07 / 08
How to Read Your Results — Mean, Median, Best and Standard Deviation
What the mean, median, best and standard deviation of your reaction time set each tell you, and how to interpret your personal progress.
When a set ends, you get four numbers. Looking at just one is easily misleading, so it's worth knowing what each one tells you.
Mean
The sum of all valid trials divided by the number of trials. It's the most familiar measure, but if your mind wanders once and you get 500 ms, the mean gets pulled up significantly.
Median
The middle value when the values are sorted by size. With 5 trials, it's the third value. Because it isn't swayed by one or two extreme values, it's the best way to see your usual ability. That's why this site's personal progress chart plots the median of each set.
Best
Your fastest trial. It's a feel-good number, but luck plays a big role. It may be a case where the random wait happened to match your expectation, so it's not a good indicator of ability.
Standard deviation
Shows how spread out your trials are from the mean. The smaller it is, the more consistently you reacted. This site uses the sample standard deviation (dividing by n−1).
For example, even if two sets both have a mean of 250 ms, the set with a standard deviation of 10 ms was similar every time, while the set with 60 ms mixed fast and slow trials. When you're tired or distracted, the standard deviation often grows before the mean does.
Worked example
| Trial | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| Reaction time (ms) | 231 | 248 | 239 | 402 | 244 |
- Mean = (231 + 248 + 239 + 402 + 244) ÷ 5 = 272.8 ms
- Median = sorted: 231, 239, 244, 248, 402 → 244 ms
- Best = 231 ms
Because of the single 402 ms in trial 4, the mean is a full 29 ms higher than the median. In cases like this, the median is more reliable.
How to read your personal progress
- Look at the trend across many sets rather than the ups and downs of a single set.
- Comparisons are meaningful only when you measure at a similar time of day, on the same device, in the same posture.
- The lowest point is shown in bold. Think back to the conditions of that set (sleep, time of day, etc.).
Go/No-go results
In Go/No-go, only trials where you responded to ● count toward reaction time statistics, and commission errors (pressing on ✕) and omission errors (missing ●) are counted separately. A good result is both fast and low in errors.