Logan:  Basic Principles of Learning

                  RESPONSE DIFFERENTIATION LEARNING

          If some dimension of the reinforcement is systematically 
          correlated with some dimension of the response, that
          dimension of the response will tend to become adjusted 
          so as to maximize the reinforcement received.

     Under typical conditions of reinforcement, the response and the
reinforcement are treated in all-or-none fashion.  That is to say,
either the response does or does not meet some very minimal criteria and 
the reinforcement is or is not given he exerted enough force on the bar 
to close and switching mechanism, and then we say that the rat was 
reinforced with a pallet of food.  In conditions of correlated
reinforcement we are interested in more than simply whether the rat 
pressed the bar; we are also interested in HOW he pressed the bar.

     As one particular example, we might arrange the conditions such 
that the harder the rat presses on the bar, the larger the reinforcement 
he is given.  We call this positively correlated reinforcement because an
increase in the response dimension is accompanied by an increase in the
reinforcement dimension.  Alternatively, we might give the rat more
reinforcement the softer he presses on the bar, in which case we would 
be using negatively correlated reinforcement.  In the former case, the 
rat will learn to press the bar very vigorously, and in the latter case, 
the rat will learn to press the bar very softly.  The effects of 
correlated reinforcement importantly enrich the Principles of Reinforcement 
by emphasizing the fact that both the qualitative and quantitative 
dimensions of a response are themselves learnable aspects of that response.

     In many situations, there is a natural correlation between the 
response and the reinforcement.  In a maze, for example, the faster
the rat runs, the sooner he gets to the goal to obtain the reward.  
If the rat is given more reinforcement for running slowly in such a
situation, he must fact the problem that, in order to obtain a large 
reward, he must delay getting it because the slow responding takes more
time.  

     Response differentiation learning may be thought of as a special case 
of stimulus discrimination where the relevant stimuli are interoceptive.  
All skeletal responses produce feedback stimuli, kinesthetic and 
proprioceptive cues from the muscles and joints conveying information
about their activity.  Insofar as a fast response "feels" different from
a slower response, the organism learns to discriminate between these
feedback stimuli and correspondingly to differentiate between these
responses.  Similarly, a forceful response feels different from a gentle
response, and a smooth response feels different from a jerky one.  It is
the extent to which we have learned to discriminate between the feedback
from the various ways of making a response that we can learn to
differentiate between them.

     There is an extremely important corollary to the Principle of 
Response Differentiation Learning: Not only CAN an organism learn
to respond in a particular way, but the organism DOES learn to respond in 
the particular way in which the response is practiced.  Extensive
experience responding slowly leads to well-learned slow responses; 
slovenly practice leads to learning slovenly performance.  There is,
however, the additional Principle of Response Generalization, which is 
that the organism can perform responses that are similar to the one
initially learned.  In the same sense that stimulus generalization means 
that learning is not confined entirely to the original stimulus,
response generalization means that learning is not confined entirely to 
the original response.  Both generalization principles imply a decrement 
in performance if the conditions are changed either to a different
stimulus situation or to a different response requirement.

     Response differentiation learning is the basis for all of our
skills.  Whether we are considering the knockout punch of a boxer or the
graceful moves of a ballet dancer, the finely-tuned response
differentiations are the result of extensive exposure to conditions
of correlated reinforcement in which the demands for excellence are
progressively increased.  What evidence there is suggests that high
standards beget outstanding performance provided only that the criteria
of success are within the reach of the person at this or her current
level of performance.  Virtuoso performance is achieved by progressively
raising the standards as improvement takes place.  An additional
practical rule of thumb is not to leave the standards at a relatively low
level for very long; the more the person practices less-than-perfect
performance, the more difficult it is to move on to higher levels of
achievement.  It is only perfect practice that makes perfect.

     The importance of response differentiation learning is not confined 
to highly skilled behaviors.  Recognizing that we learnresponses the way 
that we practice them, we must attend to the possibility that we have 
developed habits that are relatively inefficient.  Even our learning of 
the basic "three R's" may have resulted in our reading being slow, our 
writing being illegible, and our  rithmeticking' being hesitant and 
unreliable.  Although it would undoubtedly be preferable had we not 
developed such inefficient habits, our performance in these activities 
can be improved by further application of the Principle of Response
Differentiation learning.  A good place to begin is with your own name; 
be sure that you both speak and write your name clearly.

TERMS: Act, response differentiation, response generalization, correlated
reinforcement, response similarity