ABA error correction data should separate initial errors, valid corrections due, correction steps completed as written, prompt levels, retry responses, later independent probes, repeated error types, latency, duration, client withdrawal, and invalid opportunities. Families can ask for raw counts and definitions. Correction fidelity shows whether staff used the procedure; later probe performance addresses learning. Neither measure alone proves causation, generalization, safety, or acceptability.
Define correction-data units
Correction fidelity equals valid corrections completed as written divided by valid corrections due. Initial independent accuracy uses correct first responses divided by eligible first opportunities. Retry or prompted accuracy uses its own denominator. Later probe accuracy includes only the predeclared unprompted probe cohort.
Keep exclusions and versions visible
Record missing materials, wrong cues, device failures, safety stops, client withdrawal, and supervisor interruptions by reason. State the procedure version, staff, setting, item type, and review period. Data from a one-model procedure should not be pooled silently with data from a three-repetition procedure.
Research does not supply a benchmark
The Leaf study involved 28 autistic participants learning narrow tact relations. The Cariveau study involved four children and reinforcement after correction. Their procedures and outcomes cannot establish a universal fidelity or accuracy threshold. The BACB outline and Ethics Code support integrity, data, risk, and evaluation.
Add client experience
The ASHA AAC guidance supports continual communication access. Track accessible reports of clarity, frustration, effort, preference, pain, and withdrawal. Include duration and repetition so a final correct response does not erase the path to it.
A practical example
Across twelve valid corrections due, staff complete ten as written. In eight later unprompted probes, Priya responds correctly five times. Report fidelity as 10 of 12 and probe accuracy as 5 of 8. Add two staff deviations, median correction duration, and Priya's feedback without claiming the correction caused the probe result.
Questions families can use
Which cohort is due for correction? Which steps define fidelity? Are retries prompted? When do probes occur? Which invalid events stay visible? How do duration, distress, and client feedback affect the decision? Which version produced each result?
Sources
- Behavior Analyst Certification Board, BCBA Test Content Outline, 6th edition
- Leaf and colleagues, Comparing Error-Correction and Errorless-Learning Procedures
- Cariveau and colleagues, The Effects of Reinforcement Following Error Correction
- Behavior Analyst Certification Board, Ethics Code for Behavior Analysts
- Council of Autism Service Providers, ABA Practice Guidelines Version 3.0 public summary
- American Speech-Language-Hearing Association, Augmentative and Alternative Communication
Finni resources
