Open almost any skill-acquisition programme and you’ll find the same line near the bottom: mastery = 80% correct across three consecutive sessions. It’s on the protocol you inherited, the one before that, and the one you’ll write next week. Most of us have typed it a hundred times. Far fewer could say where it came from — or why three sessions and not two, 80% and not 90%.
It’s worth asking, because that little number does a lot of quiet work. It decides when a child stops working on a target and moves on to the next. Multiply it across a whole programme and it shapes the entire pace of someone’s learning. A rule with that much leverage deserves more than “it’s what the template said.”
The number is a convention, not a finding
Here’s the uncomfortable part: 80% across three sessions was never handed down by a definitive study showing it to be the optimal point to stop teaching. It’s a sensible-sounding default that propagated because it is easy to apply, easy to defend, and good enough a lot of the time. A clean threshold you can point to in an audit has obvious appeal when you’re writing twenty programmes and need them to be consistent.
None of that makes it wrong. Defaults are useful. The problem is what happens next: the default stops being a starting point and becomes a substitute for thinking. We stop asking what would tell me this skill is actually learned? and start asking have we hit 80% three times yet? Those are not the same question, and the gap between them is where learners get stuck on things they’ve clearly mastered — or moved off things they haven’t.
What mastery is supposed to mean
Strip away the number and mastery is a simple idea: the skill is usable. The learner can do it when it matters, not just when conditions are perfect. In practice that has four parts, and accuracy is only the first:
- Accurate — they get it right.
- Fluent — they get it right at a useful speed, without long hesitation. Accuracy alone is a snapshot; fluency is what makes a skill survive contact with the real world.
- Durable — it maintains over time, not just on the day you hit criterion.
- Generalised — it shows up with other people, in other settings, with other materials.
Eighty per cent across three sessions, collected in a quiet room with the same instructor and the same stimuli, speaks to the first property and barely touches the other three. It’s a proxy. The danger is mistaking the proxy for the goal — calling a skill “mastered” when all we really know is that it cleared a low bar under ideal conditions.
Why percentage-correct is a weak stand-in
Lean too hard on a percentage and a few blind spots open up.
It ignores rate. A learner who answers correctly but slowly, with visible effort each time, is not in the same place as one who responds instantly — yet both can score 80%. This is the whole argument behind precision teaching and celeration: how fast a skill is performed predicts whether it lasts and whether it’s available under pressure. (More on why rate beats raw accuracy in what a celeration chart actually shows you.)
It can be propped up by prompts. Three “good” sessions of prompt-dependent responding isn’t mastery — it’s a record of your prompting. If you’re not separating independent responses from prompted ones, the percentage can look healthy while the skill quietly leans on you.
It’s noisy when trials are few. Eighty per cent of ten trials is a different thing from eighty per cent of three. With small sessions, “80% across three” can be met by a couple of lucky guesses as easily as by genuine learning.
It assumes the data are trustworthy. A criterion is only as good as the measurement under it. If two staff would score the same response differently, you’re setting a threshold on noise — which is worth checking with an operational definition tight enough that two people agree, and the odd reliability check (the IOA calculator makes that quick).
What the evidence actually says
When researchers have compared different mastery criteria head-to-head, the same themes keep surfacing. The criterion you choose genuinely affects how well a skill maintains — it isn’t a cosmetic decision. More stringent criteria — a higher percentage, more sessions, or an added rate requirement — tend to produce better retention than leaner ones. And surveys of practising clinicians find wide variation in what’s actually used, with little standardisation behind the familiar numbers.
So the honest answer to “is 80% across three sessions right?” is: right for what? It’s a defensible middle setting. Whether it’s the correct setting depends entirely on the skill, the learner, and what the skill has to do later.
When to set the bar higher
Some skills earn a stricter criterion. Raise it when the skill is:
- Foundational or a prerequisite — something later targets will be built on. Shaky foundations compound.
- A safety or health skill — road safety, responding to one’s name, medical compliance. “Usually” isn’t good enough.
- Expected to maintain long-term without much further practice.
For those, accuracy alone won’t cut it. Require independent (unprompted) responding, add a rate or fluency aim where speed matters, spread the sessions across different days, people, and materials, and — the step most often skipped — run a maintenance probe a week or two later before you close the target. “Mastered” should mean still there when I come back to it, not hit a threshold three times in a row.
When to be more flexible
Flexibility cuts both ways, and the more common failure is being too rigid, not too loose.
If the data show clean, independent, fluent responding well before the magic count is met, trust the data and move on. Holding a learner on a target they’ve plainly got wastes instructional time, risks satiation and lost motivation, and teaches them that effort doesn’t change anything. The criterion exists to protect against moving on too early; it was never meant to keep someone drilling a skill they own.
And for genuinely low-stakes skills — ones you’ll keep practising in context anyway — a lighter criterion is often perfectly reasonable. Not every target needs the full apparatus. The skill is deciding which ones do.
A criterion that earns its place
You don’t need to abandon the 80% rule. You need to demote it from a reflex to a choice. Before you write the criterion on a new programme, run it through four questions:
- What does “got it” have to look like for this skill to be useful? Define mastery by the job the skill does, then work backwards to the number.
- Am I measuring more than accuracy? Independence always; rate or fluency where speed matters; a maintenance probe before you close it.
- Did I set this in advance, for this learner and this skill — or did I paste it?
- Will I watch the data, not just the threshold? Clearly acquired before criterion? Move on. Brittle once it’s met? Don’t close it, even if the box is ticked.
The 80%-across-three-sessions rule isn’t a mistake. It’s a starting point that quietly became a stand-in for clinical judgement. Mastery is a decision about whether a skill is genuinely usable and durable — and the criterion should serve that decision, not replace it. Keep the default if you like. Just make it something you chose, not something you copied.
Frequently Asked Questions
What is a mastery criterion in ABA?
A mastery criterion is the performance standard that decides a learner has acquired a skill and can stop targeting it — usually written as a percentage correct sustained across a set number of sessions. It is a decision rule we impose, not a property of the skill itself.
Where does ‘80% correct for three consecutive sessions’ come from?
It is an inherited convention rather than a figure derived from a single definitive study. It spread because it is easy to apply and easy to defend in an audit. That makes it a reasonable default, not a law — and not necessarily the right standard for any particular skill.
Should every skill use the same mastery criterion?
No. The right standard depends on what the skill is for and how durable it has to be. Foundational, safety, and prerequisite skills warrant a higher bar — often including independence, a rate or fluency requirement, and a maintenance check — while lower-stakes skills may need less.
When should you use a stricter mastery criterion?
When a skill has to maintain over time, generalise across settings, or support later learning. Consider requiring independent (unprompted) responding, adding a rate or fluency aim, spreading sessions across days and people, and running a maintenance probe before you call the skill mastered.