The Pontomedullary Reticular Formation Contributes to the Compensatory Postural Responses Observed Following Removal of the Support Surface in the Standing Cat

Paul J. Stapley, Trevor Drew


This study was designed to determine the contribution of reticular neurons in the pontomedullary reticular formation (PMRF) to the postural responses produced to compensate for an unexpected perturbation. We recorded the activity of 48 neurons in the PMRF, including 41 reticulospinal neurons, to removal of the support surface under each of the four limbs in four cats. The perturbations produced robust postural responses that were divided into three periods: an initial postural response (P1) that displaced the center of vertical pressure over the two diagonal supporting limbs; a secondary response (P2) during which the cat restored a tripedal support pattern; and a prolonged tertiary response (P3) that maintained a stable posture over all three supporting limbs. Most (44/48) reticular neurons showed modified activity to perturbation of at least one limb and a majority (39/48) showed changes in activity to perturbations of more than one limb. A few (7/48) discharged to perturbations of all four limbs. Discharge frequency in neurons showing increased activity during P1 was relatively high (>100 Hz in 57% of the neurons responding to perturbations of either the left or right forelimbs, lFl and rFL) and of short latency (17 ms for the lFL and 14 ms for the rFL). Discharge activity in most neurons was sustained throughout P2 and P3 but at a reduced level. These data show that neurons in the PMRF discharge strongly in response to unexpected perturbations and in a manner consistent with a contribution to the compensatory responses that restore equilibrium.


When standing balance is unexpectedly disturbed by a movement of the support surface, compensatory postural responses in the supporting limbs are generated to stabilize the body center of mass (CoM) and return it to its initial position (Coulmance et al. 1979; Diener et al. 1988; Horak and Macpherson 1996; Massion 1992; Nashner 1977; Rushmer et al. 1987). In the standing cat, postural responses following unexpected perturbations have principally been studied using two paradigms. In the first, the support surface is moved beneath the animal by producing a horizontal or rotational displacement of the whole body or the fore- or hindquarters with the four limbs in contact with the support surface (Macpherson 1988a,b; Rushmer et al. 1983, 1988; Ting and Macpherson 2004). In the second, the postural adjustments in the supporting limbs have been studied when the support surface from under one leg is unexpectedly dropped (Dufossé et al. 1982, 1984, 1985; Rushmer et al. 1987). In both paradigms, the perturbations lead to a relatively stereotypical pattern of responses that serve to displace the CoM to a stable position.

In response to single limb drops in the cat, it is the contralateral limb of the same girdle and that limb's diagonal opposite that primarily serve to maintain weight support (termed the diagonal support strategy; Dufossé et al. 1982; Rushmer et al. 1987). Dufossé et al. (1982) showed that responses in the muscles of the diagonal supporting limbs temporally preceded any change in vertical force under those limbs. In other words, the electromyographic (EMG) changes were centrally generated and not local reflexes produced by mechanical displacement of the other limbs. This led to the formulation of an important concept concerning the nature of the control signals that ensure balance under these conditions: the initial stimulus for a single-limb perturbation is responsible for the muscular activity required to brake the displacement of the perturbed (dropped) limb and the coordinated responses in the other supporting limbs (Dufossé et al. 1982). This has important implications for the neural control of posture because it suggests that a single command coordinates postural responses across all limbs, rather than responses in each limb being determined by four independent signals.

As in most activities, the full expression of an appropriate postural response to a perturbation undoubtedly depends on the coordinated activity of a number of central neural structures including the motor cortex (Beloozerova et al. 2005), basal ganglia (Horak et al. 1984), and cerebellum (Horak and Diener 1994). However, there is abundant evidence to suggest that reticulospinal neurons (RSNs) in the pontomedullary reticular formation (PMRF) might make a particular contribution to the initiation and elaboration of these postural responses. For example, lesions that involve the PMRF or the reticulospinal tract (Afelt 1974; Bem et al. 1995; Brustein and Rossignol 1998; Górska et al. 1990, 1993, 1995; Kuypers 1963; Lawrence and Kuypers 1968a,b; Lyalka et al. 2005) lead to deficits in both static (e.g., standing) and dynamic (e.g., locomotion) postural control. The functional connectivity of the PMRF is also compatible with a role in producing the complex patterns of postural responses produced by an unexpected perturbation. Neurons in the PMRF branch extensively in the spinal cord to potentially influence multiple muscles in several limbs (Matsuyama et al. 1993, 1997; Peterson et al. 1975) and microstimulation in the PMRF both in cats (Drew 1991; Drew and Rossignol 1990a,b) and in primates (Cowie and Robinson 1994; Davidson and Buford 2004, 2006; Quessy and Freedman 2004) may produce simultaneous responses in widespread parts of the body. Moreover, not only do neurons in the PMRF influence muscles in all four limbs but many also receive afferent input from all four limbs (Drew et al. 1996; Eccles et al. 1975; Fields et al. 1975; Peterson and Felpel 1971; Siegel and Tomaszewski 1983). As such, individual neurons in the PMRF have characteristics that are compatible with a contribution to producing the coordinated responses in the supporting limbs following perturbation.

There were two major goals in this study. The first was to extend our information on the biomechanical responses to removal of the support surface to determine the strategy used by the animals to restore their equilibrium. Although the changes in vertical force to perturbations have been previously studied (see earlier text), the perturbations were normally weak and the analysis concentrated on the initial responses rather than the later responses that restore equilibrium. In the present study, we used large perturbations that required pronounced and sustained postural adjustments. In addition, we recorded horizontal forces so that we could better determine how mediolateral and anteroposterior forces acted to restore equilibrium. The second goal was to determine whether neurons in the PMRF discharged in a manner compatible with a role in compensating for the postural instability produced by the perturbation. Neurons that are involved in the compensation to the perturbation should discharge not only in response to the signal indicating the presence of a perturbation but also throughout the changes in activity in the different limbs that result in the restoration of a final stable posture. Given that neurons in the PMRF receive convergent input from multiple limbs and may equally project to multiple limbs, we also wanted to determine whether neurons in the PMRF: 1) discharge to perturbations of multiple or even all four limbs and 2) in the latter case, whether the pattern of activity was similar or different for each perturbation. Given that perturbation of each limb will produce a unique strategy for restoration of posture, a similar pattern of discharge would suggest a more global or generalized contribution to the compensatory postural response, whereas different patterns of activity would be more consistent with a specific contribution to the postural strategy required for each perturbation. Last, we also measured head acceleration during the perturbations to determine whether vestibular input was generated at a sufficiently short latency to contribute to these postural responses.

Some of this work previously appeared in abstract form (Stapley and Drew 2004).


Training and task

Experiments were conducted on a total of four male cats: RS22, RS23, RS25, and RS26, weighing respectively, 6.2, 4.2, 5.5, and 6.2 kg. Cats RS22 and RS23 were previously used in studies by Schepens and Drew (2003, 2004, 2006), whereas data from cats RS25 and RS26 were previously reported in Schepens et al. (2008). These four cats were all trained to maintain and restore balance following an unexpected vertical displacement of one of the four limbs. During these experiments, the cats stood with each paw on one of four force plates, placed on supports that could be dropped in the vertical plane (see Fig. 1A). The system was pneumatically driven and under computer control. Drops of the supporting platforms were sufficient to induce a rapid and complete loss of ground support under the perturbed limb. Figure 1B illustrates the events during a trial. An audible tone (0.5 s in duration) marked the start of the acquisition period and indicated to the animal that it had to stand quietly. After a random period of between 2.0 and 2.5 s, a relay opened, triggering the pneumatic system to produce the drop of support under the limb. Data were collected for a period of ≥3 s after the drop occurred, providing a total acquisition period of about 5 s. Each limb was perturbed 10 times and the order in which the limbs were perturbed was randomized. Because the opening of the relay always produced an audible cue just before the perturbation, we slightly modified the task conditions for cats RS25 and RS26 so that 10 trials were also randomly executed during which the initial 0.5-s warning tone was given and the relay audibly opened but no limb displacement occurred (null trials). In all trials, cats were rewarded by soft food and verbal encouragement. Care was taken that the animals adopted the most natural-looking posture without straining forward or crouching backward by adjusting the position of the forelimb force plates to each cat's preferred stance distance (obtained by recording natural fore–hind paw distances during stance on the floor and during training).

FIG. 1.

A: schematic representation of the behavioral task showing the situation for removal of the support surface under the left forelimb. B: behavioral protocol. The cat received a 0.5-s warning signal 2.0–2.5 s (random duration) before the support surface was removed. Recording continued for a further 3 s. The coordinate system in A indicates the direction of positive force values in the 3 planes. AP, anteroposterior; ML, mediolateral; V, vertical.


Animals were prepared for surgery in aseptic conditions under general anesthesia (2–3% of isoflurane with oxygen using methods similar to those already described in detail; Schepens and Drew 2003, 2004). In brief, 20–24 pairs of insulated stainless steel wires were inserted into selected fore- and hindlimb, axial, and nuchal muscles. The wires were attached to a 51-pin connector cemented to the cat's cranium. In addition, in cats RS25 and RS26 a support was embedded in the acrylic to allow the attachment of three accelerometers (Model 137, Wilcoxin Research), which measured linear accelerations in vertical (V), anteroposterior (AP), and mediolateral (ML) directions. A stainless-steel recording chamber (internal dimensions 10 × 8 mm) was placed over a craniotomy in the occipital bone that provided access to the PMRF. Three microwire electodes (50 μm, Tri-ML insulated stainless-steel) were inserted into the ventrolateral quadrant of the spinal cord (L2) on the same side as the recording chamber (Drew et al. 1986). All surgical and experimental procedures described in these experiments were carried out according to principles outlined by the Council of the American Physiological Society and the Canadian Council on Animal Care and were approved by the institutional body.


After a recovery period of 1–2 wk, experiments were conducted three to five times weekly for 3–4 mo; each session lasted between 2 and 4 h. In brief, a microelectrode was lowered through the cerebellum to a position just above the brain stem. As the electrode was slowly lowered into the brain stem, stimulation (one pulse of 0.2-ms duration each 0.7 s) was applied through the spinal microwire that preliminary experiments showed produced the largest antidromic fields in the PMRF. This constant stimulus intensity was <800 μA for all four cats in the study. Isolated action potentials that discharged antidromically to stimulation of the spinal cord microwire and that fulfilled all conditions of the collision test (Lipski 1981) were classified as RSNs and were tested in the task. If a single neuron was encountered that did not discharge to the test microwire, we tested the other two implanted microwires. If neurons did not discharge antidromically to the initial stimuli then the stimulus intensity was increased until a noticeable twitch was observed in the hindlimb. The maximum current used in the four cats to achieve a twitch ranged from 700 μA to 1.8 mA; the cats never objected to this stimulation. Following completion of testing with one neuron, the electrode was further advanced into the PMRF. If no RSNs were encountered in a penetration, we sometimes recorded unidentified neurons.

Forces exerted at the ground along vertical (FV), anteroposterior (FAP), and mediolateral (FML) axes were recorded using four strain-gauge force platforms (Model ORS6-5-1, 10 × 9 cm; AMTI, Watertown, MA). Throughout the study, forces exerted by animals against the ground at each limb (FV, FAP, and FML) are represented by using the coordinate reference system illustrated in Fig. 1A. Each platform was calibrated along the three axes before and periodically during each experiment. Signals from the platforms were amplified with a gain of 2K and low-pass filtered at 500 Hz. EMG signals were amplified by a factor of between 500 and 10,000 to give a final peak-to-peak signal of ±1 V and band-pass filtered between 100 Hz and 3 kHz. Both EMG and force data were digitized at 1 kHz and stored on a computer for further analysis.

Reflective discs (7-mm diameter) were fixed to the joint centers and the tips of the third digit of each paw of the left forelimb (lFL) and left hindlimb (lHL) at the positions shown in Fig. 1A. Video recordings in the sagittal plane were made using a Panasonic video camera (Model 5100, 60 samples/s; shutter speed 1/1,000 s) positioned such that the optical axis was perpendicular to the left side of the animal. Video recordings were made only of the left side of the animal. A digital time code (SMPTE) recorded onto the recording medium and written into the header of the digitized data files at the beginning of each trial ensured that kinematic, kinetic, and EMG data were synchronized.

Data analysis

For the analysis of neuronal discharge activity, we included all trials in which the cat maintained a steady posture up to the time of the perturbation. This provided us with a range of forces and levels of EMG activity to facilitate correlation of cell activity with behavioral changes (Schepens and Drew 2004). However, in our biomechanical analyses, we used stricter criteria to obtain a representative and reproducible summary of the changes in force and EMG that result from removal of the support surface. For these analyses, trials were excluded from the analysis of the biomechanical effects of limb drops if: 1) they showed a variation in FV traces exceeding 10% of the force exerted by that limb during a period of 500 ms before the onset of the drop (the control period); 2) the difference in body weight supported by the forelimbs was ±12% of that on the hindlimbs or the weight on the left side was ±12% of that on the right side (Macpherson 1988a); and 3) the perturbed foot was replaced on the dropped platform within 1,500 ms of the drop onset.

In all trials, any offset on the force channels was removed in the first instance using a force calibration file recorded at the beginning of the experiments. Trials were also corrected on a trial-by-trial basis by using the known zero values at the dropped limb as the offset. Each trial was then displayed in an interactive window for a period 500 ms before to 1,500 ms after the onset of the drop to identify selected events (see results). Onsets and offsets of muscle activation were determined by visual inspection and corresponded to the moment when activity exceeded ±2SDs of the mean recorded during the control period. All measurements were made with respect to the onset of the drop of the platform, defined as the first detectable change in acceleration of the displaced platform. Acceleration and velocity of the platform were obtained electronically from the derivative of a displacement trace produced by a potentiometer attached to each platform. The center of vertical pressure (CVP) was calculated from the vertical forces exerted by each limb. The mediolateral component of the CVP was calculated as the percentage of the weight distributed on the left homolateral limbs as a function of the total FV. The anteroposterior component was calculated as the proportion of the force distributed under the two forelimbs as a function of the total force.

Cell discharge was analyzed to determine the peak or mean rate of discharge during different parts of the response (defined in results) to the removal of the support surface. It is customary to define changes in cell discharge activity with respect to a stable period of background activity. However, the nature of this task, in which the cat was warned of the forthcoming displacement of the platform by the audible opening of the relay, resulted in a preparatory discharge (see e.g., Fig. 12) that made it impossible to compare activity immediately following removal of the support surface with the background activity preceding the warning cue. We therefore used the following methods to define changes in cell discharge. First, we determined the background discharge from the activity occurring between 1 and 1.5 s prior to the opening of the relay. We then determined whether there was a significant change in activity (±2SD) during the period between the opening of the relay and the drop of the platform (preparatory discharge). In the absence of any preparatory discharge, the initial change in activity following the drop was determined with respect to the background activity. In the presence of a significant preparatory discharge, a significant change in discharge activity following the perturbation was defined with respect to the discharge rate at the onset of the drop ±2SD of the background activity. The two sets of horizontal lines on each of the postevent histograms (PEHs) of Fig. 12 illustrate this method. Adopting this as an objective measure agreed well with our subjective impression of whether a response occurred (see discussion).

Comparison of the discharge frequency and mean latency of the responses following perturbation was made using an ANOVA and post hoc t-test with the Bonferroni correction. Significance was set at P < 0.05.


During and at the end of the series of experimental sessions in each animal, small electrolytic lesions (20–50 μA, 10-s DC cathodal current) were made in selected locations in the PMRF. At the end of the recording sessions, the animals were placed under deep general anesthesia and perfused per aorta with formaldehyde. The brain stem was sectioned in the sagittal plane (40-μm sections) and stained with cresyl violet. Locations of penetrations and recorded neurons were plotted on standard sections of the brain stem (Berman 1968) based on the marking lesions.


General features of limb perturbations: sagittal plane kinematics

Representative sagittal-plane kinematics for the left side of cat RS25 are shown for perturbations of the two left limbs in Fig. 2. Dropping the supporting surface under the lFL or lHL (Fig. 2, A and B, respectively) produced a vertical displacement and extension of the respective limb, accompanied by a rotation of the body around the toes of the ipsilateral limb of the other girdle. The left hind- and forequarters, respectively, were seen to shift backward with respect to the initial position as the dropped limb descended and the body rotated. For drops of the right-side limbs (not illustrated), left-side kinematics showed that there were large changes mainly in flexion of the ipsilateral supporting limb of the same girdle, with the contralateral limb of the opposing girdle extending upward.

FIG. 2.

Stick figures calculated from the video recordings showing changes in sagittal plane kinematics on the left side during perturbations to the left forelimb (A) and left hindlimb (B). For clarity, individual stick figures are shown only every 60 ms starting from the frame immediately preceding the onset of the drop (solid black lines and filled circles) to 1 s after the limb perturbation (dashed lines and open circles).

General behavioral strategy: biomechanical and EMG changes

The biomechanical consequences of the removal of the support surface from under the left forelimb were obtained from 47 selected trials in cat RS25 and from 146 in cat RS26; for the left hindlimb, measures were obtained from 48 trials for cat RS25 and 137 trials from cat RS26. Measures were also made for a total of 88 trials following perturbation of the right forelimb (rFL) and for 83 trials following right hindlimb (rHL) perturbation. Because our initial analysis confirmed that changes following perturbations to the right-side limbs were reciprocal to those following left-side perturbations, they will not be described. All trials used for the biomechanical analysis satisfied the criteria outlined in methods, Data analysis. Across both cats average values of the peak velocity of the dropped platforms ranged from a minimum of 142.3 ± 14 cm/s for the lFL for cat RS25 to a maximum of 199.6 ± 35.5 cm/s for the lFL in cat RS26. Displacements always exceeded 9 cm and were sufficient to ensure loss of contact of the paw with the support surface.

Changes in the vertical force (FV) exerted by each limb following left forelimb and hindlimb drops were very consistent. On the basis of our analysis, we identified three key events (E1–E3) in the vertical force profiles from the four limbs that allowed us to define three corresponding periods (P1–P3) that we used for our analysis of cell discharge.

For the lFL perturbation (Fig. 3A), the first event (E1) was defined by the onset of the initial change in force in the supporting forelimb, in this case the rFL. The peak of the change in force in the rFL was defined as E2. The period between these two events was defined as P1 (Fig. 3C). This corresponds to the initial and rapid change in posture following the perturbation. At approximately the same time that the supporting forelimb was loaded, the hindlimb located diagonally to the perturbed forelimb, the rHL, was unloaded. The moment when this limb was again loaded was identified as E3 and the period between E2 and E3 was identified as P2 (Fig. 3C). During this period, the cat adopted a diagonal pattern of support as described by Dufossé et al. (1982). Subsequent to the reloading of the rHL, the cat maintained its posture by using a tripedal stance. The period from E3 until 1,500 ms after the perturbation was defined as P3. The changes occurring during these three periods are further detailed in the following text with respect to Figs. 4 and 5.

FIG. 3.

Representative vertical forces (FV) exerted at each limb following perturbation of the left forelimb (lFL, A) and the left hindlimb (lHL, B). A and B: the first vertical dashed line indicates the onset of the perturbation and the subsequent 3 vertical gray dashed lines indicate 3 events (E1–E3) used to define different periods of the behavioral response (P1–P3; see text). Data show the responses for a single perturbation in cat RS25. The scale on the top trace in A is valid for all traces in A and B. C: definition of P1–P3 for a perturbation to the lFL. D: average latencies (+SD) of the 3 events identified in A and B for cats RS25 and RS26. lFL, left forelimb; lHL, left hindlimb; rFL, right forelimb; rHL, right hindlimb.

FIG. 4.

Representative electromyographic (EMG) activity recorded from each of the 4 limbs, together with the center of vertical pressure (CVP) and average horizontal force vectors in response to a left forelimb perturbation in cat RS26. For each of the 4 limbs we illustrate EMG activity from a representative flexor and extensor muscle, as well as the ground reaction forces exerted in the AP and ML planes. Scales for the EMG activity are in arbitrary units. Data for each of the 4 limbs are taken from the same, individual trial. At each limb, dashed vertical lines indicate the onset of the perturbation and solid gray lines indicate events 1–3 (see text). P1–P3 indicate the corresponding periods defined in the text. The location of the CVP, from all 47 analyzed trials in cat RS26, is indicated by the different colored circles in the central illustrations. Each circle indicates the CVP from a single trial and the colors represent different parts of the response: in the top center plot, black = quiet stance (at the onset of the drop), blue = E1, and red = E2; in the bottom center plot, the E2 data are repeated (red circles), orange = E3 and purple = 1.5 s after the onset of the drop (see text). The horizontal force vectors, at the corner of the rectangles (each corner representing one limb), use the same color code as for the CVP; these force vectors represent the ensemble average of the 47 trials depicted in the positions of the CVP. ClB, cleidobrachialis; GlM, gluteus maximus; GsM, gastrocnemius, medial head; Srt, Sartorius, anterior head; Tri, triceps brachii, long head.

FIG. 5.

CVP and average horizontal force vectors in response to a left hindlimb perturbation in cat RS26. Data for the CVP are taken from 33 trials selected on the basis of the criteria set for the biomechanical analysis. The figure is organized as for the central portion of Fig. 4.

Changes in FV after removal of the support surface under the lHL were qualitatively similar to those following the FL perturbation with the diagonal support being provided by lFL and rHL (Fig. 3B) and the rFL becoming unloaded, regaining support at E3.

The averaged latencies of these three events following perturbations to either the lFL or the lHL were similar in both cats (Fig. 3D). Latencies for E1 were always <50 ms, for E2 between 135 and 165 ms, and for E3, 348–414 ms (Table 1).

View this table:

Latencies of behavioral events and selected EMGs in each of the four limbs to perturbation of the FL or HL

The postural changes in terms of EMG activity, horizontally oriented ground reaction forces and the shift of the CVP following the perturbation (from quiet stance to 1.5 s after the drop) are illustrated in Fig. 4 for perturbation of the lFL. During the quiet stance period, the CVP was situated centrally between the left and right limbs and just forward of the midline between fore- and hindlimbs and with about 60% of the weight distributed on the two forelimbs (see black circles in top center plot). A slightly higher FV was thus exerted at the forelimbs than that at the hindlimbs. At this time horizontal forces were directed outward and forward at the forelimbs and backward and outward at the hindlimbs (black lines on top center plot).

When the support under the left forelimb was removed, there was an immediate transfer of the CVP backward and to the right (blue circles and associated blue vectors, top center) so that it fell within the triangle of support of the other three limbs. However, the subsequent unloading of the limb diagonal to that perturbed, in this case the rHL, resulted in the CVP moving back to the left and being aligned along the diagonal line between the two supporting limbs (red circles, top center). During this period (P1), the cat was exerting active forces to counteract the perturbation. This is shown by the increase in the horizontal forces exerted by the rFL and the lHL and an increase in the outward force component (FML). The CVP was maintained along this diagonal (orange circles, bottom center) until the cat again began to load the rHL at Event 3. During this period (P2), the cat pushed inward with the rFL and the lHL, as can be seen by comparing the red vectors with the orange ones, to restore the classical tripedal support pattern. At the time of reloading of the rHL (E3) the CVP was again transferred back over the triangle provided by the three supporting limbs (purple circles, bottom center) and continued to move to the right for the next ≥1.5 s (P3). During P3 the CVP attained a stable position within the triangle as the cat maintained its posture until the displaced platform was replaced.

Changes in EMG activity during the perturbation were consistent from trial to trial and from cat to cat. Following the lFL perturbation, there was an initial short-latency increase in activity in the flexors and extensors of the perturbed limb as well as in the three supporting limbs (Fig. 4 and Table 1). In the perturbed limb, the latency of these initial responses in the flexor and extensor muscles ranged from 7 ms in the brachialis in cat RS25 to 27 ms in the long head of triceps brachii (lTri) in cat RS26 (Table 1). Similar short-latency responses in the flexor muscles were also observed in the rFL following the lFL drop. Responses in the two hindlimbs, one loaded and one unloaded, were slightly longer than those observed in the forelimbs.

Responses in the flexor muscles in the two forelimbs were normally of short duration and terminated at approximately the time that peak force in the diagonal supporting limbs was attained (E2; see Fig. 4). In the rHL this flexor muscle activity continued until the limb was reloaded, suggesting that the unloading was active and not simply passively caused by the change in body orientation produced by the loss of the support surface. Increased extensor muscle activity in the rFL was sustained throughout the postural response, whereas extensor muscle in the lHL was increased until reloading of the rHL was complete, suggesting that the unloading in this limb was an active process. Extensor muscle activity in the rHL was initially decreased and then showed an increase in activity.

Qualitatively similar responses were obtained following lHL perturbation (Fig. 5) with the obvious proviso that the CVP was initially shifted forward over the forelimbs. Moreover, hindlimb drops resulted in smaller changes in the horizontal forces produced by the forelimbs compared with the FL drops. Drops of the lHL equally resulted in an analogous organization of EMG to those recorded when the forelimbs were dropped (not illustrated). The earliest short-latency responses in the hindlimb flexor and extensor muscles occurred at comparable latencies to those evoked in forelimb muscles following lFL drops (Table 1). However, responses in the forelimb muscles were at a longer latency than those in the hindlimbs. As for the forelimb drops, there was a sustained increase in activity in the extensor muscles of the limb on the same girdle (rHL) as the perturbed limb (lHL) and a smaller-magnitude period of sustained activity in the extensors of the diagonal supporting forelimb. One notable difference between FL and HL drops was the inward shift of the horizontal vectors at the limb of the same girdle (rHL) at E2 and E3 during the lHL drop. This is in contrast to the forward and outward vectors produced at the rFL at the equivalent time periods for the lFL drop.

Acceleration of the head in all three planes was observed following perturbation of both the lFL and the lHL (Fig. 6, A and B). However, the magnitude of the responses was substantially larger following the FL perturbation. The largest accelerations were in the AP and the V planes following forelimb perturbation and in the AP plane following hindlimb perturbation. The latency of the onset of the head acceleration (Fig. 6C) was comparable to that observed in the forelimb flexor and extensor muscles following the lFL perturbation. However, t-tests comparing onset latencies following HL perturbation in all three planes (each cat tested separately) were significantly longer than those recorded during FL perturbations for all axes and for both cats (P < 0.001).

FIG. 6.

Five representative trials from cat RS26 of head accelerations recorded during vertical limb perturbations of the left forelimb (A) and left hindlimb (B) as well as average onset latencies of head acceleration in the 3 axes (ML, AP, and V), for left forelimb (C) and left hindlimb drops (D) both in cat RS25 and in cat RS26. The vertical dashed lines on A and B indicate the onset of the perturbation. In C and D, mean onset latencies plus 1SD are shown. Data in C and D are compiled from all measured trials.

Neuronal activity in the PMRF

Database and histology.

We recorded neuronal discharge activity from 48 cells in the left reticular formation of four cats during removal of the support surface from under all four limbs. Forty-one (41/48) of these neurons were identified as RSNs. These 48 neurons were obtained from cats used in previous experiments designed to study activity during reaching with the forelimb (Schepens and Drew 2003, 2004, 2006; Schepens et al. 2008). Most cells were recorded from cats RS25 and RS26 that were implanted specifically for these experiments; supplementary data were obtained from cats RS22 and RS23. Histological localization was possible for 45/48 of these cells. Most of these neurons (36/45) were recorded from the nucleus reticularis gigantocellularis, with others (5/45) recorded in the nucleus reticularis magnocellularis or the nucleus pontis caudalis (4/45) (Fig. 7A). The conduction velocities of this population of neurons ranged from 39 to 137 m s−1 (mean ± SD: 107 ± 22 m s−1) (Fig. 7B). Our analyses showed no noticeable differences between the properties of the 41 RSNs and the 7 unidentified cells and they are thus considered together in the following text.

FIG. 7.

A: histological reconstruction of the location of the reticular neurons recorded in this study. The neurons are plotted on a single, representative brain stem section at laterality (L) of 1.2 mm taken from the atlas of Berman (1968). Note that only the locations of the 45 cells that we could positively identify are illustrated. B: conduction velocity of those neurons identified as projecting to the spinal cord. The conduction velocity is calculated as the distance (mm) from the obex to the location of the stimulating electrodes as measured in each cat postmortem divided by the latency (ms) of the antidromic activation. 7G, genu of the facial nerve; IO, inferior olive; N, number of cells; NRGc, nucleus reticularis gigantocellularis; NRMc, nucleus reticularis magnocellularis; NRPc, nucleus reticularis pontis caudalis; NRPo, nucleus reticularis pontis oralis; PH, nucleus praepositus hypoglossi; TB, trapezoid body.

Discharge activity during P1.

Changes in activity immediately following the removal of the support surface (Period 1) were observed in 44/48 of these cells to perturbation of at least one limb. Changes in activity following drop of the support surface under the lFL were observed in 35/48 neurons. The most common effect was an increase in discharge activity (21/48 neurons, Fig. 8, A and B, Table 2), with a decrease in neuronal discharge frequency during P1 being observed in 14/48 cells (Fig. 8, C and D).

FIG. 8.

Responses of 4 different reticulospinal neurons (RSNs) to perturbation of the lFL. Data are synchronized to the onset of the drop (vertical solid line). For each cell we illustrate a postevent histogram (PEH) showing the averaged display together with raster displays of cell discharge during each of the individual trials. N, number of trials recorded for each cell during lFL perturbation. Dotted lines delimit the approximate time of the 3 periods (P1–P3) of the behavioral response used in our analysis and defined with respect to Fig. 3. Note that P1 starts shortly after the drop (Table 1 and Fig. 3) but that event E1 is not marked with a dotted line to avoid undue clutter. Filled rectangles on the raster displays preceding the onset of the perturbation indicate the time that the relay opened. Vertical force traces illustrated in D were recorded simultaneously with the cell illustrated in D.

View this table:

General patterns of modification of unit activity in response to perturbations

In most neurons (39/48), changes in neuronal activity immediately following the perturbation were observed following removal of the support surface from more than one limb (Fig. 9), although the response to perturbation to each limb was not necessarily identical (Fig. 9D). The most common response (30/48 neurons) to a drop of the right (contralateral) forelimb was an increase in discharge activity. This was observed regardless of whether the response to removal of the support surface from under the left forelimb was an increase (Fig. 9B) or a decrease (Fig. 9D). Decreases in activity following removal of the support surface from under the right forelimb were observed in only 7/48 neurons. Changes in activity to removal of the support surface from under the left and right hindlimbs were also observed in a majority of neurons (increased activity in 26/48 for the lHL and for 26/48 for the rHL; see Table 2).

FIG. 9.

AD: example of the discharge activity of 4 RSNs to perturbation of each of the 4 limbs. For each perturbation, we illustrate a PEH of the averaged discharge activity and a raster display showing trial-by-trial activity. Data are synchronized to the onset of the drop (first, vertical dashed line). The second, dotted, vertical line separates P1 and P2 from P3. E: changes in FV recorded from each of the 4 limbs to each perturbation; these data were recorded simultaneously with the cell illustrated in D.

Figure 10 uses a Venn diagram to summarize the frequency with which cells responded to removal of support from one or multiple limbs with an increase in activity during P1. Two representations are used. In the first (Fig. 10A), we illustrate the unique relationships of each cell to perturbations of each limb. This serves to illustrate that relatively few cells showed increased discharge to perturbation of only a single limb (e.g., three discharged only to perturbation of the rFL), whereas substantially more discharged to perturbations of multiple limbs (e.g., seven discharged to perturbation of all four limbs). There was no change in activity following removal of the support surface in four neurons and five others showed only decreases in activity. One cell responded only to perturbation of the lFL and rHL and could not be represented in this modified Venn diagram.

FIG. 10.

Modified Venn diagram showing the number of cells showing a significant increase in discharge frequency to perturbation of each limb during P1. A: illustration of the cells showing a response only to the indicated limb or combination of limbs. Note that only 38 cells are included because 4 cells did not discharge at all to the perturbations, 5 cells showed only a decrease in activity, and one cell was activated by perturbation of the lFL and the rHL. B: illustration of the number of times that a cell was activated by a given limb or combination of limbs, regardless of the overall pattern. In other words, a neuron responding to the lFL, rFL, and lHL would be represented only once in A in the appropriate area of the Venn diagram but would be represented 6 times in B for each of the possible combinations.

The second representation (Fig. 10B) illustrates the number of cells that discharged to a given limb, or combination of limbs, regardless of the exact pattern of discharge (e.g., 21 cells showed increased discharge to perturbation of the lFL). Increased discharge frequency in response to removal of the support from at least two limbs was frequently observed, both for limbs of the same girdle as well as between them. For example, increases in activity following removal of the support surface from either of the forelimbs was observed in 14/48 neurons and was equally observed in 20/48 neurons following removal of the support surface from either hindlimb. A similar number of cells (11 and 18) were activated by removal of the support surface from the diagonal limbs (not illustrated in Fig. 10B). Reciprocal changes in activity in the lFL and the rFL (see Fig. 9D) were observed in 11 neurons; in 10/11 of these there was a decrease in activity to drop of the lFL and an increase to drop of the rFL.

Discharge frequencies during P1 were relatively high. Indeed, averaged peak frequencies following perturbation to the lFL exceeded 100 Hz in 12/21 neurons (mean 132 ± 68 Hz, Table 3) that showed increased discharge activity following the perturbation. Discharge frequencies to perturbation of the rFL were equally intense, with the mean discharge of the 30 neurons that increased activity following the perturbation of the rFL being 137 ± 83 Hz. However, discharge frequency to the rFL perturbation exceeded that in the lFL in 12/14 neurons in which excitatory responses were observed following perturbation to both the lFL and the rFL (Fig. 11A). In contrast, cells responding to perturbation of the lHL or rHL consistently discharged at lower frequencies compared with the activity during the lFL perturbation (Fig. 11, B and C; Table 3). Discharge activity to perturbation of the rHL was greater than that to perturbation of the lHL in 13/20 cases (Fig. 11D). Overall, the mean discharge frequency following the lHL perturbation was significantly less than that to perturbation of either the lFL or the rFL.

FIG. 11.

Discharge frequency of reticular neurons during P1 to perturbation of each limb. AD: each graph plots the discharge frequency following perturbation of one limb as a function of the discharge frequency following perturbation of a second limb. The diagonal line indicates discharge frequencies that would be identical for each perturbation.

View this table:

Average discharge frequency (spikes/s) of reticular neurons in response to perturbations

Mean latencies of the increases in response activity ranged from 14 ± 6 ms for the rFL to 21± 5 ms for the rHL. An ANOVA showed a significant effect on latency as a function of the limb perturbed but post hoc t-test with Bonferroni adjustment showed only one pairwise difference, the mean latency for the rFL perturbations being significantly shorter than that of the rHL perturbation.

Responses during P2.

Following the initial large changes in neuronal discharge activity there were more modest changes in activity during the immediately subsequent period (P2) in which weight support was modified from a bipedal to a tripedal support pattern (Figs. 35). These changes in discharge activity were either the same sign (increase or decrease) as the initial change in activity, as in Fig. 8, AC, or of opposite sign, as in Fig. 8D. Note, however, that the change in discharge activity for any one neuron during P2 was normally the same for perturbation to each limb, despite differences in activity during P1. For example, 11/14 of the neurons that showed reciprocal patterns of activity to perturbations of the lFL and rFL during P1 showed increases to perturbation of either limb during P2.

Indeed, as for the short-latency responses, discharge frequency was most commonly increased during P2, frequently following perturbations of each limb. Overall, 32/48 neurons showed increased activity following removal of the support from the left forelimb during P2 and 30/48 during removal of the support surface under the right limb (Table 2). Of 48 neurons, 24 showed increased activity to removal of the support surface under either forelimb. Only one neuron showed a reciprocal pattern of activity to removal of the forelimb support. A similar pattern of activity was observed during the removal of the support under the hindlimbs (Table 2). Nineteen (19/48) neurons showed an increased level of activity to removal of the support under the left hindlimb and 20/48 to removal of the support surface under the right hindlimb; 11/48 responded to removal of the support surface under either hindlimb and 4/49 neurons showed a reciprocal pattern of activity. Eight neurons responded with an increase in activity to perturbation of all four limbs. Mean discharge rates were substantially lower than those evoked during P1 and averaged values ranged from 28 to 41 Hz (see Table 3).

Responses during P3.

Period 3 defines the time of the major compensatory responses to the perturbation. It is defined as the time from restoration of the full tripedal support pattern (Event 3) until a period 1,500 ms following perturbation onset. At the end of this period, the cat had normally obtained a stable position in which the CVP remained constant within the triangle defined by the three supporting limbs (Figs. 4 and 5).

Sustained changes in discharge activity were normally observed throughout P3 as illustrated in Fig. 9, although slightly fewer cells showed a significant increase in activity compared with P2 (Table 2). Again, the most frequent response was an increase in activity following perturbations of the lFL (29/48), with a slightly smaller number of neurons (23/48) showing increased activity to removal of the rFL support (Table 2). Seventeen (17/48) showed increased activity to removal of the support under either forelimb (e.g., Fig. 9C) and only 2/50 showed a reciprocal pattern of activity (e.g., Fig. 9D). In most cells discharge frequencies during P3 were greater to perturbation of one limb than to the other three (e.g., Fig. 9, A and B). Discharge rates during P3 were slightly lower than those during P2, with averaged rates ranging from 27 to 32 Hz (Table 3).


For the responses during each of the three periods (P1–P3), we used linear regression analyses to determine whether there was a relationship between cell discharge frequency and muscle activity. For this analysis, we calculated average cell discharge frequency during either 20-ms (P1 and P2) or 50-ms (P3) bins and integrated EMG activity during this same period for each individual trial (see Schepens and Drew 2004). Regressions were initially made from all trials, irrespective of the limb perturbed, to determine whether there was a fixed relationship between cell and muscle activity. As might be expected, given the strong short-latency responses in flexor and extensor muscles of all four limbs following the perturbation and the high discharge frequency during this period, several of the linear regressions during P1 gave relatively strong correlations with at least one muscle. Altogether, in 9/48 cells the coefficients of determination obtained (R2) exceeded 0.5 and in 14/48 cells they exceeded 0.3. In general, these 14 cells were those discharging most intensely to the perturbation during P1.

In contrast to the results obtained during P1, the values of R2 obtained for the regressions made during P2 and P3 were relatively low and none exceeded 0.5. Particular attention was paid to the extensor muscles during these two periods because these muscles make a large contribution to the restoration of equilibrium (Fig. 4). For the left lateral head of the triceps brachii (lTriL) during P2, only 2/42 regressions were >0.3 and for the rTriL, 2/43 were >0.3. During P3, only 1/50 regressions for the lTriL was >0.3, whereas for the rTriL 3/47 were >0.3. We also examined the responses evoked during P2 and P3 only by perturbation of the left forelimb to see whether these were substantially different from those obtained by grouping together the perturbations to each limb. For example, increases in the right long head of the triceps brachii (rTri) and rTriL were consistently seen throughout P3 to left forelimb perturbation (see Fig. 4). However, analyses showed that R2 was not substantially different from that when considering all perturbations together. Indeed, for only four neurons did R2 for the rTri or the rTriL exceed 0.3 (three positive correlations and one negative). Moreover, visual inspection of the correlations for both P2 and P3 showed no consistent relationships between cell activity and any given muscle or force trace. This is quite different from the regressions observed in our reaching study in which 51% of neurons showing a maintained discharge during the reach showed an R2 of >0.3 with the extensor muscle EMG in the supporting forelimb (Schepens and Drew 2004).


We applied spike-triggered averaging to all recorded cells. Postspike responses were observed in only three cells. All three of these showed postspike depression in one or two ipsilateral extensor muscles; one also showed facilitation of the ipsilateral flexor muscle, sartorius, and another also showed postspike depression in the contralateral vastus lateralis.


As mentioned in methods, the triggering of the relay responsible for activating the pneumatic system resulted in the production of an audible signal that frequently produced a modification of cell activity prior to the onset of the perturbation (e.g., Figs. 8, A and B and 9, B and D). Indeed, a significant preparatory discharge (exceeding ±2SD of the control activity) was observed in 40/48 neurons. However, although the cat received this audible signal informing it of the occurrence of a perturbation, it had no prior information as to which limb would be perturbed. Thus as might be expected, there was no difference in the sign or the magnitude of the preparatory discharge according to which limb was perturbed (see e.g., Fig. 9B). Increases in discharge during the preparatory period were observed in 36/48 neurons and decreases in only 4/48. On average, the relay preceded the perturbation by 237 ± 18 ms and the averaged latency of the change of discharge activity with respect to the onset of the relay was 82 ± 51 ms; discharge onset therefore occurred an average of 152 ms prior to the onset of the drop of the platforms. The peak magnitude of the discharge ranged from a minimum of 8 Hz to a maximum of 147 Hz (average = 62 ± 44 Hz) and, for facilitatory responses, was proportional to the magnitude of the response evoked during P1 (not illustrated).

In 25/36 cells showing a significant increase of discharge activity during at least P1, we randomly inserted null trials in which the relay was triggered but the platform was not displaced. In these trials the response rapidly decayed so that baseline activity was restored within 286 ± 48 ms following the onset of the relay. Given that the onset of the displacement occurred at an average of 237 ms following the onset of the relay, this means that the preparatory response was completely lost within 49 ms of the expected onset of the platform drop.

Despite the large changes in neuronal activity observed in many cells, there were only modest changes in EMG or force activity during this same period. This is illustrated for three cells, from three different cats in Fig. 12. Each of these cells shows clear preparatory activity subsequent to the activity of the relay and prior to the perturbation of the limb. However, clear changes in force were observed only in the mediolateral component of the forelimbs and these were relatively modest (<1 N). Moreover, the AP and ML components of the CVP showed no obvious changes in activity. Indeed, the examples shown in Fig. 12 were among the clearest examples of changes in EMG activity or force levels that we observed during the preparatory discharge. In many cells, there were no obvious changes in activity in any of the traces that we recorded and in those in which we did observe changes they were normally limited to small changes in FML with no or little change in CVP.

FIG. 12.

Responses of 3 RSNs during the time between the opening of the relay and the onset of the drop. Rasters and PEHs are illustrated together with representative EMG and force traces and are aligned to the onset of the drop. Note that all traces are compiled from all perturbations to all 4 limbs. As such, the EMG and force traces prior to the drop show the ensemble anticipatory activity and the activity following the drop is not representative of the responses to perturbation of any single limb. EMG scales are arbitrary units. The 1st set of horizontal lines on the PEHs indicates the control discharge activity (calculated prior to the activation of the relay); these values are then transferred to the discharge just prior to the perturbation (2nd set of horizontal lines) to indicate the method used to determine the presence and the latency of the discharge during P1 (see methods). Thick vertically oriented lines on the rasters indicate the opening of the relay.


To our knowledge this is the first study to examine neuronal responses in the PMRF of intact cats in response to an unexpected perturbation of body position. The results show that removal of the support surface from under a limb results in robust changes in discharge frequency that are maintained throughout the time that the cat compensates for the threat to its equilibrium. Changes in activity were observed in a majority of the cells that we examined and many of these cells showed modified discharge activity patterns in response to perturbation of more than one limb; in many cases the signs of the responses (increase or decrease) were similar for the perturbations of each limb. The results confirm that the PMRF may contribute to the compensatory postural responses that follow an unexpected perturbation, but argue against a contribution to the specification of the detailed postural response required for this compensation.

Biomechanical changes following perturbation

The changes in kinematics, vertical forces, and EMG activity produced by unexpected loss of support under one limb of the cat have been previously documented in only a few studies (Dufossé et al. 1982, 1984; Rushmer et al. 1987). Although our perturbations were much larger and faster than those of Rushmer et al. (1987), in general our results are similar in most of the measures that are comparable. As in their study, we found that the initial changes in vertical force in the unperturbed limbs was delayed by 25–45 ms following the perturbation because of the compliant nature of the body compared with a true rigid body. Similarly, the initial changes in EMG activity, in all three supporting limbs, occurred at short latency and preceded, or occurred at a similar time to, the changes in vertical force in these limbs (Table 1). As emphasized by several authors (Dufossé et al. 1984; Massion 1992; Rushmer et al. 1987) this is an important finding that indicates that the initial postural changes are initiated in response to the initial perturbation and are not simply local changes produced by subsequent perturbation of each limb. Again, as in the study of Rushmer et al. (1987), we found that the responses to the perturbations of the forelimbs produced larger changes than those produced by the hindlimbs, presumably because of the larger proportion of the weight of the cat that is carried over the forelimbs because of the weight of the head.

Our results extend those previously published both by examining the changes in horizontal force that make a large contribution to the postural changes that restore equilibrium and by determining the strategies that progressively result in a stable body posture. These analyses allowed us to operationally define three periods of restoration: an initial period (P1) during which the cat produces cocontraction of multiple flexor and extensor muscles in all four limbs, a secondary period (P2) during which the cat modifies the support pattern from one that is predominantly diagonal and bipedal, and a tertiary period (P3) during which smaller adjustments in posture are made to attain and maintain a stable tripedal support pattern. These three periods of postural adjustments are the result of complex modifications of EMG activity in all four limbs that result in the production of the appropriate horizontal forces under each foot required to restore equilibrium. We suggest that the changes in discharge activity of the reticular neurons recorded in this study contribute to these three phases of adjustment.

Neuronal recordings

The majority of the neurons that we recorded (44/48) showed a change in discharge activity to perturbation of at least one limb. In all of these 44 neurons, there was a short-latency change in activity that occurred during P1. This change in discharge activity, on average, preceded any change in EMG activity in the limb and is likely to be directly produced by the perturbation itself. The short latency of this activation (14–21 ms) is comparable to the latencies that we observed in a previous study to electrical stimulation of peripheral nerves (mean values, 12.0 for the lFL, 15.5 for the rFL, 19.0 for the lHL, and 18.0 ms for the rHL; Drew et al. 1996). We assume that the activation that we see in this study is equally the result of activation of peripheral afferents in the limbs. The short latency of this neuronal discharge makes it unlikely that the initial discharge activity can be the result of activation of vestibular afferents because our recordings show that the mean average latency of the onset of head acceleration was 15 ms for perturbations of the FLs and 57–70 ms for perturbations of the hindlimbs. In contrast to the large difference in latency of head acceleration following FL versus HL perturbation, the latencies of the cell onset latencies were more comparable, supporting the view that they were initiated by activation of peripheral afferents in the limbs. However, vestibular input could certainly contribute to later parts of the neuronal discharge pattern.

Because many of the neurons recorded in this study were antidromically activated from the lumbar spinal cord, it is probable that they contribute to a spino-bulbo-spinal (SBS) reflex pathway as originally detailed in the studies of Shimamura (Shimamura and Kogure 1979; Shimamura and Livingstone 1963; Shimamura et al. 1990) and discussed also with respect to our locomotion study (Drew et al. 1996). Given the fast conduction velocity of the axons of most of the recorded cells (Fig. 7) these reticular neurons would be able to influence EMG activity in the forelimbs within 4–5 ms of the start of the change in neuronal activity and could thus contribute to the production of even the earliest postural adjustments following the perturbation.

In this respect, it is important to note that the initial responses to the perturbation (during P1) were modulated according to the limb that was perturbed and varied from one cell to another. This is an indication that the initial change in the discharge activity is not a general signal that simply informs the reticular formation of some event that has occurred in the periphery. Rather, even at this early stage following the perturbation, the discharge activity in the cells in the PMRF is most likely adapted to contribute to an appropriate postural response specific to the limb that is perturbed. For example, in a very few cells (5/48), there was a significant response only to perturbation of a single limb (Fig. 10A). In other cells that responded to perturbation of more than one limb, discharge activity was facilitated by perturbation of one or more limbs and depressed by perturbation of another (17/48). Last, even in cells in which responses were facilitated by perturbation of two or more limbs, the discharge frequencies to perturbation of each limb could vary quite considerably (see e.g., Fig. 9). Moreover, although discharge frequencies were normally greater for forelimbs than for hindlimbs and greater for the right (contralateral) forelimb than for the left forelimb, exceptions were sometimes observed, again supporting the view that both the sign and the magnitude of the responses are determined both by the limb perturbed and the characteristics of the cell recorded. This is similar to the results obtained in our study of the effects of peripheral nerve stimulation during locomotion (Drew et al. 1996).

The initial response to the perturbation (P1) was nearly always larger than the discharge frequency at any other time following the perturbation. This presumably reflects the fact that the perturbation evoked a strong, synchronized afferent volley. We suggest that the summed activity of all of the reticular neurons activated by this perturbation serves to ensure that the CoM is displaced over the supporting limbs. This initial transfer of the CoM can probably be viewed as a default strategy that ensures that equilibrium is maintained in the face of a major threat to stability. Subsequent to this initial response, the cat begins to actively modify its CVP to improve its stability. These secondary postural adjustments are composed of two periods as detailed in results: 1) period 2, which restores the tripedal stance; and 2) a much slower period, P3, which adjusts the CVP until it reaches a stable value indicating that the CoM is now positioned within the supporting three limbs. It is likely that these secondary postural adjustments rely on slightly different mechanisms and signals than the initial postural response. This suggestion is based largely on the fact that the sign of the neuronal response evoked following the perturbation may sometimes be inversed (e.g., Fig. 9D).

The neuronal activity in any one cell during P2 and P3 normally showed the same sign of response, although the magnitude was normally greater during P2 than that during P3. The function of the discharge during these two periods would seem to be one of facilitating the transfer of the CoM to a stable position within the triangle of the supporting limbs and ensuring that weight is equally supported by each of the three supports of this tripod. Discharge activity in P2 is presumably higher than that in P3 because it is during the P2 phase that the major transition from a bipedal to a tripedal support occurs. Subsequently, once the full tripedal support is obtained the transition to a stable CoM is relatively slow and probably does not require high phasic levels of cell activity as during P1 and, to a lesser extent, P2.

One question that arises from consideration of the pattern of cell discharge during this task is the extent to which one may consider that the discharge activity in these reticular neurons specifies the postural pattern required to restore a stable posture following perturbation to different limbs. For example, it is clear that the pattern of postural activity following perturbation of one forelimb is quite different from that produced by perturbation of the other forelimb. If a given cell signals a specific postural pattern it should discharge in response to perturbation of one of the forelimbs and not the other. Although this pattern of activity was observed in some cells, a large number of cells showed increased activity during both P2 and P3 in response to perturbation of either forelimb (Figs. 9 and 10). Further, a small proportion of cells discharged to perturbation of three limbs and some to perturbation of all limbs (Fig. 9). Clearly, such cells cannot specify a complete postural pattern for each perturbation because each adjustment is different. How then do we reconcile these results?

One possibility is that these cells are specifying activity in muscle groups that show the same response to each of the perturbation, perhaps, for example, in axial muscles. We cannot rule out this parsimonious explanation. However, our linear regression analyses failed to show any strong relationships when responses to perturbations of each limb were combined, even though the muscles tested with the analyses included axial muscles such as the biventer cervicus and the longissimus dorsi. Moreover, many of these cells had receptive fields on the limbs (and responded strongly and at short latency to perturbation of these limbs) and microstimulation at these sites produced movement of the limbs, in addition to head and/or trunk movements. It thus seems unlikely that the cell discharge would be related to only a part of the muscle field available to the cell.

An alternative explanation, as we have suggested before, is contextual gating of the descending signal. During locomotion, many reticular neurons discharge during voluntary gait modifications of each of the four limbs, despite the need for different postural patterns, as in this task (Prentice and Drew 2001). We have suggested that in this situation the descending signal is modified by the excitability of neurons that form part of, or are influenced by, the central pattern generator (CPG) for locomotion. Similarly, during reaching many cells also discharge during movement of each limb (Schepens and Drew 2004, 2006; Schepens et al. 2008). In this situation, we have suggested that the descending command for the voluntary reaching movement may preset the excitability of the same populations of spinal interneurons responsible for postural changes during locomotion. In our current task, it is possible that a similar mechanism is engaged. In this case, the perturbation will provide a strong activation of peripheral afferents that may, itself, serve to modify the excitability of spinal neurons onto which the reticular neurons impinge. For example, even in the spinal cat a perturbation will result in the facilitation of ipsilateral flexors and contralateral extensors (the crossed-extensor response).

In all of these examples, we assume that there are different populations of reticular neurons with differing functions. At one extreme, some cells respond to perturbation of only one limb, to voluntary gait modifications of a single limb, or, less frequently, to voluntary reaching movements of one limb. We consider that these cells could specify a particular pattern of postural activity. At the other extreme are the reticular neurons that discharge to perturbation of all four limbs, to voluntary gait modifications of all four limbs, and to reaching of either limb. These cells may provide a more general signal specifying the time and the magnitude of the signal, with the final expression of the pattern depending on the excitability of spinal neurons preset by a CPG, a command for voluntary movement, or modification of reflex excitability. We would surmise that such cells should have wide branching axons that innervate muscle groups in different limbs (Matsuyama et al. 1993, 1997; Peterson et al. 1975).

Preparatory activity

In these experiments the cat obtained information concerning the presence of a perturbation prior to its onset. The relatively constant time between this auditory signal and the perturbation (∼240 ms) allowed the cat to quite accurately predict when the perturbation would occur. This statement is supported by the finding that in the null trials, in which no perturbation occurred, the preparatory signal decayed within roughly 50 ms of the expected arrival of the perturbation. We consider this signal to be one of preparation for a perturbation rather than a startle response because of the relatively long latency between the preparatory signal and the neuronal response. The average latency of the change in modulation was about 80 ms, whereas startle responses in reticular neurons typically begin within 3–10 ms of the stimulus (Wu et al. 1988; Yeomans and Frankland 1996). Moreover, auditory startle responses are normally considered to be induced by loud, unexpected noises (>80 dB) and habituate to repeated exposure. In these experiments, the noise was barely audible and there was no sign of habituation over several months.

As mentioned in methods, perturbations were applied in a random order and it was not possible for the cat to predict which limb was going to be perturbed. As a result the preparatory activity prior to perturbation of each limb was similar (see e.g., Fig. 9). This effectively means that the preparatory activity cannot be specific to the perturbation. Rather, we suggest that this activity is a more general preparation for a perturbation. The fact that so many cells showed this activity suggests that it should have resulted in a cocontraction of muscles in all four limbs, perhaps resulting in a stiffening of the body. This suggestion is supported by the fact that microstimulation of the PMRF of the awake cat at rest frequently results in coactivation of flexors and extensors in all four limbs (Drew and Rossignol 1990a,b). However, our analysis of changes in behavioral measures during this period rarely showed large changes in either EMG activity or force levels during the preparatory activity (Fig. 12). The reasons for this are not clear because microstimulation in the regions in which cells were recorded frequently produced strong motor responses. Possibly, the levels in EMG activity were distributed among many muscles and were not obvious in any one example. Further, if muscle activity was coactivated among flexors and extensors, the final net force activity would be unchanged. In this respect, it should be noted that our previous studies during reaching frequently showed large changes in neuronal activity prior to the movement with noticeable changes in force but sometimes in the absence of noticeable changes in EMG activity (Schepens and Drew 2004, 2006). This emphasizes the difficulty in determining preparatory activity simply on the basis of selected EMG recordings. Alternatively, it is possible that other groups of neurons were activated that inhibited movement so that no overt behavioral activity was evoked until the onset of the perturbation.

One question that is raised by the presence of this preparatory activity is whether it might have biased the cell activity in response to the perturbation. We do not believe that this is the case. First, responses during P1 normally showed a clear deviation from the preparatory activity in the preceding period (Fig. 12), suggesting that the signal produced by the perturbation was responsible for the discharge activity of the cells at this time. Second, in several cases, the sign of the response evoked during P1 was opposite that observed during the preparatory period (Fig. 9B). This suggests that the preparatory activity was a default strategy, as stated earlier, that did not prevent the production of a response specific to the limb that was perturbed. Third, responses were not qualitatively different in neurons that did not show preparatory discharges. We cannot, of course, exclude that the preparatory activity may have modified the magnitude of the response during P1 but, even so, it should have had an equal effect on the response to perturbation of each limb. Moreover, the fact that the preparatory response decayed within about 50 ms of the expected time of the relay means that there should have been no effect on the responses measured during P2 and P3.


There is a large body of work describing postural responses in animals and humans in response to perturbation of the support surface (Horak and Macpherson 1996). However, the neuronal mechanisms that underlie these postural responses have been scarcely studied and much of our information on such neuronal mechanisms has come from studies on patients with various pathologies (e.g., Horak and Diener 1994; Horak et al. 1984; Viallet et al. 1992). The results in this study strongly suggest that the PMRF contributes to the production of the compensatory postural responses that follow perturbations of the support surface and may function as a final modulatory output system for postural responses both in response to perturbations and during voluntary movements that destabilize body equilibrium (Schepens and Drew 2004, 2006). The results from this study extend and complement those from other studies implicating the PMRF in the control of posture. For example, the studies of Mori (Mori 1987; Mori et al. 1982, 1992) clearly showed that the PMRF modifies muscle tone in all four limbs and the studies of Takakusaki et al. (1994, 2003) have elucidated some of the pathways involved in this modification. In addition, as referenced earlier, neurons in the PMRF modify their activity in tasks requiring voluntary gait modifications and reaching movements and are also modified during locomotion on inclined planes (Matsuyama and Drew 2000). This role of the PMRF in the control of compensatory responses is also supported by work in the rabbit showing that damage to the ventral funiculi, in which the reticulospinal fibers course, leads to deficits in the postural responses produced in response to tilting of a support surface (Lyalka et al. 2005).

In addition to showing strong neuronal responses in the PMRF in response to perturbations, the results from the present study also speak to the question raised in the introduction concerning the nature of the control signal responsible for restoring equilibrium. In particular, the finding that all responsive reticular neurons discharged at short latency to the perturbations and discharged prior to the initial changes in EMG activity in each of the four limbs strongly suggests that the initial postural responses, in all four limbs, are the result of a single descending signal, initiated by the initial perturbation. As originally suggested by Massion and his collaborators (see Dufossé et al. 1982, 1984), this would suggest that the initial postural responses in each of the four limbs are a coordinated synergistic pattern triggered by the perturbation of the single limb. Whether this pattern results primarily from RSNs with widely branching axons that activate the ensemble of the synergistic pattern or whether it is the result of more selective activation by different populations of RSNs remains open to question.


This work was supported by a Canadian Institutes of Health Research operating grant and a Fonds de la Recherche en Santé du Québec infrastructure grant. P. Stapley was supported by a Jasper Fellowship during the performance of this study.


We thank M. Bourdeau, N. De Sylva, P. Drapeau, C. Gauthier, J. Lavoie, F. Lebel, and J. Soucy for technical assistance in the performance and analysis of these experiments and Drs. Chapman and Rossignol for helpful comments on this manuscript. We thank B. Schepens for participating in some of these experiments.


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


  1. Afelt 1974.
  2. Beloozerova et al. 2005.
  3. Bem et al. 1995.
  4. Berman 1968.
  5. Brustein and Rossignol 1998.
  6. Coulmance et al. 1979.
  7. Cowie and Robinson 1994.
  8. Davidson and Buford 2004.
  9. Davidson and Buford 2006.
  10. Diener et al. 1988.
  11. Drew 1991.
  12. Drew 1993.
  13. Drew et al. 1996.
  14. Drew et al. 1986.
  15. Drew and Rossignol 1990a.
  16. Drew and Rossignol 1990b.
  17. Dufossé et al. 1982.
  18. Dufossé et al. 1984.
  19. Dufossé et al. 1985.
  20. Eccles et al. 1975.
  21. Fields et al. 1975.
  22. Górska et al. 1990.
  23. Górska et al. 1993.
  24. Górska et al. 1995.
  25. Horak and Diener 1994.
  26. Horak and Esselman 1984.
  27. Horak and Macpherson 1996.
  28. Kuypers 1963.
  29. Lawrence and Kuypers 1968a.
  30. Lawrence and Kuypers 1968b.
  31. Lipski 1981.
  32. Lyalka et al. 2005.
  33. Macpherson 1988a.
  34. Macpherson 1988b.
  35. Massion 1992.
  36. Matsuyama and Drew 2000.
  37. Matsuyama et al. 1993.
  38. Matsuyama et al. 1997.
  39. Mori 1987.
  40. Mori et al. 1982.
  41. Mori et al. 1992.
  42. Nashner 1977.
  43. Peterson and Felpel 1971.
  44. Peterson et al. 1975.
  45. Prentice and Drew 2001.
  46. Quessy and Freedman 2004.
  47. Rushmer et al. 1987.
  48. Rushmer et al. 1988.
  49. Rushmer et al. 1983.
  50. Schepens and Drew 2003.
  51. Schepens and Drew 2004.
  52. Schepens and Drew 2006.
  53. Schepens et al. 2008.
  54. Shimamura and Kogure 1979.
  55. Shimamura and Livingston 1963.
  56. Shimamura et al. 1990.
  57. Siegel and Tomaszewski 1983.
  58. Stapley and Drew 2004.
  59. Takakusaki et al. 2003.
  60. Takakusaki et al. 1994.
  61. Ting and Macpherson 2004.
  62. Viallet et al. 1992.
  63. Wu et al. 1988.
  64. Yeomans and Frankland 1996.
View Abstract