What is latency in ABA data collection? Latency is the elapsed time from a defined start event to the onset of a defined response. A timer might start when a needed item becomes unavailable, a chosen activity ends, or an instruction finishes. It stops when the target response begins. Useful latency data require clear events, eligible opportunities, a maximum window, and consistent timing.
Latency needs two observable events
Define the event that starts the clock and the response onset that stops it. “Time to cooperate” lacks observable boundaries and can carry a value judgment. “Seconds from the end of a spoken or visual direction to the first task step” is measurable when the direction, accessible format, task step, and scoring rule are defined.
Latency can measure communication access, transition initiation, response to a safety cue, staff follow-up, or time to begin a chosen activity. The same response can have different latency across partners and settings.
Set the observation window in advance
Every opportunity needs a maximum wait. If the response does not begin within 60 seconds, record “no response within 60 seconds” or a predeclared capped value with a separate no-response flag. A literal zero means the response began at the start event. It should not represent missing data.
Also define ineligible opportunities. If the instruction was unclear, the needed communication system was unavailable, a medical concern arose, or the observer missed the start event, record the actual system or observation state.
Calculate and summarize latency
If a cue ends at 10:05:20 and the response begins at 10:05:32, latency is 12 seconds. Across several opportunities, preserve each value before calculating a mean or median.
Consider latencies of 4, 7, 9, 10, and 40 seconds. Mean latency is 14 seconds. Median latency is 9 seconds. The single 40-second value pulls the mean upward, while the median shows the middle observation. Both summaries and the range can be useful.
A fictional family-readable example
Omar is a fictional 12-year-old who wants more time to process schedule changes and an easy way to ask questions. The team defines the start event as a partner finishing a visual and spoken schedule update. The response begins when Omar moves toward the next activity, asks a schedule question through speech or AAC, or selects the agreed “more time” message.
Across six eligible opportunities, latencies are 8, 15, 11, 22, 10, and 18 seconds. Median latency is 13 seconds, the average of the ordered middle values 11 and 15. Omar uses “more time” in two opportunities, and both count as accessible responses.
One planned opportunity is invalid because the visual schedule was unavailable. The team reports 6 of 7 planned opportunities valid, or 85.7% and treats the missing visual as a system issue. The results describe observed timing and offer no proof that a support caused change.
Faster is only one possible direction
Shorter latency may be useful for a person-selected safety or participation goal. Longer latency can reflect thoughtful processing, a reasonable pause, uncertainty, inaccessible communication, pain, fatigue, or a preference to decline. Speed by itself is not quality.
Review accuracy, independence, comfort, assent, partner response, and the person’s experience. A rapid response produced under pressure may fit poorly even when the number looks better.
Keep opportunity conditions comparable
Record the start-event form, partner, setting, task difficulty, available choices, communication access, prompt, health or sensory conditions, and observation window. A ten-second latency after a familiar visual cue and a ten-second latency after an unexpected verbal instruction may describe different contexts.
If the start event changes, mark a new measurement version. Keep old and new data visually distinct until their comparability is established.
Prompts also change interpretation. Record whether the response began independently, after an ordinary support, or after a planned prompt. If prompting resets the timer, stops it, or creates a new trial, specify that rule before data collection. Preserve the original latency and prompt record so a shorter value cannot hide greater assistance.
Protect communication and response time
Keep AAC, interpreters, mobility and sensory supports, breaks, food, water, bathroom access, prescribed care, and emergency help available. Do not remove a support or repeat a direction rapidly to force a shorter latency.
Explain the observation in an accessible way. Follow the applicable consent and assent process, including the person’s way to pause or withdraw from nonemergency activity.
Check data quality
Useful measures include valid latency opportunities divided by planned opportunities, trials with synchronized timestamps, capped observations, exclusions by reason, and sampled observer agreement on both boundary events.
The Association for Behavior Analysis International basic-principles page situates measurement within higher-education behavior-analysis content. The BACB BCBA Test Content Outline, Sixth Edition identifies latency as a temporal dimension and covers validity, reliability, and representative procedure selection. Neither source determines a client’s latency goal.
Related terms
Sources
Take the next step with clarity
Whether you are finding care, growing as a clinician, or building a stronger ABA practice, Finni brings the people, tools, and support together to help you move forward.
Explore clinical roles at Finni practices