• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of jphysiolThe Journal of Physiology SiteMembershipSubmissionJ Physiol
J Physiol. Oct 15, 1999; 520(Pt 2): 591–604.
PMCID: PMC2269584

Voluntary and reflex control of human back muscles during induced pain


  1. Back pain is known to change motor patterns of the trunk. The purpose of this study was to examine the motor output of the erector spinae (ES) muscles during pain in the lumbar region. First, their voluntary activation was assessed during flexion and re-extension of the trunk. Second, effects of cutaneous and muscle pain on the ES stretch reflex were measured, since increased stretch reflex gain has been suggested to underlie increased muscle tone in painful muscles.
  2. The trunk movement and electromyographical (EMG) signals from the right and left ES during pain were compared with values before pain. Controlled muscle pain was induced by infusion of 5% saline into the right lumbar ES. Cutaneous pain was elicited by mechanical or electrical stimulation of the dorsal lumbar skin. The stretch reflex was evoked by rapidly indenting the right lumbar ES with a servo-motor prodder.
  3. The results from the voluntary task show that muscle pain decreased the modulation depth of ES EMG activity. This pattern was associated with a decreased range and velocity of motion of the painful body segment, which would normally serve to avoid further injury. Interestingly, when subjects overcame this guarding tendency and made exactly the same movements during pain as before pain, the EMG modulation depth was still reduced. The results seem to reconcile the controversy of previous studies, in which both hyper- and hypoactivity of back muscles in pain have been reported.
  4. In the tapped muscle, the EMG response consisted of two peaks (latency 19.3 ± 2.1 and 44.6 ± 2.5 ms, respectively) followed by a trough. On the contralateral side the first response was a trough (26.2 ± 3.2 ms) while the second (46.4 ± 4.3 ms) was a peak, similar to the second peak on the tapped side. Cutaneous pain had no effect on the short-latency response but significantly increased the second response on the tapped side. Surprisingly, deep muscle pain had no effect on the stretch reflex. A short-latency reciprocal inhibition exists between the right and left human ES.
  5. It is concluded that deep back pain does not influence the stretch reflexes in the back muscles but modulates the voluntary activation of these muscles.

Low back pain is a common disabling musculoskeletal disorder, whose prevention and treatment are problematic. The main reason is that current imaging techniques do not identify the source of pain in the vast majority of cases. The diagnosis of low back disorder is, therefore, often based on non-specific signs, such as deep tissue pain and altered motor patterns. Unfortunately, there is little agreement on how these patterns change with pain. While most authors agree that force generated by the muscles is usually diminished (the mechanism is unknown), electromyographic (EMG) recordings are equivocal - both hyper- and hypoactivity have been reported (Ahern et al. 1988). The contradictory results may reflect two major disadvantages connected with clinical studies. First, as a direct consequence of the difficulty of localizing the primary tissue damage, the clinical studies may involve heterogeneous populations of patients with different primary afflictions (affecting intervertebral disks, ligaments, facet joints, muscles, etc.). Second, in clinical studies there is no healthy ‘norm’ to which the patients could be compared. Data from the individual's pain-free history are rarely available and there is a certain inter-individual variability in motion patterns even within the healthy population. Uncontrolled factors such as these are responsible for the current poor understanding of the complex pathology of back pain. Experimentally produced back pain eliminates some of these uncertainties.

Kellgren's experiments in the 1930s showed that injections of hypertonic saline into muscles or ligaments cause deep pain comparable to spontaneous musculoskeletal pain (Kellgren, 1938). Paintal (1960) demonstrated that intramuscular injection of hypertonic saline activates group III afferents, which may be responsible for certain aspects of muscle pain. Although its exact mechanism of nociceptive action is not known, hypertonic saline has been successfully used to induce muscle pain in numerous studies (Stohler et al. 1988; Schwartz et al. 1993; Arendt-Nielsen et al. 1995; Svensson et al. 1997). An invaluable advantage of this method is that both the intensity and location of the pain can be controlled. By selectively targeting a chosen structure one can determine the impact of pain on the surrounding tissues and general motor patterns. An obvious candidate for studying the link between pain and altered movement is skeletal muscle. Inappropriate neural control of a given muscle has been suggested to bring about compensatory changes in neighbouring muscles disposing them to overload and injury. In our study we therefore examined the effect of experimentally elicited pain in a back muscle, the right erector spinae (ES), on the voluntary and reflex activity of both the right and the left ES.

It has also been noticed that deep somatic or visceral pain leads to local increases in muscle tone (Travell & Simons, 1983). The opinion that muscular hypertonus is a result of increased muscle stretch reflexes is frequently encountered in the literature (Berberich et al. 1987; Johansson & Sojka, 1991; Matre et al. 1998).

In general, the reflex activity of human back muscles has received very little attention in the past, compared with limb muscles. To our knowledge, only one study concerned with human back muscle reflexes and pain has been published (Kugelberg & Hagbarth, 1958). It described trunk muscle responses to painful cutaneous stimuli from the trunk region. However, the back pain syndrome is usually characterized by deep ligamentous, articular or muscular pain. Cutaneous signs, if present, occur secondarily (Travell & Simons, 1983).

The amplitude of the stretch reflex is determined by the excitability of the skeletomotor neurones, fusimotor neurones and/or muscle spindles. A classical manoeuvre to test the stretch reflex gain is the muscle tendon jerk. Since the 19th century it has been widely used in testing limb reflexes and, in isolated cases, it has also been described for the back muscles (Dimitrijevic et al. 1980; Tani et al. 1997). If the activity of α-motoneurones is held constant by maintaining a constant level of voluntary EMG activity during muscle taps, it is possible to test the hypothesis that painful input to the central nervous system leads to muscular hypertonus through increased spindle sensitivity to stretch or increased gain of central transmission.

Since both deep and superficial pain could affect muscle tone, in this study we investigated whether a stretch reflex evoked by tapping the lumbar portion of the ES is changed by superficial (cutaneous) or deep (muscle) pain.

Some of the present data have been published in a preliminary form (Zedka et al. 1998).



Five healthy volunteers (1 female and 4 males, age 20-55 years) participated in the study. They were familiarized with the protocol and signed a consent form in accordance with the requirements of the University of Alberta, Faculty of Medicine Research Ethics Board. The study was conducted according to the Declaration of Helsinki.


After preparation of the skin with ethyl alcohol, bipolar surface EMG activity from the right and left ES at the third lumbar (L3) vertebral level was monitored with self-adhesive Ag-AgCl electrodes (Fig. 1; ElectroTrace ET 301, Huntington Beach, CA, USA). The EMG signal was amplified (gain 1000), band-pass filtered (30 Hz to 50 kHz), full-wave rectified, low-pass filtered (300 Hz), and digitized by a 1401 laboratory interface (CED, UK). The EMG signal was sampled at 500 s−1 during voluntary movements and at 1000 s−1 during testing of the stretch reflexes. CED SIGAVG 6.30 software running on a personal computer was used to store and average the recordings.

Figure 1
View of subject's back

Painful stimulation

Deep (muscle) pain

A small area of skin over the right ES 3 cm lateral to the L3 spinous process was cleaned with alcohol and an intravenous catheter/needle (1.1 mm in diameter, 5.1 cm long) was inserted perpendicularly to the surface of the back, 4 cm into the muscle (Fig. 1). An intravenous infusion set attached to a 10 ml disposable syringe was filled with 5 % aqueous solution of NaCl. The needle was removed and the infusion set was connected to the flexible intravenous catheter, which remained in the muscle, secured to the skin by tape. The syringe with saline was placed in an infusion pump (Harvard Apparatus, model 22). The pump was controlled digitally by a computer, which regulated the infusion rate in a feed-forward compensation manner (Appendix, Fig. 7) in order to keep the pain approximately constant. The profile of the infusion rate over approximately 25 min was an inverse function of the mean pain rating profile obtained from two subjects during preliminary tests when 5 % NaCl was infused at several constant rates (50, 60, 100, 140, 150 and 200 μl min−1 for 12 min (Appendix, Figs. 8 and and9).9). The subjects verbally rated their pain every 15 s on a 0-10 scale, where 0 represented ‘no pain’ and 10 was ‘unbearable pain’. Pain rating profiles during the main experiment revealed that the feed-forward regulation of the infusion rate provided minimal fluctuations in perceived pain (Appendix, Fig. 10).

Figure 7
Block diagram of the pain control system
Figure 8
Pain step response to NaCl infusion
Figure 9
Infusion rate for desired pain response
Figure 10
Pain rating

Mechanical cutaneous stimulation

Superficial pain was first elicited by pressing the tip of a metal woodscrew into the lumbar skin 1 cm lateral to the tapped site. The screw was connected to a shaft in series with a spring (compliance 2.4 mm N−1). Application of constant pressure during a series of 20 taps was ensured by pressing on the shaft so that the spring was compressed by a constant amount. The radius of curvature of the screw tip was about 0.5 mm, and the skin was typically indented by about 5 mm during the procedure. After collection of 20 sweeps the pin was removed.

Electrical cutaneous stimulation

For superficial pain elicited by electrical stimulation of the skin, two surface self-adhesive gel electrodes (Chattanooga Corp., diameter 1 cm) were placed over the right ES. The cathode was centred 1.5 cm lateral to the tapped site and the anode was centred 2.7 cm lateral to the tapped site. Each tap was preceded by a 500 ms, 25 s−1 train of 0.1 ms constant current pulses. The tap was delivered 50 ms after the end of the train of stimuli. The delay was chosen to avoid interference of the electrical stimulation artefact with the EMG response to the tap. The threshold of perception of the pulses was determined for each subject in terms of pulse amplitude (mA). Responses to 20 taps were collected after painful trains of stimuli at a stimulation intensity 4 times the perceptual threshold (4T). The subjects verbally rated the pain on a 0-10 scale, where 0 represented ‘no pain’ and 10 was ‘unbearable pain’.

The perceived intensities of all three forms of elicited pain were reasonably constant over the test period. Both mechanical and electrical stimulation of the skin were perceived as pricking, burning pain confined to a small area around the site of stimulation, without spread. In contrast, saline infusion into the right ES muscle resulted in deep, dull pain spreading from the site of injection downwards to the ipsilateral superior gluteal region.

Experimental protocol

Voluntary movements

The subjects stood upright with their backs positioned close to a wall. Their hips and knees were stabilized in an extended position by firmly strapping the pelvis and the thighs to the wall. On a ‘ready-steady-go’ signal the subjects bent their trunks forward from the waist as far as they could without bending their knees (no other instruction was given). They remained in the flexed position for several seconds and then returned to the upright position. The task was performed before and during the saline infusion. Since back pain can lead to alterations in movement patterns, in addition to these trials the subjects also performed controlled movements where the velocity and range of motion in pain were the same as those without pain (Fig. 2). This was achieved by asking the subjects to follow a target immediately in front of their eyes. The target was a horizontal rod connected to a shaft pivoting in the saggital plane about the subject's pelvis. The shaft was rotated up and down at constant velocity by a servo-motor (the sacrum being the approximate pivot point of the sagittal trunk movement). Movements of identical velocity and various amplitudes within the maximal range of motion were tested to determine the ES EMG patterns before and during pain. The trunk motion was quantified using a semiconductor gyro placed on the L3 spinous process. An angular displacement curve was calculated off-line by integrating the recorded velocity signal from the gyro. As this method does not measure local segmental motion of the spine, minor changes in the movement of individual vertebral segments during pain cannot be entirely excluded. However, no obvious splinting manoeuvres or asymmetry in trunk movement were observed.

Figure 2
Voluntary flexion-extension of the trunk with controlled velocity and range of motion

Stretch reflex

Stretch reflexes of the right lumbar ES were elicited by sudden rapid indentations delivered by a prodder fixed to an electromagnetic length servo-motor, which was driven by rectangular waveforms. The end of the metal prodder was provided with a hard rubber cap. The cap was in permanent contact with the skin and the mechanical perturbation of the muscle was produced by indentations of a constant amplitude and duration (5 mm, 1 s). The displacement and the impact force of the prodder were recorded using the CED SIGAVG 6.30 software.

The subjects were seated on a stool with their feet on the ground and positioned so that the prodder was touching the skin over the right ES between the two EMG electrodes, i.e. adjacent to L3 spinous process, 2 cm from the midline. To ensure that during the whole experiment the taps would be delivered to the same site, adjustable upholstered pads positioned against each side of the body were used to restrict lateral movements of the trunk. The sitting subjects produced constant background ES EMG activity by stretching both arms forwards while watching a low-pass filtered (10 Hz) version of the rectified EMG signal on an oscilloscope in front of them. The position of the lumbar spine in the antero-posterior direction was maintained constant by monitoring the prodder force trace on the oscilloscope. Under these conditions 20 control taps were delivered.

Each session started with control trials without painful stimulation. The subjects first performed the voluntary flexion-extension exercise. Five trials without feedback were followed by five trials with controlled velocity and amplitude. Then the stretch reflexes were tested. Twenty control taps were delivered before any stimulation. After 5 min of rest, 20 taps were delivered during painful mechanical skin stimulation. When the pain completely subsided, the responses to 20 taps were recorded during painful electrical stimulation. After pain subsided, the infusion catheter was introduced into the ES. The initial insertion pain usually subsided to a zero rating 10-20 min after the catheter was in place and a second set of control taps was delivered. Only if the stretch reflex monitored on a computer screen at this point was not different from the pre-insertion control (later confirmed by off-line analysis) was the effect of muscle pain tested with 20 taps. Finally, with the infusion still running, five uncontrolled and five controlled painful trunk flexion-extensions were performed.

Data analysis

Voluntary movement

The mean amplitude of the rectified EMG signal was calculated for all phases of the task cycle (flexion, relaxation, extension). These phases were defined on the basis of discrete events in the EMG signal in non-painful conditions. EMG was chosen rather than the angular displacement signal because it was the EMG modulation that was of major interest. The start and end of the whole sequence were defined in terms of the onset and the end, respectively, of EMG activity in the injected muscle, the right lumbar ES. The transitions between flexion and extension were defined as a 1 s-long interval of the lowest EMG activity. The timing of divisions defined with respect to the right lumbar ES (no pain) was applied to all other muscles.

Stretch reflex

Control indentations of the right lumbar ES elicited an EMG response consisting of two peaks followed by a trough. This pattern is similar to that reported by Dimitrijevic et al. (1980) and Tani et al. (1997), using pulsatile tap stimuli. These authors ascribed the terms R1 and R2 to the first and second peaks, and although the latencies we observed were somewhat longer, conceivably because of a slightly longer rise time of our step indentations (12.5 ms, 0-63.2 % of maximum; NB rise times were not stated in the above references), we will adhere to the R1 and R2 terminology. In the left ES the taps to the right ES elicited a different response. Instead of the R1 peak there was a trough. This trough was followed by a peak which corresponded to R2 in the right ES. To normalize the EMG responses, mean amplitudes in the right ES were computed over the post-stimulus interval 15-35 ms for R1 and the interval 45-70 ms for R2. In the left ES, the interval for R1 was set between 25 and 45 ms, and for R2 between 45 and 70 ms (Fig. 3). These values were divided by the mean pre-stimulus amplitude calculated between -25 and 0 ms.

Figure 3
Mean EMG responses of the right and left lumbar ES to 20 taps delivered to the right ES of one subject

Student's t test (paired two-sample test for difference of means) was used to compare the means of the EMG signals within corresponding phases before and during pain. The same test was used to compare the amplitudes of the stretch reflex components between each painful condition and the control condition. The significance level was set at P= 0.05. Analysis was performed with Microsoft Excel 97 software.


Voluntary movement

Before pain

The EMG patterns of pain-free flexion- extension for one subject are shown in Fig. 4A. The right and left ES muscles participated in both flexion and extension. The muscles were activated in the early phase of flexion (sometimes preceded by initial inhibition, if resting background activity was present) and remained active until a certain lumbar angle was reached where they became silent. In general, the ES EMG ceased between 80 and 100 % of flexion. In the static flexed position there was virtually no ES EMG activity (relaxation). Trunk re-extension was accomplished with strong ES activation, which started before the movement and continued until the upright position was reached.

Figure 4
A, the angular displacement in the sagittal plane of the 3rd lumbar vertebra and EMG patterns from the back muscles during painless flexion-extension for one subject. Note the spontaneously smaller angular displacement and depths of modulation of both ...

During pain

Pain spontaneously decreased the velocity and range of trunk motion. Across all subjects, the flexion angle represented only 60-90 % of the control maximal range of motion. This change in the motor pattern was associated with a bilateral alteration in the EMG patterns, which was specific for each phase of the movement (Fig. 4B). While the mean ES EMG amplitude during trunk flexion was not significantly changed, the ES relaxation observed during maintained flexion in painless conditions was absent in pain (P < 0.05). During the extension phase the mean EMG amplitude was significantly decreased compared with the painless extension (P < 0.05, Fig. 5A). Altogether, the presence of pain resulted in a decreased modulation depth of the voluntary EMG.

Figure 5
Quantified observations from Fig. 4

When subjects in pain were guided to perform movements identical to their painless control movements, the EMG profiles showed an asymmetrical pattern. While the EMG on the non-injected side revealed no difference from the painless control trials, the injected right side showed decreased EMG modulation (Figs. 4C and and5B5B).

Stretch reflex

The mean latency (±s.d.) of R1 in the right ES across all subjects was 19.3 ± 2.1 ms and that of R2 was 44.6 ± 2.5 ms. In the left ES, the R1 trough at an onset latency of 26.2 ± 3.2 ms was followed by an R2 peak at 46.4 ± 4.3 ms.

For control trials, the mean amplitude (±s.d.) of the normalized R1 in the right ES computed across all subjects was 178 ± 37 %, while the mean amplitude of R2 was 159 ± 29 % (Fig. 6). In the left ES in control trials the mean amplitude (±s.d.) of the R1 trough was 86 ± 10 % and the amplitude of R2 was 120 ± 13 %.

Figure 6
Amplitudes of R1 and R2 responses to stretch

During mechanical cutaneous pain the amplitudes of R1 and R2 in the right ES were 162 ± 45 and 257 ± 69 %, respectively. Pain had a statistically significant effect only on the R2 amplitude, which was almost doubled compared with the controls (P < 0.05). No significant difference in EMG amplitude was seen in the left ES. The values were 86 ± 13 % for R1 and 146 ± 22 % for R2.

A similar effect was found with painful electrical stimulation. The mean EMG values in the right ES were 171 ± 27 and 247 ± 43 % for R1 and R2, respectively. Again R2 showed a significant increase (P < 0.05) while R1 did not change significantly. No significant effect of electrical stimulation was seen in the left ES. The respective values of R1 and R2 were 76 ± 10 and 118 ± 33 %.

Contrary to expectation, the stretch reflex during deep muscle pain elicited by NaCl infusion was not significantly different from the stretch reflex in control trials. The mean pain rating across subjects during infusion was 5.3 ± 0.8, which was similar to that during the mechanical (5.2 ± 0.7) and electrical (5.2 ± 0.8) cutaneous pain trials. The EMG amplitude of R1 in the right ES was 162 ± 45 % and the amplitude of R2 was 182 ± 45 %. Again, the deep pain did not influence the responses in the left ES (R1 = 84 ± 10 %, R2 = 148 ± 29 %).


Voluntary movement

The findings of this study indicate that unilateral injection of hypertonic saline into the lumbar portion of the ES produced symptoms similar to those encountered in low back pain patients. The pain was described by the subjects as deep and dull, as expected for pain originating from deep tissues. Another characteristic feature similar to spontaneous back pain was the irradiation of the pain. The pain spread from the site of injection 3 cm to the right of the L3 spinous process down the right ES into the right superior gluteal region. The thoracic portions of the right ES and the left ES were never described as painful. The gluteal pain was probably of reflex origin since spontaneous lumbar pain follows the same pattern and referred pain has been reported after administration of hypertonic saline (Kellgren, 1938; Travell & Simons, 1983; Graven-Nielsen et al. 1997). It cannot be ruled out, however, that saline spread into the gluteal region due to gravity and/or intramuscular pressure gradients since injected fluid has a tendency to diffuse through the muscle along its longitudinal fascicular planes.

The saline-induced pain resulted in specific changes in the voluntary movement pattern, similar to those observed in patients with low back pain. The most striking finding in pain is the absence of relaxation in the back muscles in full flexion. The ‘flexion-relaxation phenomenon’ was first noticed by Allen (1948), and although its absence in low back pain patients has since been reported by many authors (Floyd & Silver, 1955; Triano & Schultz, 1987; Gracovetsky, 1988; Shirado et al. 1995), its origin is still unknown. Most authors speculate that the permanent ES activation is a protective response of muscles to low back pathology. For example, it has been suggested for healthy subjects that from the point where the ES muscles become silent, passive lumbar structures, such as interspinous ligaments, etc., take over the trunk load. The function of the ES in flexion is to ensure adequate loading of these structures through the control of lumbar lordosis (Gracovetsky, 1988). An absence of the flexion-relaxation phenomenon may indicate that the injured ligaments cannot sustain these forces and have to be protected by ES contraction (Floyd & Silver, 1955; Triano & Schultz, 1987). From this point of view, it is hard to see why in the present study the painful muscles should remain contracted to reduce loading of uninjured, painless ligaments. It is possible, however, that pain in any lumbar structure (ligaments, joints or muscles) is transmitted to ES motoneurones through a common element in the central neuronal circuitry. The muscle then works in a ‘pain mode’ to protect the spine from extreme movement whenever pain is signalled, without distinguishing the tissue of origin. Candidates for such a common element are dorsal horn neurones known to receive mixed input from deep tissues (Hoheisel & Mense, 1989). Ascending signals from these neurones can interact with motor commands at higher levels as well as at the spinal level. Both reflex and voluntary mechanisms have been suggested to underlie the absence of flexion-relaxation (Schultz et al. 1985). Our present results indicate that the changes in EMG activity are not associated with an increased gain of the ES stretch reflex, which was one possible mechanism. Rather, the persistence of the changes even when subjects voluntarily overcame their natural tendency to move more slowly and over a smaller range while in pain suggests a more complex mechanism.

Similarly, the decreased activation during re-extension may reflect voluntary avoidance of pain and possible harmful consequences of strong muscle contractions (further mechanical damage or chemical damage due to contraction ischaemia). On the other hand, the observed changes could also be due to a segmental inhibition by nociceptive input of descending voluntary commands to α-motoneurones. The possibility that a decreased voluntary drive alone is not responsible for the modulation of EMG amplitude is supported by several animal studies in which noxious muscle stimulation evoked specific modulation of masticatory movements driven by electrical stimulation of descending pathways (Schwartz et al. 1993; Westberg et al. 1997).

During the dynamic phase of trunk flexion, no significant difference in the mean ES EMG amplitude was found between trials in pain and the control trials. This is in accordance with results obtained from back pain patients (Shirado et al. 1995). These authors did not further discuss this finding, even though the absence of a pain effect on the EMG activity during dynamic flexion is in contrast with its effect on other phases of the movement. Trunk flexion may require smaller voluntary ES activation since gravitational force acts in the direction of the movement and, for a given level of activation (EMG), significantly larger forces are generated during eccentric contractions. This is due to the visco-elastic properties of the muscle, which are also potentiated by its stretch reflexes. If the EMG during flexion reflected predominantly a stretch reflex rather than voluntary activity, the non-significant effect of pain on this phase would be in line with our finding that deep muscle pain has no effect on the stretch reflex. It would also suggest that pain influences descending control much more than it does segmental control.

Whatever the reason for the decreased modulation depth, it seems to deprive painful muscles of sufficient relaxation. Interestingly, lack of relaxation due to repetitive movements or sustained contractions (muscle ‘overuse’) has been considered as a possible reason for development of chronic musculoskeletal pain (Elert et al. 1989; Veiersted et al. 1990). At the same time, several studies have shown that a paucity of EMG silent periods is characteristic of painful muscles or muscles surrounding painful tissues (Stohler et al. 1988; Arendt-Nielsen et al. 1995; Svensson et al. 1997). Though the widespread view that painful muscles are hyperactive even when subjects are at rest has not been substantiated in recent controlled studies, the question is, how often during the day are painful muscles actually allowed to rest? A study of low back pain patients found almost continuous activity in their back muscles during the night (Fischer & Chang, 1985). Therefore there is a possibility that, paradoxically, the beneficial ‘splinting’ muscle activity present after acute injury could in the long term lead to chronic musculoskeletal pain. Although the pain in the present study was strictly limited to the right lumbar region and below, the spontaneously altered motor patterns were associated with a bilateral decrease in EMG modulation. Thus, the non-painful left ES was deprived of relaxation to the same extent as the painful right ES. If continuous activity leads to muscle pain, this mechanism may contribute to the gradual spread of pain to muscles in the vicinity of the original lesion.

Stretch reflex

The stretch reflex in the ES was first described in detail by Dimitrijevic et al. (1980). Tapping the ES at the level of L5 revealed a short-latency response (R1) at 12 ± 1.6 ms and a long-latency response (R2) between 30 and 50 ms, which seemed to be independent of each other. R1 was more constant than R2 and showed vibratory-induced suppression, postvibratory facilitation and facilitation during a Jendrassik manoeuvre. These observations and the reflex latency led the authors to the conclusion that the R1 response was probably oligosynaptic. In contrast, R2 was more variable and was increased during muscle vibration. Its latency and its bilateral presence during unilateral tapping suggested a polysynaptic pathway.

A further description of the ES stretch reflex was provided recently by Tani et al. (1997). The reflex was elicited by taps in the interspinous space at various thoracic and lumbar levels. R1 was more reproducible than R2 and its latency was shortest when recorded at the tapped level. The latency increased in the caudal direction but not as steeply as the R2 latency, which resulted in a greater temporal separation between R1 and R2 caudally. Also, in contrast to R1, R2 recorded at L4-5 showed a progressive decrease in latency with more rostral stimuli. The authors suggested that R2 propagates rostral from the stimulus site and may include supraspinal transmission. Based on rough calculations of conduction times, the authors implicated fast-conducting dorsal column pathways as a possible part of the loop.

There is also evidence from limb studies of segmental mechanisms contributing to long-latency responses, such as polysynaptic spinal transmission (Ghez & Shinoda, 1978), slowly conducting afferents (Kirkwood & Sears, 1974), or repetitive spindle discharge (Eklund et al. 1982). It seems likely that more than one pathway or mechanism is involved in generating long-latency reflex activity.

In connection with the present study, it is appropriate to mention that cutaneous input has also been proposed to participate in the generation of the long-latency stretch reflex. Marsden et al. (1972) showed that anaesthesia of the thumb suppressed the late response in the long flexor of the thumb. Although subsequent studies of various muscles, including the ES, have shown that local anaesthesia did not abolish the long-latency response, the possibility that the stretch reflex is modified by cutaneous input cannot be ruled out (Darton et al. 1985). Deuschl et al. (1985) showed that stimulation of cutaneous afferents evoked long-latency reflexes in the thenar muscles similar to the response to simulation of a motor nerve, suggesting that R2 may be a product of mixed proprioceptive and cutaneous input. Thus, while not necessary to generate the long-latency response, cutaneous input may play a modulatory role. Conditioning a tap-elicited stretch reflex in back muscles with trains of electrical stimuli to the saphenous nerve in decerebrate cats demonstrated a facilitatory effect on the reflex amplitude if the conditioning-test interval was longer than 20 ms (Carlson & Lindquist, 1976). In a human study dealing with cutaneous reflexes, Kugelberg & Hagbarth (1958) found facilitation of tonic ES activity by painful stimulation of the skin overlying the muscle. This response was considered to be a withdrawal reflex because activation of the ES would increase lumbar lordosis and thus remove the skin from the painful stimulus. Similar effects were observed in cats following mechanical stimulation of the skin of the trunk and electrical stimulation of nerves supplying the lumbar back skin (Carlson & Lindquist, 1976). The significant increase in R2 amplitude found in the present study reflects increased excitability of polysynaptic reflex pathways and is consistent with the concept of avoidance of painful stimulation. The noxious character of the cutaneous stimulation seems to be important since no effect was present if the stimuli were not painful.

In contrast to the facilitatory effect of skin stimulation, deep muscle pain did not change the amplitude of the long-latency stretch reflex. This fact demonstrates that the differences between deep and superficial pain observed in the sensory domain (perceptual quality, pattern of spread, etc.) can be extended to the motor output. From the present study it is not possible to identify the neural circuits responsible for the different effects of skin and muscle pain on the long-latency response. It has been established that cutaneous pain is signalled to the CNS through small diameter myelinated Aδ and unmyelinated C afferent fibres which terminate in laminae I, IIo and V in the dorsal horn and also in lamina X. Afferent fibres excited by noxious stimuli in deep tissues terminate mostly outside lamina IIo, particularly in laminae I and V (Sigiura et al. 1989). Little is known about the connections of the second-order neurones in the dorsal horns with the α-motoneurones in the ventral horns but the different termination of the primary afferents suggests separate processing of the two types of pain. The interaction between the sensory and motor systems is polysynaptic and probably occurs at all levels of the CNS hierarchy. In decerebrate rats, Wall & Woolf (1984) showed that stimulation of C fibres of ankle extensor muscles evoked activity in knee flexor nerves which persisted for several minutes after stimulation had ceased. Flexion- withdrawal responses to other noxious stimuli were sensitized for up to 90 min after the cessation of ankle muscle stimulation. Experiments of this type have not been done for back muscles.

At the spinal level, the γ-motoneurones and interneurones synapsing onto the α-motoneurones have been considered as possible links between pain input and altered motor output (Johansson & Sojka, 1991; Lund et al. 1991). γ-Motoneurones (fusimotor neurones) receive a complex input from the skin, joints and muscles. Stimuli that excite small diameter muscle afferents have been shown to increase the excitability of γ-motoneurones (Jovanovic et al. 1990; Wenngren et al. 1998; but note Mense & Skeppar, 1991). Johansson (1981) formulated a ‘final common output hypothesis’ according to which γ-motoneurones integrate descending and multimodal reflex input and transmit the result to α-motoneurones via the spindle reflex arc. According to this view, muscle spindle afferents are seen as the ‘final common output’ to α-motoneurones. Pain-evoked fusimotor sensitization of the muscle spindles would increase the gain of the stretch reflex, which in turn would result in muscle hypertonus. According to this hypothesis muscle pain elicited in the present experiment was therefore expected to change the amplitude of the EMG response to muscle tap. Surprisingly, the reflex gain remained unchanged. This implies that spindle responses to indentation were unchanged and that there was therefore no pain-induced change in fusimotor action on spindles. There are caveats to this conclusion. First, it is just conceivable that fusimotor action did indeed increase with pain, but the incremental response to muscle indentation in our experiments was unchanged either because the effects on spindle gain of increased static and dynamic fusimotor action happened to be equal and opposite (Prochazka, 1996), or because spindles were taut throughout, so their afferent responses to indentation were ‘saturated’ and therefore unresponsive to fusimotor changes (Wood et al. 1994). On balance, however, the parsimonious conclusion is that deep back pain of the type we studied does not cause significant changes in fusimotor action. This is a fairly important point, because the notion of a fusimotor involvement in chronic muscle pain is based on several lines of evidence from acute experiments and is widely entertained as being a plausible mechanism.

The demonstration of a lack of pain effect on the short latency R1 response confirmed the conclusions of previous studies in limbs that pain has no influence on the monosynaptic spinal pathway as tested through the H-reflex (Leroux et al. 1995). Although the short-latency stretch reflex evoked by muscle tap may not be exclusively monosynaptic, evidence from the limb muscles shows that the monosynaptic component is dominant. Notwithstanding presynaptic inhibitory mechanisms, the monosynaptic pathway is probably exposed to less modulation than polysynaptic pathways. However, the number of synapses involved in the short-latency transmission in human back muscles is not established. Although monosynaptic connections of ipsilateral low-threshold afferents onto back muscle α-motoneurones were demonstrated by intracellular recordings in cats (Jankowska & Odutola, 1980), an earlier report considered the monosynaptic pathway for the stretch reflex in cats as exceptional and suggested a polysynaptic route instead (Carlson, 1978). Matre et al. (1998) have reported that short-latency EMG responses to stretch of human soleus increased during administration of hypertonic saline. However, a significant increase was stated clearly only when testing a relaxed muscle. This is insufficient for valid conclusions about the reflex amplitude, since the excitability of the α-motoneuronal pool may not have been constant. If for any reason the motoneurones were closer to firing threshold during pain (e.g. increased descending drive), the stretch would elicit a larger response. Also, the fact that a similar increase was observed during administration of isotonic saline (non-painful) supports our notion that the reflex did not increase in relation to pain.

The tap to the right ES resulted in a short-latency attenuation of EMG activity in the contralateral muscle. Although reciprocal inhibition between paravertebral muscles has been observed in humans (Kugelberg & Hagbarth, 1958; Zedka et al. 1998), the pathways connecting the human axial muscles have not been satisfactorily described. Latency measurements during intracellular recordings from cat motoneurones revealed that, unlike limb antagonists, reciprocal inhibition between contralateral paraspinal muscles is not mediated through a disynaptic pathway (Jankowska & Odutola, 1980). Similar information on reciprocal inhibition between human axial muscles is not available. Even though the results of the present study were not precise enough for accurate estimates of the number of synapses involved in the inhibition, it demonstrated for the first time that the delay between ipsilateral excitation and contralateral inhibition can be as short as 5 ms, suggesting an oligosynaptic pathway.


ES pain induced with hypertonic saline produced changes in the trunk EMG and motor patterns, which were similar to those in real back pain. Deep pain changed descending motor commands but had surprisingly little effect on segmental stretch reflexes.

Pain in the right ES led to a reduction in the velocity and range of voluntary trunk motion, which is consistent with the guarding theory and the ‘pain adaptation’ model of Lund et al. (1991). The pain was associated with a more continuous EMG activity, which may lead to pain perpetuation. When subjects voluntarily overcame this guarding strategy and produced identical trunk motions before and during pain, the changes in EMG persisted. This indicates that the changes involve more than just a strategy to reduce the extent of movement. Since the altered motion pattern exposed the non-painful contralateral ES to similar overload, it could be one of the mechanisms responsible for the spread of pain to neighbouring muscles.

Increased stretch reflex gain (of either monosynaptic or polysynaptic pathways) does not seem to be the mechanism behind increased muscle tone during back pain originating from deep tissues. Our results do not support the idea that muscle spindles are sensitized by fusimotor action elicited by deep pain, at least of the type evoked by saline infusion. In contrast to the lack of effect of deep pain on the ES stretch reflex, cutaneous pain did increase the amplitude of the long-latency response, suggesting facilitation of polysynaptic reflex pathways by cutaneous receptors.


This study was supported by the Alberta Heritage Foundation for Medical Research and the Canadian Medical Research Council. We also thank Drs James P. Lund and Christian S. Stohler for suggestions regarding the infusion control system.


The infusion pump was controlled digitally by a computer which regulated the infusion rate in a feed-forward manner in order to keep the pain approximately constant. The main program was written in Delphi (Borland International Inc.) which linked to the compensating filter running in real time with the use of Matlab Simulink software (The Mathworks Inc.) and controlled the infusion pump via the computer's serial port.

Originally, our intention was to use the closed-loop system described by Zhang et al. (1993). The software for this system was kindly supplied to us by Dr J. P. Lund (Faculty of Dentistry, Department of Physiology, McGill University, Montreal, Quebec, Canada). However, due to difficulties implementing this software on our equipment we opted for a simpler, open-loop control. Preliminary tests showed that in terms of the variability of the subjects’ pain response our results with simple open-loop control were comparable to those of the above-mentioned paper. The characteristics of the open-loop system are described below.

In order to achieve a desired pain response which closely resembles a step response, the unknown response dynamics of the subject must first be identified. The input-output relationship between the pain response and a known infusion rate input was obtained in preliminary tests on two subjects. A 5 % NaCl solution was infused at six different constant rates (50, 60, 100, 140, 150 and 200 μl min−1) in each case for 12 min. The subjects verbally rated their pain every 15 s on a 0-10 scale, where 0 represented ‘no pain’ and 10 was ‘unbearable pain’. A block diagram of the set-up is shown in Fig. 7A. The input signal, a step function, was fed to the digital controller of the infusion pump resulting in a step change in infusion rate. Eight trials were averaged to obtain the mean step response which was then fitted by eye with the following transfer function (TF):

equation image

The exponential term e-45s corresponds to a delay and the other s-terms correspond to differential and integral components. The mean and modelled step responses are shown in Fig. 8.

Once the subject's response dynamics were known, a compensator was designed so that a given input signal would be propagated through the system with the net result that the desired pain response would closely correspond to the input signal. The inverse of the subject's transfer function was used as the compensation element ‘cancelling out’ the effects of the subject's delayed and sluggish response so that the output signal (i.e. perceived pain) closely followed the input signal (Fig. 7B). The infusion rate controller was therefore modelled as the inverse transfer function of the mean pain rating step response (neglecting the delay term) as given by:

equation image

The denominator term (s+ 0.03)4 was added, partly to avoid overshoot and ringing in the response and partly because the Matlab software requires that the denominator be of a higher order than the numerator. In reality, due to approximations in the identification of the subject's response dynamics, the input and output signal are not exactly the same, but do closely resemble each other. Plots of the predicted response to a step input (pain intensity level of 5) and the required control action (flow) to achieve this response are shown in Fig. 9. Compare this predicted response with the actual mean response obtained across all subjects in the experiment (Fig. 10).


  • Ahern DK, Follick MJ, Council JR, Laser-Wolston N, Litchman H. Comparison of lumbar paravertebral EMG patterns in chronic low back pain patients and non-patient controls. Pain. 1988;34:153–160. [PubMed]
  • Allen CEL. Muscle action potentials used in the study of dynamic anatomy. British Journal of Physical Medicine. 1948;11:66–73. [PubMed]
  • Arendt-Nielsen L, Graven-Nielsen T, Svarrer H, Svensson P. The influence of low back pain on muscle activity and coordination during gait: a clinical and experimental study. Pain. 1995;64:231–240. [PubMed]
  • Berberich P, Hoheisel U, Mense S, Skeppar P. Fine muscle afferent fibers and inflammation. In: Schmidt RF, Schaible HGG, Vahle-Hinz C, editors. Fine Afferent Nerve Fibers and Pain. Weinheim, Germany: VCH Verlangsgesellschaft; 1987. pp. 166–175.
  • Carlson H. Observations on the stretch reflex in lumbar back muscles of the cat. Acta Physiologica Scandinavica. 1978;103:437–445. [PubMed]
  • Carlson H, Lindquist C. Exteroceptive influences on the lumbar back muscle tone and reflexes in the cat. Acta Physiologica Scandinavica. 1976;97:332–342. [PubMed]
  • Darton K, Lippold OCJ, Shahani M, Shahani U. Long-latency spinal reflexes in humans. Journal of Neurophysiology. 1985;53:1604–1618. [PubMed]
  • Deuschl G, Schenck E, Lucking CH. Long-latency responses in human thenar muscles mediated by fast conducting muscle and cutaneous afferents. Neuroscience Letters. 1985;55:361–336. [PubMed]
  • Dimitrijevic MR, Gregoric MR, Sherwood AM, Spencer WA. Reflex responses of paraspinal muscles to tapping. Journal of Neurology, Neurosurgery and Psychiatry. 1980;43:1112–1118. [PMC free article] [PubMed]
  • Eklund G, Hagbarth KE, Hagglund JV, Wallin EU. The ‘late’ reflex responses to muscle stretch: the ‘resonance hypothesis’versus‘long-loop hypothesis’ The Journal of Physiology. 1982;326:79–90. [PMC free article] [PubMed]
  • Elert JE, Rantapaa Dahlqvist SB, Henriksson-Larsen K, Gerdle B. Increased EMG activity during short pauses in patients with primary fibromyalgia. Scandinavian Journal of Rheumatology. 1989;18:321–323. [PubMed]
  • Fischer AA, Chang CH. Electomyographic evidence of paraspinal muscle spasm during sleep in patients with low-back pain. Clinical Journal of Pain. 1985;1:147–154.
  • Floyd WF, Silver PHS. The function of the erectores spinae muscles in certain movements and postures in man. The Journal of Physiology. 1955;129:184–203. [PMC free article] [PubMed]
  • Ghez C, Shinoda Y. Spinal mechanisms of the functional stretch reflex. Experimental Brain Research. 1978;32:55–68. [PubMed]
  • Gracovetsky S. The Spinal Engine. Wien, Austria: Springer Verlag; 1988.
  • Graven-Nielsen T, Arendt-Nielsen L, Svensson P, Jensen TS. Stimulus-response functions in areas with experimentally induced referred muscle pain - a psychophysical study. Brain Research. 1997;744:121–128. [PubMed]
  • Hoheisel U, Mense S. Long-term changes in discharge behaviour of cat dorsal horn neurones following noxious stimulation of deep tissues. Pain. 1989;36:239–247. [PubMed]
  • Jankowska E, Odutola A. Crossed and uncrossed synaptic actions on motoneurones of back muscles in the cat. Brain Research. 1980;194:65–78. [PubMed]
  • Johansson H. Reflex Control of γ-Motoneurones. Sweden: Umea University Medical Dissertations; 1981. New Series no. 70.
  • Johansson H, Sojka P. Pathophysiological mechanisms involved in genesis and spread of muscular tension in occupational muscle pain and in chronic musculoskeletal pain syndromes: a hypothesis. Medical Hypotheses. 1991;35:196–203. [PubMed]
  • Jovanovic K, Anastasijevic R, Vuco J. Reflex effects on γ-fusimotor neurones of chemically induced discharges in small-diameter muscle afferents in decerebrate cats. Brain Research. 1990;521:9–94. [PubMed]
  • Kellgren JH. On the distribution of pain arising from deep somatic structures with charts of segmental pain areas. Clinical Science. 1938;4:35–46.
  • Kirkwood PA, Sears TA. Monosynaptic excitation of motoneurones from secondary endings of muscle spindles. Nature. 1974;252:243–244. [PubMed]
  • Kugelberg E, Hagbarth KE. Spinal mechanism of the abdominal and erector spinae skin reflexes. Brain. 1958;81:290–304. [PubMed]
  • Leroux A, Belanger M, Boucher JP. Pain effect on monosynaptic and polysynaptic reflex inhibition. Archives of Physical Medicine and Rehabilitation. 1995;76:576–582. [PubMed]
  • Lund JP, Donga R, Widmer CG, Stohler C. The pain-adaptation model: a discussion of the relationship between chronic musculoskeletal pain and motor activity. Canadian The Journal of Physiology and Pharmacology. 1991;69:683–694. [PubMed]
  • Marsden CD, Merton PA, Morton HB. Servo action in human voluntary movement. Nature. 1972;238:140–143. [PubMed]
  • Matre DA, Sinkjaer T, Svensson P, Arendt-Nielsen L. Experimental muscle pain increases the human stretch reflex. Pain. 1998;75:331–339. [PubMed]
  • Mense S, Skeppar P. Discharge behaviour of feline gamma-motoneurones following induction of an artificial myositis. Pain. 1991;46:201–210. [PubMed]
  • Paintal AS. Functional analysis of group III afferent fibres of mammalian muscles. The Journal of Physiology. 1960;152:250–270. [PMC free article] [PubMed]
  • Prochazka A. Proprioceptive feedback and movement regulation. In: Rowell L, Sheperd JT, editors. Handbook of Physiology, section 12, Exercise: Regulation and Integration of Multiple Systems. New York, USA: American Physiological Society; 1996. pp. 89–127.
  • Schultz AB, Haderspeck-Grib K, Sinkora G, Warwick DN. Quantitative studies of the flexion-relaxation phenomenon in the back muscles. Journal of Orthopaedic Research. 1985;3:189–197. [PubMed]
  • Schwartz G, Stohler CS, Lund JP. Proceedings of the 6th World Congress on Pain. IASP Publication; 1993. The effect of muscular pain on repetitive movements; p. 545.
  • Shirado O, Ito Y, Kaneda K, Strax TE. Flexion-relaxation phenomenon in the back muscles - a comparative study between healthy subjects and patients with chronic low back pain. American Journal of Physical Medicine and Rehabilitation. 1995;74:139–144. [PubMed]
  • Sigiura Y, Terui N, Hosoya Y, Kohno K. Distribution of unmyelinated primary afferent fibers in the dorsal horn. In: Cervero F, Bennett GJ, Headley PM, editors. Processing of Sensory Information in Superficial Dorsal Horn of the Spinal Cord. New York, USA: Plenum Press; 1989. pp. 15–27.
  • Stohler CS, Ashton-Miller JA, Carlson DS. The effects of pain from the mandibular joint and muscles on masticatory motor behavior in man. Archives of Oral Biology. 1988;33:175–182. [PubMed]
  • Svensson P, Houe L, Arendt-Nielsen L. Bilateral experimental muscle pain changes electromyographic activity of human jaw-closing muscles during mastication. Experimental Brain Research. 1997;116:182–185. [PubMed]
  • Tani T, Yamamoto H, Ichimiya M, Kimura J. Reflexes evoked in human erector spinae muscles by tapping during voluntary activity. Electroencephalography and Clinical Neurophysiology. 1997;105:194–200. [PubMed]
  • Travell JG, Simons DG. Myofascial Pain and Dysfunction: the Trigger Point Manual. Baltimore, USA: Williams and Wilkins; 1983.
  • Triano JJ, Schultz AB. Correlation of objective measure of trunk motion and muscle function with low-back disability ratings. Spine. 1987;12:561–565. [PubMed]
  • Veiersted KB, Westgaard RH, Andersen P. Pattern of muscle activity during stereotyped work and its relation to muscle pain. International Archives of Occupational and Environmental Health. 1990;62:331–341. [PubMed]
  • Wall PD, Woolf CJ. Muscle but not cutaneous C-afferent input produces prolonged increases in the excitability of the flexion reflex in the rat. The Journal of Physiology. 1984;356:443–458. [PMC free article] [PubMed]
  • Wenngren BI, Pedersen J, Sjolander P, Bergenheim M, Johansson H. Bradykinin and muscle stretch alter contralateral cat neck muscle spindle output. Neuroscience Research. 1998;32:119–129. [PubMed]
  • Westberg KG, Clavelou P, Schwartz G, Lund JP. Effects of chemical stimulation of masseter muscle nociceptors on trigeminal motoneuron and interneuron activities during fictive mastication in the rabbit. Pain. 1997;73:295–308. [PubMed]
  • Wood SA, Morgan DL, Gregory JE, Proske U. Fusimotor activity and the tendon jerk in the anaesthetised cat. Experimental Brain Research. 1994;98:101–109. [PubMed]
  • Zedka M, Knight B, Prochazka A. The effect of back pain on voluntary and reflex activation of human erector spinae muscles. The Journal of Physiology. 1998;511.P:60P.
  • Zedka M, Kumar S, Narayan Y. Electromyographic response of the trunk muscles to postural perturbation in sitting subjects. Journal of Electromyography and Kinesiology. 1998;8:3–10. [PubMed]
  • Zhang X, Ashton-Miller JA, Stohler CS. A closed loop system for maintaining constant experimental muscle pain in man. IEEE Transactions on Biomedical Engineering. 1993;40:344–352. [PubMed]

Articles from The Journal of Physiology are provided here courtesy of The Physiological Society
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Compound
    PubChem Compound links
  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles
  • Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...