What reaction time includes
In a visual task, the clock begins when the browser presents a signal and ends when it receives your response. Between those two events, the display must draw the signal, your visual system must register it, your brain must select a response, your muscles must act, and the keyboard, mouse, or touchscreen must report the input. The recorded number includes all of those stages.
Three tasks, three different demands
Compare like with like: Flash against earlier Flash runs, Choice against earlier Choice runs, and Go / No-Go against earlier Go / No-Go runs.
Flash
Wait for one visual go signal, then respond. This is the simplest protocol and emphasizes signal detection plus a prepared movement. Early clicks are recorded as false starts rather than fast reactions.
Choice
Identify a left or right signal and make the matching response. The added decision makes accuracy part of the result. A wrong-side response counts as an error.
Go / No-Go
Respond to go signals and withhold the response on hold signals. This protocol combines speed with response control, so false alarms and correct holds matter alongside reaction times.
Why the wait is unpredictable
If every signal arrived on a fixed schedule, a participant could anticipate the next frame instead of reacting to it. Splitframe varies the wait and treats premature responses separately.
How to read the numbers
| Metric | What it summarizes | How to use it |
|---|---|---|
| Median | The middle valid reaction after the times are ordered. | Use this as the primary speed summary. One unusually slow or fast frame has less influence on it than on the mean. |
| Mean | The arithmetic average of valid reaction times. | Compare it with the median. A noticeably higher mean often indicates one or more slow frames. |
| Fastest | The lowest valid time in the run. | The single quickest frame; the median is the steadier summary. |
| Spread | Sample variability across valid reaction times. | Lower spread means the frames were more consistent. |
| Faults | False starts or incorrect decisions recorded by the protocol. | Each fault is shown in the frame strip and counted for the run. |
| Drift | The difference between later and earlier frames. | A positive value means later frames were slower. |
| Precision | The share of signal decisions handled correctly. | This is especially important for Choice and Go / No-Go, where selecting or withholding a response is part of the task. |
Ten frames is the default; five and twenty are also available. When following a trend, use the same round count and the same protocol.
Make sessions comparable
- Keep the setup stable. Use the same device, display, browser, input method, viewing distance, and hand position when you care about small differences.
- Let the device settle. Close heavy background work, avoid an active download or update, and keep the tab visible. Splitframe aborts an armed frame when the page becomes hidden.
- Do a short familiarization run. The first attempt may measure learning the controls as much as reacting. Do not repeatedly restart until a favorable score appears.
- Choose a repeatable cadence. Similar time of day and similar conditions are more informative than many scattered tests. A few planned sessions beat constant checking.
- Record context. If something changes—new monitor, wireless mouse, browser update, unusual distraction—treat that as a new comparison period.
- Look at a trend. Several clean runs carry more information than one personal best. Splitframe’s history and CSV export let you inspect the underlying trials.
A practice routine
For a simple personal baseline, select one protocol and complete two or three clean runs after a brief warm-up. Stop when attention or posture changes noticeably. Repeat on a few separate days.
The daily drill sets one protocol and frame count for each UTC day.
When to discard a run
- The page was interrupted, the window lost focus, or the device visibly stalled.
- You changed input method or posture during the run.
- You clicked deliberately before the signal to explore the controls.
Local history, accounts, and exports
Without signing in, completed runs and preferences stay in this browser’s local storage. The clear control removes local history, and the CSV export produces a copy you can inspect independently. Signing in syncs validated run records across sessions.
A trial export contains timestamps, protocol names, outcomes and reaction times. Read the privacy notice for the current storage and account details.