The big idea
A learner shows stimulus discrimination when they respond one way in the presence of one stimulus and differently, or not at all, in the presence of another. A driver brakes at a red light and keeps going at a green one. A toddler says “dog” to dogs and “cat” to cats. The response changes when the antecedent changes.
Discrimination is not a talent the learner brings. It comes from a history of differential reinforcement: the response paid off in the presence of one stimulus and did not pay off (or paid off less) in the presence of another. Once that history has done its work, the stimulus controls the behavior, which is the relation B.12 calls stimulus control. Discrimination is how you see that control.
Almost every skill you teach is a discrimination. Reading, following directions, answering questions and greeting the right person at the right time all depend on responding to some stimuli and not to others.
Breaking it down
The SD and the S-delta
A discriminative stimulus (SD) is a stimulus in whose presence a response has been reinforced. An S-delta (SΔ) is a stimulus in whose presence that same response has not been reinforced. Discrimination shows up as a difference between the two: responding is frequent when the SD is present and rare when the S-delta is present.
An S-delta signals that reinforcement is unavailable. It does not signal that something bad will happen. A stimulus correlated with punishment for a response is a different thing (some texts call it a discriminative stimulus for punishment, written SDp).
Discrimination training
Discrimination training is the procedure that produces discrimination. Reinforce the response when the SD is present, withhold reinforcement when the S-delta is present, and repeat across many trials until the learner responds reliably differently. In a simple discrimination, one stimulus signals that one response will pay off. In a conditional discrimination, which response pays off depends on another stimulus, as when a learner touches the picture that matches the word the teacher says. How to teach both is covered in G.6. B.13 asks you to recognize the outcome.
Generalization gradients
Train a response in the presence of one stimulus, then test stimuli that differ from it along one dimension, such as pitch or color. Plot responding against stimulus value and you get a generalization gradient. In a classic study, Guttman and Kalish (1956) trained pigeons to peck a key lit with one wavelength and then tested others. Pecking peaked at the trained wavelength and fell off as the test colors moved away from it.
The shape of the gradient tells you how much discrimination there is:
- Steep and narrow: responding is limited to stimuli very close to the SD. Strong discrimination, tight stimulus control.
- Flat and wide: responding spreads across many values. Little discrimination, a lot of generalization (see B.14).
Discrimination training against a nearby S-delta makes the gradient steeper.
One continuum, two ends
Discrimination and generalization are relative. Responding the same way to everything is generalization; responding only to the exact training stimulus is discrimination. Good teaching aims for the right balance. A child who says “dog” to every four-legged animal has generalized too broadly. A child who says “dog” only to the family beagle has discriminated too narrowly.
Seeing it in practice
| Setting | SD | S-delta | Discrimination is shown when |
|---|---|---|---|
| Clinic | Picture of a cup with “What is it?” | Picture of a spoon with “What is it?” | Kai says “cup” to the cup picture and not to the spoon picture. |
| Classroom | Teacher says “Turn and talk” | Teacher is giving a lecture | Students talk to partners after “Turn and talk” and stay quiet during the lecture. |
| Home | A parent’s unlocked phone | The same phone, locked | Valentina, age 4, taps the phone to play videos when it is unlocked and rarely when it is locked. |
| Workplace | A weekday | A weekend | Layla texts her coworker Rashid with work questions on weekdays, when he answers, and almost never on weekends, when he does not. |
In every row the response stays the same. What changes is the antecedent and the reinforcement history attached to it. If the learner responded the same way in both columns, there would be no discrimination to see.
Common mix-ups
- Procedure vs. outcome. Discrimination training is what you do. Stimulus discrimination is what the learner shows afterward.
- An S-delta is not a punisher. It signals that reinforcement is unavailable. Nothing aversive has to be present.
- Discrimination vs. stimulus control. They describe the same relation from two angles. Stimulus control is the relation; discrimination is the differential responding that demonstrates it.
- An antecedent that changes value is not an SD. If a learner asks for water more after eating salty pretzels, the pretzels changed how valuable water is. That is a motivating operation (B.16, B.17), not a discrimination.
- No comparison, no discrimination. If every trial presents the cup, the learner can say “cup” without attending to anything. You need trials with the S-delta to show discrimination.
Exam tip: Look for two antecedent conditions and a difference in responding between them, with reinforcement in one and not the other. A scenario with one stimulus, one response and no comparison is probably testing something else.
Quick recap
- Stimulus discrimination is responding differently in the presence of different stimuli.
- It comes from differential reinforcement: reinforced when the SD is present, not reinforced when the S-delta is present.
- Discrimination training is the procedure; discrimination is the outcome; stimulus control is the relation.
- A steep generalization gradient means strong discrimination; a flat one means broad generalization.
- An S-delta signals no reinforcement, not punishment.
Further reading
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied behavior analysis (3rd ed.), chapter 17.
- Guttman, N., & Kalish, H. I. (1956). Discriminability and stimulus generalization. Journal of Experimental Psychology, 51(1), 79-88.
Key terms
- Stimulus discrimination (Discrimination)
- Responding differently in the presence of different stimuli. It results from a history in which a response was reinforced in the presence of one stimulus and not reinforced, or reinforced less, in the presence of another.
- Discrimination training
- Reinforcing a response in the presence of one stimulus, or one sample-comparison relation, and withholding reinforcement in the presence of others, until responding occurs reliably only under the intended stimulus conditions.
- Discriminative stimulus (SD, S-dee)
- A stimulus in whose presence a particular behavior has been reinforced and in whose absence it has not been reinforced, or has been reinforced less. Because of that history, the behavior is more likely when the stimulus is present. A discriminative stimulus evokes behavior; it does not elicit it.
- S-delta (SΔ, Stimulus delta)
- A stimulus in whose presence a particular behavior has not been reinforced. Because of that history, the behavior is less likely when the S-delta is present. It signals that reinforcement is unavailable; it does not signal punishment.
- Stimulus control
- A relation in which the rate, latency, duration or intensity of a behavior changes in the presence of a particular antecedent stimulus. It develops when the behavior has been reinforced in the presence of that stimulus and not reinforced, or reinforced less, in its absence.
- Generalization gradient
- A graph showing how often a trained response occurs across test stimuli that differ from the training stimulus along one dimension, such as color or pitch. A steep gradient shows strong discrimination; a flat one shows broad stimulus generalization.
Check yourself
4 questions on B.13. Choose an answer to see whether you're right and why.
Show the answer
Answer: A. Aiyana now responds differently to two stimuli because saying 'red' was reinforced in the presence of one and not the other. That is discrimination. Generalization would be the opposite pattern: saying 'red' to both cards.
Show the answer
Answer: D. Joaquín's presence has been correlated with asking going unreinforced, so he is an S-delta. He is not a punisher: nothing aversive followed Tyler's asking, it simply did not pay off. Leaving early is just as valuable to Tyler when Joaquín is on shift, so he is not an abolishing operation either.
Show the answer
Answer: D. A steep, narrow gradient means responding is restricted to stimuli very close to the one present during training, which is strong discrimination. A flat, wide gradient would show generalization. Pecking at 550 nm stayed high, so it was not extinguished.
Show the answer
Answer: C. Discrimination training is what the analyst does: reinforce in the SD and withhold reinforcement in the S-delta. Stimulus discrimination is what the learner then shows: different responding to the two stimuli. Punishment is not required, since extinction in the S-delta is enough.
Adds a step to your study record. Saved in this browser only.