Conditional discrimination

Imagine an activity whose rule is to choose the picture identical to the sample. There are two cards to choose from: a dog and a cup.

With a dog as the sample, choosing the dog is correct. With a cup as the sample, choosing that same dog is incorrect. The dog card has not changed. Its role in the activity has.

That dependency is what makes the discrimination conditional: which selection is reinforced depends on the sample as well as on the available comparisons.

First, locate the sample and the choices

In the illustration, the sample is the single card above. The comparisons are the two cards below, from which the child chooses. Both scenes use the same instruction: choose the identical picture.

The sample changes; the comparisons stay in place. The dog is correct with the dog sample and incorrect with the cup sample. The child's hands illustrate the expected choices, not recorded results.

Read the left scene first, then the right. The cup has gone from incorrect to correct without changing its appearance or position. The dog has made the reverse change. Neither comparison is always the answer.

Putting the sample above the comparisons is a layout choice, not part of the definition. Nor does conditional mean “whatever came first”: in simultaneous matching, the sample remains visible while the child chooses. In delayed matching, it disappears before the choices appear. Both arrangements can involve conditional discrimination.

Two comparisons make the example easy to read. They are not a universal recommendation for the size of a teaching array.

How this differs from simple discrimination

Use the same dog and cup cards to arrange two different tasks:

	Simple discrimination	Conditional discrimination
Arrangement	Dog and cup, without a sample.	Dog and cup, with a sample that varies.
What is reinforced?	Choosing the dog throughout this teaching arrangement.	Choosing the dog with the dog sample, and the cup with the cup sample.
What must the learner discriminate?	The dog from the cup.	The samples, the comparisons, and which selection goes with each sample.
Is choosing the dog always correct?	In this arrangement, yes.	No: it depends on the sample.

“Simple” does not mean that the dog must stay correct forever. The instructor could later reverse the contingencies. The distinction here is that no additional stimulus selects between those contingencies within the task.

Identity matching is not automatically simple discrimination. If the sample changes and the learner selects the corresponding identical picture from alternatives, the arrangement is conditional. “Sorting” or “matching shapes” is not enough information to classify a task; we need to know how it is arranged. Green (2001) explains this distinction in terms of stimulus control.

These are two kinds of arrangement, not a compulsory sequence of “teach simple first, then conditional”.

Why the consequence matters

So far the pictures show what the task asks for. Teaching also involves what follows the choice.

With the dog sample present, an instructor may reinforce choosing the dog; with the cup sample, choosing the cup receives that consequence. Over teaching, the aim is for the child’s selections to come under the control of those relationships.

This is why the technical description has four terms: conditional stimulus (the sample), comparison stimulus, selection response, and consequence. The consequence is part of the teaching arrangement even though it is not drawn in the figures. A later assessment may check the learned selection without feedback.

Arranging those four terms does not establish that they control the child’s behavior as intended. That requires observing the choices.

The sample does not have to be an identical picture

Keep the dog and cup comparisons, but replace the sample picture with a spoken word. On hearing “dog”, the learner selects the dog; on hearing “cup”, the learner selects the cup. This is auditory–visual conditional discrimination. The sound and picture do not look alike; their correspondence has to be learned.

Other conditional tasks use non-identical pictures or arbitrary symbols. Identity is one possible relation, not the definition of conditional discrimination. Grow and LeBlanc (2013) describe the auditory–visual case and its teaching arrangements.

Here, comparison means an option offered for selection. If you have read our article about Ribes, that is a different use of the word: offering comparison cards does not, by itself, establish “comparison” in his classification.

Two changes that ask different questions

The first illustration asks: when the sample changes, does the selected picture change with it?

The next asks: when the sample stays the same but the cards move, does the learner still select the corresponding picture?

The sample stays the same; the comparisons swap positions. The expected choice is still the dog. The hand changes location because the dog card has moved, not because the required relation changed.

Together, these changes help distinguish responding to the sample from always selecting one side. They also help reveal a pattern of repeating the previous choice. Neither illustration is a test protocol or evidence about a particular child.

A position preference only produces a high score if that position is correct often enough. Likewise, repeating the dog can look successful during a long run of dog trials. Vary the samples and distribute correct comparisons across positions without a predictable alternation; also avoid indicating the answer with your gaze, pointing or unequal card sizes. These are practical stimulus-control considerations discussed by Grow and LeBlanc.

What to record, and what you can conclude

In practice

Use familiar material from an existing activity. Plan opportunities for each sample and for the correct comparison in different positions, at a pace appropriate for the child.

Interleave trials for different concepts when checking independent choices, instead of completing a long run of one concept before moving to the next. Vary the order so that repeating the previous choice or following a predictable dog–cup alternation is not enough to answer correctly.

For each trial, record the sample, the comparison positions, the first selection and any help given. Separate independent selections from prompted ones.

Look for a repeated pattern: does the selected picture correspond to the sample across positions? Does the learner keep selecting a side or the previous answer? If performance changes, check the arrangement, access to the sample, fatigue and prompts before deciding what the error means.

A brief check cannot settle this on its own. One error does not establish that the sample has never controlled responding, and a short correct run does not establish mastery. Decisions need repeated observations and criteria appropriate to the activity.

Reliable selections with these cards support a specific conclusion: the learner discriminates these trained relationships under these conditions.

Generalization needs its own check with new pictures, people or settings. Changing from picture samples to spoken words may require additional teaching. Stimulus equivalence also needs additional tests of untrained relations; it does not follow automatically from accurate conditional discrimination. That distinction is central to Sidman and Tailby (1982).

The useful question is therefore concrete: with which samples, comparisons and arrangements does the learner select independently, and what remains to be checked?

Sources and further reading

Green, G. (2001). Behavior analytic instruction for learners with autism: Advances in stimulus control technology. Focus on Autism and Other Developmental Disabilities, 16(2), 72–85.
Grow, L., & LeBlanc, L. (2013). Teaching receptive language skills: Recommendations for instructors. Behavior Analysis in Practice, 6(1), 56–75.
Sidman, M., & Tailby, W. (1982). Conditional discrimination vs. matching to sample: An expansion of the testing paradigm. Journal of the Experimental Analysis of Behavior, 37(1), 5–22.
Generalization — checking performance beyond the trained material.
Errorless teaching and the ABA prompt hierarchy — planning how to fade help and check independent selections.