Interresponse time

IRT measures the gap between responses

If a response occurs at 2:10 and the next occurs at 2:42, the IRT is 32 seconds. A series of five response timestamps produces four within-series IRTs. The time before the first response and after the last response are boundary periods, rather than ordinary IRTs, unless the written method defines and labels another measure.

IRT can describe communication, practice attempts, workflow steps, health-related episodes, or staff actions. The selected response class and purpose should matter to the client and decision.

Define timestamps and boundaries first

Choose whether a response timestamp marks onset or completion. Use the same point for every response. Define how to handle:

• responses already underway when observation starts
• a response that crosses a session boundary
• breaks, blocked observation, or device downtime
• a new activity, partner, or setting
• rapid repetitions that may form one response burst
• days with no eligible opportunity

An empty record does not equal an infinite IRT or a zero. Report no observed response, no eligible opportunity, or missing observation according to the actual state.

Calculate IRT from consecutive timestamps

Suppose the same response begins at 0:40, 1:10, 2:00, and 3:20. The three IRTs are:

• 1:10 minus 0:40 = 30 seconds
• 2:00 minus 1:10 = 50 seconds
• 3:20 minus 2:00 = 80 seconds

The mean is (30 + 50 + 80) divided by 3, or 53.3 seconds. The median is 50 seconds. Preserve the individual intervals because a mean can hide a cluster of short gaps and one long gap.

IRT and rate are related under defined conditions

Shorter average gaps usually accompany a higher rate when responses form a continuing stream under comparable observation conditions. The reciprocal of a mean IRT can approximate rate when units and boundary rules fit.

The relation needs care. A 10-minute session with three responses contains two measured within-session IRTs plus time before the first and after the last. Ignoring those boundaries can make a reciprocal differ from count divided by total observed time. Pauses, bursts, changing opportunities, and irregular response patterns also matter. Report rate directly from count and observation time when rate is the target measure.

A fictional spacing example

Lila is a fictional college student who wants longer uninterrupted study periods while keeping break requests easy to use. With Lila, the team timestamps each independent break request made through speech, text, or AAC during a 45-minute study block.

Requests occur at minutes 6, 14, 19, 31, and 40. The four within-session IRTs are 8, 5, 12, and 9 minutes. Median IRT is 8.5 minutes, and mean IRT is 8.5 minutes. Request count is five and rate is 5 divided by 45, or 0.11 per minute after rounding.

The team keeps all measures and Lila’s report that the breaks felt well timed. It does not treat longer spacing as automatically better or infer that a support caused the pattern.

Summaries should match the distribution

Mean IRT uses every interval and can shift sharply because of an outlier. Median IRT identifies the middle and can describe a skewed distribution more clearly. Range, individual values, and a simple graph show variability that either single summary can hide.

For safety-sensitive or health-related responses, summarize short intervals separately when close clustering matters. A long average cannot erase a brief high-risk burst.

Protect access and context

Keep augmentative and alternative communication, interpreters, sensory and mobility supports, food, water, bathroom access, prescribed care, breaks, and emergency help available. Removing a support to lengthen an interval would change the context and could create risk.

Record opportunity, partner response, schedule, setting, health concerns, and other conditions that help interpret spacing. Pause nonemergency observation when assent is withdrawn or distress appears, following the applicable process.

Check data quality

Useful controls include valid timestamps divided by responses due for timestamps, records using the current definition, clock synchronization checks, excluded periods by reason, and observer agreement on response onset. Device precision should fit the behavior. Millisecond timestamps add little when observers cannot agree within several seconds.

Document the clock source so merged records use the same time basis.

The BACB BCBA Test Content Outline, Sixth Edition identifies interresponse time as a temporal dimension and also covers reliability, validity, and representative measurement selection. It describes exam content and does not choose IRT for an individual case.