Animal Behavior for Shelter Veterinarians and Staff. Группа авторов
Читать онлайн книгу.
(Skinner 1956). Skinner was running an experiment in which a rat pressed a lever for food. Unbeknownst to Skinner, the pellet dispenser jammed at some point during the session. Therefore, presses on the lever no longer produced reinforcement and underwent extinction. The rat didn’t immediately stop pressing the lever; instead, there was a gradual reduction in the number of lever presses before the behavior finally stopped. Skinner’s accidental demonstration of extinction highlights an important feature of the process: the behavior under extinction diminishes gradually, not in an all‐or‐none fashion. How quickly behavior decreases during extinction is a function of the schedule of reinforcement that maintained the behavior, the length of time the behavior has been in the animal’s repertoire, and whether conditioned reinforcers are delivered during extinction.
At the beginning of the extinction process, an increase in frequency or intensity of the behavior, called an extinction burst, may occur. Extinction bursts can make the use of extinction difficult when the intention is to reduce behavior. Many pet owners are familiar with a dog who persistently begs during dinner time. She might sit politely by the dining table waiting for the usual handout. The dog has a long history of reinforcement for begging at the table and has learned some methods that might increase the odds of her getting a snack. When she is ignored, she might escalate to whining. When that doesn’t work, she might start to bark. The barking can be pretty annoying, making it really difficult to continue ignoring her. If the owner gives the dog food at that point, he or she would have been relieved of the dog’s barking temporarily but would have also reinforced barking as a way to get food. In addition, it teaches the dog to be further resistant to extinction because the owner has effectively thinned the reinforcement schedule. Using an extinction procedure can take a long time, and owners must wait for the extinction burst to subside before the begging behavior is completely extinguished. When begging no longer occurs, the association between begging at the dinner table and getting food is overridden by the new learning that begging does not result in food.
Extinction and reinforcement are used in combination to teach new behaviors through a technique called shaping. In shaping, a behavior is trained by reinforcing responses with forms that are closer and closer to a final desired behavior. In the laboratory, a common scenario is for a rat to press a lever for food. However, when a rat is put in the operant chamber for the very first time, it is highly unlikely that he would press the lever since the lever‐press response and the reinforcer have yet to be associated. Experimenters must first shape the lever‐press response before they can run their experiments. As the rat sniffs around the operant chamber and looks in the direction of the lever, the experimenter delivers a food pellet. As the rat moves progressively closer, each approach is reinforced with a food pellet. The experimenter might then wait for the rat to place his paw on the lever before delivering food. And finally, the rat presses down on the lever, exhibiting the final desired behavior. As successive approximations to a lever press are reinforced, previous responses that had formerly been reinforced are extinguished.
3.4.3 Stimulus Control
An important aspect of learning is emitting certain behaviors at certain times or in certain contexts. Otherwise, energy and time are wasted emitting behaviors when the desired consequence is unlikely to happen. We answer the phone when it rings, a cat runs to the sound of the can opener, a trained scent detection dog sits when they smell the target scent, and drivers stop at intersections when the light is red. A stimulus that precedes operant behavior, called an antecedent, can become correlated with the consequences that follow behavior. When an antecedent stimulus exerts control over whether or not a behavior occurs, it is said that behavior is under stimulus control. Stimulus control explains why animals don’t engage in random behavior all the time—an antecedent stimulus that an animal experiences at any given moment signals the animal to behave in ways that are likely to produce reinforcement and avoid behaving in ways that are likely to result in punishment or extinction. In other words, the antecedent stimulus “controls” the occurrence of behavior because it signals that the behavior will be reinforced, punished, or extinguished. A thorough analysis of behavior in terms of operant conditioning usually entails looking at the ABCs: antecedents, behaviors, and consequences (see Table 3.3).
Table 3.3 The ABCs of behavior analysis.
| Description | Example | |
|---|---|---|
| Antecedent | A stimulus that precedes a response | Mailperson walks down the street |
| Behavior | The organism’s response to the antecedent | The dog barks |
| Consequence | The stimulus change that follows the behavior (addition or removal of a stimulus) | The mailperson crosses the street and thereby reinforces the dog’s barking behavior |
In technical terms, if the presentation of a stimulus reliably evokes an operant response, the stimulus is called a “discriminative stimulus.” In application, a discriminative stimulus is often called a “cue.” For a stimulus to reliably function as a discriminative stimulus, the same rules for creating strong associations apply. The cue needs to reliably and consistently signal a certain consequence if a behavior occurs. Naive trainers sometimes attempt to train their pet to sit by repeatedly saying “sit.” After saying “sit” a dozen times, the pet sits and gets a treat. Unfortunately, “sit” never becomes a reliable cue because the pet did not sit most of the time the cue was presented. However, after a few pairings of the trainer saying “sit” once and the dog sits, and the dog is unlikely to sit when the trainer refrains from saying “sit,” then the behavior can be said to be under stimulus control.
Antecedent stimuli can reliably evoke behavior after being correlated with reinforcement. Antecedent stimuli can also reliably inhibit behavior by signaling that there is no chance that reinforcement will occur following a behavior (i.e., extinction). A stimulus that signals extinction is termed s‐delta. Stimuli that signal whether a behavior will be reinforced or not are ubiquitous and very effective in guiding behavior. An “Out of Order” sign on a vending machine tells us that using the machine will not yield us any goodies. In animal behavior, research has found that dogs don’t beg from people who aren’t looking at them (Udell et al. 2011). A person not looking at the dog is effectively an “Out of Order” sign to the dog that means if the dog begs, it is highly unlikely that she will get food. If the person is making eye contact with the dog, it is more likely that food will be given.
One area of training in which “tight” stimulus control is desired is landmine detection. A non‐profit organization called Anti‐Persoonsmijnen Ontmijnende Product Ontwikkeling (APOPO) employed the use of giant pouched rats to help with de‐mining areas of Africa (Poling et al. 2010). The rats are given extensive training on detecting the odor of a landmine and emitting a behavior to indicate that they found an explosive. So that each landmine is identified and removed safely, it is imperative that the explosives have perfect stimulus control over the indicator behavior. That is, the behavior must occur each time a landmine is found, it must never occur when there is no landmine present, it must never occur in response to a different stimulus, and no other behavior should occur in the presence of the landmine (Pryor 1999). Otherwise, it would be a waste of resources to dig for a landmine that isn’t there and would put lives at risk if a landmine is missed.
3.4.3.1 Discrimination and Generalization
After some experience learning a target behavior with one person, a dog might respond to a discriminative stimulus (such as “sit”) from other people. Responding to the same cue from a different person is an instance of generalization. Conversely, if the dog does not