Interresponse time observation boundaries require separating time between successive responses from time before the first and after the last observed response. Calculate each IRT from one response onset to the next under the defined rule. Boundary periods are censored by the observation window and should not be inserted as complete IRTs. Preserve timestamps, exposure, and rate.

Define the response and timestamp rule

Interresponse time measures elapsed time between successive responses. Define the response, onset rule, offset if relevant, minimum separation, unit, observation window, and timestamp precision before collection. Onset-to-onset IRT is common, but another stated rule may fit a specific response. Use the same rule throughout the sample.

Osei's worksheet records response onsets in minutes from the start of a 12-minute observation. Each timestamp remains linked to the source record and context. The observation start and end are exposure boundaries, not response events.

Build the event and boundary record

Keep observation start, every response timestamp, observation end, missed or uncertain events, pauses, source, observer, setting, and procedure version. Sort events chronologically and inspect duplicates, clock drift, rounded times, overlapping definitions, and entries outside the observation window.

Create IRT rows only between verified successive responses. Also retain the pre-first-response boundary and post-last-response boundary in separate fields. These boundary periods may contain incomplete information about an interval that began before observation or continued after it.

Calculate Osei's complete IRTs

Response onsets occur at minutes 2, 5, and 9.

  1. First complete IRT = 5 minus 2 = 3 minutes.
  2. Second complete IRT = 9 minus 5 = 4 minutes.
  3. Number of complete IRTs = number of responses minus 1 = 3 minus 1 = 2.
  4. Mean IRT = (3 + 4) / 2 = 3.5 minutes.

The median is also 3.5 minutes, and the range is 1 minute. Report the individual IRTs because a mean alone hides their distribution, especially in longer samples.

Keep observation boundaries separate

The first response occurs two minutes after observation starts. That two-minute period is not a complete IRT because the previous response, if any, is outside the observation window. The observation ends three minutes after the response at minute 9. That final three-minute period is also incomplete because the next response, if any, is unobserved.

Label these periods as boundary or censored time under the worksheet rule. Do not insert 2 and 3 into the IRT mean. Doing so would change the question from time between observed successive responses to a mixture of complete and boundary intervals.

Compare IRT with rate carefully

Osei has three responses during 12 observed minutes:

Rate = 3 / 12 = 0.25 responses per minute

The reciprocal of mean IRT is:

1 / 3.5 = approximately 0.286 responses per minute

The values differ because rate uses all 12 minutes and three responses, while the mean IRT uses two complete gaps spanning the seven minutes between the first and last responses. Boundary time is present in rate exposure but absent from the complete-IRT denominator. Neither result is an arithmetic error. State which question each measure answers.

Handle zero, one, and uncertain responses

An observation with zero responses has a rate of zero under valid full exposure but no complete IRT. An observation with one response also has no IRT pair. Report the response count, exposure, and boundary information without assigning an infinite IRT or using the whole session as a completed gap.

If a response timestamp is missing or uncertain, the adjacent IRTs may be invalid. A missed middle response can combine two shorter gaps into one falsely long IRT. Record visibility loss, technical failure, safety response, and observation pauses. Repeated missingness can make the IRT distribution unrepresentative.

Graph events and gaps in time order

Use an event plot or time-series display that shows response onsets, observation boundaries, pauses, and each complete IRT. A companion table can list event number, timestamp, prior timestamp, complete IRT, boundary state, and source. Keep rate and IRT in separate columns with their denominators.

During a valid 12-minute observation, Osei responded at minutes 2, 5, and 9. Complete IRTs were 3 and 4 minutes, with a mean of 3.5 minutes. The first 2 minutes and final 3 minutes were retained as boundary periods. Overall rate was 0.25 responses per minute.

Lobo and colleagues provide single-case measurement and analysis context. Pritchard and colleagues and Wirth and colleagues illustrate why measurement procedures and parameters need explicit review.

Review meaning and access with Osei

Discuss whether spacing between responses is a meaningful measure for Osei and the clinical question. Maintain AAC, interpreters, mobility supports, food, water, bathroom access, prescribed care, rest, relationships, and emergency help during measurement and review. ASHA supports ongoing AAC access.

If responses occur through AAC, ensure the timestamp rule fits the person's access method and does not treat ordinary device navigation or partner delay as client failure. The responsible clinician should consider opportunity, context, burden, risk, implementation evidence, and Osei's perspective.

Keep professional sources in scope

The BACB ethics hub and CASP public summary provide professional context. The BCBA Test Content Outline includes temporal dimensions, validity, reliability, and representative measurement as examination content. These sources do not establish a universal IRT target or boundary rule.

Clinician review checklist

For interresponse time observation boundaries, confirm the decision question, response definition, timestamp rule, precision, observation start and end, event sequence, duplicates, missing events, complete IRT count, each subtraction, mean denominator, rate numerator and exposure, boundary states, pauses, rounding, graph, report language, access, client perspective, qualified decision owner, and review date.

Limits of IRT at observation boundaries

Observed IRTs omit the unknown gaps crossing the beginning and end of the observation. Short sessions, low response counts, missed events, rounded timestamps, and opportunity changes can distort the distribution. Mean IRT is undefined without a complete response pair and is not generally interchangeable with the reciprocal of session rate. IRT alone does not establish measurement validity, benefit, future spacing, treatment integrity, or cause. Preserve events, exposure, boundaries, and raw context.

Related resources

Sources