Standard deviation vs standard error: the same formula, a different question
Standard deviation and standard error start from the same calculation and end up answering completely different questions. Confusing them is one of the most common mistakes in scientific writing. The table below summarises the difference, and the sections that follow explain each point in detail.
| Feature | Standard Deviation (SD) | Standard Error (SE) |
|---|---|---|
| Symbol | σ (population), s (sample) | σx̄ or SE |
| Measures | Spread of individual data points | Precision of the sample mean |
| Answers | "How much does each value vary?" | "How close is my average to the truth?" |
| Formula | σ = √[Σ(x − μ)² / N] | SE = σ / √n |
| Sample formula | s = √[Σ(x − x̄)² / (n − 1)] | SE = s / √n |
| As n grows | Approaches the population SD (fixed) | Gets smaller (SE ∝ 1/√n) |
| Use in graphs | Show the raw data spread | Show uncertainty in the mean |
| Used for | Describing variability | Confidence intervals, hypothesis tests |
What standard deviation actually measures
Standard deviation takes every data point, measures how far it lies from the mean, squares those distances (so positive and negative deviations do not cancel), averages them, and takes the square root to put the result back in the original units. The result is a single number that tells you, on average, how far any random observation is likely to sit from the centre.
If a class of students has a mean test score of 75 and an SD of 8, most scores fall between 67 and 83 (mean ± 1 SD), and almost all fall between 59 and 91 (mean ± 2 SD). The SD describes the spread of individual scores.
What standard error actually measures
Standard error asks a different question: if you drew a new random sample of the same size from the same population, how far would its mean be from the true population mean? SE is the standard deviation of the sampling distribution of the mean — the spread you would see if you repeated your experiment infinitely many times and plotted each sample's mean.
SE is always smaller than SD because averaging reduces variability. The relationship is exactly:
SE = SD / √n
Doubling the sample size divides SE by √2 ≈ 1.41. Quadrupling it halves SE. This is the mathematical reason why larger studies give more precise estimates.
Worked example
A researcher measures the resting heart rate of 16 people. The sample mean is 72 bpm and the sample SD is 12 bpm.
- SD = 12 bpm — individual heart rates typically vary ±12 bpm from the mean.
- SE = 12 / √16 = 12 / 4 = 3 bpm — the sample mean is estimated to within ±3 bpm of the true population mean.
- A 95 % confidence interval for the mean is 72 ± (1.96 × 3) = 72 ± 5.88, or 66.1 to 77.9 bpm.
The SD of 12 describes the raw variability in the data. The SE of 3 describes the precision of the estimate of the mean. Both are correct; they just answer different questions.
When to report which
- Graphs: Use error bars showing SE when the figure is meant to show how precisely the mean is estimated (common in biomedical and experimental papers). Use SD bars when the figure is meant to show the spread of the raw data. Never mix them without labelling.
- Results text: Report the mean ± SD when describing your sample (e.g. "the sample had a mean age of 34 ± 8 years"). Report mean ± SE when emphasising precision of the estimate.
- Confidence intervals: Always derived from the SE, never from the SD.
- P-values and hypothesis tests: Based on SE, because they test the mean, not the spread.
Common mistake: using SE in place of SD
It is surprisingly common to see error bars drawn with SE but labelled as SD, or SD reported where SE is needed. The mistake makes variability look smaller than it is — SE bars are narrower, which makes the data look more precise than it actually is. Journals increasingly require authors to state which measure they used.
Relationship to confidence intervals
A 95 % confidence interval for the mean is:
mean ± 1.96 × SE
Because SE shrinks as n grows, larger samples give tighter confidence intervals — but only because the estimate of the mean gets more precise. The underlying variability (SD) does not change. A study with n = 1,000 and SD = 20 has SE = 0.63 and a very tight CI; the raw data is still just as spread out.
Summary
SD describes your data. SE describes your estimate of the mean. Both are useful. Both are derived from the same underlying calculation. The distinction is not about which number is "better" — it is about which question you are trying to answer.