ÐÓ°ÉÔ­´´

(see Graphic)
(see Graphic)
Every year thousands of people suffer strokes which cause all manner of
disturbing problems for their unlucky victims, including impaired speech or
vision, paralysis and selective amnesia. But one of the most bizarre stroke
symptoms, is ‘alien hand’ a condition where one arm makes involuntary yet
purposeful movements. In one study of this strange condition, a woman stroke
victim grasped any object that touched her hand. She even had to use her
normal hand to prise open her clamped fingers. Luckily, such patients
gradually regain normal muscle control .

Alien hand seems to be caused by damage to a small area at the front of the
brain, suggesting that this region has a role in motor control. But what is
this role? The answer requires a new theory about how our nervous system
plans and executes movement, argues Gary Goldberg, a neuroscientist at the
Temple University School of Medicine in Philadelphia. A key element in his
theory is that each decision to move a part of your body is governed by two
distinct neural pathways in the forebrain. One deals with environmental
stimuli – things seen, touched or heard, and the other deals with internal
signals about the body’s needs. When a stroke damages the ‘internal’ path,
the ‘external’ path literally gets out of hand. The arm reaches out and
grabs objects regardless of whether the patient wants them.

Goldberg backs up this theory with evidence from two sources. One involved
studies of electrical recordings and brain lesions in monkeys. The other
experiments involved looking at brain images of human subjects who had been
asked to carry out a variety of motor tasks. Goldberg concluded from these
studies that internal signals activate neurons near the middle of the
brain’s cortex (the folded sheet of neural tissue that forms the two
hemispheres), while environmental signals activate areas at the sides of
the cortex. He believes that these two neural pathways act in part by
balancing each other, the medial pathway providing the internal information
that checks the lateral pathway’s responses to environmental signals.

Such ideas do not sit easily within mainstream thinking about motor control.
Nor is Goldberg alone in holding this view: over the past decade,
researchers studying animals ranging from snails to monkeys – including our
own work on insects – have obtained results that challenge the conventional
view. This view has held ever since the last century, when the British
neurologist John Hughlings Jackson made his pioneering studies of motor
control. Its supporters consider that movements are controlled by a
hierarchical ‘chain’ of motor centres which run from the brain down through
the spinal cord. Command signals are issued from the top, in this case from
the motor centres of the cerebral cortex. These commands are then received
by motor centres lower down in the brain stem, which in turn pass these
electrical signals to nerve circuits in the spinal cord. Here signals
activate preprogrammed instructions which tell muscles to perform particular
movements. The overall picture is one of the cortex dictating every movement
we make.

One reason that this model is generally accepted is that it fits comfortably
with the orthodox view that the cortex is the nervous system’s chief
decision-maker. But another reason lies with some influential research, done
in the 1960s, on motor control in lobsters and crayfish. Kees Wiersma at the
California Institute of Technology in Los Angeles and Donald Kennedy’s group
at Stanford University found that lobsters and crayfish would make complex
tail-flip and swimming movements if individual neurons connecting the brain
and the equivalent of the spinal cord were stimulated with an electrode.
These neurons became known as command fibres and their discovery reinforced
the hierarchical model of motor control.

Command fibres may seem to be a reasonable way for a nervous system to
control simple movements like swimming, or reflex movements that need to be
fast, such as those involved in escaping from threatening situations. But
can command fibres deal with more complex and varied movements as well?

In the early 1980s, we began to have our doubts. In insects, even routine
movements like walking seemed to require a much more flexible control
mechanism; one-off commands issued by individual neurons seemed hardly
adequate. So over the past decade, we have been studying motor control in
locusts, trying to understand how the nervous systems of these invertebrates
generate the instructions needed for flying and walking.

Locusts have many advantages over mammals for this kind of research. First,
the locust nervous system has fewer neurons, with one neuron often doing
the jobs done by tens, hundreds or even thousands in a primate. Secondly,
many of these neurons can be identified easily by their unique shapes and
electrical ‘signatures’. This meant we could make reliable ‘maps’ that
showed how the neurons link up – a crucial first step in working out how
the they collaborate to control a particular movement. Indeed the locust’s
neural signatures proved so distinctive that many researchers have
identified and studied equivalent neurons in different locusts, building up
an accurate picture of how they function – something that is still
impossible in mammals.

In insects the structures that correspond to the mammalian brain, brainstem
and spinal cord are the brain, the suboesophageal ganglion and the ganglia
of the ventral nerve cord (a ganglion is a body of interconnected neurons).
The structures are linked by pairs of nerve trunks called connectives. We
studied the movements involved in flight and walking because their
repetitive nature makes them easy to analyse. The distinctive patterns of
electrical impulses that regulate these movements are produced by pattern
generators – specialised circuits of neurons in the ventral cord ganglia.

Using microelectrodes, one of us (JK) eavesdropped on messages flowing down
the connectives that join the locust’s brain and suboesophageal ganglion to
the pattern generators in the ventral cord ganglia. It soon became clear
that the pattern of information flow was too complicated to be explained by
command fibres. Many neurons in the connectives were electrically active at
the same time, some even before walking began, while others became active
as the locusts started to move. So the locust’s brain did not appear to
issue motor commands one at a time. Rather, many messages flowed
simultaneously to the ventral cord ganglia. Moreover, the firing of
particular neurons coincided with particular movements or parts of
movements, such as the swing and stance phases of stepping, or with the
timing or direction of a movement. Each neuron in the nerve trunks seemed
to convey precise information about some aspect of the movement.

Meanwhile, researchers investigating motor control in mammals have found
that, contrary to the command fibre theory, each movement in a particular
direction is not controlled by one specific neuron. These researchers
examined two areas in the brains of primates, one involved in controlling
eye movements, and the other involved in planning arm movements. They
discovered that each neuron participates in coding for movements over a
range of directions, and that each direction is coded by a particular
combination of active neurons. Even with a small group of neurons the number
of potential combinations is huge, so relatively few neurons can code for
many directions. This ‘population coding’ uses neurons much more
economically and to greater effect than assigning each neuron to one
specific task. The neurons in the locust’s nerve trunks that fire during
walking probably represent information in a similar way.

The concept of population coding is now replacing the idea of single command
neurons in theories of motor control. But even so, motor systems could still
be hierarchical: a group of powerful individuals can determine policy just
as much as a dictator can. But the notion of hierarchy is severely shaken by
observations showing that the ventral cord ganglia do not slavishly follow
the dictates of the head ganglia – some decisons are made locally. The
locust provides a simple example: it will fly if there is a stream of air
over its head but only if its feet are free of contact. Messages from the
feet to the ventral cord ganglia signalling any kind of contact will negate
‘air stream’ signals coming from the head. This makes biological sense
because locusts spend a lot of time feeding on grass stalks and do not need
to take off in response to every little breeze.

Another major flaw in the hierarchical model of motor control is that it
focuses almost exclusively on information flowing from the head to the
spinal cord (or ventral ganglia in the case of locusts). In reality,
information must also flow back from the lower motor centres to keep the
higher centres informed of what is happening on the front line: every
movement an animal makes alters its relationship with its environment, and
the brain must be constantly updated about such alterations. So the ventral
cord ganglia of the locust don’t just receive messages from the head, they
also feed information back, partly through the movements they produce. In
effect, the locust’s motor control system works more like a loop than a
hierarchical chain.

In fact, the system seems to consist of several loops that work in parallel.
When we examined the anatomies and electrical properties of neurons that
fire during walking and flying, we found evidence of two main loops. One
runs between the brain and ventral cord ganglia and is involved mostly with
planning and preparing movements. The second links the suboesophageal
ganglion with the ventral cord ganglia and is more concerned with the
detailed execution of movements. Both loops consist of several subloops,
each of which looks after a particular aspect of motor control.

Similar loops are now being found in the motor systems of mammals. A common
approach involves training monkeys to reach for an object in response to a
stimulus. Researchers can then monitor the levels of neural activity in
different parts of the animals’ brains by implanting electrodes. In the late
1980s, such experiments showed that the motor cortex (the ‘highest’ motor
centre according to the hierarchical theory) does not send signals directly
to the brainstem and spinal cord. Instead, the information first circulates
in two loops, each passing through other parts of the brain before returning
to the cortex.

The functions of these two loops are still poorly understood. But one
possibility is that they enable the brain to continually update and modify
the signals the cortex sends to the spinal cord. Further hints come from the
brain structures through which the signals pass. One loop includes the
cerebellum, a structure that is probably involved in timing and fine-tuning
movements. It was discovered when researchers were examining pathways
between the cortex and cerebellum. The second loop, described in 1990 by
researchers at the Johns Hopkins University Medical School, involves the
basal ganglia, a part of the forebrain whose function may be to sequence
movements and control coordination. This loop seems to be divided into
several subloops, each with a slightly different function.

Reaching conclusions

That motor systems use many subloops to control specific aspects of
movement is perhaps only to be expected, given the complexity of behaviours
such as walking and reaching. They require planning and preparation,
starting and stopping, and the shutting down of incompatible actions – you
can’t hop and walk at the same time. To accomplish all this, the various
loops appear to communicate with each other through nerve connections. It is
probable that whatever is decided in one loop modifies the outputs of the
others, and vice versa. This arrangement is reminiscent of the way parallel
computers work, where there is no single command centre and outputs emerge
through the interactions between the parallel circuits.

But in contrast to parallel computers, where all the processors are usually
equivalent, the various motor centres in insects and mammals have
specialised functions with different levels of complexity. Neurons in the
motor centres of the brain have more complex functions than those in the
spinal cord or the ventral cord ganglia. In our locust studies, we gained
some powerful insights into these different levels of complexity and
specialisation by correlating the activities of different neurons with
movements recorded on video.

In walking locusts, some neurons fire steadily as the animal moves. These
so-called ‘tonic’ neurons may help to keep the movement going. Then there
are neurons whose firing is related to the function of the movement; some
fire when the locust is moving forward, for example, others as it turns in a
particular direction. These neurons are found mostly in the brain and
suboesophageal ganglion. Neurons of a third type – found mainly in the
ventral cord ganglia – deal with the details of the movement. Each fires
only when a particular set of muscles is active.

Similar types of neuron activity can be detected in the brains of mammals;
when a monkey reaches for a banana, for example, or a cat starts walking. In
the monkey, the planning phase – when the brain calculates the direction
the arm must move to reach the banana – seems to take place in parts of the
cortex known as the supplementary motor area and the premotor cortex. The
premotor cortex also deals with making sure things happen in the right
sequence and direction, like the second type of neurons in locusts. Neurons
in a third area, the primary motor cortex, specify which groups of muscles
need to contract, and when. These are like the third type of locust neuron.

Viewing the motor control system as a parallel distributed system, rather
than a command hierarchy, brings experimenters new challenges. Most
conventional experiments on motor control monitor just a few neurons at a
time. This can give only a limited view if, as we believe, many neurons
throughout the loops are active simultaneously. Roy John and colleagues at
the New York Medical Center, using a metabolic method for identifying active
neurons, have estimated that one hundred million neurons are fired when a
cat makes a learned paw movement. Researchers are now developing techniques
to tackle this problem. One method uses dyes that change colour when the
neurons are electrically active, and is particularly powerful for studying
the small nervous systems of invertebrates.

At the other end of the scale, much useful information about motor control
in humans is coming from experiments using brain scanners, especially PET.
For example, Richard Frackowiak and his colleagues at the Hammersmith
Hospital in London are using PET to identify which areas of the brain become
active when people do simple tasks, involving eye-hand coordination. In one
study, they compared people who had suffered stroke damage to their motor
cortex with healthy people. Unexpectedly, they found that after a stroke,
areas that would normally be involved only in complex tasks tended to become
active during simple movements (see ‘Rescuing damaged minds from decay’,
New ÐÓ°ÉÔ­´´ supplement, 14 November 1992). These results show vividly the
benefits of a distributed motor control system. Since no function is rigidly
tied to a specific area or group of neurons, other parts of the system can
take over if the need arises.

Most of our results have suggested that the lower motor centres of the
locust, far from being passive servants of a command centre, are actively
involved in planning movements. And the same story is emerging from these
studies of humans and other animals: all suggest that when it comes to
controlling movement, nervous systems behave more like democracies than
dictatorships. Decisions about what to do and how to do it are distributed
throughout a network of neural centres rather than coming from a central
command centre.

In future, insights may also come from mathematical models and computer
simulations. With colleagues in the Institute of Applied Physics at the
University of Regensburg, we have set up an artificial neural network that
behaves like a motor control system. For instance it spontaneously changes
from one output pattern, equivalent to a movement, to another. These could
suggest how a complex control system might operate, pointing the way to
experiments to test if it really behaves that way.

Jennifer Altman is a freelance consultant and science writer specialising in
the neurosciences. She has collaborated for many years in research on
locusts with Jenny Kien, who is a lecturer in the Zoological Institute at
the University of Regensburg, Germany.

* * *

An eye for the action

Studies of eye movements in primates have provided key clues to motor
control. Each eye is moved by only six muscles and makes two basic types of
movements – saccades, to stabilise the eye when the world moves relative to
the head, and tracking, to follow a moving object. What’s more, eye
movements are controlled by a select group of neurons in an area of the
brain known as the superior colliculus.

Each of these neurons fires a series of impulses when the eye moves through
a certain angle. But the neurons are not individually tuned to specific
small angles. Instead each has a broad tuning curve, firing fastest when the
movement is through a particular angle and less enthusiastically at a range
of angles on either side. In 1988, David Sparks’s team, at the University of
Alabama in Birmingham, discovered that any one saccade is coded for by a
population of neurons. In the population, some neurons fire faster, some
slower, according to how close their preferred angles are to the desired
direction. The movement is controlled by the sum of these neurons’
activities. Conversely, a slightly different population of neurons codes for
each direction and any one neuron will participate in coding for movements
over a wide angle. Population coding of direction has also been demonstrated
for arm-reaching movements by Apostolos Georgopoulos at Johns Hopkins
University Medical School.