Logan:  Basic Principles of Learning

                  PRINCIPLE: PRF EXTINCTION EFFECT (PREE)
 
         A response that has received partial reinforcement during
         acquisition will persist longer during extinction than
         a continuously reinforced response.

     According to the Principle of Partial Reinforcement, it is not
necessary that reinforcement occur on every trial in classical or
instrumental conditioning or for every response in operant conditioning.
In operant/instrumental conditioning contexts, it is usually important to
reinforce behavior rather generously during the early stages of
acquisition simply to insure that the organism maintains a sufficient
level of performance that occasional reinforcements will be received.
From that point, it is possible to "wean lean;" that is, one can
progressively reduce the percent reinforcement even to the point where,
on the average the organism is not receiving enough nutritive
value from the reinforcer to replace the energy expended to receive it.

     This fact actually suggests a partial reinforcement extinction
effect.  That is to say, for a response to be maintained under PRF, it
must have enough persistence to carry through the runs of nonreinforced
trials that are encountered in the schedule itself.  Although the PREE
may not be surprising when viewed in this light, it has historically been
a source of great interest.  The reasoning was this: Reinforcement is
necessary to increase response strength and hence partially reinforced
responses must be weaker.  Weaker responses should, indeed, be weaker in
all regards, including their persistence, yet partially reinforced
responses persist longer in extinction.  Many theorists have puzzled over
this apparent paradox.

     Although a great many hypotheses have been advanced in an attempt to
explain the PREE, two have attracted the greatest attention.  The
aftereffects hypothesis requires recognition of the fact that a
reinforcer can serve as a stimulus for future behavior and that
nonreinforcement also has stimulus properties.  This can be seen in the
case of temporal conditioning and, in other studies, rats have been
trained to turn right if given food.  In PRF, the reinforcement or
nonreinforcement on the immediately preceding trials are a part of the
total stimulus complex in which behavior occurs.  The organism is
therefore reinforced for responding even after one ore more preceding
nonreinforced trials but organisms given CRF have no such training.  The
PREE is therefore not so much that PRF increases resistance to extinction
as that CRF leads to rapid extinction because the second trial of
extinction is the first time CRF subjects have ever been exposed to the
situation after a nonreinforcement.  Consistent with this reasoning, the
longer the run of nonreinforced trials given during PRF acquisition, the
greater the resistance to extinction.  To state this basic reasoning in
cognitive terms, it is easy for CRF subjects to know when extinction has
begun whereas, for PRF subjects, the beginning of extinction could be
just another run of nonreinforced trials yet to be followed ultimately by
reinforcement.

     The other hypothesis involves the notion of frustration.  The
frustration hypothesis has many of the same features as the aftereffects
hypothesis and is not incompatible with it.  But the contentions is that
nonreinforcement is more than simply a stimulus that persists on to later
trials.  Nonreinforcement when reinforcement is expected is frustrating
and frustration becomes conditioned to the situation.  CRF subjects are
frustrated for the first time with the onset of extinction, and the
natural response to anticipatory frustration is to quit.  Because of the
drive properties of frustration, CRF subjects quit vigorously.  In
contrast, PRF subjects have learned to continue to perform the observed
response vigorously in spite of anticipating frustration for their efforts.

     Of special practical significance is the fact that the PREE
generalizes from one situation to another.  People who have been
reinforced for persisting in their efforts to obtain one goal develop
such habits as a personality trait; others have learned to quit at the
first sign of difficulty.  Whether dogged persistence or hasty retreat is
more adoptive depends on the situation and we can never really know in
advance.  However, many gambling addicts are probably in a condition of
extinction after a lean PRF schedule.

TERMS: Conditioning (classical, instrumental, operant, temporal), effect
(PREE), frustration, hypothesis (aftereffects, discrimination,
frustration), partial reinforcement.