What Is a Confidence Interval?
Plain English first: when a study reports a 95% CI of [68, 76] for a mean exam score, it's saying the best single estimate is somewhere around 72 points, and the data support a plausible range from 68 to 76. That range isn't random — it's built from the sample mean, the sample size, and the standard error of the estimate.
Two components always make up a confidence interval. The point estimate sits at the midpoint of the interval and is the most likely single value for the parameter — usually a sample mean, proportion, or regression coefficient. The margin of error is the distance from that midpoint to either bound. A smaller margin of error means more precision, which usually comes from a larger sample or less variability in the data. You can read more about this component on the margin of error page.
Confidence intervals appear across research disciplines. In medicine, they tell you the plausible range for a drug's effect. In survey research, they frame how far an approval rating might really be from your estimate. In our full confidence intervals guide, we cover how they are calculated for different parameters — this page focuses entirely on reading and reporting them correctly.
How to Interpret a Confidence Interval
To interpret a confidence interval, identify the lower and upper bounds and check whether the interval includes a key value such as zero. A 95% confidence interval of [2.1, 8.4] means that if the study were repeated 100 times, approximately 95 of the resulting intervals would contain the true population value. It does not mean there is a 95% probability the true value lies within this specific interval.
That distinction matters. The confidence level describes the long-run reliability of the method, not the probability attached to any single interval. Once the data are collected, the interval either contains the true value or it doesn't. What "95% confidence" guarantees is that the procedure used to build it will produce intervals that capture the true value in about 95 out of 100 repetitions.
Use the five steps below to read any confidence interval systematically, regardless of whether it comes from a t-test, a regression, or a survey.
Identify the confidence level
Check whether it is a 90%, 95%, or 99% interval. A 95% confidence interval is the most common in academic research and clinical trials. Wider confidence levels (99%) produce wider intervals; narrower ones (90%) produce tighter bounds but accept more risk of missing the true value.
Read the lower and upper bounds
The interval is written as [lower bound, upper bound] or (lower, upper). Both bounds are in the same units as the parameter — if you're estimating a mean blood pressure reduction in mmHg, both numbers are in mmHg.
Locate the point estimate
The midpoint of the interval is the point estimate: (lower + upper) / 2. For [2.1, 8.4], the point estimate is 5.25. Software output usually reports the point estimate directly alongside the interval, but confirming the midpoint is a useful sanity check.
Check for the critical value
For differences and regression coefficients, the critical value is 0. For ratios (odds ratios, relative risks), it is 1. If the interval excludes that critical value entirely, the result is statistically significant at the corresponding alpha level. If it includes the critical value, you cannot rule out no effect.
State your conclusion in plain language
Write a sentence that names the confidence level, the parameter, and the bounds. Example: "We are 95% confident the true mean exam score lies between 68 and 76 points." Avoid probability language — don't write "there is a 95% chance" for this specific interval.
A 95% CI does not mean there is a 95% probability the true value falls within this specific interval. It refers to the long-run behaviour of the method — 95% of intervals built this way will contain the true value. This particular interval either does or doesn't contain it; we just can't know which.
What Does a 95% Confidence Interval Mean?
The confidence level and the confidence interval are different things. The confidence level is a property of the procedure — it's the percentage of intervals the method would produce correctly across many repetitions. The confidence interval is the specific range you calculated from your data.
Think of casting a net into the ocean 100 times to catch one fish at a fixed location. A 95% confidence level means your net will capture the fish in roughly 95 of those casts. Whether this particular cast caught it, you can't say for certain — but you know the net works 95% of the time. The interval is the net; the fish is the true population parameter.
The choice of 95% is conventional rather than scientific. Fisher proposed it as a practical threshold in the 1920s, and it stuck. The trade-off is straightforward: a higher confidence level means a wider interval.
| Confidence Level | Critical Value (z*) | Interval Width | Common Use Case |
|---|---|---|---|
| 90% | ±1.645 | Narrowest | Exploratory research, business analytics |
| 95% | ±1.96 | Standard | Academic research, clinical trials (default) |
| 99% | ±2.576 | Widest | Medical safety studies, manufacturing tolerances |
A deeper explanation of why these thresholds behave as they do is on the confidence level page, which covers the relationship between confidence level, sample size, and interval width in detail.
How to Interpret CI Results: 3 Real Examples
The three examples below cover the scenarios you're most likely to encounter: a mean from a continuous variable, an effect that includes zero, and a proportion from survey data. Each follows the same five-step logic from Section 2.
Mean Exam Score — 95% CI of [68, 76]
A study tests 120 students and finds a mean exam score of 72 points with a 95% confidence interval of [68, 76].
Reading it: The point estimate is 72. The interval runs from 68 to 76. The margin of error is ±4 points. The interval is reasonably narrow, suggesting good precision for this sample size.
✅ We are 95% confident the true population mean exam score is between 68 and 76 points.
Drug Blood Pressure Reduction — 95% CI of [−1, 7]
A drug trial finds a mean reduction in systolic blood pressure of 3 mmHg with a 95% confidence interval of [−1, 7].
Reading it: The critical value for a treatment difference is 0. This interval includes 0, meaning a true reduction of zero is consistent with the data. The wide interval (spanning 8 mmHg) also signals low precision, likely from a small sample size.
✅ Because the 95% CI [−1, 7] includes zero, we cannot conclude the drug has a statistically significant effect on blood pressure at the 5% significance level.
Approval Rating — 95% CI of [58%, 66%]
A poll of 1,000 respondents finds a 62% approval rating with a 95% confidence interval of [58%, 66%].
Reading it: The interval [58%, 66%] lies entirely above 50%, so we can say with 95% confidence that approval exceeds 50% in the population. The margin of error is ±4 percentage points — typical for a sample of this size.
✅ We are 95% confident the true population approval rating is between 58% and 66%. The entire interval exceeds 50%, supporting a majority approval finding.
Summary: When Is a Result Significant?
| CI Result | Includes Critical Value? | Conclusion |
|---|---|---|
| [2.1, 8.4] | No — excludes 0 | Statistically significant result |
| [−1.0, 7.0] | Yes — includes 0 | Not statistically significant |
| [0.5, 0.9] (ratio / OR) | No — excludes 1 | Significant — ratio differs from 1 |
| [0.8, 1.4] (ratio / OR) | Yes — includes 1 | Not statistically significant |
How to Interpret Confidence Intervals in Research
When you read a journal article, confidence intervals appear in several formats. Most papers report them inline in the results section, alongside the test statistic and p-value. In tables, each row often has its own CI column. Understanding what the output looks like in common software helps you read them without guessing.
SPSS displays CIs in the output under the heading "95% Confidence Interval for the Mean" with separate columns for "Lower Bound" and "Upper Bound." For a t-test, the CI is shown in the "Independent Samples Test" table. R reports CIs via functions like t.test() or confint(), typically as a named numeric vector: 95 percent confidence interval: 68.1 76.7. Excel does not compute CIs natively in its basic tools, but the Data Analysis ToolPak reports a "Confidence Level" value you add and subtract from the mean.
APA Format for Reporting Confidence Intervals
Mean: M = 72.4, 95% CI [68.1, 76.7]
Difference: Δ = 3.2, 95% CI [1.1, 5.3]
Odds ratio: OR = 2.4, 95% CI [1.3, 4.5]
Proportion: p̂ = .62, 95% CI [.58, .66]
Interval width is itself informative. A narrow confidence interval signals a precise estimate — the sample was large or the data consistent. A wide confidence interval signals more uncertainty, which typically traces back to a small sample size or high variability. For a deeper discussion, the confidence interval for the mean page works through how sample size and variability interact to produce wider or narrower bounds. For broader guidance on reading statistical output, see the statistical interpretation hub.
Confidence Interval vs P-Value
Both a confidence interval and a p-value address statistical significance, but they tell you different things. A p-value gives you a binary signal: is the result significant at this threshold? A confidence interval gives you that signal and more — it shows the direction of the effect, its magnitude, and the precision of the estimate.
If a 95% CI for a difference excludes zero, the corresponding two-sided p-value is less than 0.05. The two are mathematically linked — but the CI gives you effect size and direction that the p-value alone cannot.
| Feature | P-Value | Confidence Interval |
|---|---|---|
| Tells you if significant | ✓ Yes | ✓ Yes (if excludes 0 or 1) |
| Shows effect magnitude | ✗ No | ✓ Yes |
| Shows direction of effect | ✗ No | ✓ Yes |
| Shows precision of estimate | ✗ No | ✓ Yes (via width) |
| Standard APA reporting | ✓ Yes | ✓ Yes (recommended) |
This is why major journals now require authors to report confidence intervals alongside p-values. A result of p = 0.04 on its own tells you a finding is significant but nothing about whether the effect is clinically meaningful or practically large. Add a 95% CI and you immediately know the plausible range for the real effect size. More on the difference between these two approaches is covered on the p-value page. For the distinction between CI width and margin of error, see the CI vs margin of error comparison.
4 Common Mistakes When Interpreting Confidence Intervals
These four errors appear in student papers, published research, and media reporting. Knowing them by name makes them easier to catch and correct.
| Wrong Interpretation ❌ | Correct Interpretation ✅ |
|---|---|
| "There is a 95% probability the true value is in this range." | 95% of intervals built with this method will contain the true value. This specific interval either does or doesn't — we can't verify which. |
| "A wider CI is always bad." | A wider CI reflects more uncertainty, which is honest. The fix is to increase sample size or reduce measurement variability — not to report a narrower CI by choosing a lower confidence level. |
| "Two overlapping CIs mean no significant difference." | Overlap between two separate CIs does not directly test whether the groups differ. You need a proper two-sample test or a CI for the difference itself. Overlapping intervals can still correspond to a statistically significant comparison. |
| "Confidence intervals only apply to means." | CIs apply to any estimated parameter: proportions, odds ratios, regression coefficients, correlation coefficients, and more. The construction method varies; the interpretation logic is the same. |
Quick Reference: CI Interpretation Checklist
- Identify confidence level — 95% is the standard default
- Read lower and upper bounds in the correct units
- Check whether interval includes 0 (for differences) or 1 (for ratios)
- Narrow CI means high precision; wide CI means more uncertainty
- Report in APA format: M = X, 95% CI [lower, upper]
- Never say "95% probability for this specific interval"
- Don't treat overlapping CIs as proof of non-significance
Confidence Interval Calculator
Enter your sample data below to compute the 95% or 99% confidence interval for a mean. The result includes the point estimate, bounds, and a plain-language interpretation you can copy directly.
Confidence Interval for a Mean
Conclusion
A confidence interval gives you two things a p-value cannot: a sense of where the true effect probably sits, and a sense of how precisely the data pin it down. A 95% CI of [2.1, 8.4] tells you the effect is almost certainly positive, probably somewhere around 5, and large enough that even the most conservative bound (2.1) is meaningfully above zero. That's far more useful than knowing p < 0.05.
Read any interval with the five steps: find the confidence level, note the bounds, identify the point estimate, check the critical value, then write a plain-language conclusion. Avoid the probability framing — it's the single most common misstatement in research writing. And when reporting, use the APA format so reviewers and readers can extract the same information you just did.
Calculate your own confidence interval
Use our free calculator to compute exact bounds for means, proportions, and more — with plain-language output you can copy straight into your paper.
Confidence Interval Calculator →