William J. McIlvane, PhD
University of Massachusetts Medical School
Like many terms in the clinical and behavioral sciences, different people use stimulus control for different purposes relating to their interests, activities, needs, and verbal conventions. For example, some clinicians may recognize stimulus control as a name for specific kinds of behavior therapy or therapeutic procedure (e.g., for compulsive gambling; Hodgins, 2001). By contrast, behavioral scientists often use the term when describing one component of a three-term contingency relation used in analyzing the environmental control of behavior (stimulus, response, consequence; see Skinner, 1935). Still others use this term as a name for an entire subfield of scientific inquiry (stimulus control research) that encompasses analytic studies of behavior—attention, memory, executive functions, concept formation, and symbolic classification (e.g., Sidman, 2008). All of these uses are relevant for the purposes of this chapter.
A stimulus is a measurable environmental event that has a measurable effect on behavior. While a tree falling in a forest may be an event that could be measured, the falling tree is not a stimulus unless someone observes it and that observation results in reactions that would not have occurred otherwise (e.g., yelling “Watch out!”). Even if someone is present to observe the tree fall, it is not a stimulus unless a behavior occurs with respect to it. If a birdwatcher’s full visual attention was captured by a rare species, for example, an observer might judge that the birdwatcher didn’t seem to notice the tree fall (i.e., it would not be a stimulus for the latter from the perspective of the former). However, if the sound of the tree falling caused a change in the birdwatcher’s blood pressure, it would be a potentially measurable event that had a potentially measurable effect on the birdwatcher. If the effect was measured via remote sensors that detected both the sound and the change in blood pressure, the tree falling could be classified as a stimulus, by my definition, even though the on-site observer detected no behavior change.
I gratefully acknowledge the long-term support of the National Institute of Child Health and Human Development (grant numbers HD25995 and HD04147) and the Commonwealth Medicine Division of the University of Massachusetts Medical School. I also thank Charles Hamad, David Smelson, and Beth Epstein for helpful input in the formulation of this chapter.
From a more functional perspective, stimuli cannot be defined independently of behavior, and behavior cannot be defined independently of stimuli. Stimuli are defined in relation to their effects on behavior as measured directly or indicated by strong inferential processes. The two events constitute a functional unit of analysis that also includes a third term—the positive or negative consequence—when defining a reinforcement contingency (see Sidman, 2008).
Early on, Skinner (1935) defined stimuli (and responses) generically in terms of their function, much as I have done here. This emphasis on function led to the idea of further defining stimuli in terms of functional classes. If the functions of stimulus events X, Y, and Z can be shown to relate to behavior and its effects in a similar manner, then these events may constitute a functional stimulus class. There are two basic types of functional stimulus classes: those defined by shared physical features or in purely functional terms.
Functional classes defined by shared physical features have been termed “feature classes” (McIlvane, Dube, Green, & Serna, 1993) or “perceptual classes” (Fields et al., 2002). To exemplify such classes, consider a simple sorting task that is used often in behavior therapy for children with autism spectrum disorders. One might teach the child to sort both coins and plastic washers from a pool containing these items and noncircular distractors to attempt to have the feature of circularity come to control behavior. Accurate sorting alone of the items does not necessarily demonstrate that a feature/perceptual class defined by circularity has been established, because the child might merely have attended to specific features of each of the items sorted (i.e., this could be a case of rote learning and nothing more). To assess whether the child was responding on the basis of the abstract property of circularity, however, one could add new circular items (e.g., buttons) and new noncircular distractors to the pool. If buttons are also immediately sorted along with the coins and washers, one has evidence that a functional feature or perceptual class (in this case, one defined by circularity) has been established.
To assess whether the circular stimuli relate in a similar manner to environmental operations, one might change the sorting task such that buttons but not coins or washers are available in the pool, and some other noncircular items (e.g., dominoes) are instead defined as correct choices. After the child masters the new task—now avoiding the buttons but selecting the dominoes—one might add back in the coins and washers. If the child now does not select those previously correct items, then it has been shown that changing the function of one class member (buttons) spontaneously changed the functions of the coins and washers, thus providing strong evidence that a functional class has been established.
Humans and nonhumans share this ability to develop such functional classes. For example, Herrnstein (1979) showed that even pigeons can (1) be taught certain generalized concepts, such as tree versus nontree or water versus nonwater, and (2) pass tests similar to those just described. The teaching method most commonly used has been termed multiple exemplar training (MET), in which several—sometimes many—examples sharing defining physical properties are contrasted with other examples lacking those properties. For example, Herrnstein’s MET required pigeons to discriminate forty scenes containing trees from forty scenes without trees to establish the concept targeted. Normally, capable humans are quite adept at such tasks and may abstract concepts such as these from only a few examples.
Feature/perceptual classes show primary stimulus generalization, in which the behavioral effects of the stimulus class apparently extend beyond the original situation in which control was observed. This is what occurred in the earlier example: the ability of buttons to control behavior after the child was trained with coins and washers was an instance of primary stimulus generalization that verified the control by circularity. To specify a feature/perceptual class, one assumes that the individual does attend to the stimulus features specified and further assumes that the individual will respond similarly when other stimuli containing that feature are presented.
As a practical application of this feature/perceptual class analysis, consider a case of phobia: A client reports that he was severely frightened by the sudden appearance of a large rat in his bedroom. After that experience, he reports not only a phobic reaction to rats and mice, but also substantial discomfort with physically similar animals (e.g. squirrels, chipmunks, rabbits). Assuming that a feature/perceptual class exists, a therapist might first teach the client to relax and/or behave more flexibly in the presence of animals that aren’t rats; she might assume also that this MET procedure will make it easier for the client to learn to relax and/or behave flexibly in the presence of rats, the animal that caused the original fright (see chapter 18). If the procedure proves successful, it is evidence that the therapist’s feature/perceptual class analysis was correct. If not, the result suggests that the stimulus class was incorrectly or incompletely specified (e.g., a furless tail not shared by the other animals was a particularly frightening component of the rat’s overall appearance).
A functional stimulus class may also include physically dissimilar stimuli. These classes may be termed contingency (only) classes or arbitrary stimulus classes to emphasize that class membership is defined by similarity of function rather than physical similarity (see Goldiamond, 1966). To understand an arbitrary stimulus class, consider a red traffic light, a STOP sign, and a policeman’s upraised hand; all set the stage for one to step on the car’s brakes. Skinner (1935) implicitly and Goldiamond (1966) explicitly defined a functional stimulus class as having two properties: (1) stimuli must exhibit the same function(s) in the control of behavior, and (2) operations that influence the function of one member of the stimulus class must influence the function of the others. Using the traffic example, motorists fleeing an imminent disaster who observe others ignoring a policeman’s directions without apparent negative consequences are more likely to also ignore other traffic-control measures. In technical terms, a transfer or transformation of functions occurs to all members of the class, although the procedure that changes the function is applied only to a subset of its members (see chapters 6 and 7).
It is an active point of discussion in behavior theory whether arbitrary stimulus classes can be extended to account for the kinds of stimulus control commonly noted in human language and cognition (e.g., Hayes, Barnes-Holmes, & Roche, 2001; Sidman, 2000). Notably, however, cognitive neuroscience methods (e.g., functional MRI, evoked cortical potentials) are increasingly showing that procedures used in basic stimulus control research have the same or similar effects on neural activities as the language and cognitive stimulus events they are intended to model (e.g., Bortoloti, Pimentel, & de Rose, 2014).
In summary, a given stimulus or stimulus class exhibits control when any measured behavior or class of behaviors is more probable in the presence of that stimulus/stimulus class than in its absence. Whether in research or clinical applications, one should not make assumptions about the specific elements and/or properties of controlling relations. It will be most useful to specify what these are by using direct measurements or inferences based on strong empirical evidence. In addition, the concept of “more probable in the presence of a stimulus/stimulus class than in its absence” is critical to understanding stimulus control. For example, suppose that behavior X occurs with a 10 percent frequency when stimulus X is present and with only a 5 percent frequency when stimulus X is absent. If one can reliably demonstrate a frequency difference using quantitative analysis techniques (see McIlvane, Hunt, Kledaras, & Deutsch, 2016), then one can say that stimulus control has been exhibited despite the low frequency of occurrence overall. As I’ll discuss below, the frequency of occurrence of a given stimulus control relation need not indicate anything about its probable persistence or other similar concerns that a clinician might have.
Feature/perceptual classes and arbitrary classes constitute a central component to the scientific analysis of complex behavior, human and otherwise. When combined with procedures exemplified in the next section, one has a strongly evidence-based conceptual, analytical, and methodological framework within which to understand critical components of therapeutic and educational procedures broadly.
At a practical level, the clinician or educator can benefit from stimulus control/stimulus class analyses, using them to promote client success or, when confronted with the failure of applied procedures that seem well designed, to understand and perhaps ameliorate puzzling treatment failures—as one illustration from my own research program shows. We conducted a long-term program aimed at developing methods for reducing so-called impulsive responding in individuals with autism spectrum and other neurodevelopmental disorders (i.e., responding too rapidly on tasks that required participants to carefully inspect stimuli in order to discriminate them). Stimuli were presented in locations defined by square borders on a computer display, thus emulating well-established procedures from much prior stimulus control research and its applications. Our procedures were able to eliminate impulsive responding in most individuals. Nevertheless, such responding persisted in a few people despite our best efforts to eliminate it. A breakthrough occurred, however, when a member of our team suggested eliminating the borders that defined stimulus locations to further simplify the display. Although we thought these borders were irrelevant constant features of the display, eliminating them instantaneously eliminated impulsive responding.
The preceding example illustrates a more general consideration in stimulus control analysis: the controlling properties of stimuli that the researcher, teacher, or therapist deems relevant may be strongly influenced by the broader context in which those stimuli are presented. We have found stimulus class analysis particularly useful in thinking about contextual stimuli and stimulus classes that relate to the critical issue of treatment generalization, and especially the failure thereof (see McIlvane & Dube, 2003). One reason that behavior therapists may prefer to provide therapy in everyday environments in which problem behavior occurs is to minimize the likelihood that they may miss critical contextual determinants of the stimulus control of behavior. Sometimes, however, therapy must be conducted outside such contexts (e.g., when the problem behavior is dangerous or socially repugnant). In such cases, the therapist may want to design the treatment contexts to include stimuli from feature/perceptual and/or arbitrary stimulus classes that simulate natural counterparts to maximize the potential for the treatment effects to be generalized.
Simple differential reinforcement. To establish control using two formerly neutral stimuli (A versus B), one can provide positive reinforcing consequences when a targeted behavior occurs in the presence of A and deliver no such consequences when B is present. Soon, one may find the target behavior occurring more frequently in the presence of A than of B. As I noted earlier, even a small difference in differential responding indicates some measure of stimulus control. After the continued application of these contingencies, however, one might find that the individual virtually always responds to A and virtually never to B.
The first sustained efforts of applying differential reinforcement procedures in clinical and educational settings began more than sixty years ago. For example, Skinner’s The Technology of Teaching (1968) was intended for broad application in both regular and special education. His goal was to translate procedures and findings of basic research with nonhumans to such applications. Work in this tradition included the extensive development of instructional technology for normally capable populations, ranging from young children to advanced professional trainees. Other efforts to develop this technology were directed at finding effective therapeutic procedures for special populations (e.g., people with neurodevelopmental and neuropsychiatric problems; Ferster & DeMyer, 1961). In the decades since Technology of Teaching, a voluminous literature has developed, reporting many thousands of studies of reinforcement procedures for a vast range of beneficial clinical and educational applications. These studies have addressed a range of populations, including normally capable children and adults as well as individuals with a broad range of neurodevelopmental, neuropsychiatric, and other neurobehavioral deficits and disorders.
There are emerging issues in differential reinforcement–based methods for establishing stimulus control. Applied behavioral research has highlighted individual differences in response to reinforcement procedures in clinical populations. For example, it may be difficult to identify and/or maintain the potency of reinforcers for some children with autism spectrum and related neurodevelopmental disorders (see Higbee, 2009). Even if seemingly effective reinforcers have been identified, however, research tells us there is another critical consideration to the design of effective therapy: the degree to which the client’s behavior is sensitive to disparities between reinforcement schedules.
As noted, if one reinforces behaviors within a given class and extinguishes behaviors in other classes, the former will come to predominate. In everyday experience, however, one rarely (if ever) encounters situations in which desirable behaviors can be consistently reinforced, nor ones in which undesirable behaviors can be consistently extinguished. Most often, one merely hopes that (1) desirable behavior will be reinforced often (rich schedules of reinforcement) and undesirable behavior only rarely (lean schedules), and that (2) client behavior will prove sensitive to the disparity between these schedules.
My stimulus control research group has long been interested in why some individuals with neurodevelopmental disorders show good sensitivity to rich-versus-lean schedule disparities, whereas others seem almost indifferent to these schedules—even in cases in which traditional reinforcer function tests show strong evidence of reinforce potency (e.g., tests contrasting continuous reinforcement versus extinction schedules, reinforcer preference tests). We are especially interested in cases in which indifference to a rich-versus-lean schedule persists despite programmed training aimed at making the schedule disparities easy to detect (McIlvane & Kledaras, 2012).
Schedule insensitivity/indifference may be a hidden variable when children with autism spectrum disorders do not respond well to applied behavior analysis therapies (see Sallows & Graupner, 2005). An increasing number of studies reference individuals with other neurodevelopmental and neuropsychiatric disorders exhibiting deviant responses to reinforcement procedures. For example, findings from clinical neuroscience research suggest that individuals with ADHD exhibit altered reinforcement sensitivity (e.g., Luman, Tripp, & Scheres, 2010).
Nevin’s studies, and direct and systematic replications by others, have lent substantial empirical support for the momentum analysis. For example, Dube and McIlvane (2002) showed that the momentum analysis can inform procedures aimed at increasing behavioral flexibility in children with autism spectrum disorders. The target task was to reverse a previously established discrimination (a basic requirement for learning educationally relevant tasks, such as matching to sample). In cases where children experienced relatively lean reinforcement schedules in learning A+ versus B– during training, they learned B+ versus A– discrimination faster than in cases where children experienced relatively richer A+ versus B– training schedules. Viewing the literature as a whole, behavioral momentum analyses of stimulus control are a promising development that will increasingly have a beneficial impact on behavior therapy.
In this context, there is clearly a downside to the potentially beneficial relationship described previously in the discussion of stimulus control therapies. Suppressing the control of stimuli associated with RSRCDs may merely reduce their frequency temporarily. Any challenge that causes the resurgence and strengthening of stimulus control by any member of an RSRCD class may increase the probability that other class members will exert stimulus control, even in situations that do not present such challenges.
The potential for resurgence may help account for the unimpressive results of cue exposure therapy (CET) for the treatment of addictive behavior (see Martin, LaRowe, & Malcolm, 2010). In CET, addicts are exposed to a series of drug-related stimuli (e.g., MET with various exemplars of drug paraphernalia) in a setting in which the resulting cravings cannot lead to drug use. The rationale is that the extinction of these cravings should at first lead to withdrawal symptoms and ultimately to the extinction of the drug-seeking/taking behavior. There are two problems with the CET approach. First, any subsequent exposure to even a small subset of stimuli associated with the addictive behavior that leads to relapse (e.g., meeting an old friend who was involved in past drug taking) may reestablish high-probability control by other members of the stimulus class, thus defeating the intent of the CET. Second, contextual stimuli (i.e., those in familiar drug-use settings) may be an unappreciated component of the stimulus control of addictive behavior. If that is the case, CET will fail if those stimulus control variables are not addressed in therapy.
These days, one cannot open the many compendiums such as this one without seeing many citations to and discussions of evidence-based practice. For both practical and ethical reasons, clinicians and educators want to apply therapeutic and/or educational procedures that are supported by scientific evidence. In my experience, most practicing clinicians and educators tend to think in terms of broad classes of procedures (e.g., applied behavior analysis versus sensory integration/occupational therapy for autism). In this chapter, I illustrate a less commonly discussed approach to defining evidence-based practice—that is, relating therapeutic/educational procedures to scientific principles, which must undergird whatever approach one chooses. By doing so, I think one can promote behavioral development, health, and wellness, and have a secure evidential foundation on which to base one’s practice and potentially improve its effectiveness, without being captured by fads and fancies that may temporarily dominate fields.
Bortoloti, R., Pimentel, N. S., & de Rose, J. C. (2014). Electrophysiological investigation of the functional overlap between semantic and equivalence relations. Psychology and Neuroscience, 7(2), 183–191.
Dube, W. V., & McIlvane, W. J. (2002). Reinforcer rate and stimulus control in discrimination reversal learning. Psychological Record, 52(4), 405–416.
Ferster, C. B., & DeMyer, M. K. (1961). The development of performances in autistic children in an automatically controlled environment. Journal of Chronic Diseases, 13(4), 312–345.
Fields, L., Matneja, P., Varelas, A., Belanich, J., Fitzer, A., Shamoun, K. (2002). The formation of linked perceptual classes. Journal of the Experimental Analysis of Behavior, 78(3), 271–290.
Goldiamond, I. (1966). Perception, language and conceptualization rules. In B. Kleinmuntz (Ed.), Problem solving: Research, method and theory (pp. 183–224). New York: Wiley.
Hayes, S. C., Barnes-Holmes, D., & Roche, B. (Eds.). (2001). Relational frame theory: A post-Skinnerian account of human language and cognition. New York: Kluwer Academic/Plenum Publishers.
Herrnstein, R. J. (1979). Acquisition, generalization, and discrimination reversal of a natural concept. Journal of Experimental Psychology: Animal Behavior Processes, 5(2), 116–129.
Higbee, T. S. (2009). Reinforcer identification strategies and teaching learner readiness skills. In R. A. Rehfeldt & Y. Barnes-Holmes (Eds.), Derived relational responding: Applications for learners with autism and other developmental disabilities. Oakland, CA: New Harbinger Publications.
Hodgins, D. C. (2001). Processes of changing gambling behavior. Addictive Behaviors, 26(1), 121–128.
Jacob, G. D. (1998). Say good night to insomnia. New York: Henry Holt.
Luman, M., Tripp, G., & Scheres, A. (2010). Identifying the neurobiology of altered reinforcement sensitivity in ADHD: A review and research agenda. Neuroscience and Biobehavioral Reviews, 34(5), 744–754.
Martin, T., LaRowe, S. D., & Malcolm R. (2010). Progress in cue extinction therapy for the treatment of addictive disorders: A review update. Open Addiction Journal, 3, 92–101.
McIlvane, W. J., & Dube, W. V. (2003). Stimulus control topography coherence theory: Foundations and extensions. Behavior Analyst, 26(2), 195–213.
McIlvane, W. J., Dube, W. V., Green, G., & Serna, R. W. (1993). Programming conceptual and communication skill development: A methodological stimulus class analysis. In A. P. Kaiser & D. B. Gray (Eds.), Enhancing children’s language: Research foundations for intervention (pp. 242–285). Baltimore, MD: Paul H. Brookes Publishing.
McIlvane, W. J., Gerard, C. J., Kledaras, J. B., Mackay, H. A., & Lionello-DeNolf, K. M. (2016). Teaching stimulus-stimulus relations to nonverbal individuals: Reflections on technology and future directions. European Journal of Behavior Analysis, 17(1), 49–68.
McIlvane, W. J., Hunt, A., Kledaras, J. K., & Deutsch, C. K. (2016). Behavioral heterogeneity among people with severe intellectual disabilities: Integrating single-case and group designs to develop effective interventions. In R. Sevcik & M. A. Romski (Eds.), Communication interventions for individuals with severe disabilities: Exploring research challenges and opportunities (pp. 189–207). Baltimore, MD: Paul H. Brookes Publishing.
McIlvane, W. J., & Kledaras, J. B. (2012). Some things we learned from Sidman and some things we did not (we think). European Journal of Behavior Analysis, 13(1), 97–109.
Nevin, J. A. (1992). An integrative model for the study of behavioral momentum. Journal of the Experimental Analysis of Behavior, 57(3), 301–316.
Podlesnik, C. A., & Kelley, M. E. (2014). Resurgence: Response competition, stimulus control, and reinforcer control. Journal of the Experimental Analysis of Behavior, 102(2), 231–240.
Sallows, G. O., & Graupner, T. D. (2005). Intensive behavioral treatment for children with autism: Four-year outcome and predictors. American Journal on Mental Retardation, 110(6), 417–438.
Sidman, M. (2000). Equivalence relations and the reinforcement contingency. Journal of the Experimental Analysis of Behavior, 74(1), 127–146.
Sidman, M. (2008). Reflections on stimulus control. Behavior Analysis, 31(2), 127–135.
Skinner, B. F. (1935). The generic nature of the concepts of stimulus and response. Journal of General Psychology, 12(1), 40–65.
Skinner, B. F. (1968). The technology of teaching. New York: Appleton-Century-Crofts.
Snell, M. E. (2009). Advances in instruction. In S. L. Odom, R. H. Horner, M. E. Snell, & J. Blacher (Eds.), Handbook of developmental disabilities (pp. 249–268). New York: Guilford Press.
Terrace, H. S. (1963). Discrimination learning with and without “errors.” Journal of the Experimental Analysis of Behavior, 6(1), 1–27.