7 min read
What Is a Good Reaction Time? (ms Benchmarks for Online Tests)
Under 250 ms is quick and under 200 ms is genuinely fast on a browser reaction test. See the full scale, why device lag matters, and how to get a fair score.
Interpreting scores
What a reaction time actually means once you have one — reference ranges, statistical expectations, and the difference between a change you caused and a change you drifted into.
You have a number. The hard part is not getting it, it is knowing what it licenses you to claim. This topic covers reference ranges and where they come from, how to read a percentile honestly, and why a single fast attempt tells you almost nothing while an average of five tells you most of what you can know.
It is also about the two errors people make most often with these scores. The first is treating a percentile as a measured fact about other people when it is computed from an assumed distribution. The second is treating a change in your score as a change in you when it may just be a change in your setup, your sleep, or your last session.
The articles here are mostly about not over-reading a figure — which sounds unambitious, but it is the difference between a measurement you can use and one you argue with.
3 articles
6 min read
What Your Percentile Rank Actually Means
The rank on your result screen is computed from an assumed distribution, not measured from real people. Here is the arithmetic, and why a single percentile should not be read literally.
6 min read
Why Practice Improves Your Score, Then Stops
Serial testing data shows a real 13 ms gain that almost entirely arrives in the first two sessions. What that means for discarding your first attempt.
Other topics
Measurement
4 articles
How a number from a browser gets produced, and what it can and cannot tell you. Apparatus effects, timing precision, latency, and why browser scores are not laboratory scores.
Test design
4 articles
The scoring ladders, ladders, and rules behind the individual tests, written out so a score can be interpreted rather than merely reported.