An ABA discontinuous measurement comparison worksheet helps a clinician compare what continuous measurement, partial-interval recording, whole-interval recording, and momentary time sampling would preserve or distort for one defined target. It should connect the target's relevant dimension and intended decision to a small pilot, document observer burden and discrepancies, and leave the final selection to qualified clinical review rather than producing an automatic recommendation.

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

The practical problem is rarely a lack of definitions. A team may know how each method works yet still choose one because it is already in the data system, because the interval timer feels manageable, or because a percentage looks familiar. Used carefully, the ABA discontinuous measurement comparison worksheet slows that choice down. It makes the team name the behavior, the dimension of interest, the decision the data must support, and the evidence that a proposed recording method can actually produce under representative conditions.

The tool is a planning aid, not a validation study. A brief pilot can reveal an unworkable interval, ambiguous scoring rule, or predictable gap in observer availability. It cannot establish that a method is accurate for every client, setting, behavior, or decision.

Start with the dimension the decision needs

A recording method should follow the clinical question, not the other way around. If the question concerns how often discrete responses occur, count or rate may be the direct dimension. If it concerns how long a response lasts, duration may be direct. Latency, interresponse time, trials, and permanent products may fit other questions. The existing measurement-system selection guide and temporal measurement protocol guide explain those choices in more detail.

The BACB Ethics Code places responsibility on behavior analysts to select and correctly implement data-collection procedures and to use data when evaluating services. The code does not designate one interval method or supply a universal interval length. The worksheet therefore records who is responsible for the clinical choice and what local requirements also apply.

Write the decision in plain language before comparing methods. Examples include evaluating whether a teaching procedure is associated with a change in independent responding, monitoring the amount of time a safety-relevant response occurs, or describing engagement during a defined routine. Avoid vague goals such as "get better data." A specific decision makes it easier to notice when a proposed percentage does not measure the needed dimension.

Give continuous measurement a real feasibility rehearsal

Continuous measurement is not automatically feasible or superior in every service context, but it deserves more than a hypothetical dismissal. A practice-focused critique of measurement conventions argues for measuring response dimensions directly when possible and describes the limitations of relying on discontinuous observation merely because it is familiar (Current Measurement in Applied Behavior Analysis).

Run a bounded rehearsal under conditions that resemble actual service delivery. Define the observation window, competing clinical duties, device or paper workflow, environmental visibility, and any planned observer rotation. Ask the observer to record the direct dimension while still carrying out the duties that would be present during real implementation. Then document what was and was not captured. Do not record protected health information in this generic worksheet unless the organization's approved system, minimum-necessary rules, permissions, and retention controls allow it.

A rehearsal can support a statement such as "the observer missed event offsets while implementing the teaching sequence." It does not prove that continuous recording is impossible. A different observer assignment, simpler interface, narrower observation window, automated timestamp, or revised target definition may change feasibility. Preserve those possible changes rather than converting one difficult rehearsal into a permanent rule.

Keep the four candidate rules distinct

Use the comparison table only after the target is operationally defined and observers are trained to the proposed rules.

CandidateWhat the observer recordsWhat the resulting percentage representsCommon interpretation risk to discussContinuous event or time recordingEach defined response, onset, offset, or other direct dimension across the observationA count, rate, duration, latency, or other directly specified dimensionObserver demands or system limitations may produce missed or imprecise recordsPartial-interval recordingWhether the response occurred at any time during each intervalPercentage of usable intervals with any occurrenceThe percentage is not percentage duration and may overstate occurrence relative to time engagedWhole-interval recordingWhether the response occurred throughout each complete intervalPercentage of usable intervals with continuous occurrenceThe percentage is not percentage duration and may understate occurrence relative to time engagedMomentary time samplingWhether the response is occurring at the designated sample momentPercentage of usable sampled moments with occurrenceBrief or infrequent responses may be missed, and the estimate depends on the sampling arrangement

The exact rules belong in the protocol. "Occurred in the interval" is not enough. For partial-interval recording, specify whether any qualifying occurrence scores the interval and how boundary events are handled. For whole-interval recording, specify what counts as continuous occurrence and how brief interruptions are scored. For momentary time sampling, identify the sample instant and what happens if the observer cannot see the target at that instant.

Treat published findings as evidence, not conversion rules

Discontinuous methods divide observation into intervals or sample behavior at selected moments. They can ease some observer demands, but they introduce measurement error. An analysis of more than 800 sessions found greater correspondence with continuous data for intervals of three minutes or less in that dataset and found momentary time sampling outperformed interval recording across the analyses reported (LeBlanc and colleagues). Those findings do not make three minutes a safe default, prove that a shorter interval is always adequate, or establish momentary time sampling as best for every target.

A practitioner review recommends considering both the behavior dimension and the expected direction of change when designing discontinuous measurement (Fiske and Delmolino). A separate comparison found that, for the treatment-analysis datasets studied, momentary time sampling interpretations were somewhat more likely to match duration data and partial-interval interpretations were somewhat more likely to match frequency data (Meany-Daboul and colleagues). A computer simulation also demonstrates that interval-method error varies with method and event patterns (Wirth and colleagues).

These studies help a team ask better questions. They do not justify adding a universal correction factor, translating an interval percentage into duration, or declaring a direction of clinical change without reviewing the actual target, method, interval, observation conditions, and data pattern.

Complete the blank comparison worksheet

Copy the following sections into the practice's approved clinical planning system. Use a study code or other permitted identifier if a pilot record must be linked to a client. Keep the blank copy free of identifying information.

Decision and target

FieldEntryPlanning record IDDate and responsible clinicianObservable target definitionDimension the decision needsIntended clinical use of the dataExpected direction of change, if applicableObservation setting and routineClient and stakeholder priorities or concerns relevant to measurementConsent, assent, privacy, payer, or organizational conditions to confirm

Continuous-measurement rehearsal

FieldEntryDirect dimension rehearsedObservation durationObserver role and competing dutiesRecording systemEvents, onsets, offsets, or time capturedKnown missed or uncertain recordsWhat made recording easier or harderPossible workflow changes before using a discontinuous method

Candidate comparison

FieldContinuousPartial intervalWhole intervalMomentary time samplingExact observer ruleProposed interval or sampling spacingNot applicable unless the protocol uses blocksPlanned observation durationRelevant dimension preserved directly?Expected directional tendency or sampling riskCompeting duties during pilotUsable records or intervalsMissed or unobservable recordsObserver feedbackDifference from comparison recordClient or stakeholder impactOpen question

Decision record

FieldEntryMethod selected, if anyWhy it fits the stated dimension and decisionKnown limitations communicated to usersTraining and competency checkData-display label that avoids overclaimingPerson approving the protocolDate to review performance and usefulnessTrigger for earlier reconsideration

Calculate percentages without changing their meaning

For an interval method, first identify usable intervals. A missed, obscured, or invalid interval should follow the written missing-data rule; it should not silently become a nonoccurrence.

Sampled occurrence percentage = intervals or moments scored occurrence / usable scored intervals or moments x 100

If 18 of 30 usable partial intervals contain at least one occurrence, the result is 60 percent of usable intervals with occurrence. It is not 60 percent duration. If 18 of 30 momentary samples show the response at the scheduled instant, the result is 60 percent of usable sampled moments with occurrence. The same number carries a different observation rule.

For planned intervals, divide observation duration in seconds by interval length in seconds. If the result is not a whole number, state whether the protocol uses a shortened final interval, stops at the last complete interval, or changes the observation duration. Never hide that choice in spreadsheet rounding.

Pilot the candidates in representative conditions

A useful pilot resembles the workflow the team expects to use. Include the normal routine, observer role, transition demands, visibility, timing cues, data-entry system, and relevant accessibility needs. If the target varies substantially across contexts, a single calm observation may not be representative. Record what the pilot covered and what it did not.

When possible and appropriate, compare a discontinuous record with a continuous or criterion record collected from the same observation. That comparison can reveal discrepancies for that sample. It cannot turn the comparison record into unquestionable truth, eliminate observer error, or validate the method across future sessions. If two observers are involved, keep method feasibility separate from interobserver agreement. The IOA sampling guide can support that separate plan.

Review both missingness and the work that produced it. An interval may be missed because the observer was delivering a safety response, supporting communication, managing materials, or could not see the target. Those circumstances matter more than a bare completion percentage. Protect the client from a data-collection arrangement that competes with safe, dignified, and responsive care.

Read the comparison as a structured discussion

The worksheet should narrow uncertainty, not disguise it. A candidate may preserve the relevant dimension but be difficult under current staffing. Another may be feasible yet poorly aligned with the intended decision. A third may work in one routine and fail in another. Preserve those tradeoffs in the decision record.

Do not rank candidates by a single score. A low missed-interval percentage does not establish accuracy. A close match in one pilot does not establish future validity. A high observer preference does not prove that the resulting measure answers the clinical question. The responsible clinician should integrate the target definition, direct dimension, research evidence, observed discrepancies, client and stakeholder input, competence, safety, implementation conditions, and applicable requirements.

Fictional filled example

The following example is invented and contains no real client, family, provider, payer, or clinical data.

Planning question. A fictional team wants to describe engagement during a 20-minute small-group routine. "Engagement" is defined as manipulating the assigned material, looking at the current speaker or material, or making a task-related response. The team needs a measure that can support discussion of time engaged while the clinician also leads the routine.

Rehearsal. A second observer collects continuous duration during two practice observations. The lead clinician cannot reliably mark every transition while presenting instructions, but the second observer can. The team does not conclude that continuous duration is impossible. It records that continuous duration would require a separate observer or different capture system under these conditions.

Comparison. The team pilots 10-second partial intervals, 10-second whole intervals, and 30-second momentary samples against the same fictional practice video. Partial intervals score higher than continuous percentage duration, whole intervals score lower, and momentary samples fall between them in this small sample. The team records the exact discrepancies and the fact that the result may differ with another response pattern.

Decision. No method is selected automatically. The clinician schedules another representative pilot, considers whether a separate observer is justified for the decision, and asks the team to review whether momentary samples would miss the brief responses most important to the client. The record labels every result by its actual method instead of calling each one "percent engagement."

Know what this worksheet cannot decide

This worksheet cannot determine the clinically correct target, establish measurement validity or accuracy, validate observer competence, or decide whether an intervention is effective. It also cannot diagnose a condition, determine medical necessity, authorize service, satisfy a payer or regulator, or establish legal compliance. Client and stakeholder involvement, informed consent or assent where applicable, supervision, direct observation, organizational policy, privacy and security controls, and review by qualified professionals remain essential.

Revisit the selection when the target definition, intended decision, or observer duties change. Reopen it as well when data become implausible, missingness clusters, client or stakeholder concerns emerge, requirements change, or the measure no longer helps the team evaluate care. Retain the rationale and limitations with the protocol so later readers do not mistake a convenience choice for a clinical standard.

Related resources

Sources