Getting Discrimination Training Actually Right
Discrimination training in ABA is one of those procedures that sounds simple until you sit down with a client and realize most of them aren't picking up the cues the way the textbooks assume they will. You teach someone to respond differently to two or more stimuli that are clearly distinct from each other. That is the definition. The practical application is where the real work happens. The basic setup involves presenting a target stimulus alongside a comparison stimulus, reinforcing the correct response to the target, and withholding reinforcement for the incorrect choice. You repeat this across trials, gradually fading prompts and testing whether the individual can identify the correct item without assistance. Most RBTs and new BCBA interns build this correctly on their first try. Then they hit the wall when the learner starts showing stimulus bias or perseveration and doesn't know how to adjust.
A Working Example Of Discrimination Training Aba
I use a three-item color discrimination program most often as a starting point because it reveals exactly where things break down. You present red, green, and blue cards face up. The instruction is "Touch red." You wait for the response. If the learner touches red, you deliver reinforcement and move to the next trial with a fresh arrangement. If they touch blue, no reinforcement, brief pause, re-present the same instruction. The tricky part comes after about twenty to thirty correct trials. The learner may have learned to always pick the card on the left regardless of color. I have seen this happen consistently. When that occurs, the standard workaround is to reshuffle so heavily that the correct position changes on every single trial for at least fifty trials, and then intermix multiple target colors in random order rather than cycling through them predictably. This typically resolves positional bias within two to four sessions if the individual has the baseline visual discrimination skills to support it. Here is something most programs gloss over. Discrimination training works best when the stimuli differ along at least one highly salient dimension from the start. Beginners often introduce too many simultaneous differences or not enough, and either choice slows progress dramatically. If you are teaching shape discrimination, keep the colors identical and vary only the shape. Once that discrimination is solid, you can then introduce color as a secondary dimension. Sticking to one varied dimension at a time cuts the number of trials needed by roughly half compared to changing multiple features simultaneously.
Another detail people miss is the timing of reinforcement delivery. The reinforcement has to land within one to two seconds of the correct response, and it has to be paired with a clear discriminative stimulus that signals which feature mattered. Without that tight contingency, you are not building a true stimulus discrimination. You are building a habit that looks like discrimination but falls apart the moment you change the context or the materials. When discrimination training fails completely
There are scenarios where this approach simply does not work, and it is worth knowing those upfront so you do not waste weeks on a procedure that is not suited to the learner. Children with severe intellectual disabilities or profound developmental delays often cannot form visual discriminations using the standard discrete trial format. In those cases, you need to shift to more foundational work first. Pairing procedures, noncontingent reinforcement to reduce frustration, or moving into foundational matching-to-sample before attempting pure discrimination training is usually the better path. Autistic individuals who are highly sensitive to visual clutter or who have motor planning difficulties can also struggle with standard discrimination tasks. The problem is not a lack of cognitive ability. It is that the task demands are too high relative to their sensory and motor profile. A practical workaround is reducing the number of choices to two, enlarging the stimuli significantly, and using high-contrast materials. This adjustment alone can turn a program that has been stalled for months into one that makes steady progress within a week.
Get the Full Details
The biggest bottleneck I see in practice is overgeneralization failure. A learner might master discriminating between specific red and blue cards, then fail completely when presented with a different shade of red or blue in a new setting. This is not a training failure. It is an expectation issue. The program needs to include systematic generalization probes from the beginning, not as an afterthought added weeks later. Testing with novel instances of the same stimuli every third or fourth session keeps generalization on track and usually prevents the frustrating regression that happens when you finally do introduce variation. Data collection for discrimination training should track correct responses, incorrect responses, latency to respond, and prompt level required on each trial. Most agencies use a simple tally sheet or an app for this. The latency data is the one people ignore most often, and it is usually the earliest warning sign that the learner is guessing rather than actually discriminating. When average latency creeps above five seconds on correct trials, that is a reliable indicator to go back and tighten the stimulus control before adding more complexity.
