# Discriminative Stimulus (SD): Definition, Examples & S-Delta

> A discriminative stimulus (SD) is a cue in whose presence a behavior has been reinforced. Definition, S-delta, examples, SD vs. CS and MO, how to train one.

- Source: https://operantconditioning.com/discriminative-stimulus/
- Author: Ryan Martinson (https://operantconditioning.com/about/)
- Publisher: Operant Conditioning Inc.
- Published: 2026-09-13 · Updated: 2026-09-13
- License: https://operantconditioning.com/terms/#copyright (quote with attribution and a link; free to reproduce for non-commercial teaching)

---

*Antecedents · Stimulus control*

A phone rings and you pick it up; the light goes off and the rat stops pressing. Nothing forced either response. The discriminative stimulus is the antecedent that makes a behavior likely without compelling it — here is how it is defined, how it is built, and how to tell it from the two things it is most often confused with.

> **Definition**
>
> A discriminative stimulus, written SD and pronounced "ess-dee," is an antecedent stimulus in whose presence a response has been reinforced and in whose absence it has not, and which, because of that history, raises the probability of the response when it is present. Its counterpart, the **S-delta** (SΔ), is a stimulus in whose presence the same response has gone unreinforced.[1]
>
> B. F. Skinner introduced the term and the notation in *The Behavior of Organisms* (1938). An SD does not elicit behavior the way a puff of air elicits a blink; it sets the occasion for it, and only because of what has followed the behavior in its presence before.[1][2]

**In brief**

- A discriminative stimulus is defined by history, not appearance: a cue is an SD for a behavior only if that behavior has been reinforced in its presence and not in its absence.
- An SD evokes an operant rather than eliciting a reflex, and it signals that reinforcement is *available*; a motivating operation is the different thing that changes how much the reinforcer is *worth*.
- To put a behavior under an SD, reinforce it in the presence of the cue, withhold reinforcement in its absence, and keep the difference consistent until responding diverges.

## How a discriminative stimulus works

Skinner's original demonstration is still the cleanest. A rat presses a lever for food, and the food is then made available only while a light is on: presses in the light produce a pellet, presses in the dark produce nothing. At first the rat presses at about the same rate either way. Over sessions the rates separate, until the rat presses briskly when the light comes on and barely touches the lever when it goes off. The light has become a discriminative stimulus, and the response is a **discriminated operant**.[1]

Three things in that description carry the definition. The SD is an antecedent, present before the response rather than delivered after it. It works by *changing probability*: the rat can press in the dark and sometimes does, and the gap between the two rates measures how much control the light has. And its power is borrowed entirely from the consequence. Take the food away in the light and the light stops mattering within a few sessions.[1][2]

### Notation: SD, SΔ, S+, S−, and SDp

Skinner wrote the discriminative stimulus as SD and the stimulus correlated with extinction as SΔ, "S-delta."[1] The generalization literature usually writes the pair as S+ and S−. Applied texts add SDp for a stimulus in whose presence a response has been *punished* and in whose absence it has not.[3] In each case the superscript names the consequence correlated with the stimulus.

### The three-term contingency

The SD is the first term of the [three-term contingency](https://operantconditioning.com/abc-model/) — in the presence of SD, response R produces reinforcer SR — which Skinner made the basic unit of analysis in *Science and Human Behavior*.[2] The consequence explains why the behavior exists; the discriminative stimulus explains when and where it appears.

## How a discriminative stimulus is established

A stimulus becomes an SD through **discrimination training**: the response is reinforced in the presence of one stimulus and undergoes [extinction](https://operantconditioning.com/extinction/) in the presence of another until responding in the two conditions diverges.[1] The procedure is [differential reinforcement](https://operantconditioning.com/differential-reinforcement/) applied to the antecedent; the process it produces, responding differently in the two conditions, is discrimination. Both halves are necessary. Reinforcement in the light builds the response; extinction in the dark is what attaches it to the light.

The second half is easy to forget, and one experiment shows what happens without it. Jenkins and Harrison trained pigeons to peck a key with a tone sounding continuously through every session. The tone was present for every reinforced peck, yet when the birds were later tested with other pitches they pecked about equally at all of them: the tone controlled nothing. Birds for whom the tone had been on during reinforcement and off during extinction responded sharply to the training pitch and little to the others.[4] A stimulus that predicts no difference in consequences gains no control, however often it accompanies reinforcement.

Ordinary discrimination training involves errors: the learner responds to the SΔ, is not reinforced, and gradually stops. Herbert Terrace showed in 1963 that the errors can be engineered out. He trained pigeons to peck a red key and not a green one, but introduced the green key from the first session so dim and so brief that the birds never pecked it, then raised its brightness and duration in small steps. Birds trained this way made few or no errors, where conventionally trained birds made thousands.[5] The SD was established just as surely; only the way the SΔ was introduced changed. [Errorless learning and its side effects ›](https://operantconditioning.com/stimulus-control/)

Dinsmoor's review adds a requirement the contingency alone does not guarantee: the organism has to actually look at or listen to the stimulus. A cue the learner never attends to cannot become an SD.[6]

The stimulus is not special; the history is

The same light is an SD for a rat that has been fed for pressing in its presence and nothing at all for a rat that has not. Before calling a cue an SD, ask what the behavior has produced in its presence and in its absence. If the answer is "the same thing," it is not an SD yet.

## Examples of discriminative stimuli

In each row the behavior has paid in the presence of the SD and not in its absence.

| Setting | SD | Behavior | What has followed it | SΔ |
| --- | --- | --- | --- | --- |
| Everyday | Phone ringing | Picking it up and saying hello | A voice on the line | A silent phone |
| Driving | Green light | Pressing the accelerator | Getting through the intersection | Red light |
| Classroom | Teacher's raised hand (the quiet signal) | Stopping talking and raising a hand | Praise; the activity the class was waiting for begins | Teacher's hand down, working with a small group |
| Dog training | The word "sit" | Sitting | A treat, or the leash going on | Any other word, or silence |
| Retail | A lit "Open" sign | Pulling the door | The door opens | A dark sign |
| Social | A particular friend | Telling a particular kind of joke | Laughter | Your manager, who has never laughed at it |

The same event is usually an SD for one behavior and an SΔ for another: the red light is an SΔ for accelerating and an SD for braking; the raised hand is an SD for hand-raising and an SΔ for calling out. A stimulus has a discriminative function *for a particular response*, and the function has to be stated with the response attached.[2]

The "particular friend" row deserves a pause. Skinner devoted a chapter of *Verbal Behavior* to the audience as a discriminative stimulus: the person you are speaking to controls not only whether you speak but what you say, because different listeners have reinforced different things.[7] The child who "behaves differently for the substitute" is discriminating between two people with two histories of consequences.

## Discriminative stimulus vs. conditioned stimulus

The nearest neighbor is the conditioned stimulus (CS) of classical conditioning. Both are learned antecedents, but they are established by different histories and do different things.

A CS acquires its function by being paired with an unconditioned stimulus — a tone followed by food, whatever the animal does in between. Rescorla's summary of the modern view is that the animal learns the *predictive relation* between the two events: a CS works to the extent that it carries information about the unconditioned stimulus, and contiguity without prediction produces little conditioning.[8] Nothing the animal does is part of that relation. The tone predicts food whether the dog salivates or not, and the salivation that comes to follow the tone is *elicited*, on a reflex model, without depending on its own consequences.

An SD acquires its function through a contingency that runs through the behavior. The light predicts food only *if the rat presses*; without the press there is no pellet, light or no light. The response is emitted in the stimulus's presence at a rate the stimulus modulates, and Skinner's formulation was that the SD sets the occasion for the response rather than eliciting it.[1][2] An elicited response follows its stimulus with roughly fixed form and latency; an evoked operant varies in speed and force and can fail to occur at all.

| Aspect | Discriminative stimulus (SD) | Conditioned stimulus (CS) |
| --- | --- | --- |
| **Established by** | Reinforcement of a response in its presence and extinction in its absence | Predicting an unconditioned stimulus, regardless of behavior |
| **What it does** | Evokes an operant: raises its probability | Elicits a respondent: triggers a conditioned reflex |
| **Does the behavior matter?** | Yes; the reinforcer depends on the response | No; the unconditioned stimulus comes anyway |
| **Example** | Lit key: pecking now produces grain | Tone before food: salivation follows the tone |

One event often carries both functions: the click of the food magazine is a CS that elicits approach, a [conditioned reinforcer](https://operantconditioning.com/glossary/#conditioned-reinforcer) for the press before it, and an SD for going to the tray. Ask which function an event serves for which response. [Operant vs. classical conditioning in depth ›](https://operantconditioning.com/operant-vs-classical-conditioning/)

## Discriminative stimulus vs. motivating operation

The confusion that matters more in practice is with the motivating operation. In 1982 Jack Michael pointed out that "SD" was being used for two different antecedent functions. One is discriminative: the stimulus is correlated with the *availability* of a reinforcer, obtainable in its presence and withheld in its absence. The other is motivational: the event changes the *effectiveness* of a reinforcer, how much it is worth right now, and in doing so evokes the behavior that has produced it. He called the second an **establishing operation** and argued that the two should not share a name.[9]

The test is what happens to the reinforcer in the stimulus's absence. A drinking fountain is an SD for walking over and pressing the bar: water has been available there and not elsewhere. A bag of salted peanuts is not an SD for the same behavior, though it makes the behavior far more likely; water was no less available before the peanuts, what changed is how much it is worth. Likewise, the vending machine says a snack is available; six hours without food says it is worth having. Both are usually needed before the behavior occurs.[9][10]

Michael drew a consequence that still surprises people. The onset of a painful stimulus is routinely called an SD for escape, but it fails the test: the reinforcer for escape, the pain ending, is not merely unavailable before the pain starts, it does not exist. Pain onset is an establishing operation for its own removal. The SD for escape, if there is one, is whatever signals that a particular response will work, such as the lever that has ended the shock before.[9]

Michael's 1993 paper set out the two effects every establishing operation has, on the value of a consequence and on the current frequency of behavior that has produced it.[10] Laraway and colleagues completed the vocabulary in 2003: [motivating operation](https://operantconditioning.com/glossary/#motivating-operation) is the umbrella term, an *establishing* operation raises a reinforcer's effectiveness and an *abolishing* operation lowers it, and the behavioral effect is *evocative* or *abative*.[11]

| Aspect | Discriminative stimulus (SD) | Motivating operation (MO) |
| --- | --- | --- |
| **What it changes** | The probability of the response, given a history of differential reinforcement | The value of the reinforcer, and with it the probability of the response |
| **What it signals** | That the reinforcer is available now | Nothing about availability; it makes the reinforcer worth more or less |
| **Test** | Was the reinforcer obtainable in its presence and withheld in its absence? | Is the reinforcer worth more or less after this event? |
| **Example** | The drinking fountain; the vending machine; the light above the lever | Salted peanuts; six hours without food; food deprivation before a session |

A cue with an impeccable history does nothing if the reinforcer it signals is currently worth nothing, which is why "sit" fails after a bowl of treats. When a cue stops working, ask whether the problem is the cue's history or the reinforcer's value.

## Discrimination, generalization, and natural concepts

An SD rarely controls in isolation; control spreads to stimuli that resemble it. Guttman and Kalish reinforced pigeons for pecking a key lit with a single wavelength and then, in extinction, presented nearby wavelengths. Responding was highest at the training wavelength and fell away smoothly on either side, a generalization gradient.[12] Discrimination training steepens the gradient, and when the SΔ lies close to the SD on the same dimension it moves the peak away from the SΔ, the [peak shift](https://operantconditioning.com/glossary/#peak-shift) Hanson demonstrated in 1959.[13] Dinsmoor's review covers the mechanics, and both effects are treated in full on the [stimulus control page](https://operantconditioning.com/stimulus-control/).[6] For the definition, the point is that "the SD" is the center of a region of stimuli that share its control, and training sets how sharp that region is.

At the other extreme, the "stimulus" can be a category no single feature defines. Herrnstein, Loveland, and Cable reinforced pigeons for pecking at color slides that contained a tree, or a body of water, or one particular young woman, and not at slides that did not. The birds learned each discrimination and transferred it to slides they had never seen.[14] The SD here is "tree," a class with no common wavelength, shape, or size: the discriminative stimulus is as abstract as the history of reinforcement makes it.

## How to bring a behavior under a discriminative stimulus

The procedure is the same whether the learner is a rat, a dog, a child, or you. If the behavior does not yet occur, [shape](https://operantconditioning.com/shaping/) it first; a cue cannot be attached to a response that is never emitted.

1. **Choose the cue.** Pick a stimulus that is distinct, easy to perceive, under your control to present and withhold, and present where you want the behavior. "Sit" said once in a normal voice; a raised hand; one desk at one hour. Avoid a cue that is already an SD for something else.
2. **Reinforce the behavior in the presence of the cue.** Present the cue, wait for the response, reinforce immediately. Early on, reinforce every correct response and, if needed, use a [prompt](https://operantconditioning.com/glossary/#prompt) to get the response going, then fade the prompt so the cue is what remains.[3]
3. **Ensure extinction in the cue's absence.** This is the half most people skip. If the dog is fed for sitting whenever it sits, the word "sit" predicts nothing and will control nothing.
4. **Add the SΔ explicitly.** Present other stimuli — other words, a different gesture, the same word from a different posture — and withhold reinforcement when the behavior follows them. This sharpens control from "anything the trainer says" to "this word." Introduce the SΔ gradually to keep errors low, as Terrace did.[5]
5. **Thin the schedule and vary the setting.** Once responding is reliable in the cue's presence and rare in its absence, move to an [intermittent schedule](https://operantconditioning.com/schedules-of-reinforcement/) so the behavior persists, and train in other rooms, with other people, and around distractions so that control belongs to the cue rather than the kitchen.[3]

The same steps describe how a [dog trainer](https://operantconditioning.com/dog-training/) adds a cue and how a [classroom](https://operantconditioning.com/classroom/) quiet signal is taught in the first week of school. Reinforcement in the cue's presence is what people remember to arrange; extinction in its absence is what makes the cue work.

## Common mistakes

- **Calling any cue an SD.** A stimulus is a discriminative stimulus for a response only if that response has been differentially reinforced in its presence. A sign nobody has ever followed, an instruction never backed by a consequence, or a light that is always on is a stimulus, not an SD.[1][9]
- **Confusing the SD with a prompt.** A prompt is a supplementary antecedent added because the natural SD does not yet evoke the behavior: the pointed finger, the "What do you say?", the hand guiding the dog into a sit. It is meant to be faded; the SD is meant to stay.[3]
- **Confusing the SD with a motivating operation.** Hunger, thirst, pain, and boredom make behavior more likely without signaling that anything is available. If removing the event would change only how much the reinforcer is wanted, not whether it can be obtained, it is an MO.[9]
- **Treating the SΔ as a punisher.** An SΔ signals that the response will go unreinforced, not that it will be punished. Extinction and punishment are different procedures with different side effects.
- **Assuming the cue you intended is the cue that controls.** Learners discriminate whatever actually predicts reinforcement, which is often the trainer's posture, the treat pouch, or the room rather than the word. If the behavior vanishes when those change, they were the SD.

## What the evidence does not show

The literature on discriminative stimuli is among the oldest and most replicated in psychology, and its limits are worth stating.

- **It does not show that stimuli control behavior without a consequence history.** Every demonstration of an SD depends on differential reinforcement, and the control disappears when the differential is removed. A claim that a cue "triggers" a habit on its own is a claim about a history.[1][4]
- **It does not show that an SD works regardless of motivation.** A discriminative stimulus evokes a response only when the reinforcer it signals currently has value; under an abolishing operation the same stimulus evokes nothing.[9][11]
- **It does not show that the organism learned the stimulus you had in mind.** Generalization tests routinely reveal control by a feature the experimenter did not intend, and a pigeon can respond to a category with no single defining feature. What was learned is inferred from the gradient, not observed.[12][14]
- **It does not settle how much of an everyday cue's power is discriminative.** Many everyday stimuli carry eliciting and evoking functions at once, and the experiments do not say which is doing more of the work when your phone buzzes.[8]

## Key takeaways

- A discriminative stimulus is an antecedent in whose presence a response has been reinforced and in whose absence it has not. The definition is a history; no stimulus is an SD on its own.
- An SD evokes an operant by raising its probability; it does not elicit a reflex. A conditioned stimulus predicts an unconditioned stimulus regardless of behavior; an SD works through a contingency that runs through the behavior.
- An SD signals that a reinforcer is available; a motivating operation changes how much it is worth. The drinking fountain and the salted peanuts both make you drink, and only the first is an SD.
- Discrimination training — reinforcement in the cue's presence, extinction in its absence — establishes an SD, and the extinction half is what makes the cue predictive.
- Control spreads to similar stimuli and can be trained onto a category as abstract as "tree." Learners discriminate whatever actually predicts reinforcement, not necessarily the cue the teacher intended.
- To put behavior under an SD: choose the cue, reinforce only in its presence, ensure extinction in its absence, add the SΔ explicitly, then thin the schedule and vary the setting.

### Check yourself

**A dog sits every time its owner says "sit" in the kitchen, and also sits every time the owner reaches toward the treat pouch, whether or not the word is said. What is the SD?**

The reach toward the pouch, or the pouch itself. The word has never predicted a difference in consequence that the pouch did not already predict, so control has settled on the pouch. To make "sit" the discriminative stimulus, the owner needs to reinforce sits that follow the word alone and stop reinforcing sits that follow the reach without the word.

**A rat is shocked through the floor, and pressing a lever turns the shock off. A textbook calls the shock onset the SD for pressing. Is that right?**

By Michael's analysis, no. An SD is a stimulus in whose presence the reinforcer has been available and in whose absence it has been withheld. Before the shock, its removal is not withheld; it does not exist. Shock onset is an establishing operation: it makes shock termination a reinforcer and evokes the behavior that has produced it. If there is an SD for pressing, it is the lever, which signals which response will work.

**A sign in the office kitchen says "Please wash your own mug." Nobody does. Is the sign a discriminative stimulus for mug-washing?**

No. An SD is defined by a history of differential reinforcement, and mug-washing has never been reinforced in the sign's presence more than in its absence; the sign predicts no difference in consequence. It is a stimulus, perhaps a prompt, but not yet a discriminative stimulus. It could become one if washing in its presence began to produce something that not washing did not.

## Frequently asked questions

**What is a discriminative stimulus in simple terms?**

A discriminative stimulus is a cue that tells you a behavior will pay off right now, because in the past that behavior has been rewarded when the cue was present and not when it was absent. A ringing phone is a discriminative stimulus for answering; a lit "Open" sign is one for pulling the door. The cue does not force the behavior; it makes it more likely.

**What is the difference between an SD and an S-delta?**

An SD (discriminative stimulus) is a stimulus in whose presence a response has been reinforced, so the response becomes more likely when it appears. An S-delta is a stimulus in whose presence the same response has gone unreinforced, so the response becomes less likely. The lit "Open" sign is the SD for pulling the door; the dark sign is the S-delta. Discrimination training alternates the two.

**What is an example of a discriminative stimulus?**

The word "sit" for a trained dog: sitting after the word has produced treats, and sitting at other times has not, so the word raises the probability of sitting. Others: a green light for accelerating, a teacher's raised hand for going quiet, a lit key for a pigeon's peck, and a particular friend for a particular joke. In each case the behavior has been reinforced in the cue's presence and not in its absence.

**What is the difference between a discriminative stimulus and a conditioned stimulus?**

A conditioned stimulus, from classical conditioning, gets its power by predicting an unconditioned stimulus regardless of what the organism does, and it elicits a reflex such as salivation. A discriminative stimulus gets its power from a contingency that runs through the behavior, since the reinforcer arrives only if the response occurs, and it evokes an operant by raising its probability rather than triggering it.

**What is the difference between a discriminative stimulus and a motivating operation?**

A discriminative stimulus signals that a reinforcer is available: the drinking fountain, the vending machine, the lit sign. A motivating operation changes how much the reinforcer is worth: salty food, hours without eating, a hot day. Jack Michael drew the distinction in 1982. The test is whether removing the event would change the reinforcer's availability (discriminative) or only its value (motivational).

**Is a discriminative stimulus the same as a prompt?**

No. A prompt is an extra antecedent added because the natural discriminative stimulus does not yet evoke the behavior, such as pointing at the answer or guiding a dog into a sit. Prompts are meant to be faded so that control transfers to the discriminative stimulus, which stays. If the learner still needs the prompt, the discriminative stimulus has not been established.

**How is a discriminative stimulus established?**

Through discrimination training: the behavior is reinforced when the stimulus is present and not reinforced when it is absent, over many repetitions, until the behavior occurs far more often in its presence. Both halves are necessary. A stimulus that is present during reinforcement and during extinction alike predicts nothing and gains no control, which Jenkins and Harrison demonstrated with pigeons and a continuous tone.

**Who introduced the term discriminative stimulus?**

B. F. Skinner, in The Behavior of Organisms (1938), where he described the discriminated operant and introduced the symbols SD and S-delta. He held that the discriminative stimulus sets the occasion for a response rather than eliciting it, which separated operant discrimination from Pavlov's conditioned reflex. Jack Michael later (1982) separated the discriminative function of antecedent events from their motivational function.

## References

1. Skinner, B. F. (1938). *The Behavior of Organisms: An Experimental Analysis*. Appleton-Century.
2. Skinner, B. F. (1953). *Science and Human Behavior*. Macmillan.
3. Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). *Applied Behavior Analysis* (3rd ed.). Pearson.
4. Jenkins, H. M., & Harrison, R. H. (1960). Effect of discrimination training on auditory generalization. *Journal of Experimental Psychology, 59*(4), 246–253.
5. Terrace, H. S. (1963). Discrimination learning with and without "errors." *Journal of the Experimental Analysis of Behavior, 6*(1), 1–27.
6. Dinsmoor, J. A. (1995). Stimulus control: Part I. *The Behavior Analyst, 18*(1), 51–68.
7. Skinner, B. F. (1957). *Verbal Behavior*. Appleton-Century-Crofts.
8. Rescorla, R. A. (1988). Pavlovian conditioning: It's not what you think it is. *American Psychologist, 43*(3), 151–160.
9. Michael, J. (1982). Distinguishing between discriminative and motivational functions of stimuli. *Journal of the Experimental Analysis of Behavior, 37*(1), 149–155.
10. Michael, J. (1993). Establishing operations. *The Behavior Analyst, 16*(2), 191–206.
11. Laraway, S., Snycerski, S., Michael, J., & Poling, A. (2003). Motivating operations and terms to describe them: Some further refinements. *Journal of Applied Behavior Analysis, 36*(3), 407–414.
12. Guttman, N., & Kalish, H. I. (1956). Discriminability and stimulus generalization. *Journal of Experimental Psychology, 51*(1), 79–88.
13. Hanson, H. M. (1959). Effects of discrimination training on stimulus generalization. *Journal of Experimental Psychology, 58*(5), 321–334.
14. Herrnstein, R. J., Loveland, D. H., & Cable, C. (1976). Natural concepts in pigeons. *Journal of Experimental Psychology: Animal Behavior Processes, 2*(4), 285–302.

## About the author

Ryan holds a master's degree from UCLA, where he studied animal behavior in Daniel Blumstein's lab and was part of the university's Evolutionary Medicine Program, which applies findings from evolutionary biology and animal behavior to human health. He founded Operant Conditioning Inc. and built the Operant habit app (https://operantconditioning.com/app/). Every page here is written from the primary literature and cites it. How pages are checked: https://operantconditioning.com/about/#editorial-standards

## Related

- [Stimulus control](https://operantconditioning.com/stimulus-control/): Discrimination, generalization gradients, peak shift, and errorless learning in depth.
- [The ABC model](https://operantconditioning.com/abc-model/): The three-term contingency the discriminative stimulus is the first term of, plus motivating operations.
- [Operant vs. classical conditioning](https://operantconditioning.com/operant-vs-classical-conditioning/): Why a stimulus that evokes is not a stimulus that elicits.