> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sightlinebehavior.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Duration Recording
> Measure the total time behaviors occur and analyze episode length using start/stop toggles.
* Measuring how long episodes last, not just how often
* Quantifying instructional time consumed by a behavior
* Capturing meaningful variation in episode length
* Pairing with frequency to fully characterize a behavior
* Best for behaviors with extended duration: use Frequency for brief, discrete events
1. Choose Duration from the method picker
2. Define target behaviors with explicit "on" and "off" criteria
3. Optional: add definitions shown via ? during recording
4. Optional: enable activity context
Duration recording measures how long each behavioral episode lasts. You press a key to start timing when the behavior begins, and press it again to stop. Each episode records individually.
Sightline calculates total cumulative duration, percentage of observation time, number of episodes, and mean episode length: the metrics needed to show both overall impact and the shape of the behavior.
## Keyboard shortcuts
| Key | Action |
| ------- | ---------------------------- |
| `1–9` | Start/stop behavior timer |
| `Cmd+Z` | Undo last episode |
| `Space` | Pause / resume session timer |
| `?` | Show behavior definitions |
| `Cmd+E` | End session |
## Recording episodes in Sightline
During recording, each behavior appears as an on-screen toggle button showing the current state (off or timing).
**To record an episode:**
* **Press the number key** (1, 2, 3...) or click the button to start timing
* The button changes appearance (usually brightens or shows a "running" indicator) and the timer begins
* **Press the same key again** to stop timing and mark the episode complete
* The elapsed time for that episode records immediately
You can have multiple behaviors timing simultaneously (example: recording both "out-of-seat" and "off-task" during the same observation), though typically you track one per session.
While recording, a running display shows which behaviors are currently timing, the cumulative duration and percentage of observation time per behavior, and the episode count so far. If you start a timer by mistake, press **Cmd+Z** to undo the last episode.
Duration recording is more cognitively demanding than frequency because you're managing start/stop state for each behavior. A practice run — timing yourself reading a paragraph, say — helps before live use.
## Reading the results
For each behavior, a duration session produces:
* **Total cumulative duration**: sum of all episode durations
* **Percentage of observation time**: (total duration ÷ session duration) × 100
* **Number of episodes**: count of distinct episodes
* **Mean duration per episode**: total duration ÷ number of episodes
alongside an episode timeline: a visual strip showing when each episode occurred, with start time and duration on hover.
Total time is only half the picture; the shape of that time is the clinical read. A student off-task for 18 minutes could be:
* 18 episodes of 1 minute each (frequent, brief off-task bursts)
* 2 episodes of 9 minutes each (infrequent but sustained disengagement)
Same total, different problem. The first looks like distractibility, the second like sustained disengagement, and they point to different interventions. That's why episode count and mean length carry as much clinical weight as the total, and why the metric that best characterizes a behavior depends on whether it's continuous (on-task, engagement) or episodic (tantrums, crying).
Duration data is also especially useful for communicating impact to teachers and families: "the student was off-task for 35% of the observation" is more meaningful than raw counts and directly connects to lost instructional opportunity.
## Common mistakes to avoid
Don't use duration recording for single-event behaviors like hand raises or call-outs. Use [Frequency Recording](/frequency-recording) instead. Duration recording works for behaviors that have extended "on" periods.
Your operational definition must specify exactly what counts as "on" and "off." Example: "on-task means eyes and pencil engaged with materials; off-task means eyes diverted or pencil idle for more than 3 seconds." This removes observer subjectivity.
If you start a timer and forget to stop it before the session ends, you'll inflate the final episode's duration. Review your episode list after recording — check that durations look reasonable — and note any discrepancy (like an episode that lasted 25 minutes in a 20-minute session) in your report.
If you pause the session timer, active behavior timers keep running. When you resume, those timers continue. This is usually the correct behavior (you want to know how long the student was off-task *during the time the observation was active*, including pauses). But if you pause to briefly interrupt the student, note this in your report.
A duration figure means little in isolation. When possible, observe a same-age peer with identical start/stop criteria, and record 2–3 baseline sessions before an intervention so you know how much the behavior naturally fluctuates day-to-day.
## Duration paired with frequency recording
Fully characterizing a behavior often takes both methods:
* **Frequency:** How many episodes occurred?
* **Duration:** How long was each episode?
You can record both in separate sessions with the same observation configuration, then compare:
* If a behavior decreases in frequency but increases in episode duration, the intervention may be suppressing initiation but not sustaining engagement once the episode begins
* If both frequency and duration decrease, the intervention is addressing the behavior comprehensively
***
For choosing which metric should lead a writeup — and framing the figures in report-ready sentences — see [Writing Up Observation Results](/writing-up-observation-results). For choosing between methods, see [Observation Methods](/observation-methods).