Logan:  Basic Principles of Learning

                   PRINCIPLE: DIFFERENTIAL CONDITIONING

          If two discriminable stimuli are presented in an irregular 
          order, and the schedule/condition of reinforcement differs 
          depending on which stimulus is presented, the organism will 
          come to respond differentially in their presence more-or-less 
          in accord with the prevailing schedule/condition.

     The Principle of Differential Conditioning applies to all of the
basic conditioning paradigms, classical, operant, and instrumental.  In most of 
the laboratory studies of differential conditioning, one of the stimuli
is associated with nonreinforcement (extinction) and hence the organism learns 
to respond in the presence of the reinforced stimulus (S+) and not in the
presence of the nonreinforced stimulus (S-).  For this reason,
differential conditioning is sometimes called a "go-no-go" discrimination.

     The most obvious factor affecting the rapidity of differential 
conditioning is the similarity between the stimuli.  As we would expect 
from the Principle of Stimulus Generalization, the greater the similarity
between the stimuli, the greater will the response tendency that is 
accruing to S+ generalize to S- and hence the greater the difficulty in
suppressing responding to S-.  In the limiting case, where the difference
between S+ and S- is not discriminable, the condition is effectively one 
of partial reinforcement.

     In the more general case, the schedule/condition of reinforcement 
for S- is simply of lesser value to the organism.  For example, S+ may 
signal a larger reinforcer than S-, in which case differential performance 
also is obtained.  This situation, however, introduces a new principle
(contrast) that is described separately.  

     As one example of differential conditioning in the laboratory, a rat 
might be rewarded for running down a white alley and not rewarded for 
running down a grey alley.  In this case, differential conditioning would 
be reflected in how rapidly the rat runs in the two alleys and the darker 
the grey alley, the more rapidly the rat learns not to run it.
Presumably, differential conditioning invokes both the relevant principles of 
conditioning and the Principle of Experimental Extinction combined with 
the Principle or Stimulus Generalization of both processes.  However, a 
complete understanding of differential conditioning proves to be more 
complex than that.

     For example, it is known that prior nondifferential reinforcement of 
two stimuli retards subsequent differential conditioning.  Now it is clear 
that such pretraining would establish a strong response tendency directly 
to the eventual S-, but that should extinguish readily enough once 
differential reinforcement is instituted.  Among the proposed accounts of 
this finding is ACQUIRED EQUIVALENCE OF CUES.  The notion is that
nondifferential reinforcement effectively makes the stimuli more similar.
Another way of saying this is that the organism learns during pretraining 
that the difference between the stimuli is unimportant and subsequently 
fails to attend to that difference even when it is made relevant.  
Consistent with this line of reasoning is the easy-to-hard effect.  
Differential conditioning between two highly similar stimuli can be 
facilitated by initial training with stimuli differing more widely along 
the same stimulus continuum.  Apparently initial training on an easy 
problem helps focus attention on the relevant stimulus dimension.

     The inverse phenomenon, ACQUIRED DISTINCTIVENESS OF CUES, is a 
by-product of differential conditioning.  After two similar stimuli have 
been differentially reinforced, the amount of generalization between them 
is decreased.  Furthermore, subsequent discriminations between the
stimuli involving other responses will also be facilitated.  The best examples 
with humans involve verbal labels. After having first learned to identify
similar stimuli, such as Arabian and Thoroughbred horses, with different 
labels, there will be less generalization between them in the future.

     Differential conditioning begins early in life and corrects our 
natural over-generalizations.  A young infant may call all men "Daddy" at 
first, but soon learns not to do this because the response to incorrect 
visual stimuli is nonreinforced.  More generally, all of our concepts are
acquired in this way; we were reinforced for saying "dog" only in the 
presence of appropriate stimuli.  One of the difficulties in overcoming 
racial prejudices once they have been acquired is that the person has 
been nondifferentially reinforced for responding equally to
distinguishable members of the same race, at least to the extent of
applying the same label to them, and hence subsequently finds it
difficult to learn to respond to them as individuals.

TERMS: Differential conditioning, contrast (behavioral, incentive,
simultaneous), cue, easy-to-hard effect, post-discrimination gradient.