What Are Error Bars On A Graph And How To Read Them?

Error bars are the small lines extending from data points on a graph that show how much variability or uncertainty exists in a measurement. They tell you whether the difference between two data points is meaningful or just random noise. To read them correctly, you need to know what type of error bar you are looking at, because standard deviation, standard error, and confidence intervals each tell you something different about the data.

What Do Error Bars Actually Show?

Error bars are a visual summary of spread. When a researcher measures something multiple times, they rarely get the exact same number each time. The error bar captures that range of possible values.

The length of the bar represents uncertainty. A longer bar means more variability in the measurements. A shorter bar means the measurements were consistent. If two bars overlap, the difference between the points may not be statistically meaningful. If they do not overlap, the difference is more likely to be real.

But the meaning of “overlap” depends entirely on which error bar you are using. This is the most common point of confusion for readers.

Standard Deviation vs. Standard Error: Know the Difference

These two are frequently confused, even by experienced researchers. They answer different questions.

Standard deviation (SD) tells you how spread out your individual data points are from the mean. It describes the data itself. If you measured the height of 100 adults, the standard deviation tells you how much individual heights vary. This is a description of your sample, not a statement about how confident you are in the average.

Standard error (SE) tells you how precise your estimate of the mean is. It describes the accuracy of the average, not the spread of the data. The standard error is always smaller than the standard deviation because it accounts for sample size. The more data points you collect, the smaller the standard error becomes, even if the underlying data spread stays the same.

When you see a graph with error bars, look at the figure caption or methods section. It should state which one is used. If it does not say, you cannot interpret the bars correctly.

Confidence Intervals: The Most Useful Error Bar

A confidence interval is another common type of error bar. A 95% confidence interval means that if you repeated the experiment many times, the true population value would fall within the bar range 95% of the time.

Confidence intervals are more informative than standard error bars because they directly relate to statistical significance. When two 95% confidence intervals do not overlap, the difference between the points is statistically significant at roughly the 0.05 level. When they do overlap, you cannot conclude a significant difference exists.

One important nuance: if confidence intervals overlap slightly, the difference might still be significant. The rule about overlap is a rough guide, not a perfect test. For a rigorous conclusion, you need the actual p-value from the statistical test. But for everyday reading of graphs, non-overlapping 95% confidence intervals are a strong signal of a real difference.

What Do Error Bars Mean When They Overlap?

Overlapping error bars are frequently misinterpreted. Many readers assume that any overlap means “no difference.” That is not always true.

With standard error bars, overlap tells you very little about significance. Two standard error bars can overlap substantially and the difference between the means can still be statistically significant. This is because standard error bars do not directly map to significance testing.

With 95% confidence intervals, the situation is clearer. If the intervals do not overlap at all, the difference is likely significant. If they overlap partially, the result is uncertain. You need to see the actual statistical output to know for sure.

The safest approach when reading any graph: treat overlapping error bars as “inconclusive” rather than “no difference.” The data might support a difference, or it might not. The graph alone cannot tell you.

Common Mistakes People Make When Reading Error Bars

The most common mistake is ignoring what type of error bar is shown. Standard deviation, standard error, and confidence intervals look identical on a graph. Their meanings are completely different.

Another frequent error is assuming error bars show the range of “normal” values. A standard deviation bar does show the spread of typical values. A standard error bar does not. It shows precision of the mean, which is a much narrower concept.

A third mistake is treating small error bars as proof of accuracy. Small error bars only mean the measurements were consistent. They do not mean the measurement itself was correct. A scale that is miscalibrated can produce consistent but wrong values. Error bars cannot detect systematic bias.

Finally, remember that error bars only represent the variability the researcher measured. They do not capture every source of uncertainty. Measurement error, sampling bias, and unmeasured variables are not reflected in the bars.

How Sample Size Affects Error Bars

Sample size changes error bars in a predictable way. Standard deviation is relatively stable as sample size grows. It reflects the natural spread of the population, which does not shrink just because you measure more people.

Standard error and confidence intervals behave differently. Both shrink as sample size increases. This makes sense: more data points give you a more precise estimate of the true mean. The bars get tighter even though the underlying data spread remains the same.

This means a graph with very small error bars is not necessarily showing “better” data. It may simply reflect a larger sample. A study with 10 participants can have large error bars. A study with 1,000 participants can have small error bars on the same data. The small bars indicate more confidence in the average, not less variability in the measurements.

Error Bars in Different Fields: Context Matters

Different scientific fields have different conventions. In experimental physics, error bars often represent measurement uncertainty from instrument precision. In biology and medicine, they typically represent standard error or confidence intervals. In psychology, standard error is common in bar graphs.

Clinical trial graphs frequently use confidence intervals because they directly relate to treatment effects. If a drug trial shows a confidence interval for the treatment effect that does not cross zero, the effect is statistically significant. This is a standard way to present results in medical journals.

When reading graphs outside your field, check the conventions. The same graph structure can mean different things depending on the discipline. The figure legend should always clarify what the bars represent. If it does not, treat the graph with caution.

How to Explain Error Bars to Someone New to Graphs

Start with the simplest concept: error bars show how sure you are about a number. A short bar means you are fairly confident. A long bar means there is more uncertainty.

Then introduce the idea that different types of error bars answer different questions. Standard deviation describes the data. Standard error and confidence intervals describe your confidence in the average. This distinction is the core of understanding any graph with error bars.

Finally, emphasize the overlap rule cautiously. Overlap is a warning sign, not a verdict. It tells you to look deeper at the statistics before drawing conclusions.