Chapter 2. Prehension

Chapter 4 - Planning of Prehension 69
goal in terms of sensory consea~ences~, an argument made by
Schmidt (1975) and Cole and Abbs (1987). Then, mismatches between
the actual outcome and the anticipated sensory consequences
would cause reprogramming. But the error processing can also be incorporated
into the plan, as modelled with TOTES (Miller, Galanter &
Pribram, 1960). A TOTE (Test-Operate-Test-Exit) is a hierarchically
organized processing model, as seen in Figure 4.2. In order to hammer
a nail, if the hammer is up, the hand will bring it down; if down,
the hand will bring it up. The ‘Test-Nail and Hammer’ TOTE hierarchically
calls the ‘Test-Hammer and Lift’ TOTE and the ‘Test-Hammer
and Strike’ TOTE. These, in turn, would call other, lower-level
TOTES as necessary. The actual representation of the goal state for
each test box could be in terms of sensory consequences (or perhaps
in terms of the sensorimotor contingencies seen by a feedback controller).
In the Arbib CCP, motor schemas could be performed by
TOTES.
One reason to hide the details of the schemas is that numerous
spatial-temporal patterns of action can produce the same results. This
fact of movement behavior, called motor eauivalence, has to do with
the nonspecificity of motor commands. Muscles can vary their functional
roles from movement to movement. As well, there is a great
deal of variability in movements, because two similar movements are
never exactly alike. However, these variations occur within a restricted
bandwidth. A hierarchical model (recall the ballpark model of
Greene, 1972) built from a high level plan that approximates a movement
goal and then is refined at lower levels explains movement
anomalies such as motor equivalence and variability. In terms of motor
equivalence, lower level subsystems are interchangeable. In terms
of variability, a plan is implemented through one to many transitions.
A hierarchical system reduces control complexity.
Motor equivalence and variability in repeated grasping instances
raise a question about the components that make up a plan: if there is
more than one way to perform something, what is invariant in the
plan? In the Arbib CCP, there is a distinction between the motion of
the arm (left side of Figure 3.3) and the motion of the hand and wrist
3Sensory consequences are distinct from the sensory motor continpencies required
during movement execution. Unfolding of the program for movement includes
descending commands, and also contingencies for sensory information in
anticipatory feedforward control. This would allow for adjustments to unexpected
perturbations or the requirements of subsequent phases. An example of this will be
seen in Chapter 5.