INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Wiping reflexes in the spinal frog have been a model preparation to examine how the spinal cord generates and controls targeted limb movements (Berkinblitt et al. 1986a,b; Fukson et al. 1980; Giszter et al. 1989; Schotland and Rymer 1993a,b; Sergio and Ostry 1993). Spinal frogs exhibit specific movement strategies for coping with ill-posed computational problems that arise from the kinematic redundancy of the limb (Giszter et al. 1989). Spinal frogs also have been shown to use body-scheme information to regulate trajectory formation during wiping (Fukson et al. 1980). Most wiping reflexes that have been examined (e.g., to irritants placed on the back, forelimb, and contralateral hindlimb) exhibit a modular organization in which the entire behavior can be described as a sequence of discrete movement phases (Berkinblit et al. 1986a; Giszter et al. 1989; Sergio and Ostry 1993). Wipe, scratch, and withdrawal reflexes all have been used to examine the possibility that the underlying spinal circuitry generating and controlling the behavior also exhibits a modular organization (Giszter et al. 1993; Schouenbourg et al. 1994; Stein and Smith 1996; Stein et al. 1986; Tresch and Bizzi 1999; Tresch et al. 1999).

Researchers have suggested that wiping reflexes may be constructed from combinations of force-field primitives (Bizzi et al. 1991; Giszter et al. 1991, 1993a,b; Mussa-Ivaldi et al. 1990, 1994). Force-field primitives initially were identified in spinal microstimulation studies. A force field is defined here as a function mapping limb position to end-point force. Force fields were constructed by measuring isometric forces at the ankle with the limb held in a range of positions and with the same stimulus applied at each position. Primitives were defined as force fields that exhibited invariant force vector directions and magnitude balances over time, i.e., conserved structures. Only a few primitives were found for any individual frog, and each primitive converged to a different location in the limb's work space. Primitives were similar among frogs (Giszter et al. 1993). The effect of coactivating primitives could be described simply as the linear sum of the individually activated force fields (Mussa-Ivaldi et al. 1994). Theoretical studies have shown that summation and magnitude scaling of a few force-field types in principle could be used to generate a large range of force-field structures and might account for the synthesis of natural behaviors including wiping reflexes (Mussa-Ivaldi 1992, 1997; Mussa-Ivaldi and Giszter 1992). However, details of the links between primitives, pattern generators, and circuitry for reflex behaviors are still not well understood.

If primitives are combined to produce reflex behaviors, as suggested by other studies (Giszter and Kargo 2000; Giszter et al. 1993; Kargo and Giszter 2000; Tresch et al. 1999), it seems likely that feedback may adaptively regulate the phasing and amplitude of the component primitives. Thus to understand how primitives may be combined in reflex behaviors (e.g., wiping reflexes), we also must understand feedback controls and their effects during these behaviors. Existing published data are insufficient to allow such an understanding. It is critical to the development of ideas about primitives and their relations to central pattern generators to examine proprioceptive feedback effects during unperturbed wiping.

Proprioceptive feedback has been suggested to play several roles during wiping movements. First, proprioception could be used in gating, interrupting, or aborting phases of wiping movements, as suggested by data from Chepelyugina (1947) and Giszter et al. (1989). Second, feedback is expected to regulate certain muscle properties during the movement (Nichols and Houk 1976). Loeb et al. (1993) showed in spinal microstimulation studies that feedback might compensate for muscle length variations and regulate stiffness at the force-field level in the frog. Third, proprioception has been suggested to participate in trajectory formation in frogs, in the context of a lambda model of motion generation (Berkinblitt et al. 1986b; Ostry et al. 1991). Finally, Schotland showed that frogs may be able to compensate for external loads added to the limb during wiping movements (Schotland and Rymer 1993b).

In contrast, other data suggest that feedback may not be necessary during unobstructed wiping or scratching movements of lower vertebrates. In paralyzed, or "fictive," turtle preparations, scratching is elicited readily and different scratching forms can be distinguished in hindlimb neurograms without phasic proprioceptive feedback (Robertson et al. 1985; Stein 1983). Moreover, in the frog, several factors have been cited as indications for a mainly feedforward control of hindlimb movements (Loeb 1986; Simpson 1969, 1976). These include the low gain of monosynaptic stretch reflexes in reduced preparations, the relatively high ratio of muscle stiffness to limb mass, and the reduced postural requirements for the hindlimbs in frogs. However, despite these data in reduced or paralyzed preparations, it remains possible that feedback plays a substantial role during active wiping movements in the spinal frog.

In this paper, we present detailed kinematic and electromyographic analyses of the hindlimb-to-hindlimb wiping reflex. This is a specific form of hindlimb wiping where an irritant is placed on the foot of the contralateral hindlimb. We provide a detailed analysis of the wiping limb trajectory when starting from a range of initial limb configurations and test for motor-pattern variations at each position. We examine the trajectories in both intrinsic (joint space) and extrinsic (Cartesian) coordinates and relate these path variables to the underlying motor pattern. We examine the roles of proprioceptive and cutaneous feedback in regulating wiping kinematics and the underlying motor pattern by using dorsal rhizotomy (i.e., complete deafferentation) and selective cutaneous deafferentation. Such an analysis has not been reported to date. Finally, we examined interanimal variations in trajectory formation among frogs, how these variations relate to motor-pattern variations, and how these can predict an animal's susceptibility to deafferentation.

The deafferentation data show a substantial role for proprioceptive feedback in regulating wiping trajectories. In particular the initial trajectory direction, the end-point path, and the accuracy of target (i.e., irritant) contact are all affected by deafferentation. In addition, when feedback was intact, motor-pattern features clearly were regulated based on the initial limb configuration. Removal of this feedback eliminated position-dependent changes in the motor pattern and produced specific motor-pattern changes. Finally, our data revealed differences in motor-pattern construction and trajectory formation among frogs that predicted their levels of reliance on sensory feedback (i.e., the degree of deficit after deafferentation). The manner of organization and control of trajectories by the spinal cord may be relevant to voluntary movement control and trajectory formation. In particular, the systems organizing trajectories at the spinal level might represent one of the substrates of voluntary or descending control (see e.g., Bizzi et al. 1995).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Surgery

Adult bullfrogs (200-350 g) were anesthetized with a solution of 5% 3-aminobenzoic acid ethyl ester (Tricaine). Skin incisions were made on the dorsal neck region, and the muscle layers separated at the midline. The dorsal connective tissue between the skull and first vertebra was opened, and the spinal cord transected at the medullary-spinal cord junction. Gelfoam was placed in the lesion cavity and over the cord. A small hole was made in the dorsal skull to cauterize the tectum and forebrain. Gel foam again was placed to fill the opening. The surgical incision on the skin was closed with wound clips and sealed with cyanoacrylate tissue adhesive (Vet-bond).

Incisions were made on the dorsal and ventral thigh to implant a pair of electrodes (7-strand Teflon-coated stainless steel wire with 1 mm exposure under a wax ball) in biceps (BI), vastus externus (VE), rectus anterior (RA), semimembranosus (SM), and gluteus (GL), sartorius (SA), ventral head of semitendinosus (ST), adductor (AD), and rectus internus (RI). A modified patch electrode design was sometimes used to record from the thin SA muscle in smaller frogs. Patch electrode designs are presented in detail in Loeb and Gans (1986; e.g., see Fig. 10.6). Our design differed from those presented in Loeb and Gans (1986) in that the EMG wires were led back through the recording surface side of the patch and through the muscle. This arrangement reduced the need to suture all four corners of the patch into SA, which has a very thin fascial sheath. BI is a knee flexor/hip abductor. VE is mainly a knee extensor. RA is mainly a hip flexor. SM is mainly a hip extensor with some knee flexor action. GL is a hip abductor/hip flexor. SA is a hip flexor/knee flexor. The ST head we implant is mainly a knee flexor/hip extensor. AD is a hip extensor. Finally, RI is a hip extensor/knee flexor. The broader muscles such as SM and RI may have also multiple functional subdivisions of the muscle; the actions noted in the preceding text were classified by grossly stimulating the motor entry point (and see Lombard and Abott 1907). Sites of implantation were the same between frogs. Electrode separation was designed to provide clean localized multiunit EMG without cross-talk from other muscles. This more local sampling from the muscle also necessitated some multi-trial averaging to give a reasonably smooth rectified envelope. The wires were braided and led directly out of the incision. The incision was closed with wound clips and Vet-bond.

A laminectomy was performed, exposing the lumbar spinal cord. The dura was opened by careful electrocautery. Three large pairs of dorsal and ventral roots (7-9) carry the vast majority of peripheral afferent and efferent information to and from the lumbar enlargement of the frog. To deafferent the frog, dorsal roots 7, 8, and 9 on the ipsilateral side to the wiping limb were identified, lifted with a hook, frozen with a cold copper rod, and cut distal to the frozen area. These three dorsal roots carry the large majority of sensory feedback from the wiping limb to the lumbar spinal cord (Ebbeson 1976). Dorsal root 10, which was left intact, carries sacrococcygeal feedback to the spinal cord, and a small number of C type afferent fibers may remain intact in the ventral roots (Della Torre et al. 1996). Nonetheless, when the animal recovered from anesthesia, there was no reflex responsiveness (as measured by EMG and video) to cutaneous stimuli, muscle palpation or kinematic manipulations to the wiping limb. To examine injury discharge in this technique, we recorded the roots to be lesioned in some deafferentations. Freezing before cutting caused a very reduced injury discharge which was silenced in ~5-10 min rather than persisting for long periods as sometimes reported. We thus reduced the likelihood of the concomitant possible injury discharge effects on spinal organization (Blenk et al. 1996, Wall et al. 1974).

In experiments where the role of cutaneous afferents was assessed, several incisions were made across the hindlimb at sites where cutaneous nerves could be dissected, identified, and cut. Light (dragging sharp forceps) and deep skin stimulation (pinching with toothed forceps) were used to assess if receptive fields remain intact (i.e., the hindlimb yielded no reflexive withdrawal or toe movements). The skin incisions were sealed with wound clips and Vet-bond.

Video recording and EMG collection

Movements were videotaped from above, perpendicular to the plane of the wipe. Therefore wiping movements out of the plane were not detected by video but were expected to be relatively small (see Giszter et al. 1989; Schotland and Rymer 1993a; Sergio and Ostry 1993). We will begin to address the vertical components of wiping in isometric force field and other experiments described in subsequent papers. The video record was collected from a camera with a 1-ms/field shutter operating at 30 frames (60 fields)/s. Video was synchronized to the EMG collection via an external tag [i.e., light-emitting diode (LED) emission] generated from Axoscope software.

EMGs were band-pass filtered with A-M Systems Amplifiers (100 Hz and 10 kHz cutoffs) and amplified by 1,000. Selected EMGs were displayed on oscilloscope. Although the amplifier upper cutoff was 10 kHz, EMG frequency content in the frog is substantially lower and our records had little high-frequency noise. They were led into an Axotape acquisition board, sampled at 1,000 Hz, a frequency appropriate for frog EMGs (see Giszter et al. 1993), and stored using Axoscope software (Axon Instruments, Foster City, CA). Because of the reduced sampling frequency of the video, there was some error in the synchrony between EMG and video due to detection of the external tag, but this was always 17 ms for individual wipes.

Recording free-limb kinematic

The frog was supported in a sculpted orthopedic casting plastic (Orthoplast) base with two horizontal platforms to separately support the hindlimbs. The arrangement allowed the hindlimbs to be positioned in a horizontal plane with the ankle as the only part of the hindlimb in contact with the surface. Wetting the Orthoplast support and the ankle skin reduced the friction at the initiation of the wipe. A pelvic clamp secured the frog on the support base and did not interfere mechanically with the wiping movement. After positioning the limb, the wiping reflex was elicited by biphasic, 1- to 2-ms voltage pulses (range 3-10 V) at a frequency of 20-30 Hz, applied for 500-700 ms through bipolar leads attached via alligator clips to the skin on the contralateral foot. The frog was allowed 3-5 min rest between stimulations to avoid habituation. A voltage and train duration was chosen to reliably elicit the reflex throughout the course of the experiment, and once again was not changed during the experiment. After a series of afferented wipes, the frog was deafferented as detailed (see Surgery), then allowed 1-3 h to recover after the dorsal root lesions (as described in Loeb et al. 1993). The same initial positions throughout the wiping limb's work space were then tested (see Experimental design).

Analysis of kinematics

Hip, knee, and ankle positions of the wiping limb were digitized from video for all experiments. Both the onset of wiping limb movement and time of target limb contact (and thus termination of the unconstrained free-limb segment of trajectory) were determined from the recorded video. This portion of the wipe was typically between 250 and 400 ms. The digitized positions were used in S-Plus statistical software package (Mathsoft, Seattle, WA) to reconstruct the ankle trajectory and joint angles during the movement.

Measures used to describe kinematic features included measures of normalized path length, initial direction of the ankle trajectory, overshoot of the ankle trajectory, peak tangential velocity of the ankle, and peak hip- and knee-joint velocities. Normalized path length was determined by dividing the total Cartesian length of the ankle path by the straight-line Cartesian distance between the start and stop locations of the ankle. Note that this measure bears no relationship to accuracy of movement in relation to the target stimulus or the frog's body. A value of 1 would indicate a perfectly straight path (see Fig. 1). The initial trajectory direction was calculated relative to the radial direction from hip to ankle. This was determined by taking the arc-cosine of the inner product of the average initial direction (i.e., unit) vector of the wiping trajectory (a' in Fig. 1) during the first five frames of movement and the direction (i.e., unit) vector from the ankle to the hip at the first frame (c in Fig. 1). Overshoot was defined here as the maximum contralateral horizontal deviation of the ankle past the stationary hip joint. Ideally we would refer to the overshoot from the target region where the legs meet. However this varied somewhat among frogs, and the value we used referenced to the hip allowed us to directly compare wipes between frogs of different sizes after scaling. Overshoot was normalized based on the length of the limb. Finally, peak tangential velocities of the ankle and peak joint velocities were determined by differentiating interpolated position-time data. The equations for calculation of normalized ankle path NPL and initial angle  trajectory parameters are given in the following text
Where x_i and y_i represent the time-varying positions of the ankle in the horizontal plane, x_n and y_n represent the ankle position at limb contact, x₁ and y₁ represent the initial position of the ankle
Where _i represents the time varying vector of the ankle at time i, and therefore (_i  i1) is a vector representing ankle velocity, (_0  hip) represents the polar vector from hip to ankle to which the velocity is referenced. Finally

View larger version (22K): [in this window] [in a new window]   Fig. 1. Data analysis. A: hip, knee, and ankle position were digitized from video. Video camera was placed perpendicular to the horizontal plane of the wiping reflex. Kinematic measures were made from digitized records of the ankle path (see text). Normalized ankle-path length was computed by dividing the actual path (line a) by the straight line distance between the initial and final or contact positions of the ankle (line b). Initial direction of the ankle trajectory was averaged over the 1st 5 frames of movement to provide the initial direction vector (vector a'). Vector a' was expressed in a polar coordinate frame, i.e., relative to a line connecting the initial ankle position and the hip (vector c). Thus the inner product of unit vectors a' and c was used to calculate the angle of the initial trajectory. Overshoot of the ankle path was calculated as the maximal horizontal (mediolateral) distance of the ankle relative to the hip (distance d). Hip angle (H') was determined relative to the rostro-caudal axis of the body, and the knee angle (K') was internal to the thigh and calf. B: raw electromyograms (EMGs) were amplified at a gain of 1,000 times, band-pass filtered with cutoffs of 100 Hz and 10 KHz, for 8-12 muscles [rectus internus (RI) and biceps (BI) shown here in columns 1 and 2, respectively] and were A/D converted and collected on computer at 1 kHz for individual trials. These records were rectified and filtered with a 30-point tapered box car filter. EMG measures were made on the individual filtered EMGs. Collections of individual trials (8-12) for a particular condition then were averaged after they were aligned on the highly repeatable peak of RI activity.

Work-space division

In the first experimental series, the work space was divided into rostral and caudal sectors and into proximal and distal rings based on a polar coordinate system centered at the hip. This was to determine regional differences in the changes of kinematic parameters after deafferentation. Wipes where the ankle was positioned between 0 and 45° vectors from the hip were included in the rostral sector. Wipes where the ankle was positioned between 45°, and larger angles from the hip were included in the caudal sector. Wipes where the ankle was positioned greater than half the length of the limb (thigh + calf) relative to the hip were included in the distal ring. Last, wipes where the ankle was positioned less than half the length of the limb from the hip were included in the proximal ring. The normalized path length and initial trajectory direction for wiping trials that began in each region were averaged for each frog individually. These individual values were then separately pooled first for the group (n = 9) of frogs with feedback intact and second for the group after feedback was removed.

We also treated the work space as a continuous area and interpolated kinematic measures across the broadly sampled polar work space. Within an individual frog, the number and location of initial limb positions were identical pre- and postrhizotomy, therefore we could interpolate a surface representing the changes in normalized path length or initial wiping direction after deafferentation. The interpolations were done in Matlab (Mathworks, Natick, MA). The surface of changes then was displayed as a contour plot. The contour plots for each frog were normalized to leg length and combined and averaged to produce an average contour plot of the change in path length and wiping direction across the work space.

EMG analysis

EMG data from our experiments were analyzed to determine the magnitude of EMG activity of 10 hindlimb muscles during the unconstrained free-limb phase of the wiping trajectory. Raw EMG signals were rectified and filtered with a 30-point tapered box car filter (Basmajian and DeLuca 1985) off-line in S-Plus (see Fig. 1B for procedure). The magnitude of rectified and filtered EMG envelopes were determined for each wiping trial by calculating the integral over the period from the onset of BI, usually the first muscle activated during wiping, up to the offset of VE, usually the last muscle activated. Thus EMG envelope magnitude is defined here as the area under the rectified and smoothed EMG signal. The EMG envelope magnitudes for an individual muscle were averaged for all wipes starting from the same initial position. In addition to looking at EMG envelope magnitudes, we looked at the latencies between the onset of critical muscles (BI or ST, usually the first detectable EMGs, and RI/SM/AD, a later activated group of muscles). We also looked at the durations of activity of these muscles (BI, ST, RI, SM, AD) during the wipe.

Experimental design

In all experimental series, the frog began with the limb positioned at rest on a moist platform in a configuration chosen by the investigator. Frogs were tested from each location at 4-min intervals to prevent fatigue, sensitization, or habituation. As a test of this, the duration of wipes was tested for stability through a series.

For the first experimental series, wipes were elicited with the wiping limb in a number of different initial positions in the work space (12-15 positions). The target limb always was placed in the same initial posture before we elicited the wipe. The frog then received a dorsal rhizotomy of the wiping limb side and was allowed 1-3 h of rest and recovery from the lesions (Loeb et al. 1993). The same work-space positions were revisited and wiping was elicited at each. The initial limb positions were similar between frogs (n = 10 frogs) and chosen to span the reachable work space of the wiping limb.

In the second experimental series, multiple wipes (8-12) were elicited at each of three positions in the work space. These experiments identified intraanimal variability at a single position and variable motor-pattern features between positions. The positions were similar between frogs (n = 11 frogs) and included a rostral position (hip ~ 80°, knee ~ 120°), a caudal position (hip ~ 130°, knee ~ 120°), and a position that represents the normal resting posture of the wiping limb (hip ~ 105°, knee ~ 80°). After an initial afferented series of wipes, frogs received a dorsal rhizotomy, were allowed 1-3 h of rest, and the same positions were revisited for a deafferented series of wipes.

In a third experimental series, wipes were elicited with the wiping limb in a number of different initial positions in the work space (12-15 positions). The frog received a cutaneous deafferentation of the wiping limb and was allowed 1-3 h of recovery. The same work-space positions were revisited, and wiping was elicited at each. The frog then received a dorsal rhizotomy of the wiping limb side and was allowed 60-120 min recovery time, and wiping was elicited at the same work-space positions. We tested four frogs in this manner and compared changes after cutaneous and total deafferentation of the wiping limb.

In a final series, wipes were elicited with the wiping limb in a number of different initial positions in the work space (12-15 positions). After dorsal rhizotomy and a series of deafferented wipes from the same positions, the frog received a contralateral ventral root rhizotomy. This immobilized the target limb and prevented beta activation of target limb muscle spindles. We tested five frogs in this manner and compared changes after deafferentation of the wiping limb and after target limb immobilization by de-efferentation.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 45 frogs were prepared, examined, and tested to provide the kinematic and motor-pattern data described here. This population of frogs was divided into several groups for specific procedures described below.

Qualitative kinematic and motor-pattern effects of deafferentation

In the hindlimb to hindlimb form of wiping, the irritant stimulus is applied to the hindlimb contralateral to the effector limb. Both limbs move in concert. They meet at a region along the midline caudal to the cloaca. The effector (referred throughout the remainder of the paper as "wiping") limb then removes the irritant from the stimulated (referred to as "target") limb using the heel as the wiping tool.

Figure 2, A and C, shows the typical kinematic patterns evoked by the wiping stimulus in two separate spinal frogs when sensory feedback was intact. The kinematics of the wiping and target limbs (thigh and calf of both) are expressed as sequences of stick figures, and arrows mark the progression of ankle movement. Bold stick figures represent the portion of the wiping limb trajectory from movement onset up to contact with the target limb. This targeted portion of the wiping limb trajectory forms the major focus of our analysis. Because this form of wiping was evoked by irritants placed on the contralateral limb, it still can be evoked when the wiping limb is deafferented (i.e., feedback from the target limb is still intact). We exploit this feature to examine if the wiping trajectory and the underlying motor-pattern change as a result of deafferentation.

View larger version (43K): [in this window] [in a new window]   Fig. 2. Digitized records and the corresponding raw EMGs during the hindlimb-to-hindlimb wiping reflex are shown for 2 frogs. Records are shown both before (A-D) and after (E-H) complete deafferentation of the wiping limb. Bold stick figures represent the wiping limb up to target limb contact. A bold line also traces the ankle path of both limbs (arrows mark direction of movement along the path), and a circled asterisk marks the point and frame of limb contact. A and C: with feedback intact, the kinematics of the wiping limb up to target limb contact was characterized by hip extension and knee flexion. During this initial phase, frog bf38 had a more synchronous pattern of knee flexion and hip extension, whereas bf28 had a more staggered pattern of knee flexion followed by hip extension. In both, knee extension followed limb contact in an attempt to remove the electrical stimulus. B and D: intact motor pattern was characterized by 3 periods of activity, separated here by vertical lines. Movement onset and contact (*) are indicated by the vertical arrows. Periods of activity consisted of an initial period of mainly BI, ventral head of semitendinosus (ST), and sartorius (SA) activity, a 2nd period of combined RI, semimembranosus (SM), BI, ST, and SA activity, and final period of vastus externus (VE) activity. Frog bf28 had a more distinct initial period. E and G: after deafferentation of the wiping limb, hip extension occurred earlier, faster, and to a larger extent in both frogs, and knee flexion was slowed, reduced, and even reversed to extension in mid-trajectory. This altered the initial ankle path (compare G and C). This resulted in the ankle overshooting and missing the stimulus initially. F and H: deafferented motor pattern was characterized by the reduced initial activity of BI, ST, and SA, an increased and broadened activity of RI and SM during the second period of activity, and the loss of knee extensor activity at limb contact.

After deafferentation of the wiping limb, there were several qualitative changes in the trajectory of the wiping limb, which can be seen most clearly by comparing the frog in Fig. 2, C and G. First, the initial direction of the ankle trajectory (first 5 frames of movement) appeared caudally rotated relative to the region of target limb contact. Second, the entire ankle path up to limb contact appeared less straight or more curved. Third, the ankle trajectory overshot the normal target region where the limbs normally made contact. Finally, after limb contact was made, the ensuing limb extension/whisk phase appeared not as strongly activated. This phase normally serves to remove the irritant once limb contact has been made.

The kinematic changes were accompanied by qualitative motor-pattern changes. Again these can be most clearly seen by comparing the motor patterns shown in 2, D and H. First the initial period of limb flexor activation (BI, ST, and SA) became less distinct. Second, there was a collapse of phasing between this initial period and a second period of hip extensor activation (RI and SM). Third, VE activity that closely followed limb contact (second arrow; asterisk) was no longer apparent.

In comparing the two frogs shown in Fig. 2, A and C, it appeared that frogs might exhibit variations in trajectory formation and in the underlying motor pattern when feedback was intact. For instance, bf28 showed an initial wiping limb trajectory (up to target limb contact) that was produced by staggered joint motion (i.e., knee flexion followed by hip extension). In bf38, the wiping limb trajectory (up to target contact) was produced by a more synchronous pattern of knee flexion and hip extension. In addition to these minor variations with feedback intact, there also appeared to be differences in the effects of deafferentation. For example, frog bf38 showed a relatively smaller effect of deafferentation on the wiping limb trajectory than did bf28.

In the following sections, we quantify changes in the wiping limb trajectory and the motor pattern after deafferentation. We then examine if feedback related to the initial posture of the wiping limb regulates motor-pattern and trajectory features. Finally, we quantify the variations observed in wiping limb trajectory among frogs, relate this variation to motor-pattern variations, and then examine if such variation can account for a frog's susceptibility to deafferentation.

Effects of deafferentation on end-point trajectory

Because we observed variations among individual frogs in motor-pattern details and ankle path kinematics, we tested the statistics of a population of nine frogs in detail. We examined their data both as a group using lumped statistical measures and individually from different spatial positions. Figure 3A shows the unconstrained ankle paths of the wiping limb from different initial positions spanning the work space for a typical frog both before and after deafferentation. Ankle paths are shown from movement onset up to target limb contact. After deafferentation of the wiping limb, the ankle paths demonstrate the qualitative kinematic features noted in the preceding text, namely an altered initial trajectory direction, decreased path straightness, and an increased maximal overshoot. Furthermore, Fig. 3A shows that there may be systematic differences in the magnitude of the change in path straightness and the initial trajectory direction based on the initial limb position. For example note that the paths starting from the more caudally extended positions of the frog stick figure are closer to the intact path than are the paths from the more rostrally originating positions (where the hip initially is flexed and the knee extended).

View larger version (23K): [in this window] [in a new window]   Fig. 3. Frogs exhibit configuration-dependent kinematic deficits after deafferentation. A: ankle paths of the wiping limb from a number of different work-space positions are shown for frog bf11 both before (top) and after deafferentation (bottom) of the wiping limb. Only the paths up to target limb contact are shown. With feedback intact, the ankle paths were fairly straight from each position and ended at a similar location of contact with the target limb. After deafferentation, the initial direction of the ankle paths deviated caudally, and the paths became less straight and overshot the normal location of limb contact. Ankle paths that started from work-space boundaries or rostral positions in the work space appeared to have larger path changes. B: ankle paths from individual locations exhibited low variability at each tested position in the limb's work space. Paths are shown from 3 representative starting positions (1, 2, and 3) and up to target limb contact for both afferented (top) and deafferented (bottom) trials for frog bf25. Multiple trials were evoked at each. Ankle paths starting from position 2 exhibited the largest changes in curvature. C: averaged knee angle (left) and hip angle (right) changes that occurred in the wiping limb up to target limb contact are shown for frog bf25. Joint angle changes are shown for each of the 3 starting positions shown in B and for the same frog (see Fig. 1 for measurement of joint angles). After deafferentation of the wiping limb, there were early, middle, and late trajectory changes in the joint angle patterns. Early changes included an increased ratio of hip extension to knee flexion. Midtrajectory changes included the appearance of knee reversals, particularly at position 2. Late changes included an increased knee angle (less knee flexion) and an increased hip angle (increased extension) at the time of target limb contact.

We calculated the combined, mean values of several kinematic variables for the nine frogs regardless of the initial position from which the wiping limb started. The normalized ankle path, the initial direction of the ankle trajectory relative to the hip, and the maximal overshoot were determined for all wipes for each frog. There were significant changes in each of these parameters following deafferentation. First, the combined, mean normalized path length (see calculation of path length in Fig. 1; 1.06 ± 0.06; mean ± SE) became significantly increased after deafferentation (1.25 ±0.10, significance P < 0.01). Second, our measure of the initial trajectory direction was altered after deafferentation. For this measure, we used the angle between the vector from the hip to the initial position of the ankle and a vector characterizing the average initial direction of the ankle trajectory (see Fig. 1, b, for calculation). The combined, mean angle increased from 38 ± 1.38 to 66 ± 1.7°, P < 0.01. Third, the maximal overshoot of the ankle path, also relative to the hip, was increased after deafferentation from a combined, mean of 39.3 ± 2.13 to 64.2 ±1.27 (significant at P < 0.01). The values of the overshoot are in arbitrary pixel units but are normalized to the length of the frog's limb to allow comparisons among frogs. The frog calf would be ~60 units in the normalized data. Finally, the peak tangential velocity of the ankle during the unconstrained wiping trajectory was increased after deafferentation from 416.75 ± 1.5 to 483.34 ± 1.3 units/s, P < 0.01. These values of end-point tangential velocity are again normalized to the length of the frog's limb.

Consistency of kinematic changes after deafferentation across the limb's work space

To more closely examine intertrial variability in end-point kinematics within an individual frog, we chose three test positions and elicited 10-12 wiping trials at each position before and after deafferentation. The three test positions chosen were similar among the 10 frogs tested. Each position required a particular pattern of joint motion to reach the target. These positions included (see Fig. 1): position 1 (knee extended at 120°, hip extended at 130°), position 2 (knee extended at 120°, hip flexed at 80°), and position 3 (the normal resting posture of the hindlimb; knee and hip both slightly flexed 80 and 105°, respectively). Positions 1 and 2 required a larger flexion at the knee to reach the target region, caudal to the cloaca, where the stimulus is removed. In addition, position 2 required a larger hip extension relative to the other two positions to reach the same target region.

Figure 3B shows ankle path data (from movement onset up to target limb contact) at the three chosen locations. Our data demonstrated that multiple trials from a single location showed great consistency. Further, just as for the data described in the preceding text, there were systematic changes that occurred after deafferentation. The ankle path of the wiping limb was increased in length or became more curved. The initial trajectory direction was altered and caudally rotated, and the path overshoot was increased. In addition, as suggested in Fig. 3A, the magnitude of the change in both path straightness and the initial trajectory direction showed a position dependency (quantified for the group in Table 1). The normalized path length increased the largest at position 2, then 1, and the smallest at position 3. The largest changes in the initial trajectory direction were at positions 1 and 2. This suggested that changes in the initial trajectory direction were largest when the knee was initially in an extended configuration. Finally, the peak tangential velocity of the ankle trajectory was on average increased at each location after deafferentation (Table 1). However, the increase in velocity was largest for deafferented wipes starting from position 2 where the limb was both extended and rostral and key hip extensor muscles were most stretched (see DISCUSSION).

Effects of deafferentation on joint kinematics

Figure 3C shows the joint angle kinematics (hip and knee) obtained at positions 1-3 both before and after deafferentation for the frog shown in 3B. We quantified three features of the joint angle data (Table 1): ratio of hip-to-knee movement during the first 85 ms of movement, peaks of hip and knee velocity, and hip- and knee-joint angles at the time of maximum overshoot. The joint angle plots in Fig. 3C show that after deafferentation, the early part of the wiping trajectory (1st 5 frames; 85 ms) was characterized by an increased hip extension and a decreased knee flexion from each position. This is the joint-based equivalent of the external rotation of the initial ankle trajectory, as described in the preceding text. The combined, mean ratio of hip extension to knee flexion after deafferentation has the largest percentage increase at position 1 and position 2; this accounts for the larger changes in the initial trajectory direction at these positions.

In addition to the early joint angle changes, we examined joint velocity profiles for the entire trajectory up to target limb contact. The hip- and knee-joint velocity profiles changed after deafferentation. Figure 4 shows the hip and knee velocities of the wiping limb for two frogs for wipes starting from position 2. Position 2 was unique in that all frogs (n = 10) displayed a prominent knee velocity reversal (from flexion to extension) after deafferentation of the wiping limb, which was absent in the afferented condition. The mean peak hip velocity was shifted relative to movement onset and was increased the most at position 2. This change in the hip velocity profile may be causally related to the appearance of the knee reversals (e.g., via an increase in interaction torques at the knee).

View larger version (20K): [in this window] [in a new window]   Fig. 4. Averaged hip- and knee-joint velocity profiles (for 8 trials) are shown for wipes starting from position 2 for frogs bf26 and bf25. dotted lines, deafferented hip and knee velocity traces; solid lines, afferented hip and knee velocity traces. Top: knee velocities; bottom: hip velocities. With feedback intact, there is some staggering between the peak knee and hip velocities (thin lines). After deafferentation, the hip begins to extend earlier, and in frog bf25, hip extension is much faster (dotted lines). In addition, the knee begins to flex more slowly, reverses to extension in synchrony with the hip velocity profile, and finally reverses back to knee flexion (solid line). Position 2 was unique in that all deafferented frogs (n = 10) showed this knee velocity reversal. In addition, the normalized path length and peak hip velocity showed the largest increase for wipes starting at position 2 compared with the other 2 positions.

As mentioned in the preceding text, the overshoot of the ankle path relative to the hip was increased after deafferentation. An increased hip extension angle and a reduced knee flexion angle characterized the posture of the limb at the time of maximum overshoot (Table 1b). With the knee more extended during approach of the target limb, contact or collision with the target limb did not occur (i.e., was missed). Thus the hip continued extending until contact was made. In fact, for the group of frogs, the mean hip angle at maximum overshoot had a smaller variance after deafferentation than with feedback intact. This suggests that the hip joint limit might have been approached during deafferented trajectories. In addition to the decrease in knee flexion magnitude that contributed to the target overshoot, other factors may contribute including the increased hip velocity, end-point velocity, and presumably the lack of decelerative or dissipative muscle action. In contrast to the deafferented pattern, we observed that with feedback intact, the trajectory of the wiping limb in most frogs began to decelerate well before contact was made with the target limb. The limb contact or collision in the afferented frog is thus a controlled collision and is not the major source of deceleration when feedback was intact.

Effects of deafferentation across the work space of the limb

Figure 3, A and B, and the preceding analysis showed that changes in the ankle path after deafferentation, and in particular path straightness and initial trajectory direction, showed positional dependencies. To better quantify kinematic changes across the reachable work-space of the limb, we examined variations in normalized path length and initial trajectory direction from a larger range of initial limb positions (10-12 different positions as shown in 3A). We plotted how the mean change in each of these two parameters varied at initial positions spanning the work-space of the limb. Figure 5, B and C, shows contour plots of the mean change in the path straightness and the initial trajectory direction over the hindlimb work space. The coordinates for the contour plot are the initial work-space location of the ankle in polar coordinates (see 5A). The abscissa is the normalized distance of the vector from the hip to the ankle. The ordinate is the normalized angle between this vector and the mediolateral axis. The contours are the mean magnitude of change in the path straightness or in the initial trajectory direction. The lightest shading between contours represent the regions of largest change. The largest change in path straightness occurred when the ankle was positioned rostrally and distally relative to the hip (i.e., when the hip was flexed and/or the knee was extended). The largest changes in the initial direction were largely confined to regions where the knee was extended.

View larger version (61K): [in this window] [in a new window]   Fig. 5. Position-dependent changes in wiping kinematics after deafferentation. A: wiping limb's work space was divided into rings and sectors (see METHODS), and the average normalized path length and initial trajectory direction for ankle paths starting from each region were found before and after deafferentation for the group of 9 frogs. B: a gray shaded contour plot of the averaged change in the initial trajectory angle across the polar work space of the wiping limb. White coloring marks region of greatest change, black is least change. Largest changes in the initial trajectory angle were localized to regions of the work space where the limb is extended (radial distance >0.8). C: a contour plot of the averaged change in the normalized path length. Largest changes are localized to rostral regions (angle <0.4) where the hip is flexed (i.e., azimuth is small) and regions where the limb is extended (radial distance high). D: initial trajectory angle for the ankle path was changed after deafferentation from each region (0 indicates no change). Changes in the 2 sectors were roughly equal. Initial angle was significantly greater in the distal ring where the knee is extended compared with the proximal ring (P < 0.01). Vertical bars represent ±SD. E: normalized path length was changed in each region. Rostral sector where the hip is flexed had significantly larger changes than the caudal sector (P < 0.01), and the distal ring had larger changes than the proximal ring (P < 0.05) [**P < 0.01, *P < 0.05]. F: Cartesian coordinates of the initial ankle position were transformed into polar coordinates for the work space size of each frog. At each polar coordinate (open square and filled circle for instance), the value of the change in the normalized path length and initial trajectory direction was plotted. A contour plot was produced for each frog by interpolating over the sampled polarwork space. Nine individual contour plots then were averaged for this figure.

To demonstrate that the spatial variations observed in these data represented a significant trend in all tested frogs (n = 9), we looked statistically at the positional dependency of changes in the path straightness and the initial direction. We divided the work space into discrete regions as in Fig. 5D. Each region comprised a sector or a segment of one of a set of concentric rings. Sectors could be classified as rostral or caudal relative to the hip. Rings could be classified as proximal or distal relative to the hip. In each region, we calculated the mean, combined change in both kinematic parameters. Figure 5, E and F, show that there was a positional dependency in the magnitude of change in both the initial trajectory direction and the path straightness after deafferentation. For instance, the path straightness was prone to larger changes when the ankle was initially positioned in the rostral versus the caudal sector (P < 0.01) and in the distal versus the proximal ring segment (P < 0.05). The initial direction was prone to larger changes when the ankle was positioned in the distal ring segments versus the proximal ring segments (P < 0.01). The regions can be associated roughly with certain initial postures of the limb. For instance, in the rostral sector, the hip is mostly flexed while in the caudal sector, the hip is mostly extended. In the proximal ring the knee is mostly flexed, and in the distal ring, the knee is mostly extended. Thus postural-related feedback appears to play a significant role in regulating limb trajectories.

Position-dependent motor-pattern adjustments

We examined the motor patterns during wiping to quantify the extent of adjustment of motor pattern based on initial limb position. We examined these at the three test locations shown in Fig. 3B. For the muscles examined here, during wiping through the whisk phase, three periods of EMG activity could be discerned. These periods included: an initial period of limb flexor activity, a second period where hip extensors become additionally activated, and a final period of knee extensor activity that closely follows target limb contact. The first two periods of muscle activity contribute to the wiping limb trajectory up to target limb contact.

For the following set of data, raw EMGs for individual trials as shown in Fig. 2 were rectified and filtered. EMGs for multiple trials at the same position were aligned individually on the peak of the RI EMG envelope and then averaged together. We aligned EMGs in individual trials on the RI peak because of its ease of detection and its almost constant timing relative to stimulus onset between positions. This timing of the RI peak showed no significant differences between positions for individual frogs (Table 2). We analyzed, and present, only the data from frogs in which deafferentation abolished EMG variation and which therefore were free of artifact (e.g., Fig. 6C).

View larger version (54K): [in this window] [in a new window]   Fig. 6. Frogs regulate EMG features based on initial limb configuration and exhibit common EMG changes after deafferentation. A: each EMG trace represents the average of 10 trials from each position when feedback was intact for frog bf25. Vertical lines represent average time of movement onset and limb contact. Ranges for these times also are shown. Averages were constructed by aligning the peak of the RI EMG for individual trials (position 1 EMGs, bold lines, position 2, dashed lines, position 3, dotted lines) (position 1, dark shaded envelopes; position 2, light shaded envelopes; position 3, white envelopes). Flexor muscles activated in the initial phase (BI, SA, and ST) showed EMG magnitude changes based on the initial starting position of the limb. BI and SA showed large increases at position 1 (dark shaded envelopes) compared with position 3, the normal resting posture of the limb. Some of these increases occurred before movement onset (first vertical line). ST showed EMG increases at positions 1 and 2 compared with the normal resting position but mainly during limb movement. Muscles activated in the 2nd and 3rd phases (hip extensors RI and SM and the knee extensor VE respectively) showed little change in timing or magnitude with initial position in this frog. B: after deafferentation in this frog, there were several EMG changes typical of most frogs (white-shaded envelopes represent deafferented EMGs; dark-shaded envelopes represent afferented EMGs at position 1). Initial EMG envelopes of BI and SA were reduced and movement onset (first dotted vertical line) was delayed correspondingly and closer to the peak of RI. EMG envelopes of RI and SM were broadened. Peak of the ST envelope was reduced during the movement phase. Finally, VE activity that occurred at limb contact with feedback intact was absent. C: plots of the averaged areas of EMG envelopes are shown for 3 flexor muscles at positions 1-3 both before (solid lines) and after deafferentation (dotted lines). Vertical bars: SDs at each location. Note the loss of position-dependent EMG modulation of BI, SA, and ST after deafferentation. EMG data for the entire group of frogs are fully summarized in Tables 2 and 3.

In Fig. 6A, averaged EMGs for six hindlimb muscles of the wiping limb for wipes starting at positions 1, 2 and 3 are shown for a frog whose kinematic data are presented in Fig. 3, B and C. The averaged EMGs in Fig. 6A show motor features that were consistent within the group of 10 tested frogs. To combine data for a group analysis and reduce trial to trial variations, EMGs for individual frogs were normalized based on the magnitude of their averaged EMG envelope at position 3. Thus after normalization all muscles had an envelope of unity at position 3. Position 3 was close to the normal, resting posture of the hindlimb in our experiments. Normalized data from individual trials were then averaged at each position. The data shown in Fig. 6A were subject to this normalization. Combining the data in this way revealed small but consistent and significant changes in EMG between each of the positions during the targeted portion of the wipe, which are presented in detail in Tables 2 and 3.

There were consistent changes between the three limb positions in the magnitude of the EMG envelopes of two groups of muscles. The first group of muscles include BI, SA, and ST, which are all muscles that contribute to knee flexion. The combined EMG data for the 10 frogs show a significant increase (P < 0.01) of the normalized envelopes of BI and SA up to target limb contact at positions 1 and 2 compared with position 3 (see Table 3). In Fig. 6A, the averaged BI EMG at positions 1 and 2 was increased relative to the averaged EMG at position 3. In the figure, the first vertical line marks the averaged onset of limb movement and the second line marks the averaged time of target limb contact. Some of the increase in BI and SA, especially at position 1, actually occurred before the onset of limb movement in the horizontal plane. In fact, the combined EMG data showed significant increases in the magnitude of the averaged envelopes of BI and SA when these were calculated from the first detectable EMG up to movement onset (Table 3). ST, another knee flexor that would be lengthened concomitantly at positions 1 and 2, showed no significant differences in the magnitude of the averaged envelope up to movement onset. However, the averaged ST envelope in the later period up to target limb contact was increased at these positions relative to position 3 (Fig. 6A).

There were also consistent changes between the three limb positions in a second group of muscles, the hip extensors RI, SM, and AD. However, although significant, these changes were small and therefore are not shown graphically. In the wipe type discussed here, this second muscle group was involved in the free-limb motion. This group of muscles also contributed to the postcontact interaction phase, and the early part of the whisking phase that serves to remove an irritant from the target limb. In this group, variation occurred in the timing of the onset of the strong hindlimb muscles RI, SM, and AD relative to the onset of BI. It is important to emphasize that the peak timing of these muscles did not vary. There was no difference between the onset of BI and the peak EMG of RI or SM between positions. However, the onset of RI and SM relative to the onset of BI changed between positions (P < 0.05). In the combined data, the latency was shortest at position 3, then position 2, and longest at position 1 (see Table 2). The latency differences appeared to be due primarily to the broadening of the rising phase of the EMG envelopes of RI and SM rather than phase shifting of the peaks. Changes in EMG duration of RI and SM were related inversely to the latency changes, whereas envelope termination times were similar across positions in these muscles. These data suggest a role of (presumably proprioceptive) afferents, in regulating or gating the onset of this muscle group. The combined data for the 10 frogs demonstrate these trends clearly (Table 2).

A final and third muscle group becomes activated near the end of the RI/SM/AD EMG envelopes and is involved in the subsequent kinematic phase of whisk/extension. These are the knee extensors (represented by VE here). VE showed no consistent regulation of EMG activity based on the initial limb position.

To summarize, there were changes in the magnitude and timing of specific EMG envelopes based on the initial limb posture. The changes in magnitude were restricted primarily to an initially activated group of relatively smaller muscles that act to flex the knee. Changes in onset timing occurred in a group of relatively large muscles that extend the hip, but the timing of their peak and termination was unaffected. It was unexpected to us that these muscles would be delayed in the onset of their activation without other phase shifts throughout the motor pattern.

Changes in motor pattern after deafferentation

After deafferentation of the wiping limb, there were several changes in the motor patterns. These changes contribute to the altered free-limb kinematics of the placing phase of the hindlimb wipe after deafferentation. Figure 6B shows the averaged EMG envelopes for hindlimb wipes starting from position 1 before and after deafferentation of the wiping limb. Once again the plots are aligned on the peak of the RI EMG envelope. In the frog shown, deafferentation resulted in a reduced latency between the onset of BI EMG activity and the activity of a second group of muscles (RI, SM, AD mainly). The reduced latency between BI onset and RI/SM/AD onset was due to the reduction of initial BI EMG strength and concomitant delayed onset of BI and the broadening of RI and SM EMGs. In addition to this altered onset timing, the peak EMG activity of RI, SM, and AD often was increased, as is the case for bf25. Finally, there was a conspicuous absence of VE activity that normally closely follows limb contact. This loss correlates with the loss of the kinematic extension phase (not shown) after deafferentation (see Fig. 2).

We computed the integrated values of the precontact hindlimb EMG envelopes for the 10 frogs before and after deafferentation at each position. We thus focused only on the envelopes up to limb contact (those contributing to the placing phase kinematics). The mean duration of this phase of movement was similar before and after deafferentation so that any magnitude changes are not related to computing the integrated EMG values over longer or shorter movement durations. Examples of the data are in Table 3. Our first observation was that the mean, integrated BI EMG at positions 1 and 2 is reduced significantly (P < 0.01) after deafferentation. The EMG of SA is reduced significantly at position 1 as well. Second, the mean, integrated EMG of particular muscles of the second group (RI, SM, and AD) is increased significantly during the wiping kinematics up to limb contact (P < 0.01). The increased magnitude is attributable to two factors. First, the EMG envelopes of these muscles broadened to overlap more with the initial BI, SA, and ST EMGs: the combined, mean durations of RI and SM increased by 18 and 15%, respectively, at each position. Second, the peak EMG was advanced relative to BI onset: the mean latency between the onset of BI EMG and the peak EMG of RI and SM was reduced by 45.5 and 40.2%, respectively. Because these changes in RI and SM occurred before limb contact, they must be attributed either to initial postural or later on-line afferent effects. We will show in the following text that these motor changes resulting from complete deafferentation are not present after cutaneous deafferentation alone.

To summarize, there are several changes in the wiping motor patterns that contribute to the altered kinematics of the wiping limb after deafferentation. First, the regulation of BI and SA EMG activity based on the initial position of the limb is absent. Therefore knee flexor related EMGs are not increased when these muscles initially are lengthened. Second, muscles of a second group that generate large hip extensor torques contribute more to the unconstrained wiping movement. Finally, in some frogs the initial BI or SA activity may collapse completely or disappear, leading to a single, synchronous activation pattern of most muscles. Thus the combined effect of these EMG alterations resulted in a deafferented motor pattern with less differentiated periods of activity (i.e., more synchronous knee flexor and hip extensor activity). Functionally, these changes led to increases in the amount of hip extensor activity up to target limb contact and to decreases in the amount of knee flexor/hip flexor activity that contribute to trajectory formation.

Afferent modality

CUTANEOUS AFFERENTS. Changes in the motor pattern and kinematics after deafferentation were described in detail in the preceding text. The changes described thus far could be due to the absence of feedback from any one or several distinct afferent types from within the wiping limb. Further, it is also conceivable that some of the changes result from bilateral interactions of feedback or less specific effects such as the presence of asymmetric, contralateral phasic feedback from the moving target limb. In this section, we establish that the major kinematic and motor-pattern changes we remarked on in the preceding text were due mainly to the loss of ipsilateral muscular and joint afferents.

View larger version (29K): [in this window] [in a new window]   Fig. 7. Deficits due to dorsal rhizotomy are due mainly to removal of proprioceptive feedback from the wiping limb. A: ankle paths of the wiping limb from 3 different positions with feedback intact (solid lines), after cutaneous deafferentation of the wiping limb (dotted lines), and after complete deafferentation of the wiping limb (dashed lines). Ankle paths are shown only up to target limb contact. After cutaneous deafferentation, ankle paths were largely unaffected, remaining within the normal trial to trial variation (compare with Fig. 3). After complete deafferentation, ankle paths were altered: deviated initial trajectory direction, increased path length, and increased overshoot. B: averaged, rectified and filtered EMG envelopes for BI and RI when feedback was intact (dark shaded envelopes; both top and bottom panels), after cutaneous deafferentation (unfilled, dotted-line envelopes; top panel only), and after complete deafferentation (unfilled, dashed-line envelopes; bottom panel only). BI and RI represent the initial knee flexor and second hip extensor phases of the wiping motor pattern. B, top: after removal of cutaneous feedback, there were minimal changes in the knee flexor and hip extensor phases. Bottom: after complete deafferentation, the magnitude of the initial knee flexor phase was reduced and the hip extensor phase was broadened. C: shows the ankle paths from 3 different limb positions with feedback intact (solid lines), after ipsilateral dorsal rhizotomy (dashed lines), and after contralateral ventral rhizotomy (dotted lines). Dorsal rhizotomy produced the typical path changes. After contralateral ventral rhizotomy, ankle paths did not return to normal. This showed that unbalanced, phasic feedback from the contralateral limb was not the major source of trajectory change. D: averaged, rectified and filtered EMG envelopes for BI and RI when feedback was intact (dark shaded envelopes; top), after dorsal rhizotomy (unfilled, dashed-line envelopes; top and bottom), and after contralateral ventral rhizotomy (unfilled, dotted-line envelopes; bottom). D, top: after deafferentation, the initial BI EMG envelope (i.e., initial knee flexor phase) was reduced and the RI EMG envelope was broadened. Bottom: after ventral rhizotomy, EMG envelopes of both muscles were equally narrowed (see text). However, the normal EMG pattern did not emerge.

CONTRALATERAL DE-EFFERENTATION. To examine the possibility that there was a reorganization or disruption of a normal centrally driven wiping motor program by contralateral phasic afference, due to the absence of balancing ipsilateral afference, we immobilized the target limb by ventral rhizotomy. After a series of wipes in five frogs with complete ipsilaterally deafferented hindlimbs, we transected the contralateral ventral roots in the frogs and positioned the target leg in the usual starting position. This manipulation allowed us to continue to stimulate the skin of the target limb while preventing limb motion and beta-fusimotor activation of muscle spindles. If asymmetric phasic feedback from the unaffected target limb contributed strongly to the kinematic and motor-pattern changes seen after deafferenting the wiping limb, we expected that the ventral rhizotomy would restore the normal pattern or would result in additional large alterations in motor-pattern structure.