DRAGENwiki · Article
Discriminative Stimulus
Customize your experience
Sources checked. Clinical review pending.
A discriminative stimulus (SD) is something a person can see, hear or notice that signals an action can pay off right now, because that action has paid off when the signal was there and not when it was missing. [1][3]
Discriminative stimuli in 60 seconds
- It signals a payoff. An SD goes with reinforcement being available for a response when it is present, and not when it is absent. [3]
- It is learned. A signal becomes an SD because, in its presence, a response has more often been followed by reinforcement. [1]
- It needs a reason to act. An SD makes a response more likely only when the person wants, right now, what that response has brought before. That want comes from a motivating operation (MO). [1][2]
- It sets the occasion; it does not force. A reflex is pulled out by a stimulus. An SD only makes an action more likely. [1][6]
- Errors follow signals too. When a learner keeps choosing the wrong thing, a side, a position or a favorite picture may be controlling the choice. [4][5]
Jump to: What an SD is · How it is learned · SDs and motivation · Instructions · Who is in the room · The wrong signal · DRAGEN's rules · Example · Mistakes · Words to know · Need a PDF? · More from DRAGEN · Sources
This page is free, like the rest of DRAGENwiki. It builds on the four-term contingency and the main page about motivating operations, and its partner is stimulus control.
What a discriminative stimulus is
The short version. An SD is a signal that a certain action can pay off now. A ringing phone signals that saying "Hello" will reach a person. An open shop door signals that walking in will get you inside. Without the signal, the same action does not pay off. Say hello to a silent phone, and no one answers.
The research version. One paper describes an SD as a stimulus that "is associated with the availability of reinforcement for a response in its presence but not in its absence". [3] The other side of the coin is the S-delta (S∆): a stimulus that goes with reinforcement not being available for that response. [3]
A classic definition. Michael (1982) gave a definition in three parts. An SD (1) works when a certain kind of reinforcement is effective at the moment, (2) makes a certain kind of response more frequent, and (3) does so because, when the SD was present, that response was more often followed by that reinforcement. [1]
Control you can count. When people say a signal has stimulus control, they mean something you can see in the data. The same paper says that "stimulus control has been established when a response reliably occurs in the presence of an SD, but not in its absence." [3]
It is everywhere. This is not only a clinic idea. One review says that "control by antecedent stimuli, whether classified as conditional or discriminative, is ubiquitous in the natural setting." [6] Antecedent means coming before. Ubiquitous means found everywhere.
It does not force. When Skinner set operant learning apart from respondent learning, he "stressed the lack of an eliciting stimulus". [6] A tap below the knee pulls out a knee jerk; that is a reflex. An SD does not work that way. Researchers say a response is occasioned by the signal. [3] In plain words, the signal sets the occasion: it makes the response more likely, but it does not make it happen. [1]
How a signal becomes an SD
Pay off here, not there. "Stimulus control is established through discrimination training, and many everyday interactions with the environment require discriminations." [3] In discrimination training, a response pays off when the SD is present and does not pay off when it is absent. Over time, the response shows up with the signal and not without it. [3]
The learner has to notice the signal. An observing response is something a learner does "before or during a training trial" that "results in sensory contact with the discriminative stimulus". [4] Looking at the pictures and listening to the words are examples. The same authors write that the learner "must observe the relevant features of the programmed antecedent stimuli." [4]
Observing and control rise together. One review reported that stimulus control goes up and down under the same conditions as observing does. Its author concluded that more frequent and longer observing, and perhaps attention, is what separates performance during discrimination learning. [6]
Different enough to tell apart. Two features help a learner tell signals apart. "Stimulus disparity refers to the difference between stimuli and is often measured by magnitude or intensity." [3] Stimulus salience is "the extent to which an SD differs from background stimulation". [3] In plain words, disparity is how different two signals are from each other, and salience is how much a signal stands out from everything around it.
What stands out depends on the learner. Salience is not fixed. One paper calls it "a continuum along which multiple variables, such as reinforcement history or sensory sensitivity, can be manipulated". [3] In one study that paper describes, young children with autism learned to select pictures of musical instruments more efficiently when the instrument's sound, not its spoken name, was the SD. [3]
Make the payoff clear. The same paper suggests that differential reinforcement "may help increase the salience of the contingencies for each response". [3] In plain words, when the right answer and the wrong answer lead to clearly different results, the learner can more easily tell which signal goes with which result.
An SD is not a motivating operation
Two different jobs. Michael (1982) separated the discriminative job of a stimulus from its motivational job. [1] He wrote: "Operations such as deprivation have two different effects on behavior." [1] One raises how effective something is as reinforcement. The other evokes the behavior that has been followed by it before. [1] That is the job of a motivating operation. The SD's job is different: it signals where and when a behavior can pay off.
An SD needs a reason to act. Poling and colleagues write that an SD evokes behavior "only if an effective EO for the kind of reinforcement previously arranged is currently in effect." [2] An EO, or establishing operation, is an MO that makes something worth more right now.
The French fries example. In their made-up example, a child gets French fries by asking grandparents. The child is "highly unlikely to ask either grandparent for French fries soon after eating a big meal but highly likely to do so 8 hr later if no food is consumed in the interim." [2] The grandparents are the signals. The hunger is the reason to act.
Which words show up. The signal also shapes how the person asks. If the grandparent who is cooking calls them French fries, the child asks for French fries. If the grandparent is from the United Kingdom, where they are called chips, the child asks for chips. [2] The MO strongly influences whether the child asks at all. The SD strongly influences which way of asking shows up. [2]
Raise a hand or scream. The same paper describes a child who has gotten attention from one adult by raising a hand and from another adult by screaming. That child "will do neither if an EO for attention is not currently operative." [2] When attention does matter right now, which one happens "depends on which adult is present." [2]
Both, almost always. The authors conclude: "Joint control of behavior by MOs and SDs is the rule, not the exception, in everyday life and in clinical situations." [2] They add that "MOs can affect stimulus generalization around relevant SDs." [2] So how far a response spreads to similar signals can change with how much the person wants the payoff. See stimulus generalization.
Instructions are signals too
The words are the signal. When a learner is taught to follow a spoken instruction, the instruction is the SD. One teaching guide describes simple listener programs as a three-term contingency: the spoken instruction (the SD), a response such as pointing, and reinforcement. [4]
Each picture takes a turn. When a learner must point to one of several pictures, "Each of the comparison stimuli functions as the discriminative stimulus or distracter on a given trial." [4] The spoken word makes one picture the SD and the others S-deltas, but only at that moment. On another trial, the roles change. [4] This is a conditional discrimination: the right choice depends on what was said.
Fewer words, clearer signal. "Including unnecessary information in the auditory instruction introduces the risk of irrelevant features exerting stimulus control over responses." [4] In plain words, every extra word is one more thing the learner might follow instead of the key word.
Extra help can mislead. Extra-stimulus prompts add something to the SD to help a correct response, such as setting one picture closer than the others. [4] They can get in the way, because the added help, not the instruction, may end up controlling the answer. [4] The same guide says that for learners who observe irrelevant features, show overselective responding or have a history of faulty stimulus control, "extra-stimulus prompts should be avoided." [4] See prompting and stimulus prompts.
Who is in the room matters
People are signals. A parent and a therapist can be different SDs for the same behavior. [2]
A caution for functional analysis. A functional analysis (FA) tests what keeps a behavior going. Poling and colleagues walk through a made-up case. A parent has regularly given attention after a child's screaming, but a therapist has not. Then "the response is more likely to occur when the parent (an SD) is present than when the therapist is present." [2] If the parent is not there, the screaming may not show up consistently, "leading to the conclusion that the target response is not attention-maintained, which is incorrect." [2] In that case, they write, the FA would be run twice, once with the parent present and once with the therapist present. [2]
What this means for families. If a behavior happens mostly with certain people or in certain places, tell your BCBA. It can help the team plan where, and with whom, to look.
When the wrong signal takes over
Errors are not random. "Errors are behaviors under stimulus control." [5] When a learner's choices look random, something may be controlling them. The usual source is a position, such as always picking the left side (a position bias), or, when the right answer depends on what was said, one picture picked no matter what was said (a stimulus bias). [5] The paper calls a position bias "a typical side effect of discrete-trial instruction." [5]
Chance payoffs keep patterns going. If a learner always picks one side, they will still be right some of the time by chance. Those chance payoffs can strengthen the pattern. "These patterns can be adventitiously reinforced, especially during long periods of chance-level responding (including baseline)." [5] Adventitious means accidental.
Corrections can backfire. In an older study the paper describes, children could keep choosing until they picked the correct key, and the trial did not move on until they did. This built "a superstitious chain of key selections rather than a single key selection." [5] In the same study, children taught with size fading made fewer errors than children who started with small differences between the shapes. [5] In size fading, the wrong shape starts out so small that choosing it is unlikely, and it grows step by step. [5]
Keep errors low. The authors recommend "designing instruction that prevents adventitious reinforcement of maladaptive discriminated response patterns by reducing errors during acquisition." [5] In plain words, teach in a way that keeps errors low while a skill is new, so chance payoffs cannot build a wrong pattern. See errorless learning.
Signals we give without meaning to. Faulty stimulus control means that "some erroneous aspect of the instructional environment controls responding". [4] The teacher is part of that environment. One guide lists cues instructors can give by accident:
- a lilt at the end of "Stand up" and a drop at the end of "Sit down" [4]
- a glance at the bookshelf while saying "Get the book" [4]
- setting the correct picture down first or last [4]
- keeping the hands nearer to the correct or the incorrect choice [4]
Because "people are unlikely to detect subtle patterns evolving in their own behavior", the authors say instructors should monitor procedural integrity. [4] They suggest placing the hands in the same position after presenting the pictures, and rotating the spoken and visual stimuli across trials in a balanced way. [4] See procedural fidelity.
Too narrow a signal. "Stimulus overselectivity or restricted stimulus control occurs when isolated stimulus features or components are controlling behavior". [3] For example, a learner might pick a picture by a small mark in its corner instead of by what it shows. In one approach the paper describes, learners first matched a two-part sample to the choice with the exact same two parts. This differential observing response was used to reduce overselective responding. [3] See overselectivity.
Change the signal, not the learner
DRAGEN's rules for signals. These are page rules from the four pillars, not findings:
- An instruction is a signal, not an obedience test. If a learner does not respond, check the signal, the payoff, the timing and the person's needs first.
- Check the reason to act. If a signal stops working, ask whether the person wants what the response brings right now.
- Make the right signal clear. Use signals that differ enough and stand out for this learner. Say only the words that matter.
- Check yourself, kindly. Ask a colleague, or watch a consented video, to catch glances, tone, hand position and the order you set things down.
- Keep errors low. Help early, then fade the help on a plan, so chance payoffs do not build a wrong pattern.
- Answer a no. A no, a turn-away or a question gets a response, not more repetitions of the instruction.
- A BCBA reviews the plan. Any plan to build or change stimulus control is reviewed by a BCBA.
A no is information. Saying no, asking a question, or taking back a yes tells you about the task, the timing or the person's needs. It is not noncompliance, and it is not a treatment target.
Safety is a separate question. If someone could get hurt right now, handle safety first. Do not use safety as a reason to stop listening to dissent.
This page is general information, not medical or legal advice. A BCBA reviews every clinical decision.
A made-up example: Learner A
Learner A is a made-up child who is learning to point to a picture when an adult names it.
What the team saw. At the clinic, Learner A pointed to the named picture most of the time. At home, when a parent asked, Learner A often pointed to the picture on the left.
What they checked first. Was Learner A hungry, tired or done for the day? Was the activity still worth doing? The team checked the reason to act before the signal.
What a video showed. With the family's written consent, the BCBA watched a short clinic video. The instructor glanced at the correct picture just before asking. Learner A had learned to follow the glance, not the word. At home, with no glance to follow, the left side was the next best guess, and it paid off some of the time.
What changed. The instructor kept their eyes and hands in the same place on every trial. The team rotated where each picture sat in a balanced way, used the same tone for every word, and kept the instruction short. They helped early and faded the help, so Learner A could be right while learning.
What Learner A said. Partway through one session, Learner A signed all done. The instructor stopped and said, "Thanks for telling me." The session ended early. The team wrote it down as information about session length, not as a problem.
What happened next. In this made-up story, Learner A began pointing to the named picture with both the instructor and the parent. Real learners differ, and a BCBA reviews every plan.
Common mistakes
- "An SD makes a person respond." It makes a response more likely when the person has a reason to act. It does not force the response. [2][6]
- "They do it at the clinic, so they know it." The response may be under the control of a person, a place or a cue the teacher did not mean to give. [2][4]
- "Random answers mean the learner is not trying." Errors are behavior under stimulus control. Look for a side, a position or a favorite picture. [5]
- "More words make an instruction clearer." Extra words add the risk that the wrong part controls the answer. [4]
- "An SD and an MO are the same thing." An SD signals that a response can pay off. An MO changes how much the payoff is worth right now. [1][2]
- "Let them keep trying until they get it right." In one older study, a correction procedure like that built a habit of pressing several keys in a row instead of one. Keeping errors low from the start helps. [5]
Words to know
- Discriminative stimulus (SD): a stimulus that goes with reinforcement being available for a response when it is present, and not when it is absent. Also called S+. [3]
- S-delta (S∆): a stimulus that goes with reinforcement not being available for a response. Also called S−. [3]
- Stimulus control: a response reliably happens when the SD is present and not when it is absent. [3]
- Discrimination training: teaching that builds stimulus control. [3]
- Observing response: something the learner does, like looking, that puts them in sensory contact with the SD. [4]
- Stimulus disparity: how different two stimuli are, often measured by magnitude or intensity. [3]
- Stimulus salience: how much an SD differs from the background. [3]
- Faulty stimulus control: a wrong feature of the teaching setting controls responding. [4]
- Position bias: choosing by place in the array. [5]
- Stimulus bias: choosing one picture no matter what was asked. [5]
- Motivating operation (MO): an event that changes how much a consequence is worth right now, and how often the behavior related to it happens. [2]
Need a PDF?
Create it here, for free: the DRAGEN app. Use it at your next session or family meeting.
Or print one we already made, also free:
- Trial-by-trial data sheet (PDF)
- Cold probe sheet (PDF)
- 4-term data sheet (PDF)
- Treatment integrity checklist (PDF)
- Generalization plan and probe sheet (PDF)
- Every free PDF
More from DRAGEN
- Lesson: How to take 4-term contingency data
- Lesson: How to help without doing it for them: prompting and prompt fading
- Lesson: How to make a skill work everywhere: generalization
- Lesson: How to tell the four functions of behavior apart
- Free app for parents: DRAGENparent
- BCBA care: DRAGENtherapy
- Books: DRAGENbooks
- Talk with Chase: Contact
- Home: dragen.cloud
Related DRAGENwiki pages: four-term contingency · motivating operations · establishing operation · S-delta · stimulus control · stimulus discrimination · stimulus generalization · conditional discrimination · faulty stimulus control · overselectivity · three-term contingency · functional analysis · prompting · errorless learning · procedural fidelity · right to say no
What remains unknown
- You cannot see control directly. Stimulus control shows up as a difference in responding with and without the signal. [3][6]
- Salience differs by learner. Reinforcement history and sensory sensitivity both play a part, so what stands out for one learner may not for another. [3]
- Private signals. A thought or an image might work as an SD. One paper offers such an example and says the analysis "is difficult to test, but that does not render it false or useless." [2]
- Small or second-hand evidence. The paper on accidental payoffs centers on one learner. [5] The musical-instrument study is known here only as another paper describes it. [3]
- This page was last checked on 2026-11-12.
Sources
- Michael, J. (1982). Distinguishing between discriminative and motivational functions of stimuli. Journal of the Experimental Analysis of Behavior, 37(1), 149–155. https://doi.org/10.1901/jeab.1982.37-149 (abstract read at https://pmc.ncbi.nlm.nih.gov/articles/PMC1333126/)
- Poling, A., Lotfizadeh, A. D., & Edwards, T. L. (2020). Motivating operations and discriminative stimuli: Distinguishable but interactive variables. Behavior Analysis in Practice, 13(2), 502–508. https://doi.org/10.1007/s40617-019-00400-2 (read at https://pmc.ncbi.nlm.nih.gov/articles/PMC7314879/)
- Halbur, M. E., Caldwell, R. K., & Kodak, T. (2021). Stimulus control research and practice: Considerations of stimulus disparity and salience for discrimination training. Behavior Analysis in Practice, 14(1), 272–282. https://doi.org/10.1007/s40617-020-00509-9 (read at https://pmc.ncbi.nlm.nih.gov/articles/PMC7900359/)
- Grow, L., & LeBlanc, L. (2013). Teaching receptive language skills: Recommendations for instructors. Behavior Analysis in Practice, 6(1), 56–75. https://doi.org/10.1007/BF03391791 (read at https://pmc.ncbi.nlm.nih.gov/articles/PMC3680153/)
- Saunders, K. J., Hine, K., Hayashi, Y., & Williams, D. C. (2016). Adventitious reinforcement of maladaptive stimulus control interferes with learning. Behavior Analysis in Practice, 9(3), 223–229. https://doi.org/10.1007/s40617-016-0131-2 (read at https://pmc.ncbi.nlm.nih.gov/articles/PMC4999367/)
- Dinsmoor, J. A. (1995). Stimulus control: Part I. The Behavior Analyst, 18(1), 51–68. https://doi.org/10.1007/BF03392691 (abstract read at https://pmc.ncbi.nlm.nih.gov/articles/PMC2733673/)