Key Points
- The standard is set when the goal is written, not once the graph looks good - decided afterwards, it only describes whatever the data already did.
- The usual form is accuracy held over consecutive sessions, and the stronger versions also require more than one instructor, setting, or set of materials.
- Hitting the criterion moves a target into maintenance probes rather than retiring it, because performing on the day of teaching is not evidence a skill will last.
Mastery Criteria Explained
Every treatment goal names an endpoint. The mastery criterion is that endpoint stated in measurable terms - how accurate, how consistently, over how many opportunities, and under what conditions - and it belongs in the goal before teaching starts. Written in advance, it decides for the team when a target is finished; written afterwards, it tends to ratify whatever happened.
The familiar form is accuracy sustained across consecutive sessions: a proportion of independent correct responses, held for several sessions in a row rather than reached once. Consecutive is the load-bearing word, because a single strong session is as likely to be a good day as a learned skill. Rate criteria are used where fluency is the point, and duration or latency criteria where the goal is measured that way - the criterion is always stated in the same measure the data are collected in.
Stronger criteria also say who, where, and with what. A response that appears only with one instructor, at one table, with one set of materials has met a narrow standard, which is why criteria commonly require performance across more than one person or setting and why independent responses are counted separately from prompted ones. A target reaching criterion on prompted trials is prompt dependence being recorded as progress.
Setting the level is clinical judgment rather than a template. Criteria set too low close targets that do not survive outside the session; set too high, they hold a learner on a skill they already have while the rest of the program waits. Safety-relevant skills and prerequisites for later goals justify a stricter standard than a target that will get natural practice anyway, and a criterion still unmet after many sessions is information about the teaching procedure, not a reason to quietly lower the bar.
Meeting the criterion is a transition, not an ending. The target moves into maintenance probes, generalization checks continue, and the next step in the sequence opens - and a mastered target that fails a later probe comes back for re-teaching. This is what a copied, unconsidered criterion breaks: goals that close on paper while the skill does not hold up a month later.
Because the criterion is the rule the data are read against, it has to live where both the plan and the data system can see it. Treatment plans record it beside the baseline for each goal, progress reports cite it when they report a target as met, and a system that tracks every target against its own stated criterion is what lets a supervising BCBA review a caseload without recomputing each graph by hand.