Chapter 2. Prehension

Chapter 5 - Movement Before Contact 201
of the movement. It is suggested here that orienting and positioning
the palm seems to be crucial to prehension. From the reaching
perspective, the arm muscles get the palm into a ballpark of its final
location. From a grasping perspective, the fingers can preshape into a
suitable posture, based on the assumption that the palm will be aligned
correctly. Fine tuning can then occur during the second phase of the
movement, where the orientation of the palm is fine-tuned by wrist,
elbow, and shoulder muscles, while the fingers close in around the
object. Feedback mechanisms can then overcome errors in perception
during this second phase of the movement.
The level of the motor command is uncertain. Models have sug-
gested how it occurs at the joint angle level, joint torque level, and
muscle level. When starting from a goal location specified in a world
or body coordinate frame, an important computation is to transform
that goal into one of these motor levels. Computations may proceed in
an orderly fashion (Step 1 --> Step 2 --> Step 3) as Uno et al. (1989)
outlined. From a biological perspective, this is inefficient. Alternatives
are to compute motor commands more directly (Steps 4 or 5) or also to
plan in joint space (Step 1’). Whether the computation is inverse
kinematics or inverse dynamics, it is an ill-posed problem. Cost
functions and/or constraint satisfaction networks suggest ways to limit
the computation towards selecting a unique solution. Of course, the
CNS never seems to have a unique solution, evidenced by the issue of
motor equivalence and also of the exhibited variability. Thinking about
the sensory side as a distributed representation (e.g., combining two
dimensional with three dimensional information and perhaps even
velocity cues, etc.) may shed light on the motor side: it too could be a
distributed computation involving many levels of sensory
reprentations and motor commands. Arbib’s coordinated control
program is a first step in this direction.
The CNS allows parallel and redundant processing of sensory in-
formation for motor control. While sensory information (given a
modality and sub-modality as well) is needed at crucial times for
completion of a motor task, there are many ways that it can solve
problems. In the 1990s, it is still not possible to distinguish the kine-
matic features in reaching and grasping movements attibutable to
supraspinal planning processes, spinal circuitry or the inevitable con-
sequence of musculoskeletal mechanics. The CNS seems to find
simple solutions out of its multi-faceted repertoire of possible modes
of interacting with the environment.