Skip to content
K Knidox Search…
Reference · Statistics

T Distribution Table

Critical t values for every degrees of freedom, switchable between one- and two-tailed.

Test type
For a one-sample t-test this is n − 1. The matching row is highlighted.
Critical t at df = 10 (two-tailed)
α = 0.10
1.812
α = 0.05
2.228
α = 0.025
2.634
α = 0.01
3.169
α = 0.005
3.581
Student t critical values by degrees of freedom and two-tailed significance level
dfα = 0.10α = 0.05α = 0.025α = 0.01α = 0.005
conf.90%95%97.5%99%99.5%
16.31412.70625.45263.657127.321
22.9204.3036.2059.92514.089
32.3533.1824.1775.8417.453
42.1322.7763.4954.6045.598
52.0152.5713.1634.0324.773
61.9432.4472.9693.7074.317
71.8952.3652.8413.4994.029
81.8602.3062.7523.3553.833
91.8332.2622.6853.2503.690
101.8122.2282.6343.1693.581
111.7962.2012.5933.1063.497
121.7822.1792.5603.0553.428
131.7712.1602.5333.0123.372
141.7612.1452.5102.9773.326
151.7532.1312.4902.9473.286
161.7462.1202.4732.9213.252
171.7402.1102.4582.8983.222
181.7342.1012.4452.8783.197
191.7292.0932.4332.8613.174
201.7252.0862.4232.8453.153
211.7212.0802.4142.8313.135
221.7172.0742.4052.8193.119
231.7142.0692.3982.8073.104
241.7112.0642.3912.7973.091
251.7082.0602.3852.7873.078
261.7062.0562.3792.7793.067
271.7032.0522.3732.7713.057
281.7012.0482.3682.7633.047
291.6992.0452.3642.7563.038
301.6972.0422.3602.7503.030
351.6902.0302.3422.7242.996
401.6842.0212.3292.7042.971
451.6792.0142.3192.6902.952
501.6762.0092.3112.6782.937
601.6712.0002.2992.6602.915
801.6641.9902.2842.6392.887
1001.6601.9842.2762.6262.871
1201.6581.9802.2702.6172.860
∞1.6451.9602.2412.5762.807

Reject the null hypothesis when your calculated |t| is larger than the critical value in the row for your degrees of freedom. The confidence row underneath each α is the interval that column builds: α = 0.025 two-tailed is the 95% confidence level.

A t-table gives the cut-off your test statistic must beat to be significant. Find your degrees of freedom down the side and your significance level across the top: at df = 10 and a two-tailed α = 0.05 the critical value is 2.228, so any |t| above 2.228 rejects the null hypothesis.

Why the t-distribution exists

When you know the population standard deviation you can use the normal curve and its fixed 1.96 cut-off. You almost never know it. Estimating σ from the sample adds a second source of uncertainty, and the t-distribution is the shape that accounts for it: same bell, but heavier tails, so the critical values sit further out and the conclusions stay honest.

How much heavier depends on the sample. That is what degrees of freedom measure — for a one-sample test, df = n − 1. With df = 1 the two-tailed 5% critical value is a remote 12.706; by df = 10 it has fallen to 2.228, by df = 30 to 2.042, and as df grows it converges on the normal value of 1.960. A t-table is really a stack of normal-ish curves, one per sample size.

Reading the significance level

Every column heading is an area in the tail. A one-tailed α = 0.05 puts all 5% at one end; a two-tailed α = 0.05 splits it into 2.5% at each end, which pushes the cut-off further out. This is the single biggest source of wrong answers with a t-table, so the switch above relabels the whole table rather than asking you to halve anything yourself.

t = (x̄ − μ₀) ÷ (s ÷ √n)

x̄ the sample mean, μ₀ the hypothesised mean, s the sample standard deviation, n the sample size

  1. 1
    Count the degrees of freedom. A one-sample test on 15 observations has df = 15 − 1 = 14.
  2. 2
    Decide one-tailed or two-tailed. Testing only whether the mean is larger is one-tailed; testing whether it differs in either direction is two-tailed.
  3. 3
    Pick the significance level. The usual choice is α = 0.05, which for a two-tailed test leaves 0.025 in each tail.
  4. 4
    Read the cell where the row and column meet. Row df = 14 and a two-tailed α = 0.05 give a critical value of 2.145.
  5. 5
    Compare it with your test statistic. A calculated t of 2.62 exceeds 2.145, so the result is significant at the 5% level; a t of 1.80 would not be.

Critical t values at the common levels

Two-tailed values are the ones used for 95% and 99% confidence intervals. The ∞ row is the normal distribution.

dfOne-tailed α = 0.05Two-tailed α = 0.05Two-tailed α = 0.01
16.31412.70663.657
52.0152.5714.032
101.8122.2283.169
151.7532.1312.947
201.7252.0862.845
251.7082.0602.787
301.6972.0422.750
601.6712.0002.660
1201.6581.9802.617
∞1.6451.9602.576

Beyond the one-sample test

The same table serves several tests, only the degrees of freedom change. A paired t-test uses the number of pairs minus one. An independent two-sample test with equal variances uses n₁ + n₂ − 2. A confidence interval for a mean uses the two-tailed value at the matching confidence level — the 2.145 above is exactly the multiplier in a 95% interval built from 15 observations.

If your degrees of freedom fall between printed rows on a paper table, the conservative habit is to round down to the next row, which gives a slightly larger critical value and makes significance slightly harder to claim. The calculator above sidesteps that entirely: it computes the exact value for whatever df you type, so there is nothing to interpolate.

How do I find degrees of freedom for a t-test?
A one-sample or paired test uses n − 1, where n is the number of observations or pairs. An independent two-sample test with pooled variance uses n₁ + n₂ − 2. Welch’s unequal-variance test produces a fractional df from a separate formula.
What is the difference between one-tailed and two-tailed critical values?
A one-tailed test puts the whole significance level in one tail, so at df = 10 and α = 0.05 the cut-off is 1.812. A two-tailed test splits it between both tails, pushing the same α = 0.05 cut-off out to 2.228.
Why does the t-table stop and say ∞?
As degrees of freedom grow the t-distribution becomes indistinguishable from the standard normal, so the last row holds the normal values — 1.960 two-tailed at 5% and 2.576 at 1%. Past about df = 120 the difference no longer changes a conclusion.
When do I use t instead of z?
Use t when the population standard deviation is unknown and estimated from the sample, which is nearly always the case with real data. Use z only when σ is genuinely known or the sample is large enough that the two agree.
Can a critical t value be negative?
Yes. The distribution is symmetric, so the lower cut-off is the negative of the upper one. Tables print only the positive half; for a left-tailed test use the same number with a minus sign.
How does the t-table relate to a confidence interval?
The two-tailed critical value is the multiplier in the interval. A 95% confidence interval for a mean from 15 observations is x̄ ± 2.145 × (s ÷ √15), using the df = 14 two-tailed α = 0.05 value.
What if my degrees of freedom are not in the table?
On a printed table, round down to the nearest listed df — that gives a marginally larger critical value and a conservative conclusion. The tool above computes the exact value for any df, so no interpolation is needed.