Scratch-Swim Hybrids in the Spinal Turtle: Blending of Rostral
Scratch and Forward Swim
GAMMON M. EARHART AND PAUL S. G. STEIN
Department of Biology and Program in Movement Science, Washington University, St. Louis, Missouri 63130
INTRODUCTION
A task is a behavior in which an organism achieves a
particular goal. Each task may be accomplished in more than
one manner. Each different motor strategy used to perform a
particular task is called a form of that task (Stein et al. 1986b).
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.
156
Locomotion is a task where the goal of the organism is to move
its center of mass. There are several forms of overground
locomotion, e.g., forward walking versus backward walking
(Buford and Smith 1990a,b; Grasso et al. 1998; Stein and
Smith 1997; Thorstensson 1986; Winter et al. 1989). For some
rhythmic behaviors, the features of two forms or tasks are
combined into a single behavior such that each cycle expresses
features of both forms or tasks. This blended behavior is called
a hybrid (Carter and Smith 1986a,b; Mortin et al. 1985; Robertson et al. 1985; Stein et al. 1986a,b).
The turtle spinal cord can produce hindlimb motor outputs
for scratch and swim tasks. During scratch, the goal of the
turtle is to rub against a stimulated site on the shell; during
swim, the goal of the turtle is to generate force against the
water and move its center of mass. There are three forms of
scratch: rostral, pocket, and caudal; each is used to rub a
specific region of the shell (Mortin et al. 1985). There are two
forms of swim: forward swim and backpaddle (Field and Stein
1997a,b; Lennard and Stein 1977; Stein 1978). We focus on the
rostral scratch and the forward swim in the present paper.
Both the rostral scratch and the forward swim have behavioral events during which force is exerted against a substrate.
The rostral scratch event is the rub of the foot against the
stimulated site on the midbody shell bridge (Field and Stein
1997a; Mortin et al. 1985). The forward swim event is the
powerstroke, a hip extension movement with the foot held
vertically and the toes and webbing spread (Davenport et al.
1984; Field and Stein 1997a; Lennard and Stein 1977; Stein
1978; Zug 1971). During the forward swim powerstroke, propulsive force is exerted against the water. In the present study,
the turtle’s carapace was restrained; thus there was no centerof-mass movement when the swim powerstroke occurred. The
rostral scratch and the forward swim are kinematically similar
(Field and Stein 1997a,b; Stein 1983). Their motor patterns are
also similar: both have alternating hip flexor and extensor
activity and similar timing of monoarticular knee extensor
activity within the hip cycle [electromyograms (EMGs): Stein
and Johnstone 1986; electroneurograms (ENGs): Juranek and
Currie 1998, 2000). Their motor patterns differ with respect to
amplitudes of activity. Juranek and Currie (1998, 2000) have
also demonstrated interruption of fictive rostral scratch with
fictive forward swim, and vice versa.
The present paper demonstrates a novel interaction between
the rostral scratch and the forward swim. We provide kinematic and electromyographic evidence for a scratch-swim hybrid behavior with combined features of two distinct tasks,
rostral scratch and forward swim, in each of several successive
cycles of a rhythmic movement. Blending of two forms of the
0022-3077/00 $5.00 Copyright © 2000 The American Physiological Society
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Earhart, Gammon M. and Paul S. G. Stein. Scratch-swim hybrids
in the spinal turtle: blending of rostral scratch and forward swim. J.
Neurophysiol. 83: 156 –165, 2000. Turtles with a complete transection
of the spinal cord just posterior to the forelimb enlargement at the
D2–D3 segmental border produced coordinated rhythmic hindlimb
movements. Ipsilateral stimulation of cutaneous afferents in the midbody shell bridge evoked a rostral scratch. Electrical stimulation of the
contralateral dorsolateral funiculus (DLF) at the anterior cut face of
the D3 segment activated a forward swim. Simultaneous stimulation
of the ipsilateral shell bridge and the contralateral DLF elicited a
scratch-swim hybrid: a behavior that blended features of both rostral
scratch and forward swim into each cycle of rhythmic movement. This
is the first demonstration of a scratch-locomotion hybrid in a spinal
vertebrate. The rostral scratch and the forward swim shared some
characteristics: alternating hip flexion and extension, similar timing of
knee extensor activity within the hip cycle, and a behavioral event
during which force was exerted against a substrate. During each cycle,
each behavior exhibited three sequential stages, preevent, event, and
postevent. The rostral scratch event was a rub of the foot against the
stimulated shell site. The forward swim event was a powerstroke, a
hip extension movement with the foot held in a vertical position with
toes and webbing spread. The two behaviors differed with respect to
several features: amount of hip flexion and extension, electromyogram
(EMG) amplitudes, and EMG duty cycles. Scratch-swim hybrids
displayed two events, the scratch rub and the swim powerstroke,
within each cycle. Hybrid hip flexion excursion, knee extensor EMGs,
and hip flexor EMGs were similar to those of the scratch; hybrid hip
extension excursion and hip extensor EMGs were similar to those of
the swim. The hybrid also had three sequential stages during each
cycle: 1) a combined scratch prerub and swim postpowerstroke, 2) a
scratch rub that also served as a swim prepowerstroke, and 3) a swim
powerstroke that also served as a scratch postrub. Merging of the
rostral scratch with the forward swim was possible because of similarities between the sequential stages of the two forms, making them
biomechanically compatible for hybrid formation. Kinematic and
myographic similarities between the rostral scratch and the forward
swim support the hypothesis that the two behaviors share common
neural circuitry. The common features of the sequential stages of each
behavior and the production of scratch-swim hybrids provide additional support for the hypothesis of a shared core of spinal cord
neurons common to both rostral scratch and forward swim.
SCRATCH-SWIM HYBRIDS
157
same task has been demonstrated previously for scratching in
the spinal turtle (Mortin et al. 1985; Robertson et al. 1985;
Stein et al. 1986a). This is the first demonstration in the spinal
turtle of blending of two different tasks to form a hybrid. Our
data were presented previously in an abstract (Earhart and
Stein 1999).
METHODS
Surgical preparation
Stimulation and recording
Reflective markers were placed over the hip, knee, ankle, and third
toe of each turtle. A band clamp that encircled the midbody was used
to suspend each turtle in a tank of water. The water level was adjusted
so that the turtle’s limbs remained below the water but the turtle could
lift its head above the water for breathing. Rostral scratching was
evoked using the smooth glass tip of a probe to stimulate mechanically the shell bridge at the SP2 site (Fig. 1) (Mortin et al. 1985).
Force was recorded using a Grass FT-03 force transducer (Astromed,
West Warwick, RI) attached to the probe. Forward swimming was
evoked by electrical stimulation of the dorsolateral funiculus (DLF) at
the anterior cut face of the D3 segment of the spinal cord (Fig. 1)
(Lennard and Stein 1977). The cut face was accessed in the midline
channel drilled for spinalization. Distinctions between gray and white
matter were clearly visible on the anterior cut face, allowing for visual
identification of the DLF. Electrical stimulation was delivered via the
cut ends of a bipolar electrode, consisting of a pair of 75 micrometer
silver wires with enamel coating. The stimulating electrode was
FIG. 1. Illustration of stimulation sites and joint angles. Mechanical stimulation at the SP2 site evoked a rostral scratch in the ipsilateral hindlimb.
Electrical stimulation of the contralateral dorsolateral funiculus (DLF) evoked
a forward swim in the ipsilateral hindlimb. Simultaneous stimulation of the
ipsilateral SP2 site and the contralateral DLF evoked a scratch-swim hybrid in
the ipsilateral hindlimb.
placed under visual guidance, and DLF stimulation in the ranges of
2–10 V, 30 –50 Hz, and 0.4 –.6 ms pulse durations was delivered.
Stimulation of some DLF sites within the ranges specified did not
evoke a forward swim. In these cases, the electrode was moved to a
new site until stimulation within the specified ranges evoked a forward swim in the hindlimb contralateral to the side of stimulation.
Seven of the 10 turtles studied produced scratch-swim hybrids in
response to simultaneous stimulation of the ipsilateral shell bridge and
contralateral DLF. One of these seven turtles also produced scratchswim hybrids in response to sequential stimulation in which the
termination of ipsilateral shell stimulation was followed immediately
by the onset of contralateral DLF stimulation.
Movement was recorded by use of a mirror placed below the tank
at an angle of 45°. Hindlimb movements were videotaped at 60 Hz
with a shutter speed of 1/250 s. Hindlimb scratching and swimming
movements were relatively planar; thus a single camera was sufficient
for measurement of joint angles. EMGs from each muscle were
amplified, filtered (100 –1,000 Hz band-pass), and recorded on digital
audio tape (DC-5 kHz band-pass). In all turtles, the monoarticular
knee extensor, the hip flexor, and the hip extensor were recorded. In
five of the seven turtles that produced scratch-swim hybrids, the
biarticular knee extensor was also recorded. A synchronization signal,
linked to opening of the camera shutter and generated by the Peak
Event Synchronization Unit (Peak Performance Technologies, Englewood, CO), was recorded on the video and the digital audio tapes so
that movement data and EMGs could be synchronized for analysis.
Data analysis
Kinematic analyses were performed on a total of 210 cycles,
composed of 30 cycles from each of the 7 turtles that produced
scratch-swim hybrids. The 30 cycles from each turtle included 10
rostral scratch cycles in response to shell stimulation, 10 forward
swim cycles in response to spinal cord stimulation, and 10 scratchswim hybrid cycles in response to simultaneous stimulation. Episodes
were selected for analysis if they contained at least five full cycles of
a behavior. Video tape records of selected episodes were replayed,
and the X and Y coordinates for each marker manually digitized (30
Hz) using a Peak 5.2 system (Peak Performance Technologies, Engle-
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Red-eared turtles (n ⫽ 10), Trachemys scripta elegans, were studied. All surgery was performed under hypothermic anesthesia (Melby
and Altman 1974). Turtles were placed in crushed ice for at least 1 h
before the start of surgery and kept in ice throughout the procedure.
For spinalization, a channel was drilled along the midline of the
carapace over the second (D2) and third (D3) dorsal segments of the
spinal cord. The spinal cord was completely transected between the
D2 and D3 segments, a wax well was constructed around the midline
channel over the exposed spinal cord, and the well was filled with
turtle saline.
After complete transection of the spinal cord, electrodes were
implanted in selected hindlimb muscles: a biarticular knee extensor
(triceps femoris pars iliotibialis), a monoarticular knee extensor (triceps femoris pars femorotibialis), a hip flexor (puboischiofemoralis
internus pars anteroventralis), and a hip extensor (flexor cruris pars
flexor tibialis internus) (Robertson et al. 1985). We used bipolar
electrodes, each a pair of 100-␮m silver wires with enamel coating
that were glued together with Permabond 910 (National Starch and
Chemical, Englewood, NJ). For hip flexor implantation, a rectangular
opening was drilled in the ventral plastron over the pelvic region,
providing access to the hip flexor muscle belly. For the knee extensors
and hip extensor, a triangular opening was drilled in the dorsal
carapace, and muscle bellies were accessed through small incisions of
the overlying skin. Electrodes for the knee extensors and hip extensor
were threaded under the skin and into the internal cavity exposed by
the hole in the carapace. Electrode tips were implanted directly into
the muscle bellies and glued in place. Following electrode placement
in the hip flexor, the hip flexor electrode was glued to the plastron and
the hole in the plastron was sealed with wax. Following electrode
placement in the knee extensors and hip extensor, the knee extensor
and hip extensor electrodes were secured to the sides of the hole in the
carapace, which was then sealed with wax. Skin incisions overlying
the implanted knee extensors and hip extensor were closed with glue
and covered with Op-Site (Smith and Nephew Medical Unlimited,
Hull, UK). The turtle was removed from ice and warmed to room
temperature.
158
G. M. EARHART AND P.S.G. STEIN
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
wood, CO). Hip angle was defined as the angle between the thigh and
the ventral midline of the body; knee angle was defined as the angle
between the thigh and the shank (Fig. 1). Hip and knee angle values
increased as the respective joints extended and decreased as the joints
flexed. Toe trajectories were obtained by viewing videos frame by
frame and tracing the path of the third toe on an acetate sheet placed
over the video screen.
The timing of behavioral events was determined from video
records. For the rostral scratch, rub was defined as the time during
which the foot made contact with the stimulated shell site. For the
forward swim, the powerstroke was defined as the time during which
the hip was extending and the foot was held in a vertical position with
toes and webbing spread.
A total of 150 episodes of EMGs (50 scratch, 50 swim, and 50
scratch-swim hybrids) was analyzed in the 7 turtles that produced
scratch-swim hybrids. Each episode of scratch-swim hybrid was
matched with a scratch episode and a swim episode from the same
turtle. Each episode consisted of at least five cycles. EMGs were
digitized at 2 kHz using Cambridge Electronic Design 1401 plus
hardware and Spike2 software (Cambridge, UK). EMGs were rectified and averaged so that the mean of 20 successive rectified measurements was calculated, giving 100 data points per second.
For some episodes, only EMGs were analyzed. For other episodes,
both EMGs and kinematics were analyzed. For these episodes, software custom written by Dr. Gavin Perry merged kinematic and EMG
files, aligning the two data sets using the synchronization signal
recorded on each. Onsets and offsets of hip flexion and knee extension
movements were determined (Field and Stein 1997a). For all episodes, EMG burst onset and offset times for each muscle were
determined. EMG amplitudes were obtained by averaging the activity
within each burst. For each muscle, amplitudes were normalized to
scratch values, such that the amplitudes for a scratch episode were
100% and amplitudes of the related swim and hybrid episodes expressed as a percentage of the scratch values. Cycle period was
defined as the time from the onset of one burst of hip extensor activity
to the onset of the next burst of hip extensor activity. Duty cycle,
defined as the percentage of cycle period during which a muscle was
active, was calculated as burst duration multiplied by 100 and then
divided by cycle period.
Phase analyses were used to determine the timing of knee extension
onset with respect to hip movement and the timing of monoarticular
knee extensor EMG onset with respect to hip muscle activity. Dualreferent phase analyses were used to normalize the hip flexion and hip
extension phases of each behavior. Dual-referent analyses are preferred to single-referent techniques in situations where the duty cycle
of the referent is variable. Dual-referent methods normalize for active
and inactive periods of the referent; thus changes in referent duty
cycle will not produce shifts in phase values obtained (see Fig. 2 of
Berkowitz and Stein 1994b). Events occurring during the hip flexion
portion of the cycle had phase values between 0 and 0.5, whereas
those occurring during hip extension had phase values between 0.5
and 1.0.
Kinematic phase analyses were performed using the method described by Field and Stein (1997a). The phase at which knee extension
onset occurred was defined as the latency between hip flexion onset
and knee extension onset, divided by twice the duration of hip flexion.
All knee extension onsets analyzed occurred during hip flexion, so
calculations for events occurring during hip extension are not presented.
Motor pattern phase analyses are described in Berkowitz and Stein
(1994b). In the present study, the hip extensor rather than the hip
flexor was chosen as a referent because the hip flexor had very short
burst durations in the forward swim and use of the hip flexor as a
referent for the swim gave inconsistent results. Using the hip extensor
as a referent, hip extensor offset was given a phase value of 0 and hip
extensor onset a phase value of 0.5. The phase of monoarticular knee
extensor onset was defined as the latency between hip extensor offset
FIG. 2. Toe trajectories and events of the 3 behaviors. A: rostral scratch.
B: forward swim. C: scratch-swim hybrid. Asterisks denote the scratch rub
of the foot against the stimulated shell site. Thick dashed lines denote the
swim powerstroke. Hindlimb position at the scratch rub is shown in A.
Hindlimb position in the middle of the swim powerstroke is shown in B.
Two hindlimb positions, one at the scratch rub and one in the middle of the
swim powerstroke, are shown for the scratch-swim hybrid in C.
SCRATCH-SWIM HYBRIDS
and monoarticular knee extensor onset, divided by twice the duration
of hip extensor quiescence. All monoarticular knee extensor onsets
analyzed occurred during hip extensor quiescence, so calculations for
events occurring during hip extensor activity are not presented.
Because the phase data are cyclic, vector algebra techniques
(Batschelet 1981) were used to determine average phase values. Each
phase was converted to a two-dimensional unit vector with an angle of
2␲␾ and a length of 1. Unit vectors were averaged using vector
addition. The angle of the mean vector, divided by 2␲, was the mean
phase; it had values between 0 and 1. Mean angular deviation, a
measure of phase data dispersion, was also calculated.
Statistical analyses
RESULTS
Mechanical stimulation of the shell bridge at SP2 evoked
rostral scratching movements in the ipsilateral hindlimb. Electrical stimulation of the contralateral DLF at the anterior cut
face of the D3 segment evoked forward swimming movements
in the ipsilateral hindlimb. In 7 of 10 turtles, simultaneous
stimulation of the ipsilateral shell bridge and the contralateral
DLF evoked a scratch-swim hybrid, a rhythmic hindlimb behavior in which features of both rostral scratch and forward
swim were expressed in each individual cycle of the multicycle
response in the ipsilateral hindlimb. In the other three turtles,
simultaneous stimulation elicited forward swimming movements only. There were no apparent differences in the scratch
or swim behaviors of these three turtles, compared with those
of the seven that did produce scratch-swim hybrids. Data from
the three turtles that did not produce scratch-swim hybrids
were not analyzed.
Kinematics
The rostral scratch, forward swim, and scratch-swim hybrid
were distinguished from one another based on toe trajectory
and behavioral events (Fig. 2). During the rostral scratch and
the scratch-swim hybrid, the limb reached toward and the foot
rubbed against the stimulated shell site. A rub against the shell
did not occur in the forward swim. The scratch rub, which
constituted the only event of the rostral scratch and the first of
two events of the scratch-swim hybrid, is marked by asterisks
in Fig. 2, A and C. For the scratch, the toe reached its most
medial position anteriorly at the stimulated site on the shell
bridge, SP2. The toe also reached this medial anterior position
during the hybrid, but not during the swim.
During the forward swim and the scratch-swim hybrid, the
limb performed a swim powerstroke: the foot was held vertically with toes and webbing spread as the hip extended.
Spreading of the toes and webbing did not occur in the rostral
scratch. The swim powerstroke, which constituted the only
event of the swim and the second of two events of the hybrid,
is marked by thick dashed lines in Fig. 2, B and C. For the
swim, the toe reached its most medial position posteriorly at a
location caudal to the rear edge of the carapace. The toe also
reached this medial posterior position during the hybrid, but
not during the scratch.
Thus the scratch-swim hybrid displayed two events: a
scratch rub and a swim powerstroke. The toe trajectory of the
hybrid encompassed both the scratch and the swim toe trajectories, merging the two into a single smooth trajectory.
The rostral scratch and the forward swim were also distinguished from one another based on minimum and maximum
hip angles (Fig. 3) (see also Field and Stein 1997a). The hip
flexed more in the rostral scratch than in the forward swim and
extended more in the forward swim than in the rostral scratch.
Thus the minimum hip angle was smaller in the scratch than in
the swim; the maximum hip angle was larger in the swim than
in the scratch. The scratch-swim hybrid had a minimum hip
angle similar to that of the scratch and a maximum hip angle
similar to that of the swim.
Hip angle versus knee angle plots provided an additional
means of distinguishing the rostral scratch from the forward
swim. The hip angle versus knee angle plot for the scratch was
roughly triangular in shape (Fig. 4A), for the swim was ovoid
(Fig. 4B), and for the hybrid combined elements of the two
such that the base of the triangle was merged with the upper
portions of the oval (Fig. 4C).
For the rostral scratch, the hip angle versus knee angle plot
(Fig. 4A) was characterized by a horizontal base, illustrating
that the knee extended while the hip remained flexed (Mortin
et al. 1985). At the peak of this knee extension, the foot rubbed
against the stimulated site on the shell (asterisk, Fig. 4A). A
similar scratch rub occurred in the hybrid (asterisk, Fig. 4C),
and the hybrid hip angle versus knee angle plot had a horizontal base similar to that of the scratch plot. The horizontal base
and occurrence of the scratch rub were not present in the swim,
which had a hip angle versus knee angle plot with a rounded
base. Scratch minimum knee angle, denoted by the left vertical
dashed lines in Fig. 4, was smaller than swim minimum knee
angle. Scratch maximum knee angle, denoted by the right
vertical dashed lines in Fig. 4, was larger than swim maximum
knee angle. The hybrid had minimum and maximum knee
angles similar to those of the scratch.
Hip angle versus knee angle plots also illustrated the differences in hip angle minima and maxima. The bottom horizontal
dashed lines on each panel of Fig. 4 show the minimum hip
FIG. 3. Average minimum and maximum hip angles ⫾SD for the 3 behaviors. Asterisks denote significance at P ⱕ 0.0166 (Mann-Whitney U). NS, not
significant.
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
Kinematic and myographic parameters were compared across conditions using nonparametric statistics. Mean minimum and maximum
hip angles, mean EMG amplitudes, mean EMG duty cycles, and mean
period were compared across behaviors using the Mann-Whitney U
test (Hays 1994). Mean phase of knee extension movement onset and
monoarticular knee extensor EMG onset were compared across conditions using the Watson U2 test for circular data (Batschelet 1981).
Pair-wise comparisons of scratch to swim, scratch to hybrid, and
swim to hybrid were used for all statistical analyses. A Bonferroni
correction was used to account for the increase in type I error that
resulted from doing multiple tests. Because three separate pair-wise
comparisons were performed, the chosen ␣ level of 0.05 was divided
by 3. All comparisons were tested at the Bonferroni-adjusted level of
0.0166. Comparisons reported to be statistically significant had values
of P ⱕ 0.0166.
159
160
G. M. EARHART AND P.S.G. STEIN
TABLE
1.
Knee phase in the hip cycle
Knee extension movement
onset
Monoarticular knee extensor
onset
Rostral
Scratch
Forward
Swim
Scratch-Swim
Hybrid
0.29 ⫾ 0.04
0.30 ⫾ 0.04
0.30 ⫾ 0.04
0.29 ⫾ 0.08
0.31 ⫾ 0.06
0.32 ⫾ 0.05
Values are mean phases ⫾ angular deviations for knee extension movement
onset within the dual-referent hip movement cycle and monoarticular knee
extensor onset within the dual-referent hip motor pattern cycle. There were no
significant differences in phase values of the different behaviors (Watson U2).
icant when testing within each turtle or across all turtles
(Watson U2).
EMG
FIG. 4. Hip angle versus knee angle plots for the rostral scratch (A),
forward swim (B), and scratch-swim hybrid (C). Asterisks denote scratch rub
of the foot against the stimulated shell site. Thick lines denote the timing of
swim powerstroke. Arrows show direction of movement. Horizontal dashed
lines at the top and bottom of each plot show the maximum hip angle for the
swim and the minimum hip angle for the scratch, respectively. Vertical dashed
lines at the left and right of each plot show the minimum and maximum knee
angles for the scratch, respectively. Note that the hybrid (C) plot spans both the
vertical and horizontal extents of the area between the dashed lines.
angle for the scratch; the top horizontal dashed lines show the
maximum hip angle for the swim. The hybrid plot spans the
area between both of these lines, indicating that the hybrid had
hip flexion excursion similar to that of the scratch and hip
extension excursion similar to that of the swim.
Mean phase values for the onset of knee extension within the
dual-referent hip movement cycle are given in Table 1. The
rostral scratch and the forward swim had very similar phasing,
with knee extension onset occurring near the middle of the hip
flexion phase (Field and Stein 1997a). Hybrid mean phase
values were also very similar to those of the scratch and the
swim. Differences among the three behaviors were not signif-
FIG. 5. Electromyograms (EMGs) for 3 cycles each of rostral scratch (A),
forward swim (B1), and the scratch-swim hybrid (C), all with the same set of
vertical scales. B2: forward swim data in B1 at a set of vertical gains 10 times
that used for A, B1, and C. The scratch frequency was higher than that of the
swim or the hybrid; the time scale represents 0.5 s for A and 0.8 s for B and C.
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
The rostral scratch and the forward swim motor patterns
were distinguished from one another by differences in amplitudes and duty cycles (Juranek and Currie 1998, 2000; Stein
and Johnstone 1986). Figure 5 shows scratch, swim, and hybrid
EMGs with the same set of vertical scales (Fig. 5, A, B1, and
C, respectively). Figure 5B2 shows the swim data in B1 at 10
times the set of vertical gains of A, B1, and C. This higher gain
representation is provided because the monoarticular knee extensor and hip flexor amplitudes are so low in the swim that
they are difficult to see when shown with the same set of scales
as the scratch and the hybrid, as in Fig. 5B1.
EMG amplitudes differed among the behaviors (Fig. 6).
Knee extensor and hip flexor amplitudes were significantly
larger for the rostral scratch than for the forward swim,
SCRATCH-SWIM HYBRIDS
161
FIG. 6. Mean EMG amplitudes ⫾ SD for
each muscle across behaviors. Amplitudes
were normalized to scratch values, such that
scratch values were 100% and values for the
forward swim and hybrid were expressed as a
percent of the scratch values. Asterisks denote
significance at P ⱕ 0.0166 (Mann-Whitney
U). NS, not significant.
swim was 0.92 ⫾ 0.25 s, and for the hybrid was 1.00 ⫾
0.25 s. We observed differences in variability of cycle
period, however, noting that the swim had a very regular
period whereas scratch period was more volatile. The hybrid
had a stable period similar to that of the swim. Cycle period
standard deviation in the scratch was significantly larger
than in the swim or the hybrid (Mann Whitney U). There
was no difference in cycle period standard deviation between the swim and the hybrid.
The mean phase values for the onset of the monoarticular
knee extensor in the dual-referent hip extensor cycle are given
in Table 1. Rostral scratch and forward swim had very similar
phasing, with monoarticular knee extensor onset occurring
near the middle of the hip flexor phase (Juranek and Currie
1998, 2000). Hybrid mean phase values were also very similar
to those of the scratch and the swim. Differences among the
three behaviors were not significant when testing within each
turtle or across all turtles (Watson U2).
Sequential stimulation
One of the turtles in this study exhibited a substantial scratch
afterdischarge (Currie and Stein 1990) in which four to six
cycles of scratch were produced following offset of the mechanical shell stimulation. In this turtle, we were able to obtain
the scratch-swim hybrid through sequential delivery of me-
FIG. 7. Mean EMG duty cycles ⫾ SD for
each muscle across behaviors. Asterisks denote significance at P ⱕ 0.0166 (Mann Whitney U). NS, not significant.
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
whereas hip extensor amplitude was significantly larger for the
swim than for the scratch (Mann-Whitney U). Knee extensor
and hip flexor amplitudes were not significantly different between the scratch and the hybrid. The hybrid had hip extensor
amplitudes more similar to those of the swim than those of the
scratch. Hybrid hip extensor amplitudes were significantly
larger than those of the scratch or the swim (Mann-Whitney
U). In summary, the hybrid had knee extensor and hip flexor
amplitudes similar to scratch but hip extensor amplitudes similar to swim (compare Fig. 5, A, B1, and C).
Duty cycle was another means of distinguishing among the
behaviors (Fig. 7). Knee extensor and hip flexor duty cycles
were significantly higher for the rostral scratch than for the
forward swim, whereas hip extensor duty cycle was significantly higher for the swim than for the scratch (Mann Whitney
U). The hybrid had knee extensor and hip flexor duty cycles
intermediate between those of the scratch and those of the
swim. Hybrid knee extensor duty cycles were significantly
different from those of the scratch and the swim (Mann
Whitney U). Hybrid hip flexor and hip extensor duty cycles
were not significantly different from those of the swim, but
were significantly different from those of the scratch (Mann
Whitney U).
Comparisons of cycle period across the three behaviors
revealed no significant differences. Mean cycle period for
the rostral scratch was 0.89 ⫾ 0.31 (SD) s, for the forward
162
G. M. EARHART AND P.S.G. STEIN
chanical shell stimulation followed by electrical DLF stimulation. The mechanical shell stimulus was used to evoke rostral
scratch and, coincident with the offset of this mechanical shell
stimulation, electrical DLF stimulation was initiated. This sequential presentation produced a response that began with
rostral scratch, showed several hybrid cycles just after the
offset of mechanical stimulation and onset of electrical stimulation, and ended with forward swim (Fig. 8). Thus the onset of
electrical DLF stimulation at the offset of scratch stimulation
evoked a swim that combined with the scratch afterdischarge to
produce the hybrid. The average number of cycles of hybrid
obtained with sequential stimulation (n ⫽ 4.8 ⫾ 0.7 cycles)
was similar to the average number of cycles of afterdischarge
obtained in scratch episodes (n ⫽ 4.4 ⫾ 0.8 cycles).
The rostral scratch cycles were characterized by greater hip
flexion and less hip extension than the forward swim cycles.
The scratch also showed higher amplitude monoarticular knee
extensor and hip flexor amplitudes and lower hip extensor
amplitudes than the swim. The hybrid showed hallmark features of both the scratch and the swim. The hybrid had two
events, a scratch rub and a swim powerstroke. The hybrid had
a minimum hip angle similar to the scratch and a maximum hip
angle similar to the swim. Hybrid monoarticular knee extensor
and hip flexor amplitudes were similar to those of the scratch,
and hybrid hip extensor amplitudes similar to those of the
swim.
DISCUSSION
The turtle spinal cord produced blends of the rostral scratch
and the forward swim in response to simultaneous mechanical
stimulation of the ipsilateral shell bridge and electrical stimulation of the contralateral DLF at the anterior cut face of the D3
segment. The scratch-swim hybrid displayed two events: a
scratch rub of the foot against the stimulated shell site and a
swim powerstroke during which the hip extended and the foot
was held in a vertical position with toes and webbing spread.
The hybrid toe trajectory combined the trajectories of the
rostral scratch and the forward swim into a single smooth path.
This is the first demonstration of a scratch-locomotion hybrid
in a spinal vertebrate.
Rhythmic behavior may be characterized as a sequence of
stages produced by a pattern generator (Kleinfeld 1986). Each
cycle of rhythmic behavior may be represented by a linear
chain, i.e., a fixed sequence in which specific stages occur in a
particular order (Berridge et al. 1987). The formation of a
hybrid of two behaviors depends on the biomechanical compatibility of the stages of the two behaviors. If the stages of two
behaviors are very different, hybrid formation may be impossible. However, if the stages of one behavior are similar to
those of the other behavior, the two will be biomechanically
compatible for formation of a hybrid via merging of the stages
of the two behaviors.
Rostral scratch and forward swim are both rhythmic behaviors in which each cycle can be represented by a sequence of
stages. Both rostral scratch and forward swim consisted of
three stages, preevent, event, and postevent, performed in
sequence during each cycle. For rostral scratch, these stages
were prerub, rub, and postrub (Mortin et al. 1985). For forward
swim, these stages were prepowerstroke, powerstroke, and
postpowerstroke (Lennard and Stein 1977). The prerub stage of
the rostral scratch and the postpowerstroke stage of the forward
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
FIG. 8. Illustration of a scratch-swim hybrid in response to sequential stimulation. The episode began with mechanical shell
stimulation only and, at offset of the mechanical stimulation, electrical DLF stimulation was initiated and maintained (shown by
bottom 2 traces). Hip angle is presented in the top trace. Asterisks represent the timing of scratch rubs. Thick lines on the hip angle
trace represent the timing of swim powerstrokes. Top dashed line on the hip angle trace denotes maximum hip angle for the swim,
and bottom dashed line denotes minimum hip angle for the scratch. The figure begins during an ongoing scratch, 2.5 s before offset
of mechanical shell stimulation and onset of electrical DLF stimulation. The 1st 3 cycles are rostral scratch, the next 6 hybrids, and
the last few forward swims.
SCRATCH-SWIM HYBRIDS
Other examples of hybrids
Hybrids of two different tasks or of two different forms of a
task have been demonstrated previously. Carter and Smith
(1986a,b) demonstrated hybrids of two different tasks, pawshake and stepping, in normal and in spinal cats. Cats produced
several cycles of paw-shake during the swing phase of each
step cycle. Paw-shake and stepping are biomechanically compatible for hybrid formation because the paw-shake can be
performed during swing, when the foot is not in contact with
the ground, without disrupting the step cycle. Hybrids of two
scratch forms in response to two-site mechanical shell stimulation or stimulation within a transition zone have been demonstrated in the turtle. In rostral-pocket, pocket-caudal, and
rostral-caudal hybrids, the turtle rubs the shell twice during
each cycle (Mortin et al. 1985; Robertson et al. 1985; Stein et
al. 1986a,b). Two scratch forms are biomechanically compatible for hybrid formation because each form has different
timing of knee extension with the hip cycle to perform the rub.
Two scratch rubs of different forms can be performed during
each hip cycle to form a hybrid. These studies demonstrate that
spinal structures can coordinate hybrids in the absence of
supraspinal inputs. The present study adds support to this
conclusion.
Hybrids of two forms of a task have also been demonstrated
in humans. Horak and Nashner (1986) reported two forms of
postural movements: the ankle strategy and the hip strategy. In
response to horizontal perturbations, subjects use the ankle
strategy if standing on a broad support surface and the hip
strategy if standing on a support surface shorter than their feet.
When standing on a surface of intermediate length, subjects
use a blend of ankle and hip strategies, an ankle-hip hybrid.
Another example of human behavior that illustrates hybrid
formation is scratching. A human can scratch some sites on the
side of the thorax using either the side of the elbow or the hand.
These sites can also be scratched using a hybrid strategy: the
site is rubbed with the elbow and then with the hand during
each single cycle of movement. This is possible because the
two movements are biomechanically compatible. Other human
scratch movements are not biomechanically compatible for
hybrid formation. For example, a human scratches the upper
back using a form with the elbow positioned above the shoulder; a human scratches the lower back using a form with the
elbow positioned below the shoulder (Stein et al. 1986b). The
below-shoulder and above-shoulder forms require very different upper extremity orientations. Biomechanical constraints
prevent the human from reaching the upper back with the
below-shoulder form or the lower back with the above-shoulder form. A human can switch between the two forms, performing the above-shoulder form in one cycle and the belowshoulder form in the next cycle, but the two forms cannot be
incorporated into a single, smooth hybrid cycle because of their
biomechanical incompatibility.
Shared circuitry
Our concepts of how pattern generators are organized have
changed (Stein et al. 1997). Pattern generators are no longer
viewed as static, unshared circuits each exclusive to a single
behavior. In Aplysia, cerebral interneuron CC5 is known to
participate in at least six distinct behaviors (Xin et al. 1996).
Evidence from the crustacean stomatogastric system shows
that two independent networks can fuse and operate as a single
functional unit (Dickinson 1995; Dickinson et al. 1990; Meyrand et al. 1994). Svoboda and Fetcho (1996) demonstrated
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
swim were similar; both exhibited hip flexion movements
during which the foot was held in a horizontal position. The
rub stage of the rostral scratch and the prepowerstroke stage of
the forward swim were similar; both exhibited knee extension
movements that occurred while the hip was flexed. The postrub
stage of the rostral scratch and the powerstroke stage of the
forward swim were similar; both exhibited hip extension
movements, although the toes and webbing were spread during
swim powerstroke but not during scratch postrub. We speculate that the similarities of stages in these two behaviors provided biomechanical compatibility that was consistent with the
formation of the hybrid. The scratch-swim hybrid, like the
individual behaviors, had three stages: 1) a scratch prerub and
swim postpowerstroke combination, 2) a scratch rub that also
served as a swim prepowerstroke, and 3) a swim powerstroke
that also served as a scratch postrub. The similarity of stages
between the two behaviors and the merging of stages in the
hybrid support the hypothesis that the two behaviors share
some pattern-generating circuitry.
Previous work has noted that the minimum and maximum
hip angles for the rostral scratch and the forward swim are
different (Field and Stein 1997a). We have confirmed this
result and expanded on it, demonstrating correspondence of
EMG amplitudes and duty cycles with hip angle minima and
maxima. Smaller minimum hip angles in the scratch and the
hybrid were associated with higher hip flexor amplitude and
duty cycle. Greater maximum hip angles in the swim and the
hybrid were associated with higher hip extensor amplitude and
duty cycle. Higher EMG amplitudes were associated with the
behavioral events during which force was exerted against a
substrate. Higher knee extensor and hip flexor amplitudes were
associated with the scratch rub; higher hip extensor amplitudes
were associated with the swim powerstroke.
Thus rostral scratch and forward swim were distinguished
from one another by EMG amplitudes. They were not distinguished from one another by phasing of knee movement or
monoarticular knee extensor activity within the hip cycle,
characteristics that are often excellent discriminators. Rostral
scratch and forward swim had similar phasing of movement
(knee extension onset during hip flexion) and of motor patterns
(monoarticular knee extensor onset near the middle of the hip
flexor burst). In contrast, forward swim and backpaddle have
very different timings of knee extension onset; knee extension
onset occurs near the middle of hip flexion for the forward
swim but near the middle of hip extension in the backpaddle
(Field and Stein 1997a). The three scratch forms can also be
distinguished from one another by phasing of knee activity
within the hip cycle. Monoarticular knee extensor onset occurs
during the hip flexor burst in rostral scratch, during the hip
extensor burst in pocket scratch, and near the offset of the hip
extensor burst in caudal scratch (Robertson et al. 1985). In the
cat, paw-shake and locomotion can be distinguished from one
another by phasing of monoarticular knee extensor activity
within the cycle of ankle motor activity. Vastus lateralis activity coincides with ankle flexor activity in paw-shake and with
ankle extensor activity during locomotion (Smith et al. 1986).
163
164
G. M. EARHART AND P.S.G. STEIN
We thank A. Berkowitz for editorial assistance and G. Perry for software
development.
This work was supported by National Institutes of Health Grants R01-NS30786 and T32-HD-07434 and by a Promotion of Doctoral Studies Award
from the Foundation for Physical Therapy, Inc. to G. M. Earhart.
Address for reprint requests: P.S.G. Stein, Dept. of Biology, Washington
University, St. Louis, MO 63130.
Received 23 July 1999; accepted in final form 3 September 1999.
REFERENCES
BATSCHELET, E. Circular Statistics in Biology. London: Academic, 1981.
BERKINBLIT, M. B., DELIAGINA, T. G., FELDMAN, A. G., GELFAND, I. M., AND
ORLOVSKY, G. N. Generation of scratching. II. Nonregular regimes of
generation. J. Neurophysiol. 41: 1058 –1069, 1978.
BERKOWITZ, A. AND STEIN, P.S.G. Activity of descending propriospinal axons
in the turtle hindlimb enlargement during two forms of fictive scratching:
broad tuning to regions of the body surface. J. Neurosci. 14: 5089 –5104,
1994a.
BERKOWITZ, A. AND STEIN, P.S.G. Activity of descending propriospinal axons
in the turtle hindlimb enlargement during two forms of fictive scratching:
phase analyses. J. Neurosci. 14: 5105–5119, 1994b.
BERRIDGE, K. C., FENTRESS, J. C., AND PARR, H. Natural syntax rules control
action sequence of rats. Behav. Brain Res. 23: 59 – 68, 1987.
BUFORD, J. A. AND SMITH, J. L. Adaptive control for backward quadrupedal
walking. I. Posture and hindlimb kinematics. J. Neurophysiol. 64: 745–755,
1990a.
BUFORD, J. A. AND SMITH, J. L. Adaptive control for backward quadrupedal
walking. II. Hindlimb muscle synergies. J. Neurophysiol. 64: 756 –766,
1990b.
CARTER, M. C. AND SMITH, J. L. Simultaneous control of two rhythmical
behaviors. I. Locomotion with paw-shake response in normal cat. J. Neurophysiol. 56: 171–183, 1986a.
CARTER, M. C. AND SMITH, J. L. Simultaneous control of two rhythmical
behaviors. II. Hindlimb walking with paw-shake response in spinal cat.
J. Neurophysiol. 56: 184 –195, 1986b.
CURRIE, S. N. AND GONSALVES, G. G. Right-left interactions between rostral
scratch networks generate rhythmicity in the preenlargement spinal cord of
the turtle. J. Neurophysiol. 78: 3479 –3483, 1997.
CURRIE, S. N. AND GONSALVES, G. G. Reciprocal interactions in the turtle
hindlimb enlargement contribute to scratch rhythmogenesis. J. Neurophysiol. 81: 2977–2987, 1999.
CURRIE, S. N. AND STEIN, P.S.G. Cutaneous stimulation evokes long-lasting
excitation of spinal interneurons in the turtle. J. Neurophysiol. 64: 1134 –
1148, 1990.
DAVENPORT, J., MUNKS, S. A., AND OXFORD, P. J. A comparison of the
swimming of marine and freshwater turtles. Proc. R. Soc. Lond. B Biol. Sci.
220: 447– 475, 1984.
DICKINSON, P. S. Interactions among neural networks for behavior. Curr. Opin.
Neurobiol. 5: 792–798, 1995.
DICKINSON, P. S., MECSAS, C., AND MARDER, E. Neuropeptide fusion of two
motor-pattern generator circuits. Nature 334: 155–158, 1990.
EARHART, G. M. AND STEIN, P.S.G. Task blends in the spinal turtle: scratchswim hybrids. Soc. Neurosci. Abstr. 25: 1908, 1999.
FIELD, E. C. AND STEIN, P.S.G. Spinal cord coordination of hindlimb movements in the turtle: intralimb temporal relationships during scratching and
swimming. J. Neurophysiol. 78: 1394 –1403, 1997a.
FIELD, E. C. AND STEIN, P.S.G. Spinal cord coordination of hindlimb movements in the turtle: interlimb temporal relationships during bilateral scratching and swimming. J. Neurophysiol. 78: 1404 –1413, 1997b.
GRASSO, R., BIANCHI, L., AND LACQUANITI, F. Motor patterns for human gait:
backward versus forward locomotion. J. Neurophysiol. 80: 1868 –1885,
1998.
HAYS, W. Statistics (5th ed.). Fort Worth, TX: Harcourt College Publishers,
1994.
HORAK, F. B. AND NASHNER, L. M. Central programming of postural movements: adaptation to altered support-surface configurations. J. Neurophysiol.
55: 1369 –1381, 1986.
JURANEK, J. AND CURRIE, S. N. Fictive swimming elicited by electrical stimulation of the turtle spinal cord: interactions with the scratch reflex. Soc.
Neurosci. Abstr. 24: 1669, 1998.
JURANEK, J. AND CURRIE, S. N. Electrically evoked fictive swimming in the
low-spinal immobilized turtle. J. Neurophysiol. 83: 146 –155, 2000.
KLEINFELD, D. Sequential state generation by model neural networks. Proc.
Natl. Acad. Sci. USA 83: 9469 –9473, 1986.
LENNARD, P. R. AND STEIN, P.S.G. Swimming movements elicited by electrical
stimulation of turtle spinal cord. I. Low-spinal and intact preparations.
J. Neurophysiol. 40: 768 –778, 1977.
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
resetting of the swim rhythm by activation of an escape behavior, supporting the concept of shared circuitry for swim and
escape in goldfish. Soffe (1993) showed that both swimming
and struggling in the Xenopus embryo are driven by a common
set of premotor and motor neurons. Previous studies in the
turtle support sharing between left and right side scratch (rostral, pocket, and caudal) circuitry (Currie and Gonsalves 1997,
1999; Field and Stein 1997b; Stein et al. 1995, 1998a,b).
Single-unit recordings in turtle also support the concept of
sharing of elements among left and right rostral and pocket
scratch circuits (Berkowitz and Stein 1994a,b). Single-unit
recordings in cat provide support for sharing of elements for
stepping and scratching (Berkinblit et al. 1978). The demonstration in cats of paw-shake and step hybrids also supports the
concept of shared circuitry (Carter and Smith 1986a,b; Smith et
al. 1986). Thus there is strong evidence from many systems for
sharing of circuitry among different behaviors.
Stein (1983) noted kinematic similarities between rostral
scratch and forward swim. Quantitative analyses of kinematics
confirmed this observation and led to the suggestion that there
may be sharing of circuitry between rostral scratch and forward
swim networks (Field and Stein 1997a). Motor pattern similarities, such as similar timing of the monoarticular knee extensor within the hip cycle, also support the possibility of
shared circuitry (Juranek and Currie 1998, 2000; Stein and
Johnstone 1986). Additional support comes from the demonstration of interruption and resetting of an ongoing forward
swim by a brief rostral scratch (Juranek and Currie 1998, 2000;
Stein 1981).
Differences in rostral scratch and forward swim EMG amplitudes served to discriminate between the two behaviors.
Motor pattern amplitude differences between the behaviors
may be produced by changes in interneuronal recruitment from
classes active during both behaviors. A subset of neurons may
be active during each behavior, and additional neurons may be
recruited during the behavioral event of each to increase amplitudes during the time when force is exerted against a substrate. These additional neurons may be recruited from unshared populations or from within classes of neurons active
during both behaviors. Such recruitment has been demonstrated in Xenopus, where additional interneurons not active in
low-frequency swimming are recruited during struggling and
during high-frequency swimming (Sillar and Roberts 1993;
Soffe 1993).
The kinematic and myographic similarities between rostral
scratch and forward swim, as well as the similarities between
the sequentially generated stages of the behaviors, support the
hypothesis that network elements are shared between rostral
scratch and forward swim circuits. The ability of the spinal
cord to blend rostral scratch and forward swim into a single
smooth behavior as demonstrated in the present paper provides
additional support for this concept. We hypothesize that studies
using single-unit recording techniques during rostral scratch
and forward swim will reveal spinal interneurons that are
active during both behaviors.
SCRATCH-SWIM HYBRIDS
STEIN, P.S.G., GRILLNER, S., SELVERSTON, A. I., AND STUART, D. G. Neurons,
Networks, and Motor Behavior. Cambridge, MA: MIT Press, 1997.
STEIN, P.S.G. AND JOHNSTONE, K. M. Motor patterns during rostral scratching
and forward swimming in the turtle. Soc. Neurosci. Abstr. 12: 790, 1986.
STEIN, P.S.G., MCCULLOUGH, M. L., AND CURRIE, S. N. Reconstruction of
flexor/extensor alternation during fictive rostral scratching by two-site stimulation in the spinal turtle with a transverse spinal hemisection. J. Neurosci.
18: 467– 479, 1998a.
STEIN, P.S.G., MCCULLOUGH, M. L., AND CURRIE, S. N. Spinal motor patterns
in the turtle. Ann. NY Acad. Sci. 860: 142–154, 1998b.
STEIN, P.S.G., MORTIN, L. I., AND ROBERTSON, G. A. The forms of a task and
their blends. In: Neurobiology of Vertebrate Locomotion, edited by S.
Grillner, P.S.G. Stein, D. G. Stuart, H. Forssberg, and R. M. Herman.
Hampshire, UK: Macmillan, 1986b, p. 201–216.
STEIN, P.S.G. AND SMITH, J. L. Neural and biomechanical control strategies for
different forms of vertebrate hindlimb motor tasks. In: Neurons, Networks,
and Motor Behavior, edited by P.S.G. Stein, S. Grillner, A. I. Selverston,
and D. G. Stuart. Cambridge, MA: MIT Press, 1997, p. 61–73.
STEIN, P.S.G., VICTOR, J. C., FIELD, E. C., AND CURRIE, S. N. Bilateral control
of hindlimb scratching in the spinal turtle: contralateral spinal circuitry
contributes to the normal ipsilateral motor pattern of fictive rostral scratching. J. Neurosci. 15: 4343– 4355, 1995.
SVOBODA, K. R. AND FETCHO, J. R. Interactions between the neural networks
for escape and swimming in goldfish. J. Neurosci. 16: 843– 852, 1996.
THORSTENSSON, A. How is the normal locomotor program modified to produce
backward walking? Exp. Brain Res. 61: 664 – 668, 1986.
WINTER, D. A., PLUCK, N., AND YANG, J. F. Backward walking: a simple
reversal of forward walking? J. Mot. Behav. 21: 291–305, 1989.
XIN, Y., WEISS, K. R., AND KUPFERMANN, I. An identified interneuron contributes to aspects of six different behaviors in Aplysia. J. Neurosci. 16:
5266 –5279, 1996.
ZUG, G. R. Buoyancy, locomotion, morphology for the pelvic girdle and
hindlimb, and systematics of crytodiran turtles. Misc. Pub. Mus. Zool. Univ.
Mich. 142: 1–98, 1971.
Downloaded from http://jn.physiology.org/ by 10.220.33.5 on June 16, 2017
MELBY, E.C.J. AND ALTMAN, N. H. Handbook of Laboratory Animal Science.
Cleveland, OH: CRC, 1974, vol. I.
MEYRAND, P., SIMMERS, J., AND MOULINS, M. Dynamic construction of a neural
network from multiple pattern generators in the lobster stomatogastric
nervous system. J. Neurosci. 14: 630 – 644, 1994.
MORTIN, L. I., KEIFER, J., AND STEIN, P.S.G. Three forms of the scratch reflex
in the spinal turtle: movement analysis. J. Neurophysiol. 53: 1501–1516,
1985.
ROBERTSON, G. A., MORTIN, L. I., KEIFER, J., AND STEIN, P.S.G. Three forms
of the scratch reflex in the spinal turtle: central generation of motor patterns.
J. Neurophysiol. 53: 1517–1534, 1985.
SILLAR, K. T. AND ROBERTS, A. Control of frequency during swimming in
Xenopus embryos: a study on interneuronal recruitment in a spinal rhythm
generator. J. Physiol. (Lond.) 472: 557–572, 1993.
SMITH, J. L., BRADLEY, N. S., CARTER, M. C., GIULIANI, C. A., HOY, M. G.,
KOSHLAND, G. F., AND ZERNICKE, R. F. Rhythmical movements of the
hindlimbs in spinal cat: considerations for a controlling network. In: Development and Plasticity of the Mammalian Spinal Cord, edited by M. E.
Goldberger, A. Gorio, and M. Murray. Padova, Italy: Liviana Press, 1986,
Fidia Research Series, vol. 3, p. 347–361.
SOFFE, S. R. Two distinct motor patterns are driven by common premotor and
motor neurons in a simple vertebrate spinal cord. J. Neurosci. 13: 4456 –
4469, 1993.
STEIN, P.S.G. Swimming movements elicited by electrical stimulation of the
turtle spinal cord: the high spinal preparation. J. Comp. Physiol. 124:
203–210, 1978.
STEIN, P.S.G. Neural control of locomotion in the turtle. In: Advances in
Physiological Science, edited by J. Szentagothai, M. Palkovita, and J.
Hamori. Budapest: Akademiai Kiado, 1981, vol. 1, p. 149 –154.
STEIN, P.S.G. The vertebrate scratch reflex. Symp. Soc. Exp. Biol. 37: 383– 403,
1983.
STEIN, P.S.G., CAMP, A. W., ROBERTSON, G. A., AND MORTIN, L. I. Blends of
rostral and caudal scratch reflex motor patterns elicited by simultaneous
stimulation of two sites in the spinal turtle. J. Neurosci. 6: 2259 –2266,
1986a.
165