Neural Mechanisms of Leg Motor Control in Crayfish - Advancing Artificial Intelligence Through Biological Process Applications - page 25

tonic background discharge, these afferents
may supply continuous information on joint
angular position. In addition, when a move-
ment is imposed, they also perform powerful
angular movement and position detection.

Coding Actual Leg Movements in
Freely Behaving Crayfish

Isolating a sensory organ in vitro necessarily
suppresses an ensemble of perturbing and modu-
latory influences that normally have substantial
effects on its coding properties. For example,
the neuromodulator serotonin modifies CBCO
encoding in a dose-dependent manner (Rossi-
Durand, 1993). Determining the actual coding
properties of a sensory receptor thus requires an
in situ analysis of its discharge evoked by real
stimuli. For this purpose (Figures 2C1, C2), two
wire electrodes were used in vivo to record and
sort extracellular orthodromic (sensory) action
potentials from the CBCO neurogram of freely
behaving crayfish (Le Ray et al. , 2005). Spike
discrimination clearly identified an average of 15
CBCO afferents activated during actual walking
and the maintenance of posture. Compared to
vertical leg movements, the identified sensory
units were grouped into three main groups ac-
cording to the phase in which they were most
active (Figure 2D).
52% of the sensory profiles exhibited a tonic
discharge, whereas 43% were sensitive to upward
movements and 5% to downward movements of
the limb. Although all identified sensory neurons
exhibited a low discharge rate throughout the
locomotor cycle (Figure 2D, top), a finer analy-
sis revealed that they express distinct peaks of
activity and that movement coding may result
either in an increase or, surprisingly, a decrease
in their firing frequency. The former probably
characterizes phasic and phaso-tonic afferents
( Lev and Dep units in Figure 2D), whereas the
latter property was encountered in one third of
the tonic afferents ( Tonic unit in Figure 2D) in a
way that was not suggested by in vitro experi-
ments. The relative contributions of velocity and
position sensitivities are difficult to assess in vivo ,
but it appears that only a minor fraction (<14%)
of movement-sensitive neurons present purely
phasic properties.

Encoding Imposed Leg
Movement-Like Stimulation

Although stance control may be mimicked by a
'ramp and plateau' mechanical stimulation, the
smoothness of real leg movements during locomo-
tion makes the sensory activation correspond to a
sinewave-like repetition of stretch and release of
the CBCO. Thus, sinewave mechanical stimula-
tion over the angular sector covered by the joint
during actual walking has been applied to the
CBCO strand in vitro , while the activity of the
whole sensory nerve was recorded (Le Ray et al. ,
1997a). Surprisingly, only a half of the afferents
was activated ( i.e. , of the 40 sensory neurons, only
about 20 active neuron profiles were identified
using spike discrimination). In addition, within
the angular range of the imposed stimulation,
the number of activated release-sensitive af-
ferents was about twice that of stretch-sensitive
afferents.
More astonishingly, individual neurons that
specifically encode for a small angular sector
were never found, and no evident angular speci-
ficity emerged, but rather, afferents fired action
potentials over wide angular sectors (Figure 2B).
However, within these large angular ranges, peaks
of firing frequency are distinct among the sensory
neurons, with the main coding specificity of an
afferent resulting from the sum vector of all its ac-
tion potentials as determined by circular statistics,
as described in humans harm movement (Roll et
al. , 2000). Since 'ramp and plateau' stimulation
showed that every afferent codes for velocity,
differences in frequency peak suggested distinct
position sensitivities.