How to Read a T-Table
A t-table turns a sample size and a significance level into a critical t-value β the number that decides a t-test and sets the width of a confidence interval. This guide shows you how to read one, step by step, and checks every lookup against the live T-Distribution Calculator, so the value you read is one you can trust.
To read a t-table, find your degrees of freedom (df = n β 1) down the left-hand column, then find your tail area across the top β one-tailed or two-tailed, to match your test. The number where that row and column meet is your critical t-value: with df = 9 and two-tailed Ξ± = 0.05, the table gives t* = 2.262.
What Is a T-Table?
A t-table is a printed lookup for the t-distribution. Its rows are degrees of freedom, its columns are tail areas, and each cell holds a critical t-value β the t that leaves exactly that much area in the tail (or tails) of the distribution. Find your row, find your column, and the cell where they meet is the number your test or interval needs.
The table exists because, before calculators, this was how every t-test was done. You worked out your test statistic by hand, looked the critical value up in the table, and compared the two. The arithmetic has moved into software, but the table is still the fastest way to see what a test is doing β and in many exams it is the only aid you are allowed. The T-Distribution Calculator gives the same values instantly, which is why this guide pairs every table lookup with a calculator check.
The t-distribution itself is a family of bell-shaped curves, one for each degrees of freedom. With few degrees of freedom its tails are heavier than the normal curve’s, so the critical values are larger; as the degrees of freedom grow, the curve settles onto the normal curve β which is why the bottom row of a t-table, the β row, is the z-table. If the t-versus-normal distinction is new to you, the t-distribution vs normal distribution guide explains it in full.
How to Read a T-Table in 3 Steps
Reading a t-table is a three-step process. Work through the steps in order: each one fixes a choice the next step depends on.
- Find your degrees of freedom. Work out df = n β 1 for a one-sample test β the sample size minus one. With n = 10, df = 9. Then run your finger down the left-hand column of the table until you reach that row.
- Find your tail area across the top. Choose the column that matches your test: a two-tailed test at Ξ± = 0.05 uses the two-tailed 0.05 column, and a one-tailed test at Ξ± = 0.05 uses the one-tailed 0.05 column. The column header is the area in the tail β the next section explains why the two columns are so easy to confuse.
- Read the cell where the row and column meet. That number is your critical t-value, t*. With df = 9 in the two-tailed 0.05 column it is 2.262: a test statistic beyond Β±2.262 is significant at the 5% level, and a 95% confidence interval uses Β±2.262 as its multiplier.
Check the value before you use it: the T-Distribution Calculator returns the same critical value for any df and Ξ± β if the two disagree, you have the wrong row or the wrong column.
Two-Tailed vs One-Tailed Columns
This is the single most common confusion in reading a t-table, so it gets its own section. Most printed t-tables label each column twice: once with the one-tailed area and once with the two-tailed area. The same number therefore appears under two different labels β and picking the wrong one is the easiest way to get a wrong answer.
The rule is simple: a two-tailed area is split across both tails, so a two-tailed Ξ± of 0.05 means 0.025 in each tail β the same column as a one-tailed Ξ± of 0.025. At df = 9, the value 2.262 is the 95% two-tailed critical value and the 97.5% one-tailed critical value. The mapping is:
- Two-tailed 0.10 = one-tailed 0.05
- Two-tailed 0.05 = one-tailed 0.025
- Two-tailed 0.02 = one-tailed 0.01
- Two-tailed 0.01 = one-tailed 0.005
The mistake to avoid is reading the one-tailed column for a two-tailed test. At df = 9 the one-tailed 0.05 column gives 1.833, far smaller than the two-tailed value of 2.262 β and using it makes you reject the null hypothesis far more often than the 5% you intended. Whenever a table shows two header rows, read the one that matches your test: two-tailed tests take the two-tailed row, one-tailed tests take the one-tailed row. For a fuller treatment of critical values in general, see how to find a critical value.
What If Your df Is Not in the Table?
Printed t-tables skip most degrees of freedom: a typical table lists 1, 2, β¦ 30, then 40, 60, 100 and β. If yours is not printed, do not panic β and do not just pick the nearest row. Use these rules, in order.
- Use the closest SMALLER df. If your df is 24 and the table lists 20 and 30, use the df = 20 row (t* = 2.086 at 95% two-tailed). A smaller df gives a larger critical value, so your test is stricter β this is the conservative choice, and the one examiners expect. Never round up to the larger df: that shrinks the critical value and makes it too easy to reject the null hypothesis.
- Interpolate only as an approximation. You can estimate a value between two rows, but t-values do not change linearly with df, so the result is approximate β fine for a rough check, not for a result you will report.
- For df above about 100, use the β row. The β row is the z-value: at 95% two-tailed it is 1.960. Even at df = 100 the t-value is 1.984, only 0.024 above the z-value, so for large samples the t-table and the z-table agree for all practical purposes.
Or skip the table entirely: give the exact df to the T-Distribution Calculator and it returns the exact critical value with no rounding to a printed row β for df = 24, two-tailed 95%, it gives t* = 2.064, between the bracketing table rows of 2.086 (df = 20) and 2.042 (df = 30).
Reading a T-Value Backwards: Finding a P-Value Range
A t-table works in both directions. So far you have used it forwards β start with Ξ± and find t*. You can also run it backwards: start with a test statistic you have already computed and find the p-value range it falls in. The table cannot give you the exact p-value, only a bracket β but the bracket is often enough to make the decision.
Take t = 2.0 with df = 10, two-tailed. Find the df = 10 row and look across it: the two-tailed 0.10 column gives 1.812 and the two-tailed 0.05 column gives 2.228. Your t of 2.0 sits between them, so the two-tailed p-value is between 0.05 and 0.10 β not significant at the 5% level, but closer to it than the table alone suggests. The T-Distribution Calculator gives the exact number: p = 0.0734.
The same reading works one-tailed. With t = 1.0 and df = 5, the one-tailed columns give 0.920 (Ξ± = 0.20) and 1.476 (Ξ± = 0.10), so the right-tailed p-value lies between 0.10 and 0.20; the calculator gives the exact P(T > 1.0) = 0.1816. And one fact applies to every t-table: P(T < 0) = 0.5 for any df β the t-distribution is symmetric about zero, so half its area always lies below zero.
The lesson is about precision: a table brackets a p-value, a calculator computes it. Use the table to decide and the calculator to report β the worked examples at the end of this guide do exactly that.
Worked Example 1: Confidence Interval
Now put the table to work. A class of 10 students takes a test; their scores have a sample mean of 50 and a sample standard deviation of 8. Build a 95% confidence interval for the population mean.
- Degrees of freedom: df = n β 1 = 10 β 1 = 9.
- Standard error: SE = s / βn = 8 / β10 = 8 / 3.162 = 2.530.
- Critical value: in the df = 9 row, the two-tailed 0.05 column gives t* = 2.262 β the table in the next section has it, and the T-Distribution Calculator confirms it.
- Margin of error: t* Γ SE = 2.262 Γ 2.530 = 5.72.
- Interval: 50 Β± 5.72 = [44.28, 55.72]. With 95% confidence, the population mean lies between 44.28 and 55.72.
Compare that with the z-interval. With z* = 1.960 the margin is 1.960 Γ 2.530 = 4.96 and the interval is [45.04, 54.96] β narrower by 0.76 on each side. The t-interval is wider because the population standard deviation is estimated from a small sample; that extra width is the price of the extra uncertainty. With a larger sample the two intervals converge, as the comparison later in this guide shows.
Critical t-Values at 95%: Two-Tailed and One-Tailed
These are the values this guide uses β and the values you will reach for most often. Every cell is checked against the T-Distribution Calculator, and they match the NIST/SEMATECH e-Handbook table of critical values of the Student’s t distribution. Read across from your df in the left column: the middle column is the two-tailed 95% value, the right column the one-tailed 95% value. Watch how the numbers fall as df grows and settle on the β row, where t and z agree at 1.960 and 1.645.
| Degrees of freedom (df) | t critical value, two-tailed, 95% | t critical value, one-tailed (right), 95% |
|---|---|---|
| 1 | 12.706 | 6.314 |
| 5 | 2.571 | 2.015 |
| 9 | 2.262 | 1.833 |
| 10 | 2.228 | 1.812 |
| 11 | 2.201 | 1.796 |
| 19 | 2.093 | 1.729 |
| 20 | 2.086 | 1.725 |
| 29 | 2.045 | 1.699 |
| 30 | 2.042 | 1.697 |
| 100 | 1.984 | 1.660 |
| β | 1.960 | 1.645 |
Values are rounded to three decimal places and match the NIST/SEMATECH e-Handbook table. The β row is the z-table: at 95% confidence the two-tailed value is 1.960 and the one-tailed value is 1.645. Every value in this table is checked against the T-Distribution Calculator.
Worked Example 2: T-Test Decision
A second example, this time a test decision. A researcher tests whether a sample of 12 people differs from a known population mean. The significance level is Ξ± = 0.05 and the test is two-tailed. The question is the same in every t-test: does the computed statistic fall far enough out that the null hypothesis becomes implausible?
- Degrees of freedom: df = n β 1 = 12 β 1 = 11.
- Table decision: with df = 11 and a two-tailed Ξ± of 0.05, the critical value is t* = 2.201. Reject H0 when |t| > 2.201. (The one-tailed critical value for the same Ξ± would be 1.796 β and the z-value is 1.960.)
- The statistic: the student’s computed t is 2.45. Compare: 2.201 < 2.45, so the statistic falls in the rejection region β reject H0 at the 5% level.
- Exact p-value: read the table backwards to bracket it β at df = 11, 2.45 sits between the 0.05 column (2.201) and the 0.02 column (2.718), so the two-tailed p lies between 0.02 and 0.05. The One-Sample T-Test Calculator gives the exact p = 0.0322 for t = 2.45 and df = 11 β inside the table’s bracket and below 0.05, the same decision, with the precision the table cannot give.
Notice what the table did and did not do. It gave the decision threshold (2.201) and the bracket (0.02 < p < 0.05); it did not give the exact p (0.0322). That is the whole relationship between a t-table and a calculator in one example. For the wider context of test decisions, see hypothesis testing explained.
T-Table vs Z-Table
A t-table and a z-table are two versions of the same tool: both give you the cut-off that separates ‘plausible chance’ from ‘too extreme’. The difference is the distribution behind the cut-off β and the choice is usually made for you by what you know about the population.
- Use the z-table when the population standard deviation Ο is known, or the sample is large (n β₯ 30). At 95% two-tailed, z* = 1.960. The Critical Z-Value Calculator gives the exact value for any confidence level.
- Use the t-table when Ο is estimated from a small sample. At df = 9, t* = 2.262 β the larger value prices in the extra uncertainty of estimating Ο.
- As df grows, the two agree. At df = 30 the t-value is 2.042, at df = 100 it is 1.984, and at the β row it is 1.960 β the z-value itself.
In short: Ο known or large n β z; Ο estimated from a small sample β t. If you are working with a small sample and you are unsure whether Ο is known, assume it is not and use t β the wider interval is the honest one. The choice between the two distributions is explained in full in the why the t-distribution differs from the normal distribution guide.
Common Mistakes When Reading a T-Table
Most errors in reading a t-table come from skipping one of the steps above. Watch for these five:
- Using n instead of df. With n = 10 the table row is df = 9, not 10 β that gives 2.262, not 2.228. Forgetting the minus one is the most common slip of all.
- Reading the wrong tail column. A two-tailed test needs the two-tailed column. At df = 9 the one-tailed 0.05 value is 1.833, not 2.262 β using it makes you reject the null hypothesis far too easily.
- Using the z row for a small sample. The β row (1.960) is the z-value, not a t-value. With df = 9 the right cut-off is 2.262: the z row understates it by 0.302 and produces intervals that are too narrow.
- Rounding the critical value down. If the exact value falls between two printed rows, round UP to the larger (smaller-df) value β never down. At df = 29, 2.045 is the value; writing 2.04 understates the cut-off.
- Treating the table’s p-range as exact. A table tells you that p lies between 0.05 and 0.10 (t = 2.0, df = 10); it does not say p = 0.0734. Only a calculator gives the exact value β read the bracket from the table, report the number from the calculator.
What is the difference between one-tailed and two-tailed columns in a t-table?
They put the same area in different places, so the same number means two things. A two-tailed column splits Ξ± across both tails, while a one-tailed column puts it all in one. At df = 9, the value 2.262 is the two-tailed 0.05 critical value β and the one-tailed 0.025 value, or, in confidence terms, the 97.5% one-tailed value. So pick the column that matches your test: a two-tailed test uses the two-tailed column, a one-tailed test uses the one-tailed column. If your table only labels one of the two, use the mapping: two-tailed 0.05 = one-tailed 0.025. Reading the one-tailed 0.05 value (1.833 at df = 9) for a two-tailed test is the classic error β it is a much smaller cut-off than the 2.262 the test needs.
What do you do when your degrees of freedom are not in the table?
Use the closest smaller df β the conservative choice. With df = 24, use the df = 20 row (t* = 2.086 at 95% two-tailed), which gives a larger, stricter cut-off than the df = 30 row (2.042). Never round up to the larger df. For df above about 100, use the β row (1.960 at 95% two-tailed β the z-value). Or avoid the table altogether: the T-Distribution Calculator gives the exact critical value for any df β with df = 24 it returns 2.064.
How do you find a p-value from a t-table?
Find your df row, then find where your t falls between two column values: the p-value lies between the two columns’ Ξ± values. With t = 2.0 and df = 10, two-tailed, the value sits between 1.812 (Ξ± = 0.10) and 2.228 (Ξ± = 0.05), so 0.05 < p < 0.10. The table gives a range; the exact p β 0.0734 for this example β comes from the T-Distribution Calculator.
Frequently Asked Questions
What degrees of freedom do I use in a t-table?
Use df = n β 1 for a one-sample test, where n is your sample size β so n = 10 means df = 9, not 10. Each row of the t-table is a different df, and using the wrong row is the most common error in the book. For a two-sample test the degrees of freedom are different (n1 + n2 β 2); the t-table row you need is the one that matches your test.
What is the t-value for 95% confidence with 9 degrees of freedom?
For a two-tailed 95% test with df = 9, t* = 2.262. For a one-tailed 95% test with df = 9, t* = 1.833. Check the column you are reading: the two-tailed 0.05 column, not the one-tailed 0.05 column.
How do you read a t-table for a one-tailed test?
Use the column labelled one-tailed Ξ± = 0.05 β no splitting needed. For example, with df = 11 the one-tailed 0.05 value is 1.796. If your table only labels two-tailed areas, use the two-tailed 0.10 column, which is the same column: two-tailed 0.10 = one-tailed 0.05.
Can I use a t-table for a confidence interval?
Yes β the critical value is the multiplier in the interval formula, xΜ Β± t* Γ SE. With n = 10, mean 50 and SD 8, the standard error is 2.530, t* is 2.262 (df = 9), and the 95% interval is [44.28, 55.72].
What is the difference between a t-table and a z-table?
A z-table is for the standard normal distribution, used when Ο is known or the sample is large; a t-table is indexed by degrees of freedom and is used when Ο is estimated from a small sample. The t-table’s β row is the z-table: at 95% two-tailed, both give 1.960.
What do you do if your t-statistic is larger than any value in the table?
It means the p-value is smaller than the smallest Ξ± your table prints β usually 0.01 or 0.001 two-tailed. A t far beyond the 0.001 column means p < 0.001. The exact p comes from the T-Distribution Calculator β you can enter any t and df and read it off directly.
Why is the critical t-value larger than the z-value?
Because estimating Ο from a small sample adds uncertainty, and the t-distribution’s heavier tails account for it. At df = 9 the 95% two-tailed t-value is 2.262 against a z-value of 1.960 β a difference of 0.302. As df grows, the t-value falls towards the z-value: at df = 100 it is 1.984, and at the β row it is 1.960.
Where can I check a t-table value?
The T-Distribution Calculator gives the exact critical value and p-value for any df and Ξ±, and the One-Sample T-Test Calculator runs the whole test for you. Both tools agree with the values in this guide’s table.
Conclusion
Reading a t-table comes down to three moves: find df = n β 1 down the left column, pick the one-tailed or two-tailed column that matches your test, and read the cell where row and column meet. The table gives you the critical value, and read backwards it brackets the p-value β but it never gives the exact p. The worked examples show both readings in action: a t* of 2.262 sets the 95% interval [44.28, 55.72] for n = 10, and a t of 2.45 against a critical value of 2.201 rejects the null, with an exact p of 0.0322 from the calculator. When your df is missing from the table, round down to the closest smaller row, or skip the table and let the calculator give you the exact value. Either way, check the value you use before you decide.
Check Tools
Use these calculators to check every table reading in this guide:
- Critical Z-Value Calculator β find a z critical value for any confidence level or Ξ±.
- T-Distribution Calculator β find a t critical value or an exact p-value for any degrees of freedom and Ξ±.
- One-Sample T-Test Calculator β compare your test statistic with t* for your sample.
Browse the full set of calculators in Tools.