Error Correction: Getting It Right Without Getting It Wrong Again

The learner is sitting across from you. You place three pictures on the table and say, "Touch dog." The learner reaches forward and touches the cat. Now what?

This moment, the moment right after an error, is one of the most important moments in teaching. And yet it is one of the least standardized, least discussed, and most inconsistently implemented procedures in ABA programming. Everyone agrees that error correction matters. But when you watch five different therapists run the same program, you will often see five different things happen after a wrong response.

Some therapists immediately re-present the trial with a full prompt. Some say "no" and move on. Some re-present the trial three times in a row, creating a rote chain. Some panic a little, unsure of what to do, and give the error so much attention that they may end up reinforcing it.

Error correction is not just a procedural step. It is a teaching tool. And like any teaching tool, it works best when it is used with precision, consistency, and a clear understanding of why each step exists. In this post I draw on my own clinical practice and on ABA 101 (Applied Behavior Analysis 101), the training DataMTD offers through the VB-MAPP App.

Why Errors Happen (and Why That Matters for Correction)

Before we talk about what to do after an error, it helps to think about why the error occurred in the first place. Not every error has the same cause, and the cause should inform how you respond.

Some common reasons a learner makes an error:

The skill is not yet in the repertoire. The learner genuinely does not know the answer. This is an acquisition issue, and it tells you something about where the learner is in the teaching sequence.

• The skill is partially acquired but not fluent. The learner sometimes gets it right and sometimes does not. Stimulus control is not fully established.

• The learner is attending to the wrong stimulus feature. They may be selecting based on position, color, size, or some other irrelevant feature rather than the target stimulus itself.

• Motivation is low. The learner may know the answer but is not motivated to respond correctly. This is not an error correction problem. This is a reinforcement problem.

• The instruction or materials are unclear. Sometimes the error is on our end. The array is too visually similar, the instruction is ambiguous, or the stimuli are poorly selected (which, as we have discussed in a previous post, matters more than most people realize).

When you understand the source of the error, you can respond more effectively. A learner who does not know the answer needs more teaching. A learner who is not motivated needs a reinforcement adjustment. A learner who is attending to the wrong feature needs a stimulus control procedure. Error correction is most appropriate when the learner has some level of acquisition but has not yet reached mastery.

The Anatomy of a Good Error Correction Procedure

While there are variations in the literature, a solid error correction procedure for discrete trial teaching generally follows a predictable sequence. Here is the version I use most often in clinical practice, drawn from standard DTT methodology and consistent with what you will find in many VB-MAPP-aligned programs:

Signal the error neutrally. Do not say "no," do not shake your head, and do not give a big reaction. Simply block or remove the incorrect response and re-present the trial. Some clinicians use a brief, neutral statement like "let's try again," or simply pause and re-present the trial, including the direction. The key is to avoid making the error itself reinforcing through attention or emotional reaction. That includes extra comments like "You got it right last time." Skip those.

1. Prompt the correct response immediately. Re-deliver the instruction and prompt at a level that will produce the correct response. This is typically a full prompt. You are not testing the learner here. You are showing them the correct answer.

2. Deliver preferred items for the prompted correct response, but at a lower magnitude. A brief "yes" or a small acknowledgement is appropriate. (A quick note on wording: we only know that something was reinforcing if the same class of behavior occurs again under similar circumstances. So what we do at the moment is deliver preferred items with the goal of reinforcing the response.) You are not throwing a party for a prompted response. But you are keeping the learner engaged and moving forward.

3. Re-present the original trial as a transfer trial. Deliver the same instruction again, this time with the least amount of prompting you think the learner can succeed with, or preferably no prompting. If they get it right, you now have a decision to make based on the learner: you can deliver preferred items at a higher magnitude right away, or you can run some distractor trials (step 5) and save the higher-magnitude delivery of preferred items for the check trial (step 6). If the learner errors on the transfer trial, cycle back through the procedure, maybe giving a less faded prompt this time.

4. Run a distractor trial (or a few). Present one or more trials on a different, already-mastered target. This prevents the learner from simply repeating the same motor response in a chain. It breaks up the pattern and tests whether the learner can shift and return.

5. Re-present the original trial as a check trial. Deliver the same instruction again, with the same amount of prompting you used on the transfer trial. This time we are looking to see if the learner can perform the skill after a little bit of a time delay. If they get it right, deliver preferred items at a higher magnitude, with the goal of reinforcing the correct response on this check trial. This is the trial that matters most. This is where you find out if the correction actually worked. If the learner errors on the check trial, cycle back through the procedure.

If the learner continues to error across multiple correction cycles, it is time to step back and ask a bigger question: Is this target appropriate right now? Is the teaching procedure working? Is there a stimulus control issue that error correction alone will not solve?

What ABA101 Says When Errors Happen at Different Points in Intensive Teaching

ABA101 has a lesson on correcting errors in intensive teaching (DTT) that gives a helpful way to troubleshoot, based on which kind of trial the error happened on:

On a prompted trial. ABA101 puts it plainly: "A prompt should get the desired behavior to occur; if the behavior doesn't occur you're not using a strong enough prompt."

• On a transfer trial. "This could happen. Maybe you faded your prompt too quickly."

• On known skills. "Momentarily that skill is not known. Watch for this!" ABA101 says to analyze why the skill may have been momentarily forgotten. If an error occurs it needs to be corrected too, and momentarily takes precedence over the original error as that error has at least been corrected with a prompted trial and a transfer trial.

• On a check trial. "Maybe there were too many known trials run. Maybe you should provide a slight prompt. Maybe you should not attempt a check trial."

Common Mistakes in Error Correction

Even when clinicians know the steps, errors in error correction (if you will forgive the recursion) are extremely common. Here are the ones I see most often:

Prompting without repeating the SD. When a learner errors, it is tempting to jump straight to the prompt, guiding their hand to the dog without saying anything. But the learner needs to contact the instruction paired with the correct response, so re-present the SD ("Touch dog") along with the prompt. That is why step 2 is "Re-deliver the instruction and prompt," not just "prompt."

Delivering preferred items for a corrected response the same way as for an independent response. If the learner gets the same preferred items, at the same magnitude, whether they need correction or not, you are not building a differential contingency. Independent correct responses should contact richer preferred items than prompted correct responses. This differential is what drives the learner toward independence.

Saying "no" or giving a big reaction to the error. For some learners, "no" is aversive and can increase escape-maintained behavior. For other learners, the big reaction may actually function as reinforcement because it produces attention. Either way, you are adding something to the teaching arrangement that does not need to be there. Keep it neutral.

Correcting too slowly. The longer the gap between the error and the correction, the less effective the correction will be. Error correction should happen immediately. If you pause to write data, consult your binder, or figure out what to do next, you have lost the teaching moment.

Running too many consecutive correction trials on the same target. If a learner errors three, four, five times in a row on the same target, continuing to run correction trials is not teaching. Step back. Reassess. Try again later or adjust the program. It also helps to look at the overall mix of your intensive teaching session. On how often to run targets versus knowns, ABA101 says "It depends on your learner" and reminds us of "the importance of fading in demands and interspersing easy and hard tasks." Initially, you may be running new/target skills only 10% of your teaching session and known/easy skills 90% of the time. As you gain instructional motivation, that ratio can slowly shift until as much as 40% of your trials are targets and 60% are known.

Error Correction Across Teaching Formats

Everything above applies most directly to discrete trial teaching. But error correction is relevant across all teaching formats, and it looks a little different in each one.

In Natural Environment Teaching (NET), error correction needs to be faster and lighter. You are working within a motivating operation that may be fleeting. If a learner reaches for a cookie and you ask "What do you want?" and they say "juice," you do not want to run a five-step correction cycle while the cookie motivation fades. A quick prompt after a pause, a brief transfer, and then deliver the item. Preserve motivation. ABA101's steps for mand errors are a good example of that lighter approach: a slight pause, put the item out of view, come back with a prompt, and deliver the item, "but less of the item than if the request had been made correctly," the same lower-magnitude idea as step 3.

In group instruction, error correction needs to be discreet. Calling attention to a learner's error in front of peers can be stigmatizing and can reduce the learner's willingness to participate. A quiet prompt, a quick re-present, and move on.

In listener responding programs, pay special attention to position biases. If a learner consistently selects the item in the same position, your error correction will not help until you address the position variable. Rotate the array after every trial, including correction trials.

The Bigger Picture

Good error correction is not about having a rigid script. It is about understanding the principles underneath the procedure. You are trying to do three things at once: prevent the error from being practiced, show the learner the correct response, and give them an opportunity to demonstrate the correct response with reduced support.

When error correction is done well, learners move through acquisition faster, with less frustration and fewer repeated errors. When it is done poorly, or inconsistently, or not at all, learners practice errors, develop faulty stimulus control, and sometimes learn that errors are a reliable path to attention or escape.

The moment after an error is not a failure. It is an opportunity. But only if you know what to do with it.

If your team is inconsistent with error correction, start there. Watch sessions. Compare what different therapists do after an error. Talk about it in supervision. Write it into your program protocols with enough detail that a new therapist could read it and implement it correctly on day one. If you want a shared starting point, ABA 101 (Applied Behavior Analysis 101) is available through the VB-MAPP App at vbmappapp.com. It covers errorless teaching, error correction, and daily data collection, and it is a good way to get everyone on your team speaking the same language.

Because the learner's next correct response depends on what you do right now.