Design and Evaluate Procedures to Produce Simple and Conditional Discriminations
If you work in ABA, you’ve likely heard these terms—but do you feel confident explaining the difference between simple and conditional discriminations to a colleague, trainee, or parent?
Many clinicians understand the concepts in isolation but struggle to choose the right procedure for their learner or evaluate whether it’s actually working. This matters because picking the wrong approach wastes time, can create prompt dependence, and may limit a learner’s independence and dignity.
The good news: once you understand the underlying logic and follow a clear evaluation process, designing and teaching discriminations becomes predictable and measurable.
This article is written for BCBAs, clinic directors, senior supervisors, and caregivers who want to sharpen their understanding of discrimination training. We’ll walk through what discriminations are, how to tell simple from conditional, how to design and teach each type, and how to measure whether your procedure is working. We’ll also touch on the ethical foundations that anchor this work—informed consent, least-restrictive practice, and dignity.
What Is a Discrimination in ABA?
A discrimination is straightforward: a behavior occurs more reliably in the presence of one stimulus than another.
Think of a learner who reaches for a ringing phone but not a silent one, or a child who touches a red button for a toy but ignores a blue one. In both cases, the learner has learned to respond differently depending on what cue is present. That’s stimulus control—and it’s the foundation of discriminations.
In ABA, we use discriminations to teach learners to navigate their world safely and effectively. A learner who can’t discriminate a stop sign from a go signal will struggle with traffic safety. A child who doesn’t discriminate between a teacher’s “time to work” cue and a “free play” cue will miss important transitions.
Discriminations don’t happen by accident. They’re built through careful arrangement of antecedents, responses, and consequences. That’s where procedure design comes in.
Simple Discrimination: One Cue, One Rule
A simple discrimination happens when a single stimulus reliably signals whether reinforcement is available for a response. We pair an antecedent stimulus (the SD) with reinforcement, and a different stimulus (the S-delta) with the absence of reinforcement.
Here’s a concrete example: you present a green card and a red card side by side. When the learner touches the green card, you deliver a preferred reinforcer. When they touch the red card, nothing happens—just a brief pause before the next trial.
Over many trials, the green card becomes an SD (a signal that touching it pays off), and the red card becomes an S-delta (a signal that touching it does not).
That’s simple discrimination. It involves one stimulus set, one rule, and direct stimulus-response learning. The learner doesn’t have to think “if this, then that.” They learn: touch green, get reinforced; touch red, get nothing.
Simple discriminations can be taught with stimuli presented simultaneously (both at once) or successively (one at a time, in alternating trials). Either way, the logic is the same: one stimulus is the SD, one is the S-delta, and the learner’s job is to respond to the SD.
Conditional Discrimination: Context Changes the Rule
A conditional discrimination is more complex. The correct response depends on a combination of a sample stimulus (the conditional cue) and the choices available. The rule changes depending on what sample you present.
The classic example is matching-to-sample (MTS). You show a learner a sample card—say, a red square. Then you present three comparison cards: a red square, a blue square, and a yellow triangle. The learner must select the red square to earn reinforcement.
On the next trial, you show a different sample—maybe a blue square—and the learner must now select the blue comparison. The correct answer depends entirely on what sample was shown first.
That’s conditional discrimination: “If the sample is red, choose red. If the sample is blue, choose blue.” The rule itself is fixed, but which stimulus is correct changes based on context.
Conditional discriminations can involve identity matching (the comparison is identical to the sample) or arbitrary relations (the comparison is linked by a learned rule—for instance, “if you see the word ‘cat,’ touch the picture of a cat”).
The key distinction: in simple discrimination, one stimulus is the cue. In conditional discrimination, the sample stimulus sets a rule that determines which comparison is correct.
Why the Difference Matters in Practice
Choosing between simple and conditional discrimination directly affects whether your learner succeeds or gets stuck.
If a learner isn’t ready for conditional tasks, asking them to do matching-to-sample will lead to high error rates, frustration, and slow progress. If a learner has mastered conditional responding but you teach a complex skill using only simple discriminations, you’ll miss opportunities to build more sophisticated choices.
A learner’s skill level should guide your choice.
Use simple discrimination to teach direct, single-cue responses: “touch the stop sign,” “pick the red cup,” “come when you hear your name.”
Use conditional discrimination when the correct choice depends on context: “if I show you a triangle, point to the matching triangle,” “if I say ‘shoes,’ touch the shoes (not the socks).”
Before you choose a procedure, probe whether your learner has the prerequisites. Can they attend to stimuli? Follow basic instructions? Discriminate simple contrasts like red versus blue? If not, start there.
Designing a Discrimination Procedure: Key Steps
Most discrimination procedures follow a similar pattern: assess, choose stimuli, teach, measure, and fade.
Start with assessment. Give your learner a brief, unprompted trial with the stimulus set you’re considering. Can they select the target stimulus without prompting? If yes, you might already have stimulus control—or you might have the wrong stimuli. If no, you have a baseline.
Select your stimulus set carefully. For simple discrimination, choose two stimuli with clear, contrasting features. Red versus blue is better than red versus dark red. For conditional discrimination, start with only two or three comparisons, and use identity matching before moving to arbitrary relations.
Arrange clear contingencies. Decide exactly what happens after a correct response and after an incorrect response. Write this down so your team stays consistent.
Teach with prompts and reinforcement. Present the stimuli, deliver a prompt if needed to guide the learner toward the correct choice, and reinforce every correct response heavily in early trials.
Fade prompts systematically. As accuracy improves, gradually reduce prompting. If you started with hand-over-hand guidance, move to a partial physical prompt, then a gesture, then no prompt. The goal is to shift control from your prompt to the discriminative stimulus.
Measure and monitor. Track the percentage of correct, unprompted responses across trials and sessions. Aim for 80–90% accuracy across consecutive sessions. If you’re not hitting that target, check whether your reinforcer is strong enough, whether you’re fading too fast, or whether stimuli are too similar.
Probe generalization. Once your learner is accurate in the training context, test whether they can apply the discrimination to new stimuli, people, and settings. If they can’t, you may have inadvertently taught them to respond to irrelevant cues rather than the intended stimulus.
Prompting, Fading, and Transferring Stimulus Control
Prompts are essential—but temporary. The most common mistake is prompt dependence: a learner who responds perfectly with guidance but freezes without it. This happens when fading is skipped or done haphazardly.
Effective fading requires a plan. Decide upfront whether you’ll use most-to-least prompting (start strong and decrease) or least-to-most prompting (start minimal and increase only if needed). Both work; the choice depends on your learner’s history and error tolerance.
During fading, build in wait time—typically 3 to 5 seconds—before delivering a prompt. This gives the learner a chance to respond independently. If they do, reinforce heavily. If they don’t, prompt calmly.
Record the prompt level on each trial. If accuracy drops as you fade, revert to the last successful level before trying again. Fading is not a race; it’s a data-guided process.
Real-World Examples
Simple discrimination in action: A young child is learning utensil skills. You place a spoon and fork in front of him. You prompt him to pick up the spoon, and he uses it to eat applesauce—his favorite.
Over 20 trials, he starts picking up the spoon without your prompt. The spoon has become the SD. A few trials later, you introduce a knife (an S-delta). He occasionally tries it—but no food follows, so he returns to the spoon. The spoon now has strong stimulus control.
Conditional discrimination in action: A learner is learning to match colored cards. You show a sample card (yellow) for 2 seconds, then present three comparisons: yellow, red, and blue. The learner must touch yellow to earn a reinforcer.
You guide his hand toward yellow for the first few trials. By trial 15, he’s reaching for yellow without guidance. Now you change the sample to red. He hesitates—the rule has changed. You give a light gestural prompt. He selects red and receives reinforcement.
Over 50 trials across several sessions, he’s matching new colors at 90% accuracy without any prompt. Stimulus control has transferred from your guidance to the sample stimulus.
Common Mistakes and How to Avoid Them
Teaching conditional tasks too early. A learner who hasn’t mastered simple discriminations will struggle with matching-to-sample. If you notice high error rates, step back and confirm simple discrimination skills first.
Using too many similar comparisons at once. Don’t start with three nearly identical cards. Start with high contrast (red versus yellow, square versus circle). Increase difficulty as accuracy climbs.
Not fading prompts, or fading too slowly. This creates prompt dependence. Fading too quickly creates errors and demotivates learners. The sweet spot is a systematic, data-guided fade.
Measuring the wrong thing. Counting trials is not a measure of learning. Measure accuracy (percentage of correct, unprompted responses), latency (response speed), and whether the skill transfers to new stimuli and contexts.
Ethical Foundations: Consent, Dignity, and Least-Restrictive Practice
Teaching discriminations is powerful work that comes with responsibility.
Before designing any procedure, obtain informed consent from the learner’s guardian or, when applicable, from the learner themselves. Explain what you’re teaching, why, and how. Revisit consent—assent—throughout the process, especially if the learner can express preferences nonverbally. If a learner becomes upset or refuses to participate, pause and reassess.
Use the least-restrictive procedure that will work. Start with positive reinforcement, environmental arrangement, and modeling before reaching for physical prompts. If you do use physical prompts, use the gentlest ones that still work.
Ensure stimuli are meaningful and culturally appropriate. A learner who doesn’t care about the materials you’ve chosen won’t learn as quickly. Check in with families about what matters to them.
Monitor your data for adverse effects. Is the learner showing learned helplessness? Avoiding the training area? Overgeneralizing in problematic ways? If so, adjust immediately.
Evaluating Your Procedure: Data and Probes
Once you’ve designed and taught a discrimination, how do you know it’s working?
During teaching, record correct responses out of total trials, prompt level used, and reinforcer delivered. Aim for a steep upward curve in accuracy. If you see a plateau or decline, troubleshoot: Is the reinforcer still strong? Are you fading too fast? Is the learner fatigued?
As you fade, conduct unprompted probes—trials where you present the stimulus and wait with no prompting or reinforcement. These tell you whether stimulus control has actually transferred. If a learner succeeds on prompted trials but fails unprompted probes, they’re not ready to fade further.
After reaching mastery, probe generalization. Present novel exemplars, try different people, try different settings. If accuracy drops sharply, your learner may have learned to respond to incidental cues rather than the intended feature. Vary your training stimuli and context more deliberately.
Mastery criteria are typically 80–90% correct, unprompted responses across at least two consecutive sessions. Document your criteria upfront and stick to them.
Frequently Asked Questions
How many comparison stimuli should I start with in matching-to-sample?
Start with two comparisons (the S+ and one S-delta). Once accuracy is solid at 90%+, add a third. Too many comparisons at once creates confusion.
When should I use errorless learning for discriminations?
Errorless learning works well when errors would be particularly harmful or demotivating—for instance, if a learner has a history of aggression triggered by mistakes, or if the skill is safety-critical. Transition to trial-and-error with prompting as the learner gains independence.
How do I know when to move from simple to conditional discrimination training?
Probe a simple discrimination task first. If the learner is at 80%+ accuracy unprompted, they have the foundation. Then probe a short conditional task. If they show some success or learn quickly with prompting, they’re ready. If they seem lost, they need more simple discrimination practice.
What should I do if prompt fading stalls?
Return to the previous prompt level for a few more trials, then try fading again more gradually. Consider whether your reinforcer is still motivating, whether the learner is fatigued, or whether you’re introducing too many changes at once.
Key Takeaways
Simple and conditional discriminations are both teachable and measurable—but they require different procedures and serve different goals.
A simple discrimination teaches a learner to respond to one stimulus and not another. A conditional discrimination teaches a learner to adjust their choice based on context or a sample stimulus.
To design either procedure well: assess first, select clear stimuli, teach with reinforcement and thoughtful prompts, fade systematically, and measure unprompted responding. Monitor generalization to new stimuli and contexts.
Always prioritize informed consent, dignity, and least-restrictive methods. Use positive reinforcement as your primary tool, and fade prompts based on data, not guesswork.
If you’re teaching discriminations now, review your current procedures against this framework. Are you measuring the right thing? Are you fading prompts, or have your learners become prompt-dependent? Are your stimuli clear and meaningful? Is consent documented?
Small adjustments in how you design and evaluate can yield big improvements in learner independence and progress.
