Behavior Theory In Practice, Part 4 (1965)

can be learned by studying a single operant. But we are frequently interested in describing sequences of behavior under multiple stimulus control. Probably the simplest sequence or chain of behavior is one in which the successive responses have the same topography. One way to establish such a homogeneous chain is to start with the response nearest reinforcement and work backwards. This quail has had a brief history of pecking in this apparatus. When the response of pecking the white key is well established, the blue key is lighted. Pecking blue is reinforced by the white light and pecking white is reinforced with food. Then the red key is added. and finally the green. Each SD light functions as a conditioned reinforcer for the preceding response in the chain. This quail is believed to be the first bird recorded in history. It is the quail of the Bible and is pictured on Egyptian reliefs. In this homogene, the reinforcer at the end of the chain is a token which can be traded in for candy or a toy at the end of the experimental session. So although the chain we are measuring is the sequence of responses on telegraph keys, the reinforcer that maintains the behavior comes much later and follows a different set of responses including gathering up the tokens and bringing them over to the table. It is relatively easy to equate the responses within homogeneous chains and that is the main reason for studying them in the laboratory. But most sequences of behavior outside the laboratory in chloriety of responses. They comprise heterogeneous chains of behavior. This pigeon is really performing a heterogeneous chain. As we have seen, such a chain starts with magazine training when the light and the sound of the feeder become SDS for the bird to approach the food magazine and eat. We then use the stimula connected with the feeder to condition the pecking response. We can say that the feeder stimula are conditioned reinforcers for pecking because they shape pecking. And we can say they are SDS for head lowering because the bird does not lower his head until they are presented. We can add another response to this kind of a heterogeneous chain using the SD for pecking as a conditioned reinforcer to shape the new response. The red disc will be the SD for pecking and we will use it to shape the response of turning a circle. A black disc in the window will be the SD for turning. We will shape the turning response, reinforcing successive approximations of a full turn. Since the first approximation was in a clockwise direction, that is the direction we will condition. His head is turned farther than before, but his feet haven't moved. Keep watching his feet. Last time his head turned more than halfway. Now we hope he'll turn his body. Once a bird turns more than halfway, you can usually bring him the rest of the way when you present the SD. Now he's turning 3/4 of the way by himself. Now he makes most of the turn before the SD is presented. This will be the first complete turn. This pigeon is working under a three component heterogeneous chain. He pecks the bell, turns a circle, and pecks the triangle on a variable ratio schedule averaging 10 responses. He was first conditioned to peck the black triangle, the last response in the chain. Then turning was shaped by using the triangle as a condition reinforcer. So turning was added to the front of the chain. After this response was well established, the bird was trained to peck a bell. A blank window was the SD for pecking the bell, and a circle was the SD for turning. Then a fourth response was added. The bird was trained to climb on a perch to reach the bell. The bell is raised and the chain breaks down. Since the bird does not peck it, the response of reaching in the direction of the bell is reinforced. Raising the bell even this small amount has generalized effects. The clockwise turn is completely disrupted. Eventually, he turns in the correct direction and receives the SD for pecking. The components of a chain closest to reinforcement are the most stable. Pecking the triangle is not affected. After going through the chain six times with the bell in this position, the performance is fairly stable and the bell is moved to the end of the box where the perch will be placed. The pecking response is disrupted and so is turning. Unfortunately, he's turning the wrong direction. For the next approximation, the perch was placed in the box under the bell, and again, the bird's behavior was disrupted. In 5 minutes, there were no approximations of pecking the bell, so it was lowered to the height you see here. After another 1 and 1/2 minutes, he makes a pecking motion in the direction of the bell. And this response is reinforced with the SD for turning. He pecks this pattern, and the first turn is in the wrong direction. The final pecking response is not disrupted by the introduction of the perch. During the next 5 minutes, pecking the bell was reestablished. Then the bell was raised to this height. Here is the next major approximation. Again, turning is disrupted and the criterion of a full clockwise turn is relaxed in favor of presenting the primary as possible. The next three times the bird placed his foot on the perch. This is the next approximation. Turning is still disrupted and the guests have to turn in the wrong direction. But climbing the perch is now established in the chain. Some two hours of training later, the chain looks like this. Probably most operant chains are established with a discriminative stimulus for each successive response. But here there is no separate SD for climbing the perch. The situation was structured so that the easiest way to reach the bell was by climbing the perch. In other cases, SDS that are present when a chain is learned may be gradually dropped out. When this dog obedience exercise is performed in competition, the handler may give only four commands: stay, get it, out, and heal. Get it. Out. Heel. Good boy. Good boy, Nick. Good boy. The command get it serves as the SD for several responses that had separate SDS during training. >> Getting up, taking the hurdle, finding the dumbbell, picking it up, returning over the jump, approaching the handler, and sitting >> out heel. That's a good boy. Good boy. >> When the chain was learned, there was an SD for each response. In the final performance, the completion of one response seems to serve as the SD for the next response in the chain. In this exercise, the chain contains alternative responses. The dog is sent out between two jumps. The next SD depends on the judge, who waits until the dog is seated before telling the handler which jump he wants the dog to take. >> The handler then says, "Over," pointing to the jump designated by the judge. The dog takes the jump and with no further commands approaches the handler and sits in front of her. The SD and the response of heal follow. >> Good boy. Good boy, Nick. Good boy. Paradigm for this chain might look like this. Of course, the handler's behavior is also a chain controlled partly by SDS from a judge. She gives no command to her dog until the judge gives her the appropriate SD. The interactions between a blind man, his guide dog, and the environment illustrate the control of behavior by several stimuli that at any given moment may be SDS for incompatible behavior. As they approach the corner, the verbal behavior of the man is an SD for the dog to go forward, >> but to curb the dog to stop and sit. >> The dog's stopping is an SD for the man to feel about with his foot. And when he finds the curb, he again tells the dog to go forward. Now the dog obeys the command and moves out. And now the man walks ahead, responding to tactual SDS from the dog and harness. The training of a seeing eye dog includes punishment as well as reinforcement. But the punishment always comes from things in the environment that the dog must learn to avoid, not from the trainer. >> Forward. Here, the trainer walk and hits it, making as much noise as possible. This is an aversive stimulus for the dog. And on future occasions, the dog avoids coming that close to a car. >> Here, she blocks the trainer and doesn't let him move from the curb. Now, when the traffic stops, she moves right out. Traffic training is relatively easy because the dog has a reliable instinct for self-preservation. But how would you train a dog to avoid something like an awning that isn't an obstacle to the dog itself? How would you make a conditioned aversive stimulus out of something that wasn't even a stimulus? The trainer walks right into the awning and makes a lot of noise. Then he backs off and tries again. That's a good dog. >> In just one trial, the dog avoids the awning and leads her trainer wide around it. In training a seeing eye dog, it is imperative that the trainer behave as though he were blind. If he anticipates the danger, he will be signaling the dog rather than the other way around. >> Forward. Good dog. Good dog. Good girl. >> Here, the dog will be trained to avoid pitfalls such as a hole in the street. The trainer walks right into the hole and stumbles, creating confusion to make the hole an aversive stimulus. it. >> On the next trial, she forces him away from the hole and is praised for this avoidance behavior. >> Good job. We have examined the behavior of a number of species both inside and outside of the laboratory. And we have become increasingly aware of those basic principles of behavior which obtain between organisms and their environment at various levels of sophistication yet with remarkably consistent results. While operant conditioning is sometimes called trial and error learning, a better name might be trial and success learning.