Chapter 5 Stimulus Control & Generalization PDF

# Chapter 5 ## Stimulus Control and Generalization * Joseph E. Spradlin * Jennifer L. Simon * Wayne W. Fisher A teenage boy may readily acquiesce to his mother's request for a hug and a kiss in the privacy of their home but vehemently refuse the same request in front of his peers. We often yell an...

# Chapter 5 ## Stimulus Control and Generalization * Joseph E. Spradlin * Jennifer L. Simon * Wayne W. Fisher A teenage boy may readily acquiesce to his mother's request for a hug and a kiss in the privacy of their home but vehemently refuse the same request in front of his peers. We often yell and cheer at sporting events, sit quietly but periodically sing during a church service, and whisper when we wish to convey information to one individual without being overheard by others. These examples illustrate how the behavior of humans and other species often changes, depending on the circumstances. When our behavior changes in response to such environmental circumstances, we say that it is under stimulus control. Stimulus control is highly relevant to applied behavior analysts because most behavior is under some degree of stimulus control. As such, in this chapter we discuss a broad range of phenomena that fall within the topic of stimulus control. These include discriminations that are learned through direct training or experience, such as simple discriminations and conditional discriminations; they also include ones learned through generalization processes, such as stimulus generalization, stimulus equivalence, and recombinative generalization. Researchers have used the term *stimulus control* in many ways, which have broad connotations. We may define stimulus control in terms of the changes in the probability of a form or rate of a behavior that occur due to the presentation of a stimulus. Defined in this way, stimulus control would include discriminative, eliciting, and reinforcing functions of stimuli. We have restricted the discussion for this chapter primarily to *discriminative control* - the stimulus control that develops when we present positive reinforcement or withdraw negative reinforcement contingent on a response in the presence of a stimulus. It's difficult to discuss the topic of *stimulus discrimination* without also discussing its countereffect - *generalization*. Researchers have used the term *generalization* in many ways. Basic researchers working with nonhumans in the laboratory typically refer to *primary stimulus generalization*. For example, Jenkins and Harrison (1960) conditioned a response to occur when they presented a 1,000-cycles-per-second (1,000-cps) tone and showed that the response also occurred when they presented a 1,500- or a 670-cps tone. Applied behavior analysts have used the term *generalization* in a much broader sense. Stokes and Baer (1977) defined generalization as "any occurrence of relevant behavior under nontraining conditions (i.e., across subjects, settings, people, behaviors, and/or time) without the scheduling of the same events in those conditions as had been scheduled in the training conditions" (p. 350). We use the term generalization in the current chapter in a slightly different way. Generalization for us includes more rapid learning of new discriminations, based on the learning of similar discriminations in the past. Individuals continuously contact contextual or background stimuli, such as visual, auditory, tactile, gustatory, and olfactory stimuli. These stimuli can be external or produced by the individual's own body. Whether a specific stimulus gains control over a specific response depends on many factors. **First**, it depends on the *saliency* of the stimulus, or how different the stimulus is from background stimuli. **Second**, a stimulus must be associated with differential consequences to gain control over a specific response. **Third**, the response must be a part of the individual's behavioral repertoire for a stimulus to gain control over a response. ## Establishing Control by a Single Stimulus Establishing stimulus control requires a salient stimulus, a somewhat controlled environment, and a target response that is part of the individual's repertoire. We can bring the target response under the control of a salient stimulus if: 1. A reinforcing consequence immediately follows each target response that occurs in the presence of the stimulus. 2. Reinforcement of target response does not occur in the absence of the stimulus. 3. There are no other conditions correlated with reinforcement and nonreinforcement. We want to ensure that the relevant stimulus is controlling the target response, and that target responding is not based on temporal patterns. For example, target responding could be based on time rather than the presentation of the stimulus if we present a stimulus once every minute for 30 seconds. Therefore, we should present the stimulus on a variable-time schedule. In some situations, stimulus presentation may function as *reinforcement* for target responding that occurs before stimulus presentation, particularly when we present the stimulus on a response-independent schedule. The sequence is as follows: 1. Target response 2. Stimulus presentation 3. Target response 4. Reinforcement Stimulus presentation functions as *reinforcement* for target responding before stimulus presentation when we deliver reinforcement for the target response in the presence of the stimulus. Such reinforcement may impede *discriminative responding*. Therefore, we should delay the presentation of the stimulus if a target response occurs just before the scheduled presentation of the stimulus, to allow for extinction of the target response during the absence of the stimulus. We use the term *change-over delay* or *momentary differential reinforcement of other behavior* to refer to delayed stimulus presentation contingent on responding before stimulus presentation. We should determine whether target responding is under the control of the *discriminative properties* of the relevant stimulus after target responding occurs primarily when the stimulus is present and not when it is absent. One possibility is that reinforcement may have developed discriminative properties, because reinforcement of target responding is likely to be followed by more reinforcement, and nonreinforcement is likely to be followed by more nonreinforcement. In this case, target responding may occur more after reinforcement delivery than following a response that was not reinforced if the delivery of the reinforcer functions as a *discriminative stimulus*. We can be more confident that the relevant stimulus controls target responding if responding begins immediately after stimulus presentation and stops immediately after stimulus termination. Even when we use a salient stimulus and have established control over a response, other similar stimuli also may control the response. *Stimulus generalization* allows for the reinforcement of responses in the presence of stimuli that initially may have been difficult to condition. For example, Fulton and Spradlin (1971) initially established control over a button-pressing response to a 70-decibel, 500-Hertz tone, to assess the hearing of children with intellectual developmental disorder. Control occurred for a target 70-decibel, 500-Hertz tone for most participants, and less intense tones also controlled responding. Responding to less intense tones than the target 70-decibel tone allowed the researchers to provide reinforcement of responses to lower-intensity tones. Generalization occurred to tones of still lower intensity after researchers provided reinforcement for responses to lower-intensity tones. The tone maintained stimulus control over button pressing due to stimulus generalization and reinforcement of responses to tones with progressively lower intensity until the tone reached a threshold level, which is the magnitude level at which responding is no longer *discriminative*. Researchers have conducted most studies of simple stimulus control like those described above with auditory and visual stimuli. We also can use these procedures to establish control with tactile, gustatory, and olfactory stimuli. Although control with a single stimulus has few parallels in individuals' daily lives, single-stimulus control is useful for conducting hearing evaluations for people with intellectual developmental disorder and for teaching dogs and rats to find people, narcotics, and explosives by smell. ## Differential Stimulus Control by Stimuli Presented Successively Although the mere detection of a stimulus is important under some conditions, stimulus control more often involves a discrimination between two stimuli: 1. A *positive stimulus* correlated with reinforcement. 2. A *negative stimulus* correlated with nonreinforcement. The researcher alternates the positive stimulus (such as a red light) with a negative stimulus (such as a green light) of the same intensity and presentation duration to establish a *successive discrimination*. We use the term *successive discrimination* because the positive and negative stimuli are present at different times. However, discriminative responding under this condition can be slow and may not occur for all experimental participants. *Fading* is likely to establish a more rapid discrimination among the stimuli, because we can maximize the difference between the positive and negative stimuli initially and then decrease the differences between stimuli gradually. Methods for increasing the difference between the stimuli include maximizing the *saliency* of the positive stimulus (such as a bright red light) and minimizing the *saliency* of the negative stimulus (such as a faint green light), and presenting the positive stimulus for a long duration and the negative stimulus for a brief duration. Maximizing the differences between the stimuli increases the probability of responding in the presence of the positive stimulus and not responding in the presence of the negative stimulus. *Fading* initially requires a salient positive stimulus (such as a bright red light) and a nonsalient negative stimulus (such as a faint green light). We establish responding in the presence of the positive stimulus and not in the presence of the negative stimulus. We then increase the intensity of the negative stimulus (such as increasing the brightness of the green light) until the intensity of the positive and negative stimuli is equal, and responding occurs only in the presence of the positive stimulus. We then gradually increase the duration of the negative-stimulus presentation until the durations of positive- and negative-stimulus presentations are equal, and responding occurs only in the presence of the positive stimulus. ## Differential Stimulus Control by Two or More Simultaneously Presented Stimuli A simple *simultaneous discrimination* involves the discrimination between two or more stimuli presented at the same time. ## Conditional Stimulus Control The discriminations we have discussed previously have been either simple successive or simple simultaneous discriminations, which are essential in daily life. However, many discriminations that individuals make during their daily activities are *conditional discriminations*, in which reinforcement of a response in the presence of a stimulus depends on the presence or absence of other stimuli. An example of a conditional discrimination is passing the salt when someone asks you to pass the salt, and passing the bread when someone asks you to pass the bread. That is, the discriminated behavior is conditional because the positive stimulus (e.g., salt, bread) changes depending on the request. *Simultaneous identity matching* is a very simple *conditional-discrimination* procedure. Researchers have studied this procedure widely in laboratories with humans and nonhumans. In the laboratory, the typical procedure involves presenting a visual sample stimulus (e.g., the numeral 2) to which the experimental participant must respond. The researcher then presents two or more comparison stimuli (such as 2 and 3, or 2, 3, 4, and 5) after the participant's response to the sample. One of the comparison stimuli (e.g., 2) is identical to the sample stimulus; the remaining stimulus (e.g., 3) or stimuli (e.g., 3, 4, and 5) differ from the sample stimulus. The disparity between the correct comparison stimulus and the other stimuli may be large or small. The researcher presents different stimuli as samples from trial to trial; thus the correct comparison stimulus is *conditional* on which sample is present. Identity-matching experiments have involved simple trial-and-error procedures and fading procedures. The researcher may present only the single comparison stimulus that matches the sample stimulus, or the other comparisons may be blank on the first few trials of an identity-matching-to-sample task using a fading procedure. The researcher may begin to fade in the nonmatching stimulus or stimuli after a few trials. The nonmatching stimuli become more visible until the intensity of the nonmatching stimuli matches that of the sample stimulus with each successive trial of correct responding. Participants typically match new stimuli on their first presentation after they have matched a few comparison stimuli to samples; that is, they exhibit *generalized identity matching*. *Generalized identity matching* is not exhibited readily by nonhuman participants and some participants with intellectual developmental disorder. Researchers have used video modeling, error correction, and fading to teach participants with autism spectrum disorder and intellectual developmental disorder *generalized identity matching* on a computer screen using a video modeling, error correction, and fading from a simple tabletop sorting task compound identity matching on a computer screen with participants with autism spectrum disorder and intellectual developmental disorder to facilitate generalized identity matching. We consider *generalized identity matching* as another example of *generalization*. Preacademic workbooks use identity matching extensively to teach letter and number discrimination to students. Usually, the workbook presents the sample letter or number at the left margin, and the choice letters or numbers in a row to the right of the sample. The student's response is to mark the correct choice. This workbook task is an example of identity matching; however, it is not an ideal teaching technique, because it involves delayed reinforcement for correct responses. *Simultaneous identity matching* requires discrimination of the sample stimuli from the remaining comparison stimuli. It neither requires nor ensures *successive discrimination* among sample stimuli, because the sample stimulus remains available throughout the trial. However, it is a *delayed-matching-to-sample* procedure if the teacher removes the sample after the onset of the comparisons and probably requires *successive discrimination* of the sample stimuli. ## Equivalence Classes Configurations of shared physical properties determine many stimulus classes (e.g., balls, cars, cats, humans, men, women, red objects). The actual configurations of shared physical characteristics determining class membership have been the focus of research by psycholinguists, cognitive psychologists. However, shared physical properties do not define many important stimulus classes (e.g., lawyers, letters, medical doctors, numbers, tools, toys). *Whether each member is substitutable, and whether they evoke new, untrained responses in certain contexts, define the members of these classes.* *For example, we may define toys as a stimulus class because they are items that children manipulate, and we may store them in a toy box. In addition, a child is likely to engage in exploratory and novel play behavior without any direct training when he or she finds a new item in the toy box. Medical doctors are a stimulus class because we call them doctor, and any member with the appropriate credentials may practice medicine. In addition, people are much more likely to follow the health-related advice of someone called doctor than someone called waiter.* Sidman (1971) established an equivalence class using a symbol-matching procedure with a 17-year-old student with microcephaly and intellectual developmental disorder. The student selected 20 pictures when the researcher presented their corresponding dictated words, and could name 20 pictures when the researcher presented the pictures before the study. However, he did not name the 20 printed words related to the pictures, select printed words in response to their dictated names, select printed words that named the pictures, or select the pictures when presented with the printed words. Sidman trained selection of printed words when he presented corresponding dictated words to the student. He conducted probes after training to test whether the student would select printed words when given the respective pictures, and whether the student would select the pictures when given the respective printed words. Not only did the student select the printed words when Sidman presented dictated words, but the BC and CB relations between pictures and printed words emerged, regardless of whether Sidman presented the printed words as sample or comparison stimuli. In addition, the student named many of the printed words after the initial AC training. This training established 20 stimulus classes; each class consisted of the spoken word, the printed word, and the pictures. ## Stimulus Control Based on Recombination of Stimulus-Response Components One of the remarkable characteristics of human behavior is the degree to which responding to complex stimuli occurs without previous direct experience. For example, young children develop *generalized imitation* so that their behavior can approximate that of a model, even though they have never had direct training on imitating the specific model (e.g., Baer, Peterson, & Sherman, 1967; Peterson, 1968). In addition, individuals may respond appropriately to specific verbal instructions they have never encountered previously . *Recombinative generalization* occurs when a student recombines responses targeted during training in novel ways. For example, a student may "push car" and "drop glass" if we teach him or her to "push glass" and "drop car". ## Some Concluding Remarks Nearly every act throughout the day requires some sort of stimulus discrimination, and teaching each discrimination by direct reinforcement would be an impossible task. Yet we make many discriminations throughout the day, which allow us to respond appropriately in a complex world. The pages of this chapter perhaps provide a less puzzling account of the acquisition of such a vast repertoire of discriminations. In our discussion of simple successive discriminations, we have noted that even though we might condition a single stimulus to control a response, other, physically similar stimuli could also control that response. Fulton and Spradlin's (1971) research on auditory stimulus control demonstrated that if a student learns to press a button in response to a 500-Hertz, 70-decibel tone, pressing the button also may occur in response to tones with other frequencies and volumes. *Hence extending the stimulus control across a total range of frequencies and volumes accessible to human hearing is easy. Therefore, more learning occurs than what we teach directly even in simple successive discriminations.* However, the *equivalence paradigm* provides even more examples of how an extensive repertoire of discriminations can emerge from very little teaching. In the hypothetical number example, researchers only teach three conditional discriminations before the nine additional conditional discriminations, and potentially three stimuli names emerge. Saunders, Wachter, et al. (1988) taught seven conditional discriminations, and a total repertoire of 56 conditional discriminations emerged. *In other words, they taught seven conditional discriminations, and 49 emerged.* *Recombinative generalization* provides an additional example of how a little training produces an extensive repertoire. Striefel et al. (1976) taught 31 noun-verb instructions to a student with intellectual developmental disorder, and 113 emerged without training. The recombination of letter-sound units makes it possible for students to respond appropriately to almost any new printed English word after being taught only a limited number of letter-sound units. In short, research on primary stimulus generalization, stimulus equivalence, and recombinative generalization provides examples of how behavioral repertoires are acquired rapidly and suggests methods for the effective teaching of such vast repertoires.