Measuring time-based dimensions: duration, latency, and IRT

Counting tells you how often behavior happens, not how long it lasts, how fast it starts, or how it is spaced. For many goals, time is the dimension that matters.

Why it matters

Imagine a learner whose tantrums drop from four per day to two, while each one now lasts 40 minutes instead of 5. A count says things improved; the family's evening says otherwise. Or think of a teenager who follows every instruction, but only after a two-minute delay. Their compliance count is perfect, and the problem is invisible.

Time-based measures let you see these patterns. As a BCBA you will choose them, write the rules for using them, and judge whether the resulting data are trustworthy.

Key ideas

• Two time-based dimensional quantities. Every response takes up some amount of time (temporal extent) and happens at some point in time relative to other events (temporal locus).
  • Duration measures temporal extent.
  • Latency and IRT measure temporal locus.
• Duration can be summarized two ways.
  • Total duration per session: add up the time across all occurrences. Use it when the concern is how much of the session the behavior fills, or when responses are hard to separate (on-task behavior, play with peers).
  • Duration per occurrence: time each episode separately and report the mean or median. Use it when you care about how long each episode lasts, such as tantrums or time spent at a work station.
• Latency is about responsiveness. It runs from the onset of an antecedent stimulus, usually an instruction or cue, to the onset of the response. Goals can aim to shorten latency (starting tasks after a direction) or lengthen it (pausing before answering an impulsive question).
• IRT is about spacing. It runs between two successive responses of the same behavior, typically from the end of one response to the start of the next. Shorter IRTs go with a higher rate, and longer IRTs go with a lower rate. IRT is also the measure behind a spaced-responding DRL schedule, which reinforces a response only when enough time has passed since the last one.
• Clear onset and offset rules are not optional. Every time-based measure starts and stops a clock. If observers disagree about when a behavior begins or ends, their numbers will not agree.
• These are continuous measures. Like count and rate, duration, latency, and IRT capture every occurrence during the observation period, so they are not estimates the way interval data are.

Definitions

• Duration: the amount of time from the onset to the offset of a response.
• Response latency: the time from the onset of a stimulus to the onset of the response it is meant to occasion.
• Interresponse time (IRT): the time between two successive responses of the same class. Some researchers time IRT from the onset of one response to the onset of the next, so check how a study defines it.
• Temporal extent: the fact that every response occupies some amount of time.
• Temporal locus: the fact that every response occurs at a certain point in time relative to other events.

Examples from a session

Each example names exactly when the clock starts and stops, which makes the data usable by a second observer.

Common mistakes / exam traps

• Timing from instruction to completion. Starting the clock at the cue and stopping it when the task is finished mixes latency and task duration into one number. If it improves, you cannot tell whether the learner started faster, worked faster, or both. Record them separately.
• Choosing count for behavior that varies in length. If episodes are long, uneven, or hard to separate, a count hides most of the change. Duration usually fits better.
• Using duration for brief, discrete responses. Timing a 1-second hand raise adds little; count or rate is the better fit.
• Forgetting that the summary matters. Total duration and mean duration per occurrence can move in opposite directions. Fewer but longer episodes can raise the mean while total duration falls, because total duration equals count times mean duration per occurrence.

Quick check

Related topics

• Measuring occurrence: count, rate, and percentage (C.3)
• Writing operational definitions (C.1)
• Continuous vs. discontinuous measurement (C.5)
• Choosing a measure that captures the critical dimension within real-world limits (C.9)
• Differential reinforcement (DRA, DRO, DRL, DRH) with and without extinction (G.2)