Session Two: Approaches to the study of learning / Radical behaviorism

Instructor notes written by J. David Perry, Ph.D., Indiana University.

Instructor notes: Approaches to the study of learning
Instructor notes: Radical behaviorism
Learning activities

Instructor notes: Approaches to the study of learning

Some fundamental questions about learning

There are no simple answers to these questions. In this course, we'll examine several different theories or research approaches that have attempted to address these questions. We're not looking for the "best" one. If the experts agreed on which was best, we would just study that one.

The approach we'll take is that each of these theories illuminates a different aspect of the teaching/learning process, and each may be useful in understanding a particular situation.

Assumptions and contextual issues for P540

No single learning theory is adequate to account for all aspects of learning. In fact, it is quite possible that learning is not a single entity at all, but an assortment of phenomena that we lump together. (Humans are good at inventing categories.)

A theory of learning does not automatically prescribe the best way to teach. As we'll see throughout this course, instructional principles are not always easily derived from learning theories. Some learning theories--behaviorism, for example--have had elaborate instructional theories developed in their wake; others have had relatively little direct impact on teaching. One of our main tasks in this course will be to see whether we can, collectively or individually, determine useful educational applications for the theoretical approaches we examine.

What people learn through formal instruction is only a very small subset of what they know. It's important to remember that, while our concern in this course is primarily on learning theories as they apply to instructional settings, most learning takes place in the "natural" world.

Some important definitions:

Cognition: Refers to all the processes by which sensory input is transformed, reduced, elaborated, stored, recovered, and used; includes such hypothetical stages or aspects as sensation, perception, imagery, retention, recall, problem-solving, and thinking. (I.e., how we receive information from the outside world through our senses; how we organize it and use it.)

Theory: A set of interrelated concepts or constructs, definitions, and propositions that present a systematic view of phenomena for the purpose of explaining, predicting, and controlling the phenomena. (For example, a set of concepts such as attention, perception, memory, motivation, development, and how they interact with one another, allowing us to understand and predict learning or behavior.)

Learning theory: A set of constructs linking observed changes in performance with what is thought to bring about those changes.

About theories:

Questions lead researchers to conduct systematic observations, on the basis of which plausible answers can be constructed. What sort of observations would you conduct, or what kind of data might you collect for each of the three questions above?

Systematic observations lead to theories, with the purpose of explaining, predicting, and controlling. My theory might be that there's a little man inside the coke machine, and if he likes the way you look, he accepts your dollar, and if not he rejects it. (I didn't say it was a good theory.) Or, my theory is that taking notes increases learning even if you never look at the notes again, because it contributes to actively organizing the material in your long-term memory.

Theories don't give us "the truth of the matter," only a conceptual framework for making sense of the data collected so far. This may seem like an odd statement. However, time and again, in both the natural and social sciences, a theory widely taken to be "true" is eventually replaced with another one judged to have greater explanatory power.

A particular theory stems from a particular perspective: thus, theories carry "worldviews." For example, a theory that says we can explain and control all human behavior in terms of stimuli and responses obviously embodies a very different view of the world than one that says we construct knowledge through negotiation and consensus with others.

Different disciplines approach phenomena with different assumptions and beliefs. For example, an anthropologist and a psychologist would likely take different approaches, even if apparently studying the same thing.

Two apparently competing theories may not even be directed at the same phenomena. What's important to one theory may not even be noticed in another. In this course, one of our important tasks will be to understand what aspects of learning are addressed by each theoretical approach.

Although theories are not built in a precise or predictable way, there are generally stages that can be discerned in the theory-building process. (See Driscoll, page 7.)

• What kinds of assumptions and beliefs will you bring to the question?
• What specific questions would you start with?
• What sort of observations or data collection would you use?
• How would the results of your data collection help you in the next step of building your "theory"?

To actually begin such a study, assuming I had access to a very cooperative class, I might ask to randomly copy notes from a different group of students at each class session, so that I had samples from all students at various points throughout the semester. I would then find a way to categorize the notes according to how well they were organized. I would also have students complete a survey to tell me how they used their notes outside of class. At the end of the course, I would look to see if there was any relationship between how notes were organized, how students used them, and their performance in the course. If I believed that I did see such a relationship, I would have a more specific question to start with for my next study.

Classifying learning theories

In this course, we will also use a slightly less esoteric way of organizing learning theories. The approaches we will discuss can be classified as primarily concerned with behavior (without regard for mental processes), cognition (including perception, memory, and problem-solving), motivation (how motivation and beliefs affect learning), cognitive development (how cognitive ability changes with maturation or experience), and instruction (how one can deliberately influence learning).

Back to Top

Instructor notes: Radical behaviorism

From the 1950s through the 1970s radical behaviorism, as espoused by B. F. Skinner, dominated theories of learning and instruction. Since that time, behaviorism has been replaced almost entirely by cognitive and other approaches. Almost, but not quite entirely. It is still possible to see vestiges of the behaviorist approach to instruction in some forms of computer-based learning and occasionally elsewhere. The greatest stronghold of behaviorism in education is probably in classroom management. Many classroom teachers are familiar with using reward systems, time-out, and other behaviorist approaches to encourage students to pay attention, stay in their seats, and so on. Although this is a use of behaviorist principles in an educational setting, it is not really the same as using behaviorism as the basis for learning or instructional theory. Although I believe that behaviorism should be considered a historical, rather than contemporary, approach to understanding learning, it is still worth knowing about because of its recent, and to some degree continuing, influence.

Important ideas leading up to Skinner

• Associationism, and Ebbinghaus's experiments with memorizing lists of nonsense syllables. Why did Ebbinghaus conduct his learning experiments with nonsense syllables? And, in retrospect, why was that perhaps not such a great idea?
• Thorndike's interest in the association between the environment and action (or behavior), rather than just the association between ideas.
• Pavlov's "discovery" of  classical conditioning, where a previously "neutral" event (e.g., ringing a bell), when paired with a biological stimulus (e.g., food) can come to elicit the same physiological response (e.g., salivation).

Some concepts associated with the classical conditioning paradigm:

• Higher-order conditioning: when a conditioned stimulus is paired with another previously neutral stimulus, it too can acquire the ability to elicit a response. So, if Pavlov's dogs were conditioned to salivate when he rang a bell, then he began flashing a bright light at the same time, eventually the light alone might come to elicit the same response.
• Extinction: when the conditioned stimulus is presented over a sufficient period of time without the unconditioned stimulus, the conditioned stimulus eventually loses its ability to elicit the response. (I.e., ring that bell enough times without presenting food, and I'll stop salivating.)
• Counter-conditioning: Changing the response to a conditioned stimulus by pairing it with a different stimulus. In the case of Baby Albert, you might pair a pleasant stimulus with the white rat, just in case extinction alone doesn't work. Or, in the case of Pavlov's dogs, you might pair an appetite-suppressing stimulus with the conditioned stimulus (if you could think of something that suppresses a dog's appetite!).

Skinner's operant conditioning

Watson (1913) introduced the notion of behaviorism: that psychology should be concerned only with the objective data of behavior, since we can't really know what's going on in the mind. B. F. Skinner followed Watson's lead in emphasizing behavior, rather than thoughts, feelings, attitudes, etc., as the "proper" subject matter of psychology. The interesting difference in Skinner's view is that he was less concerned about the relationship between environmental stimuli and responses; rather, he was mainly interested in behavior (or a response) and its consequences. Attempts by psychologists to link all behaviors to stimuli from the environment had become very cumbersome, and required postulating a lot of intervening mental processes. Just imagine starting only with stimuli that elicit physiological responses (heat, loud noises, hunger, etc.), and trying to account for complex human behavior like building a cathedral or writing a poem!

Skinner said that the environment provides us with "cues" that serve as antecedents for behavior; i.e., they set the conditions for it to occur. But it's the results or consequences of our behavior which make that behavior more or less likely to occur in the future, and so it was these results that were more interesting to him.

Thus, Skinner distinguished two classes of behavior:  respondent and operant. Respondent behavior is the kind we perform automatically in the presence of some stimulus from the environment. Operant behavior is "emitted" without necessarily following any particular stimulus. For example, babies grasp at things and babble, just because that's what babies do. To Skinner, it would be the consequences of their grasping or babbling behavior that are important.

According to Skinner, the basic "unit" of behavior analysis is a discriminative stimulus, followed by an operant response, followed by a contingent stimulus. This is commonly diagrammed something like this:

S (discrim) --> R --> S (conting)

Remember that the discriminative stimulus sets the stage for the response, but does not automatically elicit it in the way that an object moving toward your eye can elicit a blink. For example, robins peck for worms when they're on the ground, but not when they're flying. So, the ground is a discriminative stimulus for pecking behavior. The pecking behavior (the R in our diagram) may be followed by getting a worm. If so, the worm is a new stimulus, which is contingent on the pecking behavior, and which may serve to reinforce that behavior.

Humans, too, are very sensitive to discriminative stimuli. We know that very different sets of behaviors are appropriate for different settings. For example, the behaviors that are appropriate at a football game are quite different from those that would be appropriate at an opera. While Skinner had some interest in the role of discriminative stimuli in setting the stage for behavior, he was more interested in the consequences of behavior; that is, in the response-contingent stimulus that follows a response.

Consequences of a response

Skinner noted that the consequences of a response (the contingent stimulus) can be either satisfying or unsatisfying to the person or organism making the response. When the consequence is satisfying, it is said to reinforce the behavior; that is, to make it more likely to occur in the future. When the response is not satisfying, or aversive, it is said to punish the behavior; that is, to make it less likely to occur in the future. It's important to remember that we can't always guess what consequence will be satisfying to someone. The final test of whether a consequence is reinforcing to someone is whether or not it increases the behavior of interest. Suppose, for example, that we try to get a child to go to bed on time by reading to her for half an hour each night. No matter how much we believe that the child enjoys being read to, if this does not increase her going-to-bed-on-time behavior, then, in Skinnerian terms, we must conclude that this was not a sufficient reinforcer for her.

Operant conditioning would be a fairly simple concept if it were concerned only with positive and negative consequences of a behavior or response. But it's a little more complicated than that. Skinner reasoned that we could think not only about situations in which a satisfying or aversive outcome was presented following a response, but also about situations in which it was removed following the response.

Suppose, for example, that every time your dog sniffs around on top of the kitchen stove, he finds a chicken leg or something equally tasty to eat. It is likely that his stove-sniffing behavior would increase, because it has been reinforced by finding food there. Now suppose that you get tired of this and you stop leaving food on top of the stove where the dog can reach it. This is a situation where a satisfying outcome is removed, rather than presented, following the response. And the likely consequence (you hope) is that the stove-sniffing behavior would decrease, through the process of extinction.

Here's another example. Suppose every time a student asks a question in class, the teacher responds in a curt or abrupt way. This may function as a punishing consequence which decreases the question-asking behavior. Then, a new teacher who is more open to questions takes over the class. The punishing or aversive consequence has been removed, so the student may increase the question-asking behavior.

If we "cross" the two possible functions of a contingent stimulus (satisfying or aversive) with the two possible conditions of presenting or removing the stimulus after the response, then we get the table with four cells that you see on page 39 of the text.

To analyze behavior a la Skinner, it is important to understand these four possible situations. In particular, note that both positive reinforcement and negative reinforcement (the upper left and lower right cells in Driscoll's table) serve to increase or strengthen behavior. Positive reinforcement does so by presenting a satisfying outcome after the response; negative reinforcement does so by removing an aversive outcome after the response. Also, both punishment, and response cost, timeout, and extinction (the upper right and lower left cells) serve to decrease or weaken behavior. Punishment does so by presenting an aversive consequence following the behavior. The others do so by removing a satisfying consequence following the behavior. Note: presenting an aversive outcome is sometimes referred to as Punishment I, while removing a satisfying outcome is referred to as Punishment II.

Learning new behaviors and maintaining established ones

• Shaping involves the reinforcement of successively closer approximations of some target behavior. For example, if you want to teach your dog to fetch the newspaper, you might begin by rewarding him when he approaches the door, then only when he goes out onto the sidewalk, then only when he approaches the newspaper, then only when he picks it up, then, finally, only when he performs the target behavior of bringing the paper to you.
• Chaining teaches complex behavior by reinforcing the performance of simpler behaviors which are then strung together in the proper sequence to make up the more complex behavior. For example, if I want to teach someone to assemble an electronic component from a kit, I might first teach the individual steps of assembly (with appropriate reinforcement after the completion of each task), then reinforce the learner only when she can asse
  a particular module, then finally only when she can assemble the entire component.
• Discrimination learning is the process of learning to discriminate between settings in which a particular behavior will or will not be reinforced. For example, if I'm learning to work simple arithmetic problems, I must know to add in the presence of a plus sign (a discriminative stimulus), but not to add in the presence of a minus sign.
• Fading is the gradual withdrawal of prompts or cues (discriminative stimuli) that guide the performance of a complex behavior. For example, when I'm first learning to play chords on a guitar, I will need lots of prompts, such as diagrams of the chord, a picture of someone with their fingers in the proper position, or a teacher who actually places my fingers in the proper position. Later, I will need fewer and fewer prompts to be able to form the chord correctly.

To this point, we have assumed that a behavior is reinforced every time it is performed. However, behaviorists found that this need not be the case. Once a behavior has been established, reinforcement "schedules" other than one reinforcement for every performance turn out to be more effective at maintaining high response rates. 

Behaviorism and education

One of the most lasting influences of behaviorism in education is in the use of principles of behavior management. Many teachers have used ideas such as "time out" and token economies to reward on-task classroom behavior and discourage disruptive behaviors.

Behaviorism also had a pervasive influence on instruction and instructional design, although that influence has faded considerably over the past two decades. Contemporary computer-assisted instruction is a direct descendant of teaching machines and programmed instruction, both of which were attempts to create a technology of teaching in accordance with behaviorist learning principles. Instruction created according to behaviorist principles (especially programmed instruction, early computer-based instruction, and some contemporary computer-based tutorials and drill-and-practice) typically emphasized these characteristics:

• Specific, observable learning outcomes. The target behaviors needed to be specified so that the instructor or designer could provide appropriate reinforcement when they were achieved.
• Individualized learning. Each learner worked independently, at his/her own pace. This permitted "reinforcement" to occur at the moment the learner completed a particular objective.
• Frequent reinforcement. The material to be learned was broken into very small "chunks", usually somewhere between a paragraph and a page of text. This was so that the learner could be reinforced frequently while working through the material.
• "Knowledge of correct results" as a reinforcer. The usual pattern of this kind of instruction was to present a small amount of reading or a problem, then to pose a question. In behaviorist terms, the reading or problem was a discriminative stimulus; when the student answered the question, they were emitting a response; and reinforcement that followed the response was finding out whether they had answered the question properly.
• Fading of prompts. In more sophisticated behaviorist instruction, the learner would be provided with less and less guidance or prompting as he/she proceeded through the material.

One of the reasons that this kind of instruction has largely disappeared is that it emphasized lower-order skills such as memorization. It is very difficult to teach higher thinking skills such as synthesis and evaluation in this instructional format.

Learning activities

Learning Activities for Session Two