Many Board-Certified Behavior Analysts (BCBAs), assistant analysts (BCaBA), and registered behavior techs (RBTs) serve those diagnosed with Autism Spectrum Disorders (ASD), a pervasive developmental disorder characterized by deficits in language and social skills as well as repetitive or restrictive behavior patterns. In addition to these characteristics, some individuals with ASD may also display an abnormality in their attending commonly referred to in behavioral literature as stimulus overselectivity or overselective stimulus control. It wouldn't be ABA if we didn't have at least two terms floating around to refer to the same thing.
What is Stimulus Overselectivity?
The term stimulus overselectivity comes to us via Lovaas and Schreibman (Lovaas & Schreibman, 1971) to describe a type of "over-generalizing" responding to stimuli that should not evoke that response. In these cases, responding is under the anctedent control of only one or a reduced number of stimuli instead of all relevant stimuli components that would, under better conditions, control the behavior. To help understand this, imagine your own behavior at a stop light: with correct training, you have learned that "green" means go, "red" means stop, and "yellow" means slow down and prepare to stop. If you exhibited stimulus overselective responding, you might "stop" at the presence of any illuminate light, and not discriminate between which color is illuminated. Your behavior is under the control of only one of the relevant stimuli - that a traffic light is present - and not all relevant stimuli: that a traffic light is present AND which light is currently illuminated (red, yellow, or green).
Let's look at another example: you may teach a student to identify a picture of a clock, and then observe that this student begin identifying pictures of beach-balls and dinner plates as "clocks", perhaps because they share stimulus features, such as "roundness" or visual similarity to a circle. Additionally, you may teach a student to identify forks as silverware, and then observe this student referring to any utensil used for eating as a "fork." That student's behavior is under the antecedent control of only ONE of the relevant stimulus features: the "roundness" and not under the other pertinent stimuli that might make a difference in your perception of beach balls, dinner plates, and clocks. A subset of relevant stimuli has control over your students behavior!
Typically developing children will display some of these characteristics and readily acquire the ability to make the discrimination between beach balls, dinner plates, and clocks with a few exposures to the stimuli. This type of discrimination is not limited to visual stimuli - consider also that a discrimination is being made in a scenario where a student is holding a microwaveable dinner and a parent vocalizes. In this case, the stimuli is audible or auditory - the learner must hear what their parent says, and respond accordingly: for instance, maybe Dad says "go put it in the fridge" or "go put it in the microwave" - the behavior evoked (either putting it in the fridge, or putting it in the microwave) depends on the learners ability to parse at least two environmental stimuli: the presence of the item to put away (the microwaveable meal) and the auditory stimulus in the form of the parent's command. If you think about your own life, you'll quickly realize that you're making discriminations constantly - if you had a deficit in making these types of simple discriminations, your day-to-day life and ability to acquire new learned behaviors would be dramatically effected. Over the next 5 minutes, observe your own behavior as you browse the web and identify the discriminations you make and how those control your behavior. When you open up your phone, you select an app from a wide array of app icons on the screen. If you find my writing very boring, you may want to close your browser window by selecting the "x" (and not the resize or minimize icons). After just a few minutes you'll realize the majority of your behavior requires a discrimination of some sort.
Conditional vs Simple Discriminations
The most basic type of discrimination you can make is sometimes referred to as a simple discrimination. In this type of task, there are three core elements: an antecedent, a response, and a consequence to that response. You may have heard this called the operant contingency, an "S-R-S" contingency, or even more informally as the ABC's of Behavior. Simple discriminations can come in two flavors: successive and simultaneous, which refer to how the antecedent stimuli are arranged. The "correct" behavior contacts reinforcement, whereas the incorrect behavior does not contact reinforcement.
In a conditional discrimination task, that same simple discrimination (antecedent-response-consequence) contingency exists with the addition of a second-order discriminative stimulus. A first-order discriminative stimulus serves as the primary cue to engage in a behavior, while a second-order discriminative stimulus further restricts this behavior to a more narrow set of conditions. Responses in the presence of both first and second-order discriminative stimuli (SDs) are reinforced whereas responses in the absence of first or second-order stimuli are not reinforced.
Conditional discriminations are a common component in many types of tasks that a learner might encounter at a behavioral clinic. (Green, 2001). Behavior analysts often target conditional discriminations in their treatment plans, and many common behavioral assessments, such as the VB-MAPP and the ABLLS-R, provide specific developmental milestones for a learner's conditional discrimination making skills. Throughout any given day, most neurotypical individuals could, conceivably, make thousands of conditional discriminations, from identifying the correct letters to press to spell words using a keyboard, to selecting the soft drink you want from the vending machine outside your apartment, to identifying what things you should and should not say to your significant other while they're rambling on about their favorite TV show. As success at making conditional discriminations is foundational to successful performance in many other skill areas, it's no wonder why this is such a frequent target by BCBAs and educators.
A simple discrimination task involves three core elements: an antecedent, a response, and a consequence to that response. You may have heard this called the operant contingency, or even the ABC's of Behavior. In a conditional discrimination, that same simple discrimination (antecedent-response-consequence) contingency exists with the addition of a second-order discriminative stimulus. A first-order discriminative stimulus serves as the primary cue to engage in a behavior, while a second-order discriminative stimulus further restricts this behavior to a more narrow set of conditions. Responses in the presence of both first and second-order discriminative stimuli (SDs) are reinforced whereas responses in the absence of first or second-order stimuli are not reinforced.
There exists a classic example of a conditional discrimination in animal behavior: a pigeon may learn to peck a button to receive food items as a reinforcer. However, food will only be dispensed in the presence of a green light. Over time and under typical conditions, the pigeons' behavior will come into control of both the presence of the button and the illumination of the green light. Should either not be present, the pigeon will not attempt to peck to obtain food.
A more familiar conditional discrimination that (if you're like me, anyway) proves challenging even among neurotypical adults is remembering the names of other adults. In a large office setting, you might have to contend with making dozens of conditional discriminations in the morning. When you see a co-worker, you may have to quickly recall their name for any number of reasons. Correct recall of a person's name produces social reinforcement; incorrect recall does not (and, in fact, may contact social punishment).
Incidence of Stimulus Overselectivity
The incidence rate of stimulus overselectivity has been described as very high in the population of those diagnosed with ASD. If you are someone who has worked in the field of Applied Behavior Analysis (ABA), chances are you've at some time encountered an individual who has exhibited stimulus overselectivity. Since the original formalization of the term by Lovaas & Schreibman in 1971, many analysts have reported very high incidence rates in a variety of research published in the Journal of Applied Behavior Analysis. It should be noted that, in 1971, the majority of individuals diagnosed with autism were on the more severe end of the spectrum as it is understood today. In 2013, the DSM-V redefined ASD to encompass the previous diagnoses of Asperger Syndrome, Pervasive Developmental Disorder Not Otherwise Specified, and Childhood Disintegrative Disorder. As the criteria for diagnosis broadened and our understanding of autism has refined, many individuals now meet the criteria for ASD and it is possible that the original prevalence rates suggested by early literature are out of date.
In 2015, Dr. Sarah Rieth (BCBA-D) set out to update our understanding of how common stimulus overselectivity really is with an updated examination of prevalence in young children with ASD (Rieth SR, Stahmer AC, Suhrheinrich J, Schreibman L., 2015). Dr. Rieth and her team identified 19% of children (sample n = 42) as displaying signs of stimulus overselectivity after direct testing in a match-to-sample format training exercise. 19% is significantly lower than what was previously suspected (numbers vary from paper to paper, and this author drew no consensus other than it was 'higher than 19%'). While 19% is lower, this is still nearly a 1 in 5 incidence rate! If 20% or more learners display difficulty acquiring conditional discriminations due to stimulus overselectivity, that is more than common enough to warrant serious demand for a behavioral technology to efficiently address this issue.
Reducing Stimulus Overselectivity: Some Available Technologies
In the field of behavior analysis, the term technology has a different connotation than that encountered in common parlance. In this setting, technology is used to describe an intervention or procedure (and not iPhones or fancy TVs). You, the reader, may be interested in identifying technologies available to us as applied practitioners who have clients with difficulties learning to make conditional discriminations due to stimulus overselectivity. As follows is a very brief review of some (and not all!) of the literature available in the Journal of Applied Behavior Analysis. As always, read the articles yourself for more information and to verify that my interpretation of the researcher's findings is true and correct.
Extra-Stimulus Prompts
Extra-stimulus prompting is an additional prompt presented simultaneously to the stimuli being trained. For example, pointing to a picture of the word "cat" in an array of three target words, such as "cup", "cat", and "car" would be a common strategy to teach an individual with autism to identify the correct stimuli from an array of similar items (in this case, sight words begininning with the letter c). The idea is that by pointing to the correct item, the learner may be able to naturally identify the relevant differences in the stimuli by seeing the "correct" item in comparison to the "incorrect" stimuli. This supposes a lot, and the early literature generally found this approach ineffective (Koegel and Rincover, 1976; Schreibman, 1975) - to be fair, many participants are identified as "stimulus overselective" only after extensive training with extra-stimulus prompts has been conducted to not effect.
Differential Observing Responses
A Differential Observing Response (DOR) usually involves the learner engaging in some sort of requirement to observe a part or all of the stimuli before they can respond. An example could include asking a participate to name an item out loud before selecting in from an array (Constantine and Sidman, 1975; Gutowski and Stromer, 2003), or - when participates are unable to name stimuli - using a compound matching or sorting task such as that described by Farber, Dube, Dickson in 2016. In this study, a correct "sample" is presented and the learner must find the match in an array of slightly different stimuli. For example, a sample might include the word "AT" and the learner would be tasked with identifying an identical match from an array of three words: "AT", "AG", "AF".
Within-Stimulus Modifications
Another approach involves exaggerating the relevent differences between stimuli to increase the likelihood of a learner attending. For example, a learner displaying difficulty with the sight words "hat", "cat", and "mat" may benefit from having the first letter highlighted or increased in size to make it more salient. As the student makes progress, the letters can be faded to normal systematically as the learner demonstrates mastery at each progressive step. There is evidence to suggest that within-stimulus prompts are more effective than extra-stimulus prompts (Schreibman, 1975) as the learner is actively observing a difference between the stimuli when within-stimulus prompting is used as opposed to responding to an external cue - such as a therapist's pointing gesture - when extra-stimulus prompts are used.
Schedule of Reinforcement Modifications
Thinning a schedule of reinforcement may also be effective at reducing stimulus overselectivity by reducing or eliminating the reinforcement the learner may contact for "guesswork." One simple modification would be to use a thin variable ratio schedule so that the learner continues to respond (i.e., you avoid ratio strain) but accuracy becomes more important in order to maximize contact with reinforcers. (Fisher et al., 2014; Koegel, Schreibman, Britten, and Laitinen, 1979). This may also function to increase the saliency between engaging in a given behavior in certain antecedent conditions by spacing opportunities out in time. A short delay (20-30s) following incorrect trials and a reinforcing stimulus following correct trials would create a contingency where the density of reinforcer contact is heavily dependent on accurate responding.
Conclusion
Stimulus overselectivity is a commonly identified problem in the practice of analysts working with children diagnosed with ASD. While the prevalency of stimulus overselectivity remains unclear, it has been recently estimated at 19% although further information is needed. While there are many treatment approaches available in the literature, it always comes down to each unique learner's differences. As with everything in behavior analysis, systematic analysis and data collection of a variety of the aforemention procedures may be necessary to identify the ideal intervention for your learner.