T-Test Distribution Graph: How to Read and Interpret It

Charts & pivot tables · Updated November 22, 2026

A t-test distribution graph is a picture of the t-distribution: a symmetric, bell-shaped probability curve centered at zero. It helps show how a t-test statistic compares with values expected under the null hypothesis, and how the areas in the tails correspond to p-values or critical values.

What a t-distribution graph shows

The horizontal axis shows possible t-statistic values, from negative values on the left through zero to positive values on the right. The vertical axis shows probability density, not the probability of one exact t value. The curve’s total area is 1, so probability is represented by area across an interval or region under the curve.

A graph may display one curve or several. Each curve represents a t-distribution with a specified number of degrees of freedom (df). A curve is often drawn with vertical lines marking an observed t statistic, a critical t value, or both. Shading may indicate a one-tailed or two-tailed p-value region.

  • Center: The curve is centered at t = 0, the result expected when a test statistic reflects no difference in the direction being tested.
  • Tails: Large positive or negative t values are less likely under the null model and lie farther out in the tails.
  • Shaded area: Depending on the graph, shading can mark the p-value region or the rejection region defined by a significance level.

Because density is not itself probability, the height of the curve at t = 2 is not the probability that the statistic equals 2. Probabilities come from areas under the curve.

Why a t-test uses this distribution

A t-test evaluates evidence about a population mean or a difference between means. It calculates a t statistic by expressing an estimated difference in units of its estimated standard error. In a one-sample test, for example, the statistic is the difference between the sample mean and the hypothesized mean divided by the standard error of the sample mean.

If the null hypothesis and the test’s assumptions are appropriate, the statistic follows a t-distribution under the null hypothesis. The distribution accounts for uncertainty from estimating the population standard deviation using sample data. That extra uncertainty matters especially with smaller samples, and gives the t curve heavier tails than the standard normal curve.

The graph is therefore a reference model, not a plot of the raw observations. A histogram of measured data answers a different question: what values occurred in the sample? A t-distribution graph answers what test-statistic values would be expected if the null hypothesis were true, subject to the test’s assumptions.

How degrees of freedom change the curve

Degrees of freedom determine the shape of a t-distribution. In common one-sample and paired t-tests, df equals the sample size minus one. For a conventional pooled two-sample t-test, df is based on the two sample sizes and equals their sum minus two. Other procedures, including Welch’s two-sample test, calculate df differently; Welch’s value can be fractional.

With low df, the curve is more spread out and has heavier tails than a normal curve. That means extreme t values are less surprising under the reference distribution, all else equal. As df increases, the t-distribution becomes narrower around its center and approaches the standard normal distribution. It does not abruptly turn into a normal curve at a fixed sample size.

For example, a graph with df = 5 will generally show heavier tails than one with df = 30. The same observed t value can consequently have a different tail area, and thus a different p-value, under the two distributions. Degrees of freedom are part of interpreting a t-test result, not merely a technical label on a chart.

How to read the axes and shaded areas

Start by identifying the degrees of freedom and the test direction. Then locate the observed t statistic on the horizontal axis. A value near zero lies close to the center; a value far from zero lies farther into a tail. Whether the distance is considered evidence depends on whether the test is one-sided or two-sided.

For a two-tailed test, results at least as extreme as the observed statistic can occur in either direction. If the observed statistic is positive, the p-value is the area to its right plus the matching area to the left of its negative. For a symmetric t-distribution, that is the combined area beyond the absolute value of the statistic. For a one-tailed test, the p-value is the area in the direction specified by the alternative hypothesis.

A graph may instead shade a rejection region. If a two-sided test uses significance level α = 0.05, the total rejection area is 0.05, split into 0.025 in each tail. The critical t values mark the boundaries of those areas. A statistic beyond either boundary is in the rejection region. The significance level and test direction should be chosen before examining results, rather than selected to make a result look significant.

How the graph connects to a test result

The t statistic, degrees of freedom, and alternative hypothesis together determine the p-value. The p-value is the probability, assuming the null hypothesis and model assumptions, of obtaining a statistic at least as extreme as the one observed. It is not the probability that the null hypothesis is true, nor does it measure the size or practical importance of an effect.

A critical-value approach and a p-value approach express the same decision rule when they use the same significance level and test. For example, in a two-sided test at α = 0.05, the critical-value method compares the absolute t statistic with the positive critical value. The p-value method compares the p-value with 0.05. The graph makes the relationship visible: the farther the statistic falls into the relevant tail or tails, the smaller the p-value.

A confidence interval for a mean or mean difference is closely related. For a two-sided test at level 0.05, a corresponding 95% confidence interval excludes the null value exactly when the test rejects that null, provided the interval and test use the same method and assumptions. The interval adds information about plausible effect sizes and precision that a p-value alone does not provide.

Worked example: a two-sided test

Suppose a one-sample test compares a population mean with a hypothesized value. A sample of 11 observations has a mean of 52, a sample standard deviation of 3, and is tested against a hypothesized mean of 50. The estimated standard error is 3 divided by the square root of 11, about 0.905. The t statistic is (52 − 50) / 0.905, about 2.21, with 10 degrees of freedom.

On a two-tailed graph, mark t = 2.21 and its mirror point t = −2.21. The p-value is the combined area beyond those two marks under the df = 10 curve. It is approximately 0.052, so it is just above 0.05. At the 5% significance level, this result does not meet the usual rejection threshold. That does not establish that the population mean equals 50; it means the sample does not provide sufficient evidence to reject that particular null hypothesis at the chosen level.

This example also illustrates why a graph needs the correct curve and tail definition. Shading only the area to the right of 2.21 would represent a one-sided calculation, not the two-sided test described. A one-sided alternative must be justified and specified in advance.

Choosing the right graph for the test

Before interpreting a displayed curve, check what test it represents. A one-sample t-test, paired t-test, pooled independent-samples test, and Welch’s test may use different degrees of freedom or standard-error calculations. A generic curve without df or tail information can explain the general idea, but it is not enough to reproduce a specific p-value.

  • One-sample test: Compares one sample mean with a specified value.
  • Paired test: Tests the mean of within-pair differences, such as measurements before and after treatment. Its df is the number of pairs with usable differences minus one.
  • Independent samples: Compares means from separate groups. Welch’s test does not assume equal population variances; the pooled version does and uses a different df calculation.

When reading software output, confirm whether the reported p-value is one- or two-sided, whether the test is Welch or pooled, and which df the program used. Different valid methods can produce slightly different curves, critical values, and p-values for the same observed data.

Assumptions and common misunderstandings

A t-distribution graph does not guarantee that a t-test is appropriate. Observations should be independent in the way the study design requires. For a paired test, the analysis concerns the differences within pairs, so those differences should be examined. For small samples, severe skew or influential outliers can make the usual t-test inference unreliable. With larger samples, the sampling distribution of the mean is often more robust to moderate departures from normality, but this does not excuse dependence or serious data problems.

Equal variances are required for the traditional pooled independent-samples t-test, but not for Welch’s test. A second common mistake is treating statistical significance as a measure of importance: a small effect can be statistically significant with sufficiently precise data, while a meaningful effect may be estimated imprecisely in a small study. Report an effect estimate and confidence interval alongside the test result when possible.

Frequently Asked Questions

What does the bell curve on a t-test graph mean?
The curve describes the reference distribution of t statistics expected under the null hypothesis, for the test’s degrees of freedom. It is not necessarily a graph of the sample’s raw data.

Why does a t-distribution have heavier tails than a normal distribution?
A t statistic uses an estimated standard error based on sample data, adding uncertainty. This makes extreme values more likely under a low-df t-distribution than under the standard normal curve. As df rises, the difference shrinks.

What does the shaded area represent?
It may show a p-value area beyond the observed statistic, or a rejection region set by a significance level. Read the graph’s labels and identify whether one or both tails are shaded before interpreting it.

How do I find the p-value from a t-distribution graph?
Use the observed t statistic, its degrees of freedom, and the stated alternative hypothesis to identify the appropriate tail area. A graph can give an approximate visual value; statistical software or a reliable t-distribution table gives a more precise one.

Is the t-distribution the same as the normal distribution?
No. Both are symmetric and centered at zero, but the t-distribution has heavier tails, especially at low degrees of freedom. It approaches the standard normal distribution as degrees of freedom increase.

Related

Try it: the Percentage Formula Generator: percent of, percent change, increase and decrease, with the formula.

More in Charts & pivot tables

Sources: Function behaviour and menu paths: Microsoft Support documentation for Excel. Method & sources

Spotted an error in this guide? Let us know.