Go
back to Behaviorism Articles & Chapters
In
Search of Cumulative-Hierarchical Learning
Richard W. Malott1
Behavior Analysis Program
Department of Psychology
Western Michigan University
Download
Word version of this article
Yes, we can say, CRF is really FR 1 (we can say, continuous reinforcement
is really fixed ratio reinforcement where the ratio of response per
reinforcer is 1 to 1). But if such labeling is not a reduction
to absurdity, it’s at least a reduction to triviality. Continuous
reinforcement shares none of the properties of fixed ratio reinforcement,
such as the pause after reinforcement followed by the rapid acceleration
of responding up to a hell-bent-for-leather, maximum rate. Continuous
reinforcement does not capture the spirit, the essence, of fixed-ratio
reinforcement.
Yes, we can say, a pigeon’s food reinforced key peck is
really a “mand” for food. But such labeling trivializes
the concept of “verbal behavior.” The reinforced key peck,
that simple, ubiquitous form of the operant shares none of the rich
properties of verbal behavior, the properties that make verbal behavior
so unique—verbal behavior’s participation in Hockett’s
linguistic productivity, a participation that allows the verbal organism
to make appropriate novel responses to novel stimuli, verbal behavior’s
grammatical nature. And, it is a short distance from saying, the
pigeon’s reinforced key peck is a mand, to saying, the
non-verbal child’s head banging is her effort to COMMUNICATE
her needs and wants to you, to communicate that she wants you to pay
more attention to her and to stop making such difficult demands of
her. It is with good reason that we classify pigeons as non-verbal
organisms.
And, maybe we can say, shaping, chaining, and mediation are really
examples of cumulative-hierarchical learning, but none of those
three behavioral processes capture the essence, the importance, of
cumulative-hierarchical learning. I think equating them trivializes
the important, but complex concept cumulative-hierarchical learning.
Why?
The first step to answering this question, was a review of some of
Staats (1977) earlier writing on what he then simply called hierarchical
learning: “. . . behavioral skills a child should have
before he should attempt a new learning task” (p.287). “.
. . the child in many instances must acquire one repertoire of skills—if
not several—before he is able to move on to the learning of
a more advanced repertoire” (p. 287). “The child learns
skills that determine how rapidly he will learn new skills”
(p. 287). “The child is involved in a progression of learning
that moves from the acquisition of basic repertoires of skills to
the acquisition of more advanced skills based upon the earlier learning.”
(p. 288). And I also reviewed Hixson’s (200?) quotation from
(Rosales-Ruiz and Baer, 1997, p. 533): “A behavioral cusp…is
any behavior change that brings the organism’s behavior into
contact with new contingencies that have even more far-reaching consequences.”
The next step was to use these quotes to construct a definition of cumulative-hierarchical learning: the learning of a skill or value
that affects learning in a wide variety of other contexts (by value I mean reinforcers and aversive conditions).
And it is this wider variety of contexts that makes cumulative-hierarchical
learning such an exciting, theoretically profound, and utilitarian
concept. The wider variety of contexts is Staats’ new
learning task, more advanced repertoire, new skills, and more
advanced skills; and more clearly, it is Rosales-Ruiz and Baer’s
more far-reaching consequences.
For example, consider the autistic child’s behavior of looking
at the trainer. This may be merely part of a simple behavioral chain:
SD1 (the trainer says, “look at me,”)
R1 (the child looks at the trainer)
That looking response always produces the same reinforcer/discriminative
stimulus.
Sr/D2 (the trainer says, “do this”; and then touches his
own nose)
This second SD is always the occasion for reinforcement of the same
response (the child’s touching her nose).
In this case, to call looking at or attending to the trainer an example of cumulative-hierarchical learning seems to trivialize
the concept.
But, the child’s response of looking at the trainer could be a behavioral cusp in an example of cumulative-hierarchical learning:
Suppose the skill of looking at the trainer affects the child’s
performance or learning in a wider variety of contexts, such as imitating
eye touching, hand clapping, toy playing, shoe tying, going to the
lunch room, playing the piano, and drawing. In this case, learning
the single behavioral cusp of looking at the trainer is clearly an
example of cumulative-hierarchical learning because it allow the child
to learn new, more advanced skills (e.g., shoe tying); the skill of
looking at the trainer participates in a wider context than the simple
nose-touching chain; the looking-at-the-trainer cusp has far-reaching
consequences.
And the look-at-the-trainer cusp participates in even more cumulative-hierarchical
learning when it produces SDs for a wider range of instructions than,
“do this” (e.g., “put same with same,” “wash
your hands,” and “put on your coat, so we can go to the
playground”).
And, just as we trivialize cumulative-hierarchical learning by offering
simple behavioral chaining as an example, I think we add to the trivialization
by offering shaping and even simple response mediation as examples.
Consider response mediation. A mediating response is a response
that is correlated with an SD, with the result that the stimuli produced
by that response will acquire the evocative function of the SD and
evoke the response in the absence of that SD. Mediating responses
may be learned in one situation (the presence of the original SDs).
And they may result in the attainment of otherwise missed reinforcers
in another situation (when the original SD is no longer present).
But it is not clear to me that these mediating responses always participate
in cumulative-hierarchical learning.
As an example, the Hixson cited work by, Lowenkron (1998) on delayed
matching to sample, where he established a set of responses that mediated
between a corresponding set of sample stimuli and the correct selection
of delayed comparison stimuli. This training in the use of these mediating
responses facilitated the use of a novel set of mediating responses
to mediate between a corresponding set of novel sample stimuli and
the correct selection of delayed comparison stimuli. That the earlier
mediation training facilitated or even eliminated the need for later
mediation training is an excellent example of cumulative-hierarchical
learning. But, as Hixson implied, by itself, the use of the original
mediating responses with the original delayed-matching-to-sample stimuli
is not an example of cumulative-hierarchical learning; we have cumulative-hierarchical
learning only when the subsequent set of mediating responses facilitated
performance with a novel set of delayed-matching-to-sample stimuli.
The learning is hierarchical in that learning the response-mediation
components facilitated learning the original delayed matching performance.
And it is cumulative in that learning the response-mediation process
for one set of delayed identity-matching stimuli, facilitated the
learning of a novel set.
Though cumulative-hierarchical learning may involve something like
chaining, shaping, or response mediation, not all chaining, shaping,
and response mediation are examples of cumulative-hierarchical learning.
References
Hixson, M. D. (200?) Response Mediation and Cumulative-Hierarchical
Learning. The Analysis of Verbal Behavior, ?, ??
Hockett, C.F. (1960). Logical considerations in the study of animal
communications. Animal Sounds and Communications, 7, 392-429
.
Lowenkron, B. (1998). Generalization of delayed identity matching
in retarded children. Journal of the Experimental Analysis of Behavior,
50, 163-172.
Rosales-Ruiz, J. and Baer, D.M. (1997). Behavioral cusps: A developmental
and pragmatic concept for behavior analysis. Journal of Applied Behavior
Analysis, 30, 533-544.
Staats, A.W. (1977). Child learning, intelligence, and personality:
Principles of a behavioral interaction approach (revised edition).
Kalamazoo: Behaviordelia.