A t-value is statistically significant when it is large enough that the difference you observed in your data is unlikely to have occurred by chance alone. In most fields, this means the t-value produces a p-value below 0.05. As a quick rule of thumb, a t-value with an absolute value greater than about 2.0 is typically significant when your sample size is at least 30 or so, though the exact threshold depends on your sample size and the type of test you’re running.
What a T-Value Actually Measures
A t-value is essentially a signal-to-noise ratio. The “signal” is the difference between the value you observed in your sample and the value you’d expect if nothing interesting were happening (the null hypothesis). The “noise” is the variability in your data, adjusted for sample size. A large t-value means your signal is strong relative to the noise. A small one means the difference you found could easily be explained by random variation.
The formula for a one-sample t-test makes this concrete: you take the difference between your sample average and the expected population average, then divide by the standard error. The standard error is simply the standard deviation of your sample divided by the square root of the sample size. So a t-value gets larger when the observed difference is bigger, when the data is less variable, or when you have more observations. All three push the ratio in the same direction: more confidence that the difference is real.
How T-Values Connect to P-Values
A t-value by itself doesn’t tell you whether your result is significant. It needs to be compared against a threshold, called a critical value, that comes from the t-distribution. If your t-value exceeds the critical value, the corresponding p-value falls below your chosen significance level, and you reject the null hypothesis.
Most researchers set the significance level at 0.05, meaning they accept a 5% chance of incorrectly concluding that a real difference exists. Some fields use stricter thresholds of 0.01 (1%) or more lenient ones of 0.10 (10%), but 0.05 is the default in most published research. When the p-value lands below this threshold, the result is declared statistically significant. When it lands above, the conclusion is that the data don’t provide enough evidence to reject the null hypothesis.
The “Greater Than 2” Rule of Thumb
You’ll often hear that a t-value greater than 2 (or less than negative 2) is significant. This shortcut works reasonably well for two-tailed tests at the 0.05 level when your sample is moderately large. For example, with 30 degrees of freedom, the critical t-value for a two-tailed test at the 0.05 level is 2.042, according to NIST’s t-distribution tables. That’s close enough to 2 that the rule of thumb holds.
But with smaller samples, the critical value climbs higher. At 13 degrees of freedom, for instance, the critical value rises to about 2.16. With very small samples (under 10), it can exceed 2.3 or even 2.5. As your sample grows into the hundreds, the critical value drifts down toward 1.96, which is the value from the normal distribution. So the “greater than 2” shortcut is a useful mental check, not a substitute for looking up the actual critical value.
Degrees of Freedom Change the Threshold
The exact critical t-value depends on degrees of freedom, which reflect how much independent information is in your data. For a one-sample t-test, degrees of freedom equal the number of observations minus one. If you measured 18 people, you have 17 degrees of freedom. For an independent samples t-test comparing two groups, degrees of freedom equal the total number of observations across both groups minus two. Two groups of 15 and 12 give you 25 degrees of freedom.
For a paired t-test, where you measure the same subjects twice and analyze the differences, degrees of freedom equal the number of pairs minus one. Twelve pairs give you 11 degrees of freedom. Fewer degrees of freedom mean a wider, flatter t-distribution with heavier tails, which pushes the critical value higher and makes significance harder to reach. This is the statistical penalty for having less data.
One-Tailed vs. Two-Tailed Tests
Whether you’re running a one-tailed or two-tailed test also changes what counts as significant. A two-tailed test splits the 0.05 significance level across both ends of the distribution, putting 0.025 in each tail. This tests whether your result is significantly different from the expected value in either direction. A one-tailed test puts the entire 0.05 in one tail, testing only whether the result is greater than (or less than) the expected value.
The practical effect: a one-tailed test has a lower critical value than a two-tailed test at the same significance level. A result that falls just short of significance in a two-tailed test might cross the threshold in a one-tailed test. This is why one-tailed tests should only be used when you have a strong reason, established before collecting data, to test in only one direction.
T-Values in Regression Analysis
T-values aren’t limited to comparing group means. In regression analysis, every predictor variable in your model gets its own t-value, which tests whether that variable has a real relationship with the outcome or whether its apparent effect could be zero. The null hypothesis for each predictor is that its true coefficient (its slope) is zero, meaning it contributes nothing to the model.
The t-value for a regression coefficient is calculated by dividing the estimated slope by its standard error. If the t-value exceeds the critical value for your degrees of freedom (which in regression equals the number of observations minus the number of predictors minus one), you conclude that the predictor has a statistically significant relationship with the outcome. You’d want to confirm this before using the regression equation to make predictions.
Statistical Significance vs. Practical Significance
A significant t-value tells you that a difference probably exists. It does not tell you that the difference matters. This distinction between statistical significance and practical significance trips up many people. As one widely cited paper in the Journal of Graduate Medical Education put it, statistical significance is the least interesting thing about the results; the magnitude of the effect is what matters.
Statistical significance depends on both the size of the effect and the size of the sample. With a large enough sample, even a trivially small difference will produce a significant t-value. A study of 10,000 people might find that a new teaching method improves test scores by 0.3 points on a 100-point scale, with a highly significant t-value and a tiny p-value. That result is real in a statistical sense but probably meaningless in practice.
This is why researchers report effect sizes alongside t-values and p-values. One common measure, Cohen’s d, divides the difference between group means by the standard deviation. Unlike the t-value, Cohen’s d doesn’t grow with sample size. It tells you how large the difference actually is in standardized terms. A Cohen’s d of 0.2 is generally considered small, 0.5 medium, and 0.8 large. A significant t-value paired with a small effect size should temper your enthusiasm about the result, while a significant t-value with a large effect size gives you real reason to pay attention.