Latency vs Interresponse Time (IRT) in ABA: Definitions, Examples and Practice Questions
Latency is the time from a cue to the response; IRT is the time between two responses. One worked timeline, when each matters, and 8 practice questions.
By Praxis Notes Team Editorial team
Updated 18 min read
On this page
Latency is the time from the onset of a stimulus, such as an instruction, to the start of the response. Interresponse time (IRT) is the time between two consecutive responses of the same class, usually timed from the end of one to the start of the next (Cooper, Heron & Heward, Applied Behavior Analysis, ch. 4).
This page is for RBT and BCBA candidates first and for practitioners choosing a measure second. Its center is one fictional session scored for latency, duration and IRT on a single timeline, followed by when each measure matters clinically and 8 original practice questions. More explainers are in Behavior Analysis Concepts.
What is the difference between latency and IRT?
Latency starts the clock at a stimulus and IRT starts it at the previous response; both stop it when the next response begins. Cooper, Heron and Heward (Applied Behavior Analysis, ch. 4) class both as measures of temporal locus, meaning when a response happens relative to another event, while duration measures how long a response lasts.
Figure 1 marks where each timer starts and stops on one fictional timeline.

| Question | Latency | Interresponse time (IRT) | Duration (for contrast) |
|---|---|---|---|
| What it measures | Time from a stimulus to the start of the response | Time between two consecutive responses of the same class | How long one response lasts |
| Timer starts | When the stimulus begins (instruction, cue, event) | When one response ends (most clinical data sheets) | When the response starts |
| Timer stops | When the response starts | When the next response starts | When the response ends |
| Needs a stimulus? | Yes | No, only two responses | No |
| Typical goal | Shorten it (respond to instructions) or lengthen it (wait, delay problem behavior) | Lengthen it (space out a too-frequent behavior) or shorten it (fluency) | Lengthen (on task) or shorten (tantrums) |
| Summarized as | Mean, median and range of latencies | Mean, median and range of IRTs | Total or mean duration |
A quick test on any scenario: if the question names a cue and a response, it is latency; if it names two instances of the same behavior and the gap between them, it is IRT; if it asks how long one instance lasted, it is duration. Duration and frequency are compared in our page on frequency vs duration recording.
What is latency in ABA?
Latency in ABA is the elapsed time between the onset of a stimulus and the start of the response that follows it (Cooper, Heron & Heward, ch. 4). The stimulus is usually an instruction or another discriminative stimulus, but it can be any defined event, such as a timer going off or a peer saying hello. Latency, also called response latency, answers how quickly a learner initiates a behavior after a direction.
Fictional example: the RBT says “Get your shoes” at 0:00 and the learner starts walking toward the shoes at 0:06. The latency is 6 seconds. How long the learner then takes to put the shoes on is a separate measure (duration), not part of the latency.
To record latency:
- Write down the stimulus and the response onset exactly as the BCBA defined them (“timer starts when the instruction ends”; “response starts when the learner stands up”).
- Start the timer when the stimulus starts, as defined.
- Stop the timer the moment the response begins, not when it is finished.
- Record the seconds for that trial right away, with any prompt used.
- If no response occurs within the time limit in the program (for example, 30 seconds), record it the way the program says, usually as “no response,” not as a number you guessed.
The Praxis Learn RBT lesson on continuous measurement (A.1) gives the short version, with a quick check.
What is interresponse time (IRT) in ABA?
Interresponse time (IRT) is the time between two consecutive instances of the same response class (Cooper, Heron & Heward, ch. 4). On clinical data sheets and in many RBT training materials it is timed from the end of one response to the start of the next, which is how the Praxis Learn continuous measurement lesson defines it. The term is also written inter-response time or inter response time; all three mean the same measure.
Fictional example: an episode of vocal stereotypy ends at 10:05:10 and the next one starts at 10:05:40. The IRT is 30 seconds. No instruction is involved; IRT only needs two responses.
Is IRT measured from the end or the start of the first response?
IRT is usually timed from the end of one response to the start of the next in applied settings, but some research measures it from the onset of one response to the onset of the next, as the Praxis Learn BCBA lesson on temporal dimensions (C.4) notes. For brief responses, such as a key press or a single word, the two give almost the same number. For responses that last several seconds, they differ, so check which one your data sheet or study uses before comparing numbers. The worked example below shows both.
Where does the timer start and stop for each measure?
The latency timer starts at the stimulus, the IRT timer at the end of the previous response, and the duration timer at the start of the response itself; scoring one fictional session on a single timeline shows all three.
Part 1, latency. During a teaching block, the RBT gives five instructions. The learner’s latencies are 6, 4, 12, 3 and 5 seconds.
- Mean latency: 30 ÷ 5 = 6 seconds
- Median latency: 5 seconds
- Range: 3 to 12 seconds
Cooper, Heron and Heward note that latency and IRT are typically reported as the mean, median and range. Here the single 12-second trial pulls the mean above the median; the median and range keep one long trial from hiding what most trials look like.
Part 2, IRT and duration. During a 2-minute free-play observation (1:00 to 3:00), the RBT records four episodes of hand flapping:
| Episode | Starts | Ends | Duration | IRT, end to start (from previous episode) | IRT, start to start (from previous episode) |
|---|---|---|---|---|---|
| 1 | 1:00 | 1:04 | 4 s | n/a (first episode) | n/a |
| 2 | 1:19 | 1:21 | 2 s | 1:04 → 1:19 = 15 s | 1:00 → 1:19 = 19 s |
| 3 | 1:51 | 1:57 | 6 s | 1:21 → 1:51 = 30 s | 1:19 → 1:51 = 32 s |
| 4 | 2:07 | 2:09 | 2 s | 1:57 → 2:07 = 10 s | 1:51 → 2:07 = 16 s |
What the session gives, measure by measure:
- Frequency and rate: 4 episodes in 2 minutes, a rate of 2 per minute.
- Duration: total 14 seconds; mean duration 3.5 seconds per episode.
- IRT (end to start): 15, 30 and 10 seconds; mean 18.3 seconds, median 15, range 10 to 30.
- IRT (start to start): 19, 32 and 16 seconds; mean 22.3 seconds, median 19, range 16 to 32.
Figure 2 draws the same observation with both IRT conventions.

Two details students miss. Four responses give only three IRTs, because the first response has nothing before it. And the time from 2:09 to the end of the observation is not an IRT, because no next response closed it.
How are IRT and rate related?
IRT and rate move in opposite directions: when IRTs get shorter, more responses fit in the same time, so rate goes up; when IRTs get longer, rate goes down. Cooper, Heron and Heward describe IRT as a direct measure of temporal locus that is functionally related to rate.
IRT shows something rate hides: the pattern. In the session above, the four episodes came at uneven gaps of 15, 30 and 10 seconds. If the same four episodes had come exactly 30 seconds apart, the rate would still be 2 per minute, but the IRTs would be 30, 30 and 30 seconds. Same rate, different behavior: bursts versus steady spacing. That pattern is the main reason to collect IRT instead of only counting.
On an exam, a question about “how much time passes between instances” points to IRT; a question about “how many times per minute” points to rate.
When does latency matter clinically?
Latency matters whenever how soon a behavior starts is the problem, not whether it happens at all. A learner who follows every instruction, but only after 40 seconds, can score 100% on a measure that only records whether instructions are followed, yet still have a real problem in a classroom or on a street crossing.
Common latency targets in ABA programs:
- Following instructions and transitions (“Line up”, “Clean up”)
- Responding to name or to a greeting
- Safety responses, such as stopping or coming back when called
- Starting a task or the first step of a chain after the instruction
- Latency to problem behavior, where the goal is a longer latency
How do you decrease latency in ABA?
Latency often decreases when easy, high-probability instructions come just before the target instruction. In our reading, checking motivation first, reinforcing faster responses more than slower ones, giving the instruction clearly and once, and prompting with a time-delay procedure that is then faded also help. The BCBA chooses the procedure.
The research support is strongest for the high-probability instruction sequence: in Mace and colleagues’ (1988) studies, giving several easy instructions just before a hard one increased compliance and decreased compliance latency and task duration (Mace et al., 1988). The Praxis Learn glossary covers the high-probability request sequence, and our page on behavioral momentum and high-p sequences gives examples.
What is a latency-based functional analysis?
A latency-based functional analysis measures the time from the start of each test condition to the first instance of problem behavior, instead of counting responses for the whole session; shorter latencies suggest a stronger response in that condition.
Thomason-Sassi, Iwata, Neidert and Roscoe (2011) compared latency with rate across functional analyses and proposed latency as an index of response strength for behavior that is too dangerous to let continue, or that ends the session when it occurs, such as elopement (Thomason-Sassi et al., 2011). The session can end after the first response, which limits exposure to risky behavior.
When does IRT matter clinically?
IRT matters when the spacing between responses is the problem, or when you want to reinforce responses that are spaced out without stopping the behavior entirely.
- Differential reinforcement of low rates (DRL). In a spaced-responding DRL schedule, a response is reinforced only if a set minimum IRT has passed since the previous one. Lennox, Miltenberger and Donnelly (1987) used a 15-second spaced-responding DRL to slow rapid eating in three adults; eating dropped to target levels after a prompting procedure was added (Lennox et al., 1987). DRL is on both outlines (RBT D.3, BCBA G.2); see our differential reinforcement glossary.
- Behavior that is acceptable at a lower rate. Raising a hand, asking questions or asking for a break are fine, just not every few seconds. Lengthening IRTs keeps the behavior and spaces it out.
- Stereotypy and other repetitive behavior. When the learner’s team and family have chosen reducing it as a goal, IRTs show whether episodes are spreading out or still coming in bursts, which a session total can hide.
- Fluency. For an academic skill such as answering math facts, shorter IRTs between correct answers mean faster, more fluent responding.
How do you record and summarize latency and IRT data?
Record every latency or IRT in seconds as it happens, then summarize the session as a mean, median and range. A minimal data sheet for latency has one row per trial:
| Date (fictional) | Trial | Stimulus (as defined) | Latency (s) | Prompt | Notes |
|---|---|---|---|---|---|
| Day 1 | 1 | “Get your shoes” | 6 | None | |
| Day 1 | 2 | “Line up” | 4 | None | |
| Day 1 | 3 | “Clean up” | 12 | Gesture | Toy in hand |
For IRT, record the start and end time of every episode (as in the timeline above) and calculate the gaps afterward; that also gives you frequency, rate and duration from the same record.
Recording errors that make latency and IRT data unreliable:
- Starting the latency timer late, after the instruction is finished, when the definition says from the start of it (or the reverse)
- Stopping the latency timer when the response is completed instead of when it begins, which adds duration to latency
- No written definition of when a response “ends”, so each observer times IRT differently
- Writing data from memory after the session
- Recording “slow” or “fast” instead of seconds
Two observers check each other with interobserver agreement; for latency and IRT, Cooper, Heron and Heward (ch. 5) describe mean latency-per-response and mean IRT-per-response IOA, which compare each pair of timings. Our page on IOA formulas shows the duration-per-occurrence calculation these follow. Graph each session’s mean or median latency on a line graph and read it with the guide to level, trend and variability.
Which latency and IRT mistakes cost points on the exam?
The costliest mistakes mix up where each timer stops: latency stops when the response starts, IRT runs between two responses, and duration covers one response. The practice questions below test each of these:
- Latency vs duration. Latency ends when the response starts. If a question gives you a start time and an end time for the behavior, that difference is duration.
- IRT vs duration. IRT is the gap between two episodes; duration is the length of one episode.
- IRT vs latency. Latency needs a stimulus. If there is no instruction or cue in the scenario, only repeated behavior, it is IRT: seconds between instances of nose tapping is IRT, while how quickly a learner initiates a behavior after a direction is latency.
- IRT vs rate. Longer IRTs mean a lower rate. A question about how many responses per minute is about rate; one about time between instances is about IRT.
- Trials are not IRTs. In discrete-trial teaching each response follows a new instruction, so the time from instruction to response is latency, even though responses repeat.
- Counting IRTs. Four responses give three IRTs.
How do latency and IRT show up on the RBT and BCBA exams?
Latency and IRT are named in RBT task A.1 (continuous measurement) and BCBA task C.4 (temporal dimensions of behavior). Task codes below are from the RBT Test Content Outline (3rd ed.) and the BCBA Test Content Outline (6th ed.).
| Exam | Task | Wording (short) | How latency or IRT appears |
|---|---|---|---|
| RBT (3rd ed.) | A.1 | Implement continuous measurement procedures (e.g., frequency, duration, latency, interresponse time) | Pick and record the right measure for a scenario |
| RBT (3rd ed.) | A.6 | Calculate and summarize data in different ways (e.g., rate, mean duration, percentage) | Mean or median latency, rate from a record |
| RBT (3rd ed.) | D.3 | Implement differential reinforcement procedures (e.g., DRO, DRA, DRI, DRL, FCT) | DRL reinforces responses after a minimum IRT |
| BCBA (6th ed.) | C.4 | Measure temporal dimensions of behavior (e.g., duration, latency, interresponse time) | Definitions and calculations |
| BCBA (6th ed.) | C.9 | Select a measurement procedure to obtain representative data that accounts for the critical dimension of the behavior and environmental constraints | Choose latency or IRT when timing is the problem |
| BCBA (6th ed.) | G.2 | Design and evaluate differential reinforcement (e.g., DRA, DRO, DRL, DRH) | Setting and thinning a DRL IRT criterion |
Domain A has 13 of the 75 scored RBT questions, and Domain C has 21 of the 175 scored BCBA questions. Older materials use earlier codes: the RBT Task List (2nd ed., used through 2025) put continuous measurement under A-2, and the BCBA Task List (5th ed., used through 2024) had C-4, which kept its wording as C.4 in the 6th edition.
For study, the Praxis Learn lessons are continuous measurement (RBT A.1), calculating and summarizing data (RBT A.6), temporal dimensions (BCBA C.4) and selecting a measurement procedure (BCBA C.9). The RBT exam study guide covers the whole outline.
Practice questions on latency and IRT
These 8 items are original and fictional. Each answer names the task it practices.
1. A.1: Starting to work
An RBT says “Start your worksheet” and the learner picks up the pencil 9 seconds later, then works for 4 minutes. What is the latency?
- A. 4 minutes
- B. 9 seconds
- C. 4 minutes and 9 seconds
- D. It cannot be measured because there was no prompt
Answer and rationale
B. Latency runs from the instruction to the start of the response, 9 seconds. The 4 minutes of work is the duration of the response, so A and C add a different measure. A prompt is not needed; the instruction is the stimulus. Task A.1.
2. A.1: Spacing between questions
A BCBA wants to know how much time passes between a learner’s questions during group work, because he asks one every few seconds. What should the RBT record?
- A. Interresponse time between questions
- B. Duration of each question
- C. Latency from the teacher’s instruction to each question
- D. Whole interval recording
Answer and rationale
A. The time between two consecutive questions is interresponse time. Latency (C) would time from an instruction to a question, which does not measure the spacing between his questions, and duration (B) is how long each question lasts. Whole interval recording (D) is a discontinuous method and would not give the time between questions. Task A.1.
3. A.1: Where the timer stops
During a latency program, an RBT starts the timer at the instruction “Wash your hands” and stops it when the learner turns off the water. What went wrong?
- A. Nothing; the RBT recorded latency correctly
- B. The timer should have started when the learner reached the sink
- C. Latency cannot be measured for a chained skill like handwashing
- D. The timer should have stopped when the learner began washing
Answer and rationale
D. Latency stops at the start of the response. By stopping at the end, the RBT recorded latency plus duration in one number, so A is wrong. B moves the start point away from the stimulus, and C is false: the latency to the first step of a chain can be measured. Task A.1.
4. A.6: Summarizing latency
Across five trials, a learner’s latencies are 3, 5, 4, 20 and 3 seconds. What is the median latency?
- A. 3 seconds
- B. 4 seconds
- C. 7 seconds
- D. 20 seconds
Answer and rationale
B. Ordered, the latencies are 3, 3, 4, 5 and 20, so the middle value is 4 seconds. The mean is 7 seconds (C), pulled up by the single 20-second trial, which is why the median and range are reported alongside it. Task A.6.
5. A.1: Counting IRTs
An RBT records five episodes of hand mouthing in a 10-minute observation. How many interresponse times can be calculated?
- A. 4
- B. 5
- C. 6
- D. 10
Answer and rationale
A. Each IRT sits between two consecutive episodes, so five episodes give four IRTs. Five (B) would count a gap before the first episode, which has no previous response. Task A.1.
6. C.4: IRT and rate
Over several weeks of treatment, the mean IRT between a learner’s episodes of vocal stereotypy increases from 20 seconds to 90 seconds, with sessions of the same length. What most likely happened to the rate of vocal stereotypy?
- A. It increased
- B. It stayed the same
- C. It decreased
- D. It cannot change, because IRT and rate are unrelated
Answer and rationale
C. Longer gaps between episodes mean fewer episodes fit in the same session, so rate goes down. A reverses the relation, and D is wrong because IRT is functionally related to rate. Task C.4 (BCBA); RBT candidates practice the same idea under A.1.
7. G.2: Choosing a procedure
A learner raises his hand about every 20 seconds during class. The team wants him to keep raising his hand but no more than about once every 2 minutes. Which procedure fits?
- A. DRO with a 2-minute interval, reinforcing any 2 minutes without hand raising
- B. Extinction of hand raising until it stops occurring during class
- C. Latency recording from the start of class to the first hand raise
- D. Spaced-responding DRL, gradually raising the minimum IRT toward 2 minutes
Answer and rationale
D. A spaced-responding DRL reinforces a hand raise only when enough time has passed since the previous one, so the behavior stays but the IRTs lengthen; starting near his current IRTs and raising the criterion step by step keeps reinforcement available. DRO (A) reinforces the absence of the behavior, and extinction (B) would remove a behavior the team wants to keep. Recording latency (C) measures and does not change behavior. Task G.2 (BCBA); RBT D.3.
8. C.9: Choosing the measure
A learner’s elopement is dangerous, and staff must block it as soon as it starts, so it can occur only once per session. The BCBA wants to compare conditions in a functional analysis. Which measure fits best?
- A. Rate of elopement per session
- B. Total duration of elopement
- C. Latency from the start of each session to the first elopement attempt
- D. Interresponse time between elopement attempts
Answer and rationale
C. When the behavior can happen only once per session, rate (A) and IRT (D) have almost nothing to measure, and blocking keeps duration (B) near zero. Latency to the first response still differs between conditions, which is the logic of a latency-based functional analysis (Thomason-Sassi et al., 2011). Task C.9 (BCBA); RBT candidates meet functional analysis under B.3.
For more Domain A items, take the free RBT Data Collection and Graphing practice exam; BCBA candidates can use the Measurement, Data Display and Interpretation practice exam. If interval methods are next on your list, read partial interval, whole interval and momentary time sampling.
Sources
- Behavior Analyst Certification Board. RBT Test Content Outline (3rd ed.). 2023, updated September 2024. PDF
- Behavior Analyst Certification Board. BCBA Test Content Outline (6th ed.). 2022, updated September 2024. PDF
- Cooper, J. O., Heron, T. E., & Heward, W. L. (2020). Applied Behavior Analysis (3rd ed.). Pearson. Chapter 4, Measuring Behavior; Chapter 5, Improving and Assessing the Quality of Behavioral Measurement.
- Mace, F. C., Hock, M. L., Lalli, J. S., West, B. J., Belfiore, P., Pinter, E., & Brown, D. K. (1988). Behavioral momentum in the treatment of noncompliance. Journal of Applied Behavior Analysis, 21(2), 123–141. PMC
- Thomason-Sassi, J. L., Iwata, B. A., Neidert, P. L., & Roscoe, E. M. (2011). Response latency as an index of response strength during functional analyses of problem behavior. Journal of Applied Behavior Analysis, 44(1), 51–67. PMC
- Lennox, D. B., Miltenberger, R. G., & Donnelly, D. R. (1987). Response interruption and DRL for the reduction of rapid eating. Journal of Applied Behavior Analysis, 20(3), 279–284. PubMed
All examples and data on this page are fictional. Sources last checked on October 29, 2026.
Updated October 29, 2026: rewritten with a worked timeline, clinical uses, exam task codes from the current outlines and 8 practice questions, and merged with our latency documentation guide and our latency, IRT and duration guide.
Reviewed by the Praxis Notes team on October 29, 2026.