Most students know that p < 0.05 means the result is "significant." Fewer can explain what that actually means. Fewer still can explain it correctly. The p-value is probably the most cited and most misunderstood number in academic research. This article covers what it is, what it is not, and how to report it properly in a thesis or journal manuscript.
What a P-Value Actually Means
The p-value is the probability of obtaining results at least as extreme as the ones you observed, assuming the null hypothesis is true. That sentence is worth reading twice, because every word in it matters.
The null hypothesis (H0) is typically a statement of "no difference" or "no relationship." If you are comparing exam scores between two teaching methods, the null says the two methods produce the same results. If you are testing whether study hours predict GPA, the null says there is no linear relationship between the two. The p-value answers a conditional question: given that the null is true, how likely is it that random sampling alone would produce data as extreme as what you collected?
Here is a concrete example. You compare test scores between two groups and find a 5-point difference. The p-value of 0.03 means: if there were truly no difference between these groups, you would see a gap of 5 points or more in only 3 out of 100 random samples. That 3% is not the probability that the groups are the same. It is the probability of seeing your data (or more extreme data) in a world where the groups are the same.
R.A. Fisher introduced the convention of using 0.05 as a threshold in the 1920s. He described it as a convenient benchmark for flagging results worth a second look. It was never intended to be an absolute boundary separating "true" from "false." Fisher himself wrote that no single threshold should be applied mechanically across all research. Yet almost a century later, 0.05 is treated as gospel in most university statistics courses and thesis guidelines.
Five Things a P-Value Does Not Mean
The first and most common misinterpretation goes like this: "p = 0.03 means there is a 3% chance the null hypothesis is true." This is wrong. The p-value is calculated under the assumption that the null is already true. It cannot circle back and tell you the probability that the null is true or false. Determining the probability of a hypothesis being true requires Bayesian methods and a prior probability distribution, neither of which a standard frequentist p-value provides. The 0.03 tells you something about the data, not about the hypothesis.
The second misinterpretation: "p < 0.05 proves my hypothesis." A small p-value means the observed data would be unusual under the null hypothesis. It does not confirm the alternative hypothesis. Your result could be explained by confounding variables you did not control for, by measurement error, by sampling bias, or by simple chance (which, at a 5% threshold, will happen 1 in 20 times even when the null is true). Rejecting the null is not the same as proving the alternative. It is, at best, a reason to take the alternative seriously.
Third: "p = 0.06 means the result is not real." This is the tyranny of the threshold. A result with p = 0.048 gets published, cited, and celebrated. A result with p = 0.052 gets buried, revised, or quietly dropped from the manuscript. The difference between these two numbers is trivial, yet the professional consequences are enormous. In 2016, the American Statistical Association published a formal statement (Wasserstein & Lazar, "The ASA Statement on P-Values: Context, Process, and Purpose") explicitly warning against this kind of binary thinking. A p-value of 0.06 still represents reasonably strong evidence against the null. The cutoff at 0.05 is a convention, not a law of nature.
Fourth: "A smaller p-value means a bigger effect." This is wrong in a way that causes real damage in applied research. The p-value conflates effect size with sample size. A trivial difference of 0.3 points on a 100-point scale can produce p < 0.001 if the sample is large enough (say, n = 10,000). The test has so much statistical power that it detects differences no one would care about. Effect size measures like Cohen's d, eta-squared (η²), or odds ratios are what tell you how large the effect actually is. A tiny p-value with a tiny effect size is a statistically significant nothing. (For more on how effect sizes differ between t-tests and ANOVA, see our comparison of the two tests.)
Fifth: "A non-significant result means there is no effect." This confuses absence of evidence with evidence of absence. If you tested 15 participants per group and found p = 0.18, you cannot conclude that no difference exists. Your sample may have been too small to detect a real but modest effect. This is a problem of statistical power. A study with 80% power to detect a medium effect (d = 0.50) at α = 0.05 needs about 64 participants per group. With only 15 per group, your power drops below 25%, meaning you would miss even a real medium effect three times out of four. Non-significance in an underpowered study tells you almost nothing.
Statistical Significance vs. Practical Significance
This is where many theses fall short. A statistically significant result can be practically meaningless, and a practically meaningful result can fail to reach statistical significance.
Consider a study with 5,000 participants that finds a new teaching method raises exam scores by 0.8 points on a 100-point scale (p = .01, d = 0.04). The result is statistically significant. The effect size is negligible. No school administrator would redesign a curriculum for less than one point of improvement. The large sample gave the test enough power to flag a difference that exists only in a technical, mathematical sense.
The reverse happens too. A pilot study with 20 participants per group finds an 8-point difference in exam scores, but the wide confidence interval and small sample produce p = .09. The effect size is d = 0.65, which is medium-to-large. An 8-point improvement is the kind of result that would change classroom practice. But the study "failed" the significance test because it was too small to provide precise estimates.
This is why effect sizes and confidence intervals matter more than the p-value on its own. A confidence interval of [0.2, 15.8] around that 8-point difference tells you the true effect is probably positive but could be quite small or quite large. That is useful information. The binary "significant or not" label throws away most of it. Most journals and thesis committees now expect effect sizes reported alongside p-values. If your thesis only reports p-values, expect questions from any methods-literate committee member.
Reporting P-Values in APA-7 Format
APA 7th edition has clear rules for how p-values should appear in text. Report exact p-values to two or three decimal places: p = .032, not p < .05. The one exception: when p is smaller than .001, report p < .001 rather than writing out a string of zeros. Do not use a leading zero before the decimal (write p = .032, not p = 0.032), because a p-value cannot exceed 1.00.
Always pair the p-value with the test statistic and degrees of freedom. For a t-test: t(38) = 2.24, p = .031, d = 0.72. For ANOVA: F(2, 87) = 6.42, p = .003, η² = .13. For chi-square: χ²(2, N = 150) = 8.91, p = .012, V = .24. For correlation: r(48) = .41, p = .003. In all of these, the letters representing test statistics (t, F, r, p, N) should be italicized. Nothing should be bolded.
A model sentence for a thesis results section might look like this: "Participants in the intervention group reported significantly lower anxiety (M = 28.4, SD = 5.1) than the control group (M = 33.7, SD = 6.3), t(78) = 4.32, p < .001, d = 0.97." Notice that the sentence includes the group means, standard deviations, test statistic, degrees of freedom, exact p-value, and effect size. That is what a complete statistical report looks like. A sentence that says only "the result was significant (p < .05)" is incomplete by current standards. (If you are still deciding which test to run in the first place, start with our guide to choosing the right statistical test.)
Common Mistakes We See in Student Theses
After reviewing over 300 student projects, certain patterns show up again and again. The first is writing "the results were very significant" or "highly significant." APA does not recognize degrees of significance. A result either crosses your predetermined threshold or it does not. The word "very" adds nothing except the impression that you are not sure what significance means. Write "the difference was statistically significant" and let the effect size convey how large the difference actually was.
The second common error is rounding p = .047 to "p < .05." Report the exact value. When a reviewer sees "p < .05" in a thesis, they wonder what the actual number was and why you did not report it. Rounding in that direction creates an impression that you are stretching a borderline result. The same applies to rounding p = .0003 up to "p < .001" when the exact value would be more informative. At the lower end, p < .001 is the accepted convention, but above .001, always give the exact figure.
Third: reporting only the significant results. Publication bias exists in professional research, and the incentive to hide null results is well documented. In your own thesis, selective reporting is worse. Your supervisor and committee expect to see every test you ran, including the ones that returned non-significant results. Omitting them is not just bad practice; some universities treat it as a form of academic misconduct. If you ran 12 comparisons and 2 came back significant, report all 12. The non-significant results are part of the story.
Fourth: never computing an effect size. A thesis that reports only p-values will face questions from any committee member who stays current with methodology. Cohen's d for group comparisons, η² or partial η² for ANOVA, Cramér's V for chi-square, and r or R² for correlations and regression are all standard expectations. Without them, your reader cannot judge whether a statistically significant result actually matters.
If you want to avoid these mistakes entirely, Academic Stats Agent generates APA-7 formatted results with exact p-values, effect sizes, and confidence intervals automatically. It runs real statistical tests on your data and produces publication-ready output, so the formatting is never something you need to worry about.
Key takeaway: A p-value tells you how surprising your data would be if nothing were going on. It does not tell you the probability that your hypothesis is correct, and it says nothing about how large the effect is. Report the exact value, pair it with an effect size, and let the reader decide.
Frequently Asked Questions
What does p = 0.05 mean?
A p-value of 0.05 means there is a 5% probability of observing results at least as extreme as those measured, assuming the null hypothesis is true. It does not mean there is a 95% chance the alternative hypothesis is correct.
Can a p-value be exactly 0?
No. A p-value can be extremely small but never exactly zero. When software such as SPSS displays .000, it is rounding a very small value; report it as p < .001 rather than p = .000.
Does p < 0.05 prove my hypothesis is true?
No. A significant p-value only indicates the data are unlikely under the null hypothesis. It says nothing about the size or practical importance of the effect, which is why APA style requires reporting an effect size alongside the p-value.