Interresponse time in ABA (IRT) is the elapsed time between the end of one defined response and the beginning of the next defined response. If a learner finishes one vocal response at 10 seconds and begins the next vocal response at 22 seconds, the IRT is 12 seconds. IRT answers the question, “How much time passed between responses?” It does not answer how long a response lasted or how long it took to begin after an instruction.
IRT is useful when the spacing or pace of repeated behavior matters. A behavior analyst may examine whether responses are becoming closer together, farther apart, or more evenly distributed across an observation period. The value is only meaningful when the response definition and the two timing boundaries are clear. This guide explains the formula, worked examples, data-collection rules, comparisons with other ABA measures, and common BCBA exam traps.
What interresponse time means in ABA
Interresponse time is a continuous temporal measure of the spacing between consecutive responses. The first response has an offset, the next response has an onset, and the interval between those two events is the IRT. For a repeated vocal behavior, the response might be each clearly defined utterance. For a discrete motor response, it might be each completed button press, hand raise, or instance of a target action.
The response definition has to state what counts as an onset and what counts as an offset. “Talking” is usually too broad if the observer cannot tell when one response ends and the next begins. A stronger definition might specify that one response is a clearly audible word separated from the next word by at least a defined pause, or that one completed hand raise ends when the hand returns below the specified position. The exact boundary depends on the target and the protocol.
IRT can help answer questions such as:
- Are repetitive responses occurring with shorter or longer pauses?
- Is a skill becoming more fluent, with responses closer together?
- Is a behavior decreasing in rate because the pauses between responses are increasing?
- Does an intervention change the spacing of behavior even when the total count is similar?
A short IRT means the next response began soon after the prior response ended. A long IRT means more time passed. Neither direction is automatically desirable: shorter IRTs may reflect fluency for a skill, while longer IRTs may be the goal when the behavior is unsafe, disruptive, or too frequent. Interpret the number against the behavior goal, accuracy, prompting, and context.

IRT formula and a 12-second example
The basic formula is:
IRT = onset time of Response 2 − offset time of Response 1
Suppose Response 1 ends at 10 seconds and Response 2 begins at 22 seconds. The calculation is 22 s − 10 s = 12 s. The 12 seconds is the interresponse time. It is not 22 seconds, because the timer does not start with the antecedent or the beginning of the observation period. It is also not the time from Response 1 onset to Response 2 onset unless that different interval is explicitly defined by the protocol.
Another example: a learner completes a sorting response at 34 seconds and begins the next sorting response at 42 seconds. The IRT is 42 − 34 = 8 seconds. If the first response began at 28 seconds and lasted six seconds, the duration was six seconds, but the IRT was eight seconds. These are separate dimensions even though both values come from the same observation.
When response boundaries are close together, observers should use a timer or data system with enough precision for the decision being made. The team should document whether times are rounded to whole seconds, tenths, or another unit. Rounding rules should be consistent across observers and sessions.
Worked IRT data example
Imagine that the target is a discrete response with a clear offset. An observer records four consecutive response pairs during one observation. The table separates the boundary times so another person can reproduce each calculation.
| Response pair | Prior response offset | Next response onset | IRT | Calculation |
|---|---|---|---|---|
| 1 → 2 | 10 s | 22 s | 12 s | 22 − 10 |
| 2 → 3 | 29 s | 34 s | 5 s | 34 − 29 |
| 3 → 4 | 41 s | 49 s | 8 s | 49 − 41 |
| 4 → 5 | 57 s | 72 s | 15 s | 72 − 57 |
The mean IRT is (12 + 5 + 8 + 15) ÷ 4 = 10 seconds. The median is also 10 seconds because the ordered values are 5, 8, 12, and 15. Reporting the individual values along with a summary is useful because the mean alone can hide a long pause or a cluster of rapid responses. A graph can show whether the spacing is changing across sessions, while the raw values preserve the timing detail.
If the response definition or observation conditions change, do not compare the numbers as though they were collected under the same rule. A 10-second IRT for one response definition may not be equivalent to a 10-second IRT for another. Note prompts, materials, task demands, and any planned consequence that could affect the opportunity to respond.
How to record IRT step by step
IRT recording is more than pressing a stopwatch between two visible events. The measurement plan should make the boundaries and the treatment of ambiguous cases explicit before data collection begins.
- Define one response. State the observable onset and offset, including how separate responses are distinguished from one extended response.
- Choose the observation window. Identify when recording begins and ends, and whether the observer will record every eligible response in that window.
- Record the first response. The first response establishes the starting boundary for the first possible IRT, but it does not itself produce an IRT.
- Mark each response offset and the next response onset. Calculate the elapsed gap using the same timing unit and rounding rule.
- Document context. Record prompts, interruptions, materials, response quality, and other protocol variables that affect how the response is defined.
- Repeat the rule across observers. Practice borderline examples so the team agrees about when a response ends and the next one begins.
For a skill that should become fluent, the team might track IRT together with accuracy. A shorter pause is not an improvement if responses become incorrect. For a behavior targeted for reduction, the team may look for longer IRTs, but should also review the total count, duration, intensity, and whether the behavior changed form.
First response, no next response, and interruptions
Several edge cases can change the meaning of an IRT data set:
- First response: there is no prior response within the observation window, so the first response has no preceding IRT. Do not use time since the session began unless the protocol is measuring latency or another interval.
- Last response: if no next response occurs before the session ends, there is no completed IRT. Record “no next response” or a planned censored value rather than quietly entering zero.
- No response: no IRT can be calculated when there is only one or no target response. A no-response code is different from an IRT of zero seconds.
- Interrupted observation: if materials are removed, the learner leaves, or the protocol pauses, mark the interruption and apply the prewritten rule for whether the response sequence resumes.
- Overlapping or continuous behavior: if an action has no clear offset, define a reliable boundary or choose a different measurement dimension. Do not manufacture an offset simply to obtain an IRT.
Some data systems use a maximum observation window after a response and label the next IRT as censored when the next response does not occur. That can be useful for analysis, but the rule must be stated in advance and kept separate from an observed IRT. The exact coding method should follow the written measurement protocol and the decision the team needs to make.
IRT vs. latency, duration, rate, and interval recording
Many ABA measures involve time, but their start and stop boundaries differ. The continuous measurement in ABA guide provides broader context. Use the comparison below to identify the question before choosing a method.

| Measure | Start and stop boundaries | Question it answers | Common mistake |
|---|---|---|---|
| IRT | Prior response offset → next response onset | How far apart are responses? | Starting at the instruction |
| Latency | Antecedent or opportunity → response onset | How long until the response starts? | Waiting until response offset |
| Duration | Response onset → response offset | How long does one response last? | Measuring the gap between responses |
| Frequency or rate | Count responses, optionally divided by time | How many responses occur, and how fast? | Assuming count reveals spacing |
| Interval recording | A defined observation interval and scoring rule | Did behavior occur under this interval rule? | Treating an interval percentage as exact IRT |
Latency and IRT are especially easy to confuse. Latency starts with an antecedent, instruction, or opportunity and ends when a response begins. IRT starts with one response and ends when the next response begins, using the prior response’s offset as the starting boundary. For a side-by-side latency example, see the latency in ABA guide. For the count-versus-rate distinction, see the frequency recording guide.
Data quality and treatment decisions
IRT data are only as reliable as the boundaries used to create them. Two observers can record the same session and produce different values if one treats a pause as the end of a response while the other treats it as part of the same response. Operational definitions, practice videos or live examples, and paired observations help reveal these disagreements.
A practical quality review should:
- Compare observers’ response-onset and response-offset decisions before comparing arithmetic.
- Include short, long, interrupted, prompted, and borderline response sequences in practice checks.
- Inspect unusual IRTs against session notes instead of deleting them automatically.
- Keep no-next-response and censored cases identifiable rather than mixing them with observed values.
- Graph IRT with the direction of the treatment goal clearly labeled.
- Pair IRT with count or rate, duration, accuracy, and intensity when one timing variable could mislead the decision.
For example, an intervention could increase the average IRT while also lowering the total number of responses, which may be desirable for a repetitive behavior. In another case, a longer IRT could reflect confusion or missing opportunities rather than progress. The analyst should interpret the time measure alongside the operational goal and the conditions under which responses occurred.
BCBA exam traps and quick review
On a BCBA-style question, underline the exact event that starts the interval and the exact event that ends it. If the stem asks for time between the end of one response and the beginning of the next, choose IRT. If it asks for time from an instruction to response onset, choose latency. If it asks how long one response continues, choose duration. If it asks how many responses occur per unit of time, choose rate.
- Trap 1: using the next response’s onset minus the prior response’s onset. That is not the standard offset-to-onset IRT definition used in this guide.
- Trap 2: counting the time from the instruction to the first response as an IRT. That is a latency when the instruction is the start event.
- Trap 3: entering zero when no second response occurs. No next response is not the same as an observed zero-second gap.
- Trap 4: treating a shorter IRT as automatically better. The desired direction depends on the behavior and treatment goal.
- Trap 5: forgetting that response definitions control the timing boundaries. A vague response cannot produce dependable IRT data.
Quick review: IRT is the elapsed time from the offset of one defined response to the onset of the next defined response. Write the response boundaries, apply the formula consistently, define the rule for the first and last response, preserve no-response cases, and compare IRT with latency, duration, frequency/rate, and interval recording before selecting an answer.
Want more timed practice? Try the free BCBA mock exam and check the explanation after each item to confirm that you identified the measurement dimension before calculating anything.





