5°; 250°/s) and hold for 150 ms. The changes in BWS and plantar-flexion perturbations were presented pseudorandomly so that the ratio between control and perturbed trials was 2:1. The plantar flexions were presented mid and late stance (30 and 50% of stance, respectively), the second step with the left leg in the new BWS condition, to avoid any accommodation to the new BWS condition. Trials were collected until ≥30 perturbed steps in each BWS condition were collected. Additionally, in the same pseudorandom manner as for the plantar flexions, rapid dorsiflexions (5°, 400°/s, hold 150 ms) were presented at the same instance in stance in the same 16 subjects. One subject developed blisters and was unable to finish the experiment. This data set was excluded from the analysis. None of the remaining 15 subjects reported discomfort.

Tizanidine suppression of group II afferent pathway

It has previously been hypothesized that the unload response induced by rapid plantar-flexion perturbations in mid stance may be highly influenced by the cessation of group II length sensitive afferents (Sinkjaer et al. 2000). To evaluate the contribution of group II afferents on the unload response, a fast acting drug tizanidine hydrochloride (Sirdalud) was used in combination with the rapid plantar-flexion perturbations. Tizanidine is a centrally acting antispastic drug that blocks the α2-adrenergic receptors. The effect of tizanidine has been investigated by recording spinal focal field potentials resulting from electrical stimulation of the quadriceps nerve in the cat (Bras et al. 1989, 1990; Jankowska et al. 1998; Skoog 1996). Although tizanidine was shown to strongly depress the potentials from group II afferents mediated through di- and trisynaptic pathways, no effect was shown on group I afferents (Bras et al. 1990; Hammar and Jankowska 2003; Jankowska et al. 1998, 2000).

Evidence for the effect of tizanidine in humans is based on indirect evidence from its depressive effect on the MLR response of the stretch reflex. The human MLR has been investigated using several methods both in standing (e.g., Chaix et al. 1997; Marque et al. 1996; Schieppati and Nardone 1997) and in walking (e.g., Grey et al. 2001). Evidence that the MLR is mediated through polysynaptic group II afferent connections can be summarized as follows: 1) the latency of the MLR is more affected by nerve cooling than the SLR, indicating slower velocity afferents (Grey et al. 2001; Schieppati and Nardone 1997); 2) the MLR is less affected by the velocity of the muscle stretch than the SLR during walking (Grey et al. 2001); 3) ischemic depression of the group I afferents during walking significantly depresses SLR magnitude, whereas MLR remains unchanged (Grey et al. 2001); 4) electrical stimulation of the quadriceps muscle reveals a significantly higher stimulation threshold for MLR than what would be expected from group I afferents (Marque et al. 1996); and 5) the latency of the earliest response of stimulation of the group II afferents is comparable to a disynaptic coupling (Chaix et al. 1997; Marque et al. 1996). In summary, there is good evidence to suggest that the MLR is an indicator of the autogenic oligosynaptic group II afferent connections to the motor neuron pool. Thus it is a marker for the effect of tizanidine on the interneurons mediating the di- and trisynaptic feedback of group II afferents. For a more thorough comparison of the human and cat reflex interneuronal systems see Jankowska and Hammar (2002).

Eight subjects (three male, five female; mean age: 30 yr, range: 24–41 yr; mean weight: 71 kg, range: 61–76 kg) were given an oral dose of tizanidine hydrochloride (150 μg · kg−1, similar to Grey et al. 2001). The depression of the group II afferent pathway was assessed by monitoring the Sol MLR response, which is known to be mostly mediated by group II afferents (Grey et al. 2001; Schieppati and Nardone 1997). The group II afferent pathway was considered depressed when the amplitude of the MLR response during standing had decreased by >20% with respect to the initial values that occurred about 1 h after administration of the drug. Plantar-flexion and dorsiflexion perturbations were investigated in late stance at full body weight prior to and after the group II afferent pathway was depressed as outlined earlier. The perturbations were pseudorandomly presented so that soleus, the unload response, and stretch reflex response were measured simultaneously.

Data analysis

Data analysis was carried out off-line. Individual trials were removed on visual inspection of the ankle trajectory significantly deviating from the ensemble average. The EMG records were rectified and low-pass filtered at 40 Hz (first-order Butterworth filter) to extract an amplitude envelope. Individual records of each trial were ensemble averaged to create a single set of records per condition and subject. The EMG activities were not normalized because intersubject variability was low; the coefficient of variation 17%.

Each trial in the data records of the BWS changes without ankle perturbations encompassed seven to nine consecutive steps after the BWS was changed. The data were parsed into single steps from the first left heel contact to the next. Each step was then classified according to the BWS condition and step number following the BWS change. To compare the effect of the two body weight support conditions on the soleus muscle activity, the area under the curve of the ensemble averages was calculated and divided by the duration of the stance phase (Grey et al. 2002; Stephens and Yang 1999).

The rapid plantar-flexion perturbations produced unload responses (i.e., reduction in EMG) in the soleus locomotor activity (see Fig. 1B, second step). The onset latency of the unload responses was defined as the first major deflection from the ensemble-average control records for the specific BWS condition within a window of 30–80 ms immediately after the onset of the plantar-flexion perturbation. Typically, the unload response was between 50 and 150 ms long; thus a 50 ms wide window of analysis was chosen starting at the onset of the depression. The unload response was quantified as the area between the ensemble-average EMG of the control and perturbed trials of each BWS condition (Fig. 2, bottom row; see shaded areas).

Fig. 2.

Ensemble averaged (n > 30) soleus electromyographic (EMG) activity and ankle excursions for a typical subject are presented for 5 and 30% BWS (see column headings). Dorsiflexion perturbations (top row) and plantar-flexion perturbations (bottom row) are elicited in mid and late stances (dotted vertical lines). The reflex and unload responses (shaded areas) are calculated as the difference between the ensemble average of the control (thick gray line) and the perturbed trials (thin black traces).

The rapid dorsiflexion perturbations produced stretch reflex responses in all subjects. Because the long latency response is not present in all subjects, only the spinally mediated short latency response (SLR) and medium latency response (MLR) responses were analyzed. The SLR and MLR were quantified by calculating the difference in area under the curve between the control and perturbed ensemble-averaged soleus EMG over a 15 ms window centered on the peak of the SLR and MLR bursts, each of which was determined within windows placed 35–60 and 50–90 ms, respectively, after the onset of the perturbation (Fig. 2, top row; see shaded areas).


To check for an effect of changing the BWS a two-way repeated measures ANOVA (rmANOVA) with main factors BWS condition and step number (i.e., the number of step cycles after BWS change) was used on the normalized area under the curve of the stance-phase soleus muscle activity.

A two-way rmANOVA was used to determine the effect of BWS and time in stance (mid or late stance) on the unload response and ankle dorsiflexion. Similarly, two-way rmANOVAs were used with main factors BWS and time in stance to test for changes in reflex magnitude and onset latency. For all ANOVAs, Geisser–Greenhouse adjustments were made if the covariance matrix sphericity assumption was violated (denoted by GG following the F-test). When the result of an ANOVA test was statistically significant, Tukey–Kramer post hoc tests were used to determine whether there were any group differences. The velocity and amplitude of the plantar-flexion perturbations were not normally distributed; therefore these measures were analyzed using the Wilcoxon nonparametric signed-rank test with a Bonferroni adjustment. The differences in mid and late stance phase dorsiflexion amplitude and velocity for the BWS conditions were analyzed, using paired t-test with Bonferroni adjustments. All statistical tests were conducted using a significance level of P = 0.05. Group mean and SD values are reported as means (SD).


Increasing the amount of body weight support decreased the load on the subjects' supporting limb and decreased the leg muscle activity and shortened the stance phase. Figure 3A shows muscle activity in a typical subject for the second step of the left leg after the transition to the new BWS condition (i.e., the BWS was changed from 5 to 30% or from 30 to 5% two steps earlier). The duration of the stance phase shortened significantly with increased BWS [5%: 780 ms (SD 58); 30%: 770 ms (SD 58); F(1,16) = 26.68, GG, P < 0.001, Tukey–Kramer, P < 0.001]. However, no significant difference in stance phase duration was found between the seven consecutive steps with the same BWS condition [F(6,16) = 1.67, P = 0.13; Fig. 3B]. The stance phase soleus activity decreased with BWS [i.e., 28.5 μV (SD 6.2) at 30% BWS compared with 32.3 μV (SD 6.8) at 5% BWS]. A significant interaction effect was also observed between BWS condition and step number [F(6,16) = 4.49, GG, P < 0.001]. Post hoc Tukey–Kramer tests showed less soleus activity for each step at 30% BWS compared with each step at 5% BWS (P < 0.001). Tukey–Kramer tests also indicated that the first step at 5% BWS had greater muscle activity than did the fourth and seventh steps at 5% BWS (P = 0.008 and P = 0.006, respectively). No other comparisons were significantly different at the 0.05 level.

Fig. 3.

A: ensemble averages of ankle excursion and soleus (SOL) and tibialis anterior (TA) muscle activity for the second step at 5% and at 30% BWS in a typical subject (n = 20, thick and thin traces, respectively). Bars indicate stance and swing phase of the 5% BWS condition used to calculate the mean soleus activity. B: duration of stance and both mean soleus and tibialis anterior activity during stance for the first 7 steps after BWS change (n = 17); error bars indicate 1SD. BWS decreases the soleus and the TA activity, but no trend with respect to step after BWS change could be noted. The duration of stance is decreased in time when the subjects are supported by 30% body weight compared with 5%. However, no change in stance duration for consecutive steps at the same BWS condition was found.

The TA muscle showed a similar tendency as the soleus muscle with respect to the BWS. rmANOVA did not reveal any significant interaction effect [F(6,16) = 1.95, GG, P = 0.11] or significant differences in the main effect of step [F(1,16) = 1.39, GG, P = 0.23]. However, the activity measured during stance of the TA was significantly decreased as the load on the supporting limb was decreased [16.1 μV (SD 7.5) and 13.3 μV (SD 7.4) for 5 and 30% BWS, respectively; rmANOVA, F(1,16) = 23.87, P < 0.001].

In summary, extensor and flexor muscle activation was decreased with the increased BWS, in agreement with previous studies (Grey et al. 2002; Stephens and Yang 1999), and no significant trend in the muscle activation profiles was observed within the first seven steps, although some steps showed smaller activity than that of the first step in the 5% BWS condition.

Unload response

Because the change in soleus activity with respect to BWS could stem from several afferent and central pathways, the afferent contribution to the activity was evaluated by introducing rapid plantar-flexion perturbations in mid and late stances. A clear unload response, i.e., a depression in the soleus activity following unloading of the ankle extensors, was observed in the four conditions: mid and late in stance at the two levels of BWS in all but three subjects (for a typical example, see Fig. 2). These subjects lacked a well-defined unload response in the mid stance at 30% BWS condition and, so as not to risk overemphasizing any effect of the BWS in midstance, these data were excluded. Thus for the unload response 12 subjects were analyzed.

The velocity of the plantar-flexion perturbations ranged from 203 to 347°/s. The perturbation at 30% BWS was slightly, albeit significantly, faster than the perturbation at 5% BWS (median difference: 8.4 and 8.7°/s, Wilcoxon, P = 0.002 and P = 0.005 for mid and late stances, respectively). The amplitude of the perturbations did not significantly differ between the conditions (mean: 5.6°). Furthermore, the ankle position at the onset of the perturbation was examined to estimate the amount of group II afferent activity removed by the plantar-flexion perturbations. The interaction was significant [time in stance × BWS; rmANOVA F(1,11) = 16.91, P = 0.002] and post hoc Tukey–Kramer tests revealed that for the perturbations presented in late stance the ankle position was significantly more dorsiflexed at 30% BWS than at 5% BWS (difference −2.8°, P < 0.001). In contrast, the ankle rotation prior to the midstance perturbation was not significantly affected by the level of BWS (P > 0.10).

Because the rapid plantar-flexion perturbations will stretch the tibialis anterior, there is a potential for inducing a stretch reflex in this monoarticulate ankle flexor. The plantar flexions were adjusted for the soleus responses, which might explain why in the majority of cases no stretch reflex was elicited in the TA. However, the midstance plantar-flexion perturbations at 5% BWS elicited significantly increased TA activity in 5 of the 12 subjects. In two subjects there were distinct SLR and MLR peaks for the midstance 5% BWS condition. Only one subject showed increased TA activity in response to the plantar-flexion perturbations, at 30% BWS, preventing a quantitative comparison of these responses.

The response to the plantar-flexion perturbation was much more consistent in the soleus muscle. The plantar-flexion perturbations elicited unload responses (i.e., depression in the soleus following the rapid plantar flexion) at a latency of 62 ms (SD 10) to the perturbation and no significant differences were found between the groups. The two-way rmANOVA on the magnitude of the response revealed significance for the two main factors, BWS condition [F(1,11) = 34.23, P < 0.001] and time in stance [F(1,11) = 13.0, P = 0.004], but no significant interactions [F(1,11) = 4.45, P = 0.059]. The magnitude of the response was significantly greater in late stance [0.88 μV (SD 0.42)] compared with midstance [0.41 μV (SD 0.21), Tukey–Kramer, P = 0.004]. More interestingly, the unload response magnitude was greater when the subjects were less supported by the harness; i.e., the response at 5% BWS was greater than the response at 30% BWS [0.74 μV (SD 0.38) and 0.55 μV (SD 0.41) for 5 and 30% BWS, respectively, Tukey–Kramer, P < 0.001; see Fig. 4].

Fig. 4.

The unload response evaluated in mid and late stances (dark and light, respectively) during 5 and 30% BWS (n = 12). The main effect of BWS is represented by the nonfilled bars. The unload response was significantly decreased as the level of BWS was increased; main effect of BWS: ***P < 0.001. Error bars indicate 1SD.

Because the interaction effects of BWS and time in stance were close to significant, post hoc tests were performed on the groups to investigate whether the unload response behaves differently in late stance compared with that at midstance. The unload response in midstance showed a significant decrease with increased BWS [0.54 μV (SD 0.18) and 0.28 μV (SD 0.13) for 5 and 30% BWS, respectively, Tukey–Kramer, P = 0.0019]. Although the unload response in late stance was also decreased by the increased BWS the difference was not significant due to the higher variance of the response [0.94 μV (SD 0.43) and 0.83 μV (SD 0.42) for 5 and 30% BWS, respectively, Tukey–Kramer, P = 0.233].

Group II afferent pathway depression using tizanidine

Before the tizanidine-induced group II suppression, the SOL stretch reflex had distinct SLR and MLR responses in all but one subject. The data records from this subject (male; age: 41 yr) were excluded as were the data from a subject (female; age: 24 yr) where no distinct unload response was found prior to ingestion of tizanidine. Approximately 60 min after the administration of tizanidine the MLR response decreased in all subjects, whereas the SLR response remained unchanged (see Fig. 5). The SLR showed a tendency to increase [before: 0.64 μV (SD 0.23); during: 0.68 μV (SD 0.27)], but the difference was not significant [one-way rmANOVA, F(5,1) = 0.43, P = 0.54]. However, the MLR response before the depression [0.45 μV (SD 0.21)] was significantly greater than that during the depression of the group II afferent pathway [0.28 μV (SD 0.14), one-way rmANOVA, F(5,1) = 7.79, P = 0.039]. The depression of the MLR response shows that the group II afferent pathway was significantly depressed at the time of the investigation. The unload response measured intermittently with the stretch reflex was unaffected by the group II afferent pathway depression [before: 1.23 μV (SD 0.65); during: 1.31 μV (SD 0.57), one-way rmANOVA, F(5,1) = 0.68, P = 0.45]. Also, the muscle activity during the nonperturbed trials, measured over 100 ms starting from the moment of the perturbation, did not significantly change with the tizanidine depression [before: 4.53 μV (SD 2.22); during: 4.29 μV (SD 2.00), F(5,1) = 1.05, P = 0.35].

Fig. 5.

Influence of group II afferent depression on the unload response. The group II tizanidine induced depression was simultaneously assessed together with the unload response. The short latency reflex (SLR), medium latency reflex (MLR), and unload response were measured in late stance during walking. Difference traces of a representative subject are presented for dorsiflexion (A) and plantar-flexion (B) perturbations before and during tizanidine induced group II depression (thin and dotted lines, respectively, n > 28). Horizontal dotted lines indicate the level of 2SDs from baseline. As evidence from the significant decrease in MLR response (C, group data of 6 subjects, P = 0.039, indicated with *), the tizanidine did depress the group II afferent pathway while not affecting the short latency component of the stretch reflex (P = 0.54). The unload response [before: 1.23 μV (SD 0.65); during: 1.31 μV (SD 0.47)] was not significantly affected by the depression of the group II afferent pathway (P = 0.44). Error bars indicate 1SD.

Stretch reflex response

To determine the influence of BWS on the SLR and MLR burst magnitude, ankle dorsiflexion perturbations were elicited at the same instances in stance as the plantar-flexion perturbations. All perturbations were performed on the second step with the new load condition as described earlier for the unload response investigation. The dorsiflexion perturbations elicited short- and medium-latency responses in all but one subject. This subject lacked a clear MLR burst for the midstance perturbation at 30% BWS and was thus removed from the analysis. Furthermore, the peak SLR and MLR bursts were not well defined in two subjects; thus the analysis windows could not be reliably placed, although a general increase in activity was seen. Therefore these two subjects were also excluded from further analysis, leaving 12 subjects for the stretch reflex analysis.

The median dorsiflexion velocities for the perturbations ranged from 357 to 403°/s and the amplitudes from 4.8 to 5.2° (see Fig. 6, C and D). It was not possible to maintain a constant perturbation trajectory when the load on the perturbed limb was changed by the BWS. This resulted in a small but significant increase in perturbation velocity for the 30% BWS condition: midstance: 20.6°/s (SD 24.2, P = 0.007); late stance: 42.5°/s (SD 37.5, P = 0.001). Similarly, the perturbation amplitude in late stance was larger in the 30% BWS condition but not in midstance: midstance: 0.08° (SD 0.18, P = 0.070); late stance: 0.22° (SD 0.15, P < 0.001). However, the amplitude of the perturbation increased slightly from mid to late stance, 0.23° (SD 0.39, P = 0.009), whereas no change in overall perturbation velocity was found between mid and late stance perturbations, 18.3°/s (SD 67.9, P = 0.2).

Fig. 6.

Bar graph of the SLR (A), the MLR (B), and boxplots of the perturbation velocity (C) and the perturbation amplitude (D) for the mid and late perturbations are shown for the 2 BWS conditions (n = 12). The nonfilled bars represent the main effect of BWS (group mean for SLR and MLR and group median for perturbation velocity and amplitude). Error bars represent SD in SLR and MLR. As the BWS was increased (i.e., less load was exerted on the supporting limb) no significant difference was found in the SLR magnitude. However, the MLR magnitude was significantly decreased for the 30% BWS compared with when more load was exerted on the limb at 5%. This decrease in MLR response cannot be explained by perturbation parameters because both the perturbation velocity and amplitude were increased with the higher level of BWS. Significance is represented by *P < 0.05, ***P < 0.001.

The SLR peak and MLR peak latencies were 56.1 ms (SD 3.9) and 86.5 ms (SD 3.9), respectively. The peak latencies were not affected by the BWS condition. However, latencies of the SLR and MLR peaks in late stance were significantly smaller than those in midstance [SLR: 54.0 ms (SD 2.6) vs. 58.2 ms (SD 4.0), F(1,11) = 11, 24, P = 0.006; MLR: 85.1 ms (SD 3.3) vs. 88.0 ms (SD 3.9), F(1,11) = 5, 67, P = 0.036].

Box plots of the MLR and SLR are shown in Fig. 6. The SLR and MLR responses were square root-transformed to get a normal distribution of data to allow for parametric statistics. No significant interactions between BWS condition and time in stance were found. The short latency response was unaffected by either the BWS condition [F(1,11) = 2.64, P = 0.13] or time in stance [i.e., mid or late stance, F(1,11) = 0.64, P = 0.44]. This indicates that it is unlikely that spinal motoneurons' recruitment gain was changed for the two BWS conditions. However, as the perturbation velocity was slightly increased, at the higher level of BWS, there is a risk that the faster perturbation velocity would increase the SLR, occluding any decrease caused by the BWS. Therefore we tested whether the SLR at 30% BWS was significantly different from the SLR at 5% BWS after compensating for the faster dorsiflexion velocity. The relationship between the magnitude of the SLR response and the dorsiflexion velocity, reported by Grey et al. (2001), was used to calculate a predicted response by adjusting the measured values at 5% BWS for the effect of the faster dorsiflexion velocity. If the level of BWS affects the short latency reflex response, the distribution of the difference between the measured SLR at 30% BWS and the predicted SLR should not be centered around zero. Both the upper and lower 95% confidence intervals for the slope of the regression reported by Grey et al. (2001) were tested using independent Wilcoxon signed-rank tests for the mid and late stance perturbations. No significant differences between the measured and predicted SLR in mid and late stances were found, indicating that the SLR response was unaffected by the BWS (mid: Plower 95%CI = 0.3 and Pupper 95%CI = 0.11; late: Plower 95%CI = 0.23 and Pupper 95%CI = 0.42).

In contrast, the MLR was significantly affected by both the time in stance [F(1,11) = 21.43, P < 0.001] and the level of BWS [F(1,11) = 5.14, P = 0.044]. The magnitude of the MLR was significantly greater in late stance than that in midstance (Tukey–Kramer: P < 0.001); more interestingly, however, the MLR response decreased from 0.47 to 0.42 μV as the BWS increased from 5 to 30% (Tukey–Kramer, P = 0.045). This decrease is unlikely to be due to the perturbation parameters because both the dorsiflexion amplitude and velocity increased with the higher level of BWS (see Fig. 6, C and D).


The aim of this study was to investigate the contribution of load and length feedback to the corrective stretch reflex response and to the afferent-mediated component of the locomotor EMG. Transient changes in body weight support (BWS) were used to alter the load on the supporting limbs and thereby change the afferent feedback. Additionally, the transient nature of the changes in BWS would limit the extent to which the CNS could adapt to the change in afferent feedback. The background muscle activity was significantly decreased at the higher level of BWS with respect to when the supporting limbs were more loaded, indicating that the magnitude of the BWS changes was sufficiently large to induce changes in the locomotor activity. This result is in agreement with previous investigations using sustained body weight support (Grey et al. 2002; Harkema et al. 1997; Stephens and Yang 1999). The response to rapid plantar-flexion perturbations (i.e., the unload response) was studied at the different levels of BWS to investigate whether the changes in background activity were afferent-mediated and to investigate which afferents contribute to the change in background activity. When BWS was transiently reduced, the magnitude of the unload response increased. In addition, the tizanidine-induced group II afferent pathway depression did not produce a significant decrease of the unload response. This indicates that the unload response may be less sensitive to the group II afferent pathway than has previously been thought (Sinkjaer et al. 2000) and it suggests that the response is potentially more sensitive to the load-sensitive changes signaled by the group Ib pathway (see following text). Furthermore, our investigation found a modulation of the medium latency component of the stretch reflex with the load on the limb, whereas no significant change in the short latency stretch reflex response was found. The MLR is mostly mediated by group II afferents (Grey et al. 2001; Schieppati and Nardone 1997), but the BWS induced modulation of the response indicates that force feedback likely converges with feedback from group II length sensitive afferents to form the MLR response. This finding contrasts previous findings (i.e., Grey et al. 2002), which did not find any difference in MLR response with respect to sustained BWS. This discrepancy between the current findings and those reported by Grey et al. (2002) may reflect the importance of using transient changes instead of sustained BWS when studying reflex activity because a potential adaptation of the spinal circuitry may occur.

Body weight support

Sustained decreased loading is known to decrease muscle activity levels in both healthy (Grey et al. 2002; Stephens and Yang 1999) and spinal cord injured subjects (Harkema et al. 1997). The response to a rapid change in BWS and body loading has also been shown to induce transient changes in locomotor output (Bachmann et al. 2008; Stephens and Yang 1999). In line with previous studies using sustained BWS, the transient weight changes induced in this study shortened the stance phase and reduced the mean muscle activity over this time. Because subjects are likely to adapt to the level of BWS it was important to measure the effect on the unload response as early as possible after the change in BWS. Due to limitations in the experimental setup we were unable to produce reliable plantar-flexion perturbations during the first step after the change in BWS condition. Therefore the overall activity level of the soleus muscle was investigated during the first seven steps after the BWS was changed. Because the activity level of the first and second step did not significantly change, we assumed that no significant adaptation had taken place and that it would be fair to make the investigation on the second step. Furthermore, no trend in the activity level was detected over the first seven steps that were measured. Similarly, Bachmann et al. (2008) reported no change in muscle activity for three consecutive steps after the BWS was changed from 25 to 45% of body weight. The decreased limb loading may still have triggered an adaptation of descending drive and of the integration of afferent information in the present study, but presumably it required a longer time to influence the activity in stance. Such adaptation could for example be the cause of larger knee flexion at touchdown found during several consecutive steps after removing a sustained BWS (Lau et al. 2008).

In the current experiment, the background locomotor activity was decreased as the BWS was increased. Possible explanations for this decrease in background activity are at least twofold: a direct effect of reduced afferent drive or potentially a decreased recruitment gain of the motoneuronal pool. Changes in recruitment gain have been described theoretically by Kernell and Hultborn (1990) to arise from an uneven distribution of inhibitory and excitatory drive in the motoneuronal pool. For the human TA, changes in recruitment gain have been reported after cutaneous afferents were electrically stimulated (Nielsen and Kagamihara 1993). A decrease in the recruitment gain of the motoneuronal pool would affect the short latency stretch reflex, given that no change in presynaptic inhibition occurred. Similarly, the H-reflex would be decreased by decreased motoneuronal recruitment gain. In the current investigation no change in the magnitude of the SLR was found for BWS levels of 5 and 30% and in both standing and walking conditions no change in H-reflex excitability has been detected for 0, 25, and 50% BWS (Field-Fote et al. 2000; Knikou et al. 2009; standing and walking, respectively). This indicates that the recruitment gain of the motoneurons was most likely not significantly changed by the increase in BWS. Although the BWS perturbation was delivered transiently, the subject was exposed to the new body weight condition for ≤2 s before the rapid plantar-flexion perturbation. Therefore we cannot rule out the possibility that some rapid supraspinal adaptation may have produced premotoneuronal changes that were not investigated in this study. Nevertheless, we believe the more parsimonious explanation for the change in background activity is the direct effect of a decreased afferent feedback.

Unload response

Because the contribution from length sensitive afferents on the locomotor activity has not been fully investigated we chose not to use the BWS transition to investigate the importance of limb load on the locomotor muscle activity but a separate measure (i.e., the unload response). The rapidly changing forces on the ankle when transiently changing the BWS alters the ankle dorsiflexion trajectory and thus cannot be used to distinguish between stretch/length sensitive and force sensitive afferent feedback. The proprioceptive contribution to the locomotor activity in stance can nevertheless be estimated by the unload response, which effectively measures the amount of locomotor activity removed by inducing a rapid plantar flexion. The proprioceptive contribution here characterized by the unload response showed a clear decrease as the load on the supporting limb was decreased with the BWS. Previous investigations have shown that the unload response is induced by the removal of afferent feedback from the triceps-surae muscle–tendon complex and not a result of reciprocal inhibition following a stretch of the TA or the depression of the largest diameter group I fibers (Sinkjaer et al. 2000). Cutaneous and muscle afferents from the foot and ankle are not essential for the unload response because a complete anesthetic block at the level of the malleoli did not influence the amplitude of the unload response (Grey et al. 2004). Additionally, the unload response has been shown to remain after ischemic block or the largest group I afferents (Sinkjaer et al. 2000) and, furthermore, the onset latency of the unload response was shown to be longer than the SLR and shorter than the MLR (Grey et al. 2004). In summary, two afferent modalities may contribute to the unload response and thus also to the afferent mediated enhancement of the locomotor activity—i.e., the mostly length sensitive group II afferents and the force sensitive group Ib afferents.

The observed MLR decrease following tizanidine indicates that the group II afferent pathway was partly suppressed. This result is similar to that reported previously for standing (Nardone and Schieppati 1998; Schieppati and Nardone 1997) and walking (Grey et al. 2001). Thus we can be certain that the group II afferent pathway was also suppressed when the plantar-flexion perturbations were delivered. The lack of effect on the unload response may suggest that group II afferents do not contribute significantly to the late-stance background locomotor EMG. However, this result must be interpreted cautiously because it is possible, although we believe unlikely, that the tizanidine induced depression of the group II afferent pathway may have a different effect on the facilitatory response evoked by muscle stretches compared with disfacilitation induced by the plantar-flexion perturbations. It is also possible that the tizanidine induced decrease in group II feedback is compensated for by upregulating other afferent pathways and that it is this compensatory feedback that is removed with the plantar-flexion perturbation. However, we believe that the more plausible explanation is that the modulation of the unload response with BWS is influenced by the changes in load related afferent firing rates as the load on the limb was changed. This result parallels a previously demonstrated correlation between the unload response and the tension on the Achilles tendon during walking on an inclined or declined treadmill (Grey et al. 2007).

Stretch reflex

Previously investigated load related effects on the stretch reflex did not find a significant change in relation to sustained BWS of 30% body weight (Grey et al. 2002). It was hypothesized in the current study that a reorganization or reweighting of the afferent feedback in the spinal cord could have occurred due to the sustained change in loading, thus obscuring a change in stretch reflex response. Although we did not observe any changes in the background locomotor activity for the first two steps of the new loading condition, we did find a small but significant load related decrease in the medium latency burst of the stretch reflex. However, no significant change was found for the short latency component of the stretch reflex, even when compensating for the difference in perturbation velocity using the regression reported by Grey et al. (2001).

Grey et al. (2001) indicated that group II and possibly group Ib afferent fibers could contribute to the MLR response. Through several experimental paradigms, the group II afferent pathway has been shown to be a major contributor to the MLR response in standing (e.g., Chaix et al. 1997; Marque et al. 1996; Schieppati and Nardone 1997) and in walking (e.g., Grey et al. 2001). In the current study, a small but significant load dependence was noted for the MLR magnitude. The MLR decrease observed together with the decreased supporting limb load at 30% BWS is unlikely to be explained by changes in group II activity because both the muscle stretching velocity and amplitude increased at the higher level of BWS. This shows that for the MLR response the two afferent pathways of group II and group Ib may converge into a combined corrective response while walking. Similarly in cats, excitatory group I convergence on the last order group II interneurons was present in >60% of the recorded interneurons (Edgley and Jankowska 1987). From a control perspective, the combined information of load and stretch would likely be beneficial to compensate for fast perturbations of the step. The corrective action taken has the potential to better compensate for the perturbation if it were dependent not only on the stretch of the muscle but also on the force of the perturbation.

The amplitude of the MLR response increased in late stance compared with midstance. There are several factors that could have influenced the gain of this pathway, such as presynaptic inhibition of the afferent terminals and changes in excitability of interneurons from both descending and/or other afferent pathways (for review, see Jankowska 1992). Therefore the potentially increased gain of the MLR may be attributed to changes in descending influence and/or influence from other proprioceptive pathways that change their activity during the course of the stance phase. However, since the perturbation amplitude was slightly increased, about 4%, between mid and late stances we are unable to separate the effects of the perturbation from other potential adjustments of the disynaptic group II afferent pathway.

For the unload response, the modulation with BWS shows that changes in load can be compensated by spinally mediated intrinsic circuitry that will modify the locomotor output. In overground walking, a similar modulation of locomotor activity was shown to covary with the Achilles tendon load and muscle–tendon length (af Klint et al. 2008). In addition, indirect evidence during overground (af Klint et al. 2009, 2010) and treadmill walking (Grey et al. 2007) suggests that this modulation most likely stems from a positive load contribution to the locomotor activity of the soleus. A similar excitatory heterogenic Ib-afferent feedback from the gastrocnemius onto the soleus has also been observed by electrical stimulation (Faist et al. 2006). However, it remains to be determined whether the modulation stems from a heterogenic and/or an autogenic feedback from the triceps surae.


This work was supported by grants from the Spar Nord Foundation and the Obel Family Foundation.


No conflicts of interest are declared by the authors.


View Abstract