Free · 2026 · All 9 IOA types, no sign-up IOA calculator: interobserver agreement for every ABA data type
IOA (interobserver agreement) is how ABA teams check their data: two observers independently record the same behavior at the same time, and IOA is the percentage of agreement between their two records. It is the most common way to judge whether data are believable and consistent enough for a supervisor to base decisions on. An IOA of 80% or higher is commonly cited as the minimum acceptable level, with 90% or higher preferred (Cooper, Heron & Heward, Applied Behavior Analysis, 3rd ed.). This calculator covers the 9 standard IOA methods: total count, mean count-per-interval, exact count-per-interval, trial-by-trial, interval-by-interval, scored-interval, unscored-interval, total duration and mean duration-per-occurrence IOA. Pick the type that matches your data, edit the worked example or enter both observers' records, and it shows the formula and every step of the math. It's free, runs in your browser and needs no sign-up. Below it: every formula in one table, the common mistakes, and how IOA shows up on the RBT exam (task A.8).
Studying for the RBT exam? Take the free 10-question RBT check It scores you in each content area, including Data Collection and Graphing, with no sign-up.
Interobserver agreement calculator
Frequency / event recording
Discrete trials
Interval recording
Duration
Total count IOA
When to use it: A quick check on frequency data from a whole session, when you only need the two totals.
Smaller count ÷ larger count × 100
Edit the example below or enter your own data. Each observer records on their own sheet at the same time; enter both records here.
Total count IOA
90%
At or above 90%, the level often preferred (a common guideline, not a rule)
How it was worked out
- Smaller count ÷ larger count = 18 ÷ 20 = 0.9
- 0.9 × 100 = 90%
Rounded to 1 decimal place at the end.
The facts · checked November 18, 2026
IOA in ABA, explained
What IOA is
Interobserver agreement (IOA) is the degree to which two observers who measure the same behavior, independently and at the same time, report the same values. It is reported as a percentage. High IOA makes the data more believable and shows the behavior definition is clear enough for two people to apply it the same way. It does not prove the data are accurate: two observers can share the same mistake. Low IOA means the data can't be trusted yet, usually because the definition is unclear, an observer needs retraining, or the observers weren't recording the same thing.
The 9 types of IOA and their formulas
| IOA type | Formula | Data type | When to use it | Worked example |
|---|---|---|---|---|
| Total count IOA | Smaller count ÷ larger count × 100 | Event recording (frequency) | A quick check on frequency data from a whole session, when you only need the two totals. | Observer 1 counts 20 hand raises, observer 2 counts 18: 18 ÷ 20 × 100 = 90%. |
| Mean count-per-interval IOA | Each interval: smaller ÷ larger. Then the mean of those × 100. An interval where both counted 0 = 100%. | Event recording split into intervals | Frequency data recorded in short intervals, when you want to know the observers agreed on when the responses happened, not just the total. | Five intervals: 3/3, 0/0, 2/1, 4/4, 1/2 give 100%, 100%, 50%, 100%, 50%. (100 + 100 + 50 + 100 + 50) ÷ 5 = 80%. |
| Exact count-per-interval IOA | Intervals with exactly the same count ÷ total intervals × 100 | Event recording split into intervals | The strictest check for interval counts: an interval counts only if both observers recorded the same number. | Same five intervals: the counts match exactly in intervals 1, 2 and 4. 3 ÷ 5 × 100 = 60%. |
| Trial-by-trial IOA | Trials agreed ÷ total trials × 100 | Discrete trials (correct or incorrect) | Discrete-trial teaching or any trial-based data, where each trial is scored correct (+) or incorrect (−). | 10 trials; the observers disagree only on trial 5. 9 ÷ 10 × 100 = 90%. |
| Interval-by-interval IOA | Intervals agreed (both + or both −) ÷ total intervals × 100 | Partial interval, whole interval, momentary time sampling | Interval recording or time sampling when the behavior happens at a moderate rate. | 10 intervals; the observers disagree only on interval 4. 9 ÷ 10 × 100 = 90%. |
| Scored-interval IOA | Among intervals where at least one observer scored +: intervals both scored + ÷ those intervals × 100 | Interval recording, low-rate behavior | Low-rate behavior, where most intervals are empty and interval-by-interval IOA would look high just because both observers left them blank. | The interval-by-interval data above: intervals 1, 4 and 8 have a +, and both scored + in 2 of them. 2 ÷ 3 × 100 = 66.7%, against 90% interval-by-interval. |
| Unscored-interval IOA | Among intervals where at least one observer scored −: intervals both scored − ÷ those intervals × 100 | Interval recording, high-rate behavior | High-rate behavior, where most intervals are scored and interval-by-interval IOA would look high just because both observers marked them. | 10 intervals of a high-rate behavior: intervals 3, 7 and 10 have a −, and both scored − only in interval 3. 1 ÷ 3 × 100 = 33.3%, against 80% interval-by-interval. |
| Total duration IOA | Shorter total duration ÷ longer total duration × 100 | Duration recording | Duration data, when you only need the two session totals (same unit for both, such as seconds). | Observer 1 times 120 s of tantrums, observer 2 times 105 s: 105 ÷ 120 × 100 = 87.5%. |
| Mean duration-per-occurrence IOA | Each occurrence: shorter ÷ longer. Then the mean of those × 100. | Duration recording, each occurrence timed | Duration data where each occurrence is timed, so long episodes can't hide disagreement on short ones. | Four occurrences timed 10/12, 25/25, 8/6 and 40/36 s give 83.3%, 100%, 75%, 90%. Their mean is 87.1%. |
Conventions used on this page: in mean count-per-interval IOA, an interval where both observers counted 0 is 100% agreement for that interval; in mean duration-per-occurrence IOA, a row where both observers have 0 isn't an occurrence and is left out. Total count and total duration IOA can't be calculated when both totals are 0. Results are rounded to 1 decimal place. The same timing formulas work for latency and interresponse time data.
Common IOA mistakes
- Observers who aren't independent. If the observers can see each other's sheets, talk during the session or compare notes before they finish, the agreement is inflated. Each one records alone.
- Records that don't line up in time. Both observers have to watch the same session at the same time, start together and use the same interval length (one shared timer or synced timers). Otherwise interval 4 on one sheet isn't interval 4 on the other.
- Trusting total count IOA too far. Two observers who each count 10 responses get 100% total count IOA even if they counted different responses. Count-per-interval IOA, or trial-by-trial IOA, shows where they disagreed.
- Interval-by-interval IOA for rare or constant behavior. When most intervals are blank (or most are scored) on both sheets, agreement looks high almost automatically. Use scored-interval IOA for low-rate behavior and unscored-interval IOA for high-rate behavior.
- Dividing by the wrong number. Scored-interval IOA divides by the intervals where at least one observer scored +, not by every interval. Unscored-interval IOA divides by the intervals where at least one observer scored −.
- Mixing units. For duration IOA both observers need the same unit, such as seconds. 2 minutes against 110 seconds has to become 120 s against 110 s first.
- Ignoring observer drift. Over weeks, an observer's use of the behavior definition can slowly change. Regular IOA checks, plus retraining when agreement drops, catch it.
IOA on the RBT exam
The RBT Test Content Outline (3rd edition) puts this under Data Collection and Graphing, task A.8: “Describe the risks associated with unreliable data collection and poor procedural fidelity”. The exam is about why it matters more than the arithmetic: if the data are wrong, the supervisor may stop a program that works, keep one that doesn't, or make changes the learner doesn't need. IOA is how those data problems get caught. Expect questions on what IOA is, what low agreement means, and what to do about it (report it to your supervisor; never estimate or fill in data from memory).
To run an IOA check, record the session on two copies of the same sheet: make them with our free printable ABA data sheets. To practice scoring intervals the way a second observer would, try the interval recording trainer, or answer free Data Collection and Graphing questions.
Data Collection and Graphing is 13 of 75 scored RBT questions
It's area A of the RBT exam: measurement, summarizing data, reading graphs and the risks of unreliable data. Try free questions from that area, or take the 10-question RBT check for a score in every area, in about three minutes, with no sign-up.
IOA calculator FAQ
What is IOA in ABA?
IOA (interobserver agreement) is the percentage of agreement between two observers who independently record the same behavior at the same time. Supervisors use it to judge whether data are believable and consistent enough to base decisions on. High agreement does not prove the data are accurate (two observers can share the same mistake), but low agreement shows the data can't be trusted yet.
How do you calculate IOA?
Pick the method that matches how the data were recorded. For two totals (count or duration), divide the smaller by the larger and multiply by 100. For trials or intervals, divide the number of agreements by the number of trials or intervals compared and multiply by 100. For example, observers who agree on 18 of 20 intervals have 18 ÷ 20 × 100 = 90% interval-by-interval IOA.
What is a good IOA percentage?
An IOA of 80% or higher is commonly cited as the minimum acceptable level, and 90% or higher is often preferred (see Cooper, Heron & Heward, Applied Behavior Analysis, 3rd ed., chapter 5). It is a guideline, not a fixed rule: the supervisor also looks at the type of IOA used, how complex the behavior is, and what the data will be used for.
How often should IOA be collected?
A common recommendation is IOA in at least 20% of sessions, and more often 25% to 33%, spread across the phases and conditions of a program. Your supervising BCBA sets the schedule for each learner.
Which IOA should I use for interval recording?
Interval-by-interval IOA works for behavior that happens at a moderate rate. For a low-rate behavior, where most intervals are blank, scored-interval IOA is the stricter choice; for a high-rate behavior, where most intervals are scored, unscored-interval IOA is. Interval-by-interval IOA can look high in both cases just because the observers agree on the many easy intervals.
What is the difference between IOA and procedural fidelity?
IOA checks the data: do two observers' records of the same session agree? Procedural fidelity (also called treatment integrity) checks the teaching: was the plan carried out exactly as written? RBT task A.8 covers the risks when either one is poor, because the supervisor can no longer tell whether the plan works.
Does an RBT calculate IOA?
An RBT is often one of the two observers. The supervising BCBA plans the IOA checks, reviews the result and decides what to do about low agreement, such as clarifying the behavior definition or retraining. On the RBT exam, IOA comes up under task A.8, the risks of unreliable data collection.
Is the data I enter saved anywhere?
No. The calculator runs in your browser, and the numbers you type are not sent to QuizPrep or saved. Reloading the page brings back the examples.
More free RBT practice
85-question RBT mock exam, ABA data sheets, interval recording trainer, RBT flashcards and the RBT study guide.
Sources, checked November 18, 2026: Cooper, Heron & Heward: Applied Behavior Analysis (3rd ed., 2020), chapter 5, Improving and Assessing the Quality of Behavioral Measurement · BACB: RBT Test Content Outline (3rd ed.). For exam study and data practice, not clinical guidance: your supervising BCBA decides how and when IOA is collected for each learner. QuizPrep is an independent study tool, not affiliated with or endorsed by the BACB.