An ABA latency data sheet records how much time passes between a defined starting event and a defined response. The number is meaningful only when the cue, timer rule, response definition, observation endpoint, and opportunity coverage travel with it. A row that says “12 seconds” but does not preserve what started and stopped the clock cannot be audited later.

The copyable worksheet below keeps raw opportunity records separate from calculated summaries. It also gives no-response, invalid, and interrupted opportunities their own codes instead of forcing every row into a numeric latency.

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

Treat this as a working draft under a qualified clinician’s measurement plan. The sheet does not select a target, cue, wait time, mastery threshold, prompt procedure, or treatment. A shorter latency is not proof of learning, willingness, compliance, or clinical improvement.

Define the clock before the first opportunity

Latency has two boundaries. The start boundary might be a clearly delivered direction, the presentation of an item, or another event identified in the plan. The stop boundary is the first observable response that meets the written definition. Both boundaries need enough detail that trained observers can apply them consistently.

The current RBT Test Content Outline, 3rd edition lists latency and interresponse time among continuous measurement procedures for the examination beginning in 2026. The outline also includes objective documentation, timely escalation of data irregularities, competence, supervision, and confidential-information requirements. Those examination topics do not choose a latency protocol for an individual.

A proposed measurement-selection model for problem behavior describes latency as elapsed time from a specified environmental event to a response and emphasizes matching the procedure to the dimension of interest. The paper supports stating both boundaries. Its discussion neither validates this blank form nor makes latency suitable for every question.

Measurement contract

Complete one contract for each version of the sheet. If the cue, target, timer, prompt rule, endpoint, setting, or observation purpose changes, close the old version rather than mixing unlike records.

Contract fieldVersioned definitionClinical or operational questionObservable response definitionValid opportunity or cueTimer start eventTimer stop eventPlanned observation endpointNo-response rule at the endpointPrompt, repetition, or cue-withdrawal ruleInterrupted or invalid opportunity ruleTiming device and resolutionSetting, routine, and schedule versionObserver role and training checkCommunication and accessibility supportsVersion, effective date, and reviewer

Write the endpoint before data collection. “No response by 60 seconds” is a bounded observation, not necessarily a 60-second response latency. Repeating a cue may begin a new opportunity, continue the same opportunity, or invalidate the row depending on the approved contract. The sheet cannot decide that rule after the result is visible.

Opportunity codes

Use a short code dictionary so blanks do not hide what happened.

CodeMeaningNumeric latency?Enters valid-opportunity denominator?RDefined response occurred before the endpointYesYesNRNo defined response before the planned endpointNo; retain endpointYesXOpportunity was interrupted or contract was not followedNoNoURequired cue, access, or observation was unavailableNoNo?Record needs timely clarificationNot until resolvedNot until resolved

An NR row is not zero seconds. An X or U row is not a failed response. If a local record system uses different codes, include its dictionary with any export.

Blank opportunity-level record

Plan the opportunities first when the clinical question requires a schedule. Record the cue and response as observed, not as reconstructed from memory at the end of the session.

OpportunityDate and timeContextCue delivered exactly?Start timestampResponse or endpointStop timestampCodeLatencyPrompt or interruption noteObserver12345678

Keep the raw timestamps when the authorized workflow permits them. A displayed latency can be rounded for review while the underlying difference remains reproducible. If device clocks or exported timestamps disagree, preserve the discrepancy and route it rather than selecting the preferred value.

Track opportunity coverage

A clean average can still represent an incomplete schedule. An ABA latency data sheet should show where opportunities were planned, actually available, and validly observed.

Planned blockIntended contextPlanned opportunitiesValid RValid NRInvalid or unavailableCoverage note

Coverage percentage equals valid opportunities divided by planned opportunities, multiplied by 100. Report the contexts and supports represented as well. A high percentage in one narrow routine does not show that the sample represents other people, tasks, places, or times.

Keep no-response records visible

The observation endpoint limits what is known. If the target did not occur before a 60-second endpoint, the record shows latency greater than 60 seconds within that opportunity. The eventual latency after observation stopped remains unknown. Substituting 60 as though the response occurred at that instant changes the record.

How censored no-response opportunities enter a formal analysis depends on the prespecified question and method. A simple mean of observed response latencies may be useful descriptively, but it excludes the no-response rows and must say so. Report the response count, valid opportunities, no-response count, endpoint, and coverage beside any average.

Research on response latency during functional analyses examined a specific assessment use and found an inverse relation between latency and response rate in its studies. That evidence does not make every latency measure a functional analysis, authorize a hazardous exposure, or establish a universal conversion between latency and rate.

Calculation panel

Review validity before calculating. Do not exclude a valid long latency merely because it changes the average.

SummaryCalculationResultPlanned opportunitiesCount scheduled rowsValid opportunitiesR plus NROpportunity coverageValid / planned x 100Responses observedCount of RNo response by endpointCount of NR, with endpointResponse-occurrence percentageR / valid opportunities x 100Sum of observed response latenciesAdd numeric R latenciesMean observed-response latencySum / count of R rowsMedian observed-response latencyMiddle ordered R value(s)Range of observed-response latenciesMinimum to maximum R value

Name the population behind each denominator. “Mean latency 15.0 seconds among six observed responses” is different from “mean latency across seven valid opportunities,” which cannot be calculated by ordinary substitution when one row is right-censored.

Fictional completed example

Case L, the target, and every value below are fictional. For arithmetic only, a valid cue is a card placed flat on a marked surface while the observer says one defined phrase. The response is the first contact of the participant’s index finger with the card. The fictional plan ends observation at 60 seconds and retains no response as NR.

Eight opportunities were planned. Six produced defined responses at 8, 12, 17, 25, 10, and 18 seconds. One valid opportunity reached the 60-second endpoint without the response. One opportunity was invalid because the card fell before the cue was complete.

OpportunityCodeNumeric latencyNote1R8 sValid cue and response2R12 sValid cue and response3R17 sValid cue and response4R25 sValid cue and response5NRNot recordedNo response by 60-s endpoint6R10 sValid cue and response7XNot recordedCard fell during cue8R18 sValid cue and response

There are seven valid opportunities out of eight planned, so opportunity coverage is 7 divided by 8, or 87.5%. Six responses among seven valid opportunities gives 85.7% response occurrence when rounded to one decimal place. Keep that occurrence percentage separate from latency.

The six numeric response latencies total 90 seconds. Their mean is 90 divided by 6, or 15.0 seconds. In order, the values are 8, 10, 12, 17, 18, and 25; the median is the average of 12 and 17, or 14.5 seconds. The range is 8 to 25 seconds.

The NR row is reported as no response by 60 seconds. Entering 60 into the numeric set would produce a different mean of 21.4 seconds and falsely imply that a response occurred at the endpoint. Entering the invalid row as zero would create another unsupported result.

Review what the summary can and cannot say

The completed example supports a bounded description: six defined responses were observed among seven valid opportunities, and their observed latencies had a mean of 15.0 seconds. Those data cannot show why the response occurred, whether the cue was appropriate, whether the person preferred the activity, whether prompting was effective, or whether change was socially or clinically important.

Compare records only when the cue, response, endpoint, prompting, timing resolution, contexts, and opportunity rules are sufficiently aligned. If they are not, keep them in separate versions and discuss the difference rather than merging the rows.

Corrections and version history

Date and timeRow or fieldOriginal valueCorrected valueEvidence and reasonAuthorReviewer

Preserve the original value and an audit trail. A late clarification may resolve a transcription error, but memory alone should not invent a missing timer boundary. Record a version change when the cue, endpoint, response definition, timer, observer workflow, or support arrangement changes.

Privacy, access, and client rights

The BACB ethics hub points readers to the current codes, and the Ethics Code for Behavior Analysts addresses competence, client involvement, documentation, confidentiality, and risk within its scope. The BACB regulates its certificants and applicants; the code is not a substitute for an organization’s complete legal or records program.

The CASP ABA Practice Guidelines access page describes Version 3.0 as guidance for planning, implementing, and evaluating ABA treatment for autism and explains its licensing conditions. This page relies only on the public summary, not on inaccessible licensed procedures.

Maintain ordinary communication and accessibility supports. ASHA’s AAC practice portal states that AAC users should always have access to their communication tools or devices. Do not remove communication, mobility, sensory, medical, food, water, bathroom, or safety supports to manufacture a latency opportunity.

Store identifiable records only in approved systems. HHS risk-analysis guidance explains that covered organizations should assess risks to the confidentiality, integrity, and availability of electronic protected health information. No spreadsheet, paper form, or consumer timer becomes compliant merely because it appears in this worksheet.

Stop and route concerns

Pause the timing task and follow the current safety, clinical, supervisory, and emergency process when participation changes, distress appears, an ordinary support is unavailable, the cue cannot be delivered as defined, the timer fails, a response boundary is ambiguous, or immediate care is needed. Do not extend a waiting period simply to obtain a number.

Route unresolved records promptly. A qualified reviewer should decide whether the opportunity can be clarified, must remain invalid, belongs to another version, or requires a different measurement approach.

Related resources

Sources