Conditional discrimination is a type of learning in which the correct response changes depending on which cue, or “sample stimulus,” is present at the time. In applied behavior analysis (ABA), it involves four components: a sample stimulus, a set of comparison stimuli, a response, and a consequence. Unlike a simple discrimination, where a single stimulus signals what to do, conditional discrimination requires a learner to pay attention to one stimulus and then use that information to pick the right option from a set of choices.
This concept sits at the heart of many skills taught in ABA therapy, from receptive language and matching tasks to answering questions and reading. Understanding how it works, and where it breaks down, is essential for anyone studying or practicing ABA.
How It Differs From Simple Discrimination
In a simple discrimination, a learner responds to one stimulus. A child sees a red card and picks it up because picking up red cards has been reinforced. The task only requires noticing one thing.
Conditional discrimination adds a layer. Now the correct choice depends on a changing cue. A therapist places three pictures on a table and says “touch the cat.” The spoken word “cat” is the sample stimulus. The pictures are the comparison stimuli. The child must listen to what was said, scan the pictures, and select the one that matches. If the therapist then says “touch the dog,” the correct picture changes even though the array looks the same.
This distinction matters practically because simple discrimination training alone does not build all the skills a learner needs for conditional tasks. Specifically, it does not require attending to the sample stimulus. A child who has only practiced picking out one picture at a time may struggle when multiple items appear and the instruction keeps changing, because the task now demands tracking two sources of information at once.
Everyday Examples
Conditional discrimination is not limited to therapy tables. It shows up constantly in daily life. Naming the color of an apple requires attending to the apple itself (the object in front of you) and to the question someone asks (“What color is that apple?”). Both stimuli together control the correct answer. If someone instead asked “What fruit is that?” you would need to shift your response even though you are looking at the same object.
A more complex example: a child is asked four different questions across a conversation. “What’s your mother’s name?” “What’s your brother’s name?” “When’s your mother’s birthday?” “When’s your brother’s birthday?” Answering each one correctly requires discriminating between two people and two question types. A child who only tunes into part of the question might say a name when asked for a birthday, or give the brother’s name when the mother’s was requested. Correct performance depends on tracking every element of the verbal cue.
Even something as simple as answering “What time is lunch?” involves conditional discrimination. The correct answer, say “noon,” requires responding to both the word “time” and the word “lunch.” A person who only responds to “lunch” might say “sandwiches.” A person who only responds to “time” might say “nine o’clock.” Neither answer reflects full conditional control.
The Four Components
Every conditional discrimination task has the same basic structure:
- Sample stimulus: The cue that tells the learner which response is correct on this particular trial. In receptive labeling, this is usually a spoken word or instruction.
- Comparison stimuli: The set of options the learner chooses from. These might be pictures, objects, written words, or actions.
- Response: The learner’s selection, such as pointing, touching, or handing over the correct item.
- Consequence: Reinforcement for a correct response, or a correction procedure for an incorrect one.
For a response to count as a true conditional discrimination, the learner must attend to the sample stimulus and then use it to guide their selection from the comparisons. If a child picks the right picture but wasn’t actually listening to the instruction (maybe the correct item was always on the left side of the table), the behavior is under faulty stimulus control, not genuine conditional control.
Where It Shows Up in ABA Therapy
Conditional discriminations are among the most commonly targeted skills in early intensive behavioral intervention. Receptive labeling (“touch the ___”), matching identical items, and sorting tasks all rely on this skill. So does much of early reading instruction, where a learner sees a written word and matches it to a picture or spoken word.
Match-to-sample (MTS) is the standard format for teaching these skills. The therapist presents a sample, then an array of comparisons, and the learner selects the match. This format is used across a wide range of programs, from identifying objects and colors to more advanced academic content.
Two broad approaches exist for introducing conditional discriminations. In the simple-to-conditional method, therapists first teach the learner to discriminate among comparison stimuli in isolation (simple discrimination) before adding the sample stimulus that changes across trials. In the conditional-only method, all stimuli are presented from the start, and the learner practices the full conditional task from the beginning. The rationale for the conditional-only approach is that it requires the learner to use all the necessary skills right away, including attending to the sample stimulus, rather than building habits that may not transfer.
Common Learning Challenges
When learners struggle with conditional discrimination, the problem usually traces back to stimulus control. The most common issues are stimulus overselectivity and position bias.
Stimulus overselectivity happens when a learner focuses on only one feature of the task and ignores others. For instance, a child might select a picture based on where it is sitting in the array rather than on what the therapist said. The behavior looks correct on some trials by coincidence, but it is not under the control of the sample stimulus.
Position bias is a specific form of this: the learner consistently picks the item on the left, or in the middle, regardless of the instruction. Error analysis can catch these patterns. Practitioners track which stimulus or position the learner selects across trials, circling selections on data sheets to reveal whether the child is consistently gravitating toward a specific spot or a specific item regardless of what was asked.
Other patterns include “win-stay” responding, where a learner keeps selecting whatever was reinforced on the previous trial even though the sample stimulus has changed. Identifying these error patterns early allows therapists to adjust their teaching before the learner practices incorrect responses dozens of times.
Teaching Strategies That Work
Prompt delay procedures are widely used to teach conditional discriminations. The basic idea is to start by immediately showing the learner the correct answer (a zero-second delay, sometimes called simultaneous prompting), then gradually wait longer before prompting so the learner has a chance to respond independently.
Two variations are common. In progressive prompt delay, the wait time increases in small steps, say one second, then two seconds, then three, based on the learner’s performance. In constant prompt delay, the therapist waits the same fixed interval (often two seconds) on every trial until the learner masters the skill. Both methods reduce errors while still building independence.
Prompting hierarchies also play a role. Most-to-least prompting starts with the most intrusive help (like physically guiding the learner’s hand) and fades to less help over time. Least-to-most prompting does the reverse, starting with minimal assistance and adding more only if the learner makes errors. Research comparing these approaches has found that most-to-least prompting tends to produce fewer errors, but least-to-most prompting sometimes leads to faster learning. A hybrid approach, most-to-least with a built-in delay, gives the learner a brief window to respond independently before the prompt is delivered, combining error reduction with opportunities for independent responding.
Trial-and-error teaching, where the learner is shown a correct response once and then left to figure it out with only feedback after each attempt, is another option. It is common in special education settings, though it typically produces more errors than prompt-based methods.
The Link to Stimulus Equivalence
Conditional discrimination training can produce learning that goes far beyond what was directly taught. This phenomenon, called stimulus equivalence, was first described by researcher Murray Sidman in the early 1970s while teaching basic reading skills to an individual with severe intellectual disabilities. It has since become one of the most important concepts in the experimental analysis of human behavior.
Here is how it works. Suppose a learner is taught to match spoken words to pictures (hearing “cat” and selecting a picture of a cat), and then taught to match spoken words to written words (hearing “cat” and selecting the printed word CAT). Without any additional training, the learner may now be able to match written words to pictures and pictures to written words. These untrained relationships emerge on their own.
Three properties define a true equivalence relation. Reflexivity means a stimulus matches itself (a picture of a cat matches a picture of a cat). Symmetry means trained relations work in reverse (if hearing “cat” leads to selecting the picture, then seeing the picture leads to selecting “cat”). Transitivity means relations chain together (if A relates to B, and B relates to C, then A relates to C without direct training). When all three properties appear, the stimuli form an equivalence class, and any new information taught to one member of the class transfers to the others automatically.
This is why conditional discrimination training is considered one of the building blocks of language and cognition in behavior analysis. Teaching a relatively small number of relations can generate a much larger network of understanding, dramatically expanding what a learner can do without needing to teach every connection individually.