An ABA interresponse time data sheet begins with event timestamps, not a list of already calculated gaps. Interresponse time, often shortened to IRT, describes elapsed time between successive defined responses under a stated boundary convention. The source events, observation windows, missing records, and calculation rule must remain visible so another reviewer can reproduce the result.

This worksheet keeps the event log and the adjacent-response calculations in separate layers. That separation matters when an event is missed, two observation windows are separated by a long break, or the response definition changes.

Clinicians & ABA Professionals / Data, Outcomes and Clinical Decision-Making.

Use the form only within a qualified clinician’s measurement plan. The worksheet cannot select the response, define a reinforcement schedule, set a desirable spacing threshold, or decide treatment. A long or short gap alone does not establish function, cause, progress, risk, or clinical significance.

State what counts as an adjacent pair

The current RBT Test Content Outline, 3rd edition names interresponse time as a continuous measurement procedure in the 2026 examination content. The same outline addresses observable descriptions, documentation, data irregularities, competence, supervision, and confidential information. It is a training and examination reference, not an individualized protocol.

The measurement-procedure selection model notes that IRT can be calculated from event recording and stresses choosing measures that represent the relevant dimension. That general point supports preserving the event sequence. No target gap, schedule parameter, or specific worksheet follows automatically from it.

Decide which timestamp convention is in use. This page’s fictional example uses response onset to the next response onset. Another approved plan may define a different boundary for a particular question. Never place values from different conventions in one series without a clear, reviewed transformation.

Measurement contract

Contract fieldVersioned definitionClinical or operational questionObservable response definitionEvent timestamp boundary: onset, offset, or otherRule for simultaneous or overlapping responsesObservation-window start and stopFirst event after window startLast event before window stopCross-window gap ruleMissing or uncertain event ruleTiming device, clock source, and resolutionPlanned contexts and coverageObserver role and training checkCommunication and accessibility supportsVersion, effective date, and reviewer

Close the version when any field that changes the meaning of a pair changes. A new device with coarser resolution, a revised response boundary, or a different observation-window rule can make the resulting IRTs unlike the earlier series.

Raw event timestamp log

Record events in time order within each valid observation window. Give every event a stable identifier so the derived pair can point back to its two source rows.

WindowEvent IDDateObservation elapsed time or timestampResponse met definition?Context and support noteObserverValidity note

Retain valid zero-event windows in the coverage table even though they produce no IRT pairs. An empty event log without a coverage record cannot distinguish “no responses occurred” from “nobody observed.”

Derive IRTs from adjacent valid events

WindowPair IDEarlier event ID and timeLater event ID and timeLater minus earlierInclude?Exclusion or review reason

For an onset-to-onset convention, subtract the earlier onset from the next onset in the same valid window. Do not subtract the observation start from the first response and call it IRT; that value is first-response latency. Do not subtract the last event of one session from the first event of tomorrow unless a qualified plan explicitly defines continuous observation across that entire gap and the coverage actually exists.

Preserve observation coverage

WindowPlanned start and stopActual valid observationContextEvent countIRT pairsMissing or interrupted coverage

An observation with one event contributes an event count but no within-window IRT. An observation with four valid events contributes three adjacent pairs. Keep those structural facts visible rather than interpreting the lack of a pair as an infinitely long gap.

Research on procedures and accuracy in discontinuous measurement illustrates that the collection procedure affects the resulting data. IRT requires a valid sequence of event records; an interval grid that only says whether an event occurred cannot reconstruct exact within-interval event times.

Missing events can change two gaps

If one event between two recorded events was missed, the apparent gap combines two true IRTs. Do not divide the large value in half or insert an estimated event. Mark the affected pair for review, preserve the source timestamps, and use the approved missing-data rule.

Timing resolution matters as gaps become short. A device that records only whole minutes cannot support second-level IRTs. Simultaneous timestamps may reflect genuinely coincident events, a response-definition problem, or clock precision; the contract must determine how to retain and review them.

The broad single-case methods review by Lobo and colleagues provides research-design and measurement context. A descriptive IRT series does not become experimental evidence through that citation, nor does the review establish a universal sample size or replace direct graph review and qualified interpretation.

Calculation panel

SummaryCalculationResultPlanned observation windowsCount scheduled windowsValid observation windowsCount completed under contractEvent countCount valid source eventsValid within-window IRTsCount included adjacent pairsTotal IRTAdd included gapsMean IRTTotal / number of included IRTsMedian IRTMiddle ordered IRT value(s)IRT rangeMinimum to maximum included gap

Use the pair count, not the event count, as the denominator for mean IRT. Five included gaps sum to 235 seconds; their mean is 235 divided by 5. Dividing by seven source events would answer no defined question.

Fictional completed example

Case T, the response, and all times are fictional. For the demonstration, the defined event is the onset of one contact with a tabletop switch after at least two seconds without contact. The fictional contract uses onset-to-onset IRT within each valid observation window and excludes gaps between windows.

Window A begins at elapsed time zero and contains valid onsets at 15, 55, 95, and 160 seconds. Its three within-window IRTs are 40 seconds, 40 seconds, and 65 seconds. Window B begins with a new elapsed-time clock and contains onsets at 20, 50, and 110 seconds. Its two within-window IRTs are 30 seconds and 60 seconds.

WindowEarlier to later eventCalculationIRTAA1 to A255 - 1540 sAA2 to A395 - 5540 sAA3 to A4160 - 9565 sBB1 to B250 - 2030 sBB2 to B3110 - 5060 s

The five valid IRTs total 235 seconds. Their mean is 235 divided by 5, or 47.0 seconds. Ordered values are 30, 40, 40, 60, and 65 seconds, so the median is 40.0 seconds and the range is 30 to 65 seconds.

The 20 seconds from the start of Window B to event B1 is first-response latency, not IRT. The elapsed real-world time between A4 and B1 is also excluded because observation did not continue across the boundary. Neither number disappears: each belongs in the appropriate boundary or coverage record instead of the five-gap series.

Interpret spacing without inventing a cause

This example supports a limited sentence: seven defined response onsets across two valid windows produced five within-window IRTs with a median of 40.0 seconds. Those gaps do not show why the responses were spaced that way, whether an intervention caused the pattern, or whether the spacing was desirable.

Inspect the timestamp plot, response count, observation duration, contexts, supports, and phase information beside the IRT summary. A mean can hide alternating short and long gaps. A missed event can inflate a gap. A session boundary can create a false extreme. These are reasons to review the record, not automatic reasons to delete an inconvenient value.

Corrections and version control

Date and timeEvent or pairOriginal valueCorrected valueSource evidence and reasonAuthorReviewer

Correct source events before regenerating affected pairs. Preserve the original timestamp, correction evidence, author, reviewer, and downstream effect. If one event moves, the pair before it and the pair after it may both change.

Professional, privacy, and access boundaries

The BACB ethics-code hub identifies the current codes, while the Ethics Code for Behavior Analysts addresses competence, documentation, confidentiality, client involvement, and risk for covered certificants and applicants. Organizational forms and records workflows still require their own review.

The public CASP ABA Practice Guidelines page describes the scope and licensing of its Version 3.0 treatment guidelines. This worksheet does not reproduce licensed content or claim that CASP endorses its fields.

Preserve communication and accessibility. ASHA’s AAC resource says AAC users should always have access to their tools or devices. Do not withhold communication, medical, sensory, mobility, relationship, food, water, bathroom, or safety supports to create or lengthen an IRT.

Identifiable timestamps may reveal routines and service details. HHS guidance on risk analysis describes assessing risks and vulnerabilities to electronic protected health information for regulated organizations. Follow the organization’s authorized collection, access, retention, export, and disposal process; this page does not declare any storage medium compliant.

Stop and route concerns

Stop measurement and use the current safety or emergency process when care is needed, participation changes, distress appears, ordinary supports are unavailable, the observation boundary is lost, or the timer fails. Never delay assistance to preserve an adjacent pair.

Route clock drift, missing event rows, ambiguous definitions, duplicate timestamps, cross-window records, or inconsistent devices to the responsible supervisor. Keep the uncertainty visible until a qualified reviewer resolves it.

Related resources

Sources