A normal distribution is a continuous probability distribution whose bell curve shape is symmetric around a central mean, with values becoming steadily less likely the farther they sit from that center. Statisticians also call it the Gaussian distribution, after Carl Friedrich Gauss, and the two names describe exactly the same curve. Every normal distribution bell curve is fixed by just two numbers — the mean and the standard deviation — and the same 68-95-99.7 rule applies no matter what those two numbers are. This guide covers the formula, the standard Gaussian distribution, how to read a normal distribution chart, and a fully worked probability calculation.

Key takeaways

Point Details
Two parameters define the whole curve The mean (μ) sets the center and the standard deviation (σ) sets the spread; nothing else changes a normal distribution’s shape.
The empirical rule gives fast benchmarks About 68% of values fall within 1σ of the mean, about 95% within 2σ, and about 99.7% within 3σ — true for every normal distribution.
Probability is area, not height The density function f(x) can exceed 1; only the area under the curve between two points is a probability.
Z-scores convert any normal to the standard one z = (x − μ) / σ maps any normal distribution onto the standard Gaussian distribution (μ = 0, σ = 1), where tables and calculators are calibrated.
Not all data is normal Skewed, heavy-tailed, or multi-peaked data needs a different model or a transformation before normal-distribution methods apply.

Quick Checklist: How to Find a Normal Distribution Probability

Before you look anything up, confirm you have the right numbers and the right question. These five checks keep a normal-distribution probability calculation from going wrong in a way that is easy to miss.

Quick checklist: finding a normal distribution probability

  • Confirm the mean (μ) and standard deviation (σ) Every calculation depends on these two numbers; mixing up σ with the variance (σ²) is the most common entry error.
  • State which direction you need P(X < x), P(X > x), or P(a < X < b) each use the standard normal CDF differently.
  • Convert to a z-score z = (x − μ) / σ turns your specific problem into a lookup on the standard Gaussian distribution.
  • Read the cumulative probability Use a z-table, software function, or the calculator below to get Φ(z), the area to the left of z.
  • Sanity-check against the empirical rule If your interval is roughly ±1σ, ±2σ, or ±3σ, your exact answer should land close to 68%, 95%, or 99.7%.

The sections below walk through each of those steps in detail, with a calculator and a fully worked example.

What Is a Normal Distribution?

A normal distribution is a continuous probability distribution that is symmetric about its mean: the left and right halves of the curve are mirror images, and the mean, median, and mode all sit at the same central point. The NIST/SEMATECH e-Handbook’s entry on the normal distribution lists a skewness of exactly 0 and a kurtosis of exactly 3 for the distribution, which is the formal way of saying the curve is perfectly symmetric with no systematic lean toward either tail.

The distribution is also called the Gaussian distribution, after the German mathematician Carl Friedrich Gauss. The MacTutor History of Mathematics archive at the University of St Andrews documents Gauss’s 1809 two-volume treatise, Theoria motus corporum coelestium, on estimating and refining the orbits of celestial bodies. “Normal distribution” and “Gaussian distribution” refer to exactly the same curve and are used interchangeably in statistics.

A random variable follows a normal distribution when three conditions hold together: the curve is symmetric about the mean, the mean/median/mode coincide at the center, and the tails approach but never touch the horizontal axis. No real dataset is perfectly normal, but many — heights, exam scores, measurement errors, blood pressure readings — are close enough for the normal distribution to be a useful model, largely because the central limit theorem pulls the distribution of sample means toward normal even when the underlying data are not. The central limit theorem article covers that convergence in depth.

What Shapes the Bell Curve: Mean and Standard Deviation?

The bell curve normal distribution shape is governed entirely by two parameters, and nothing else about the curve can change independently of them.

The mean (μ) sets the location. It is the single peak of the curve and the point where the left and right halves split evenly — exactly 50% of the area sits on each side. Shift μ, and the whole curve slides along the number line without changing shape.

The standard deviation (σ) sets the spread. A small σ produces a tall, narrow bell with short tails: values cluster tightly near the mean. A large σ produces a wide, flat bell with long tails: values spread out further. Two normal distribution curves can share the same mean and still look completely different if their standard deviations differ — this is why two exam score distributions with an identical average can have very different numbers of students at the extremes.

How μ and σ change the bell curve
Parameter Controls Effect of a larger value
Mean (μ) Location of the peak Shifts the entire curve right along the x-axis
Standard deviation (σ) Width of the bell Flattens and widens the curve; tails extend further

For a quick visual check, the inflection points — where the curve switches from curving downward to curving upward — always sit at exactly μ − σ and μ + σ. That is a fast way to read σ directly off a plotted normal distribution chart without doing any arithmetic.

What Is the Normal Distribution Formula?

The shape of every normal distribution curve is governed by its probability density function (PDF):

f(x) = (1 / (σ · √(2π))) · exp( −(x − μ)² / (2σ²) )

Where μ is the mean, σ is the standard deviation, π ≈ 3.14159, and exp(·) is the exponential function. The NIST/SEMATECH handbook gives this exact formula and notes that the special case μ = 0, σ = 1 is called the standard normal distribution — the subject of the next section.

The exponent −(x − μ)² / (2σ²) is zero only when x = μ, which is where the PDF reaches its peak. As x moves away from μ, the exponent grows more negative and the PDF value shrinks toward zero, producing the bell taper. Crucially, f(x) is a density, not a probability — for the standard normal distribution, f(0) ≈ 0.399, a number that can exceed 1 for a narrow enough σ and is never itself a probability. The probability that X falls in a range [a, b] is the area under the curve between those two points, found with the cumulative distribution function (CDF) rather than the PDF directly — in practice, that means a z-table, software, or the calculator below, not hand integration.

What Is the Standard Gaussian Distribution, and How Do You Convert to Z-Scores?

The standard Gaussian distribution — also called the standard normal distribution — is the specific case where μ = 0 and σ = 1, written Z ~ N(0, 1). It matters because any normal distribution can be converted to it through standardization, which is what makes a single set of tables and software functions work for every possible mean and standard deviation.

Given a value x from a normal distribution with mean μ and standard deviation σ, its z-score measures how many standard deviations above or below the mean it sits:

z = (x − μ) / σ

For example, with test scores normally distributed at μ = 70 and σ = 10, a score of 85 has:

z = (85 − 70) / 10 = 1.5

That score sits 1.5 standard deviations above the mean. Looking up Φ(1.5) — the cumulative probability at z = 1.5, as tabulated in NIST/SEMATECH’s cumulative distribution function table for the standard normal distribution — gives 0.9332, meaning about 93.3% of scores fall below 85. Reading the table itself (rows, columns, and the two common formats) is covered step by step in the z-table guide; this section is about the standardization move itself, which is what makes that lookup possible for any mean and standard deviation at all.

What Does the Empirical Rule (68-95-99.7) Tell You?

The empirical rule gives three probability benchmarks that hold for every normal distribution, regardless of its specific μ and σ.

The empirical rule for any normal distribution
Range Approximate probability
μ ± 1σ 68.27%
μ ± 2σ 95.45%
μ ± 3σ 99.73%

In a dataset of adult heights with μ = 175 cm and σ = 7 cm, the rule says about 68% of people fall between 168 cm and 182 cm, about 95% between 161 cm and 189 cm, and nearly everyone — 99.7% — between 154 cm and 196 cm. A value more than 2σ from the mean occurs only about 5% of the time under a normal distribution; one beyond 3σ is rare enough (roughly 1 in 370 observations) that many quality-control systems treat it as a signal something changed. The NIST/SEMATECH handbook’s discussion of variables control charts documents exactly this use: standard control limits are set at the process mean ± 3 standard deviations, directly borrowing the empirical rule’s 99.7% benchmark to flag unusual process behavior.

The rule is specific to the normal distribution — it does not hold for a skewed distribution, where the percentages within one or two standard deviations can be very different. Chebyshev’s inequality offers a weaker, distribution-free bound (at least 1 − 1/k² of any dataset falls within k standard deviations of the mean) that holds everywhere but gives much looser estimates than 68-95-99.7 does for data that actually is normal.

How Do You Read a Normal Distribution Chart?

A normal distribution chart plots probability density on the vertical axis against the value on the horizontal axis, producing the bell curve. The single rule that makes the chart readable: area corresponds to probability, not height.

The total area under any normal distribution curve equals exactly 1. The tallest point is always at the mean, which splits the area exactly 50/50. The probability that X falls between two values a and b is the shaded area between them, computed from the standard normal CDF, Φ(z):

P(a < X < b) = Φ(z_b) − Φ(z_a)

where z_a = (a − μ) / σ and z_b = (b − μ) / σ. OpenStax’s Introductory Statistics, chapter 6 works through this same area-based logic for the standard normal distribution before extending it to any μ and σ. A useful rule of thumb when scanning normal distribution curves visually: shading one bell-width on either side of the peak captures roughly two-thirds of the area, and shading two bell-widths captures nearly all of it — the empirical rule, read directly off the shape of the chart.

Where Does the Normal Distribution Show Up in Practice?

The normal distribution appears across science, industry, and statistics itself for a specific reason: many measurements are the sum of many small, independent sources of variation, and sums of independent random effects converge toward a normal shape as the central limit theorem predicts. Adult height, blood pressure, measurement error, and manufacturing tolerances all approximate normal distribution curves for this reason.

Inferential statistics leans on the same fact from the other direction. The t-test, z-test, ANOVA, and linear regression all assume that residuals are approximately normal, and confidence intervals for a population mean rely on the sampling distribution of the mean being approximately normal for large samples — the normality-tests guide covers how to formally check that assumption with Shapiro-Wilk, Anderson-Darling, and related tests rather than eyeballing a histogram. When the assumption holds, these methods produce accurate p-values; when it fails badly — strong skew, heavy tails, multiple modes — a transformation or a nonparametric alternative is usually the safer path.

A Worked Example: Probability a Test Score Falls Between 60 and 90

Setup. A standardized test is normally distributed with mean μ = 70 and standard deviation σ = 10. Find the probability that a randomly selected student scores between 60 and 90.

Normal distribution probability process Four connected steps: identify the bounds 60 and 90; convert each to a z-score; look up the cumulative probabilities; subtract to get the interval probability of 0.8185. 1 Identify thebounds a = 60, b = 90. 2 Convert toz-scores z_a = (60−70)/10 =−1.00; z_b = (90−70)/10= 2.00. 3 Look up Φ(z) Φ(−1.00) = 0.1587;Φ(2.00) = 0.9772. 4 Subtract P(60<X<90) = 0.9772 −0.1587 = 0.8185.
Figure 1. Four steps from the raw bounds to the interval probability for the test-score example.
  1. Identify the bounds. Lower bound a = 60; upper bound b = 90.

  2. Convert to z-scores.

z_a = (60 − 70) / 10 = −10 / 10 = −1.00
z_b = (90 − 70) / 10 =  20 / 10 =  2.00
  1. Look up the cumulative probabilities from a standard normal table (or the calculator below):
Φ(−1.00) = 0.1587   (15.87% of scores fall below 60)
Φ( 2.00) = 0.9772   (97.72% of scores fall below 90)
  1. Subtract to get the interval probability.
P(60 < X < 90) = Φ(2.00) − Φ(−1.00) = 0.9772 − 0.1587 = 0.8185

About 81.85% of students score between 60 and 90.

Verification using the empirical rule. The interval [60, 90] runs from −1σ to +2σ around the mean. Because the normal distribution is symmetric, half of the ±1σ band (half of 68%, or 34%) plus half of the ±2σ band (half of 95%, or 47.5%) covers exactly the −1σ-to-+2σ span: 34% + 47.5% = 81.5%, close to the exact 81.85% computed above. Decision: 81.85% of students fall in this range, so a score of 60 or 90 is not unusual on its own — only scores well outside roughly 50 to 90 (beyond ±2σ) would be worth flagging as atypical.

Try the Calculator

Enter mean = 70, standard deviation = 10, x = 60, upper bound = 90, and select P(x < X < upper) to reproduce this calculation. The calculator returns 0.8186 rather than 0.8185, because it computes Φ(z) directly instead of reading it from a table rounded to four decimal places — the small gap is rounding, not an error.

Calculator

Normal Distribution Calculator

Enter values and compute the result.

How Do You Calculate Normal Distribution Probabilities in Excel, Python, or R?

Software removes the table-lookup step once you have the right mean, standard deviation, and bound.

Excel:   =NORM.DIST(x, mean, standard_dev, TRUE)     returns P(X ≤ x)
         =NORM.S.DIST(z, TRUE)                       returns Φ(z) for a standardized z

Python:  from scipy.stats import norm
         norm.cdf(x, loc=mean, scale=sd)             # P(X ≤ x)
         norm.cdf(z)                                  # Φ(z) for the standard normal

R:       pnorm(x, mean = mu, sd = sigma)              # P(X ≤ x)
         pnorm(z)                                      # Φ(z) for the standard normal

Microsoft’s documentation for NORM.DIST and NORM.S.DIST covers the Excel syntax above, including the cumulative-vs-density flag. SciPy’s scipy.stats.norm reference documents cdf, along with pdf, ppf (the inverse CDF), and the location/scale parameterization Python uses instead of separate mean and standard-deviation arguments. R’s documentation for the normal distribution functions covers dnorm, pnorm, and qnorm, which play the same density, CDF, and inverse-CDF roles. For an interval probability in any of the three, compute the CDF at both bounds and subtract — the same Φ(z_b) − Φ(z_a) logic from the worked example above, just without the manual table lookup.

Common Mistakes with the Normal Distribution

  • Assuming all data is normal. Income, survival times, and stock returns are typically skewed or heavy-tailed. Inspect a histogram or a normal probability plot before applying normal-distribution methods, and run a formal check with a normality test rather than relying on a visual guess alone.
  • Confusing the PDF with a probability. f(x) is a density and can exceed 1; it is never itself a probability. Probability comes only from the area under the curve between two points.
  • Misreading σ as the variance. The formula and the z-score both use σ (standard deviation), not σ² (variance). If you only have the variance, take its square root first.
  • Applying the empirical rule to non-normal data. The 68-95-99.7 benchmarks are specific to the normal distribution; for a skewed distribution, the actual percentages within one or two standard deviations may differ substantially.
  • Forgetting the continuity correction. When using the normal distribution to approximate a discrete distribution such as the binomial distribution, a ±0.5 continuity correction improves accuracy — for example, approximate P(X = 10) with P(9.5 < X < 10.5), not a direct density lookup.

Statohub’s Take on the Normal Distribution

Statohub’s position is that the normal distribution earns its central place in statistics honestly — it is the limiting shape the central limit theorem predicts for sums of independent effects, not a default assumption to reach for out of habit. The practical discipline is to treat “approximately normal” as a claim to check, with a plot and a formal test, rather than something to assume because the math downstream gets easier if it is true.

Put the Normal Distribution to Work With Statohub’s Tools

The calculator above handles one-tailed and interval probabilities directly — enter your mean, standard deviation, and bound, and it returns the area under the curve. For a standalone version with the same options, open the full normal distribution calculator, or browse every probability tool on Statohub’s calculators hub. For the surrounding concepts, the Probability & Distributions hub covers discrete distributions such as the binomial and continuous ones such as the exponential, both of which the normal distribution connects to, and the probability formula guide covers the rules this distribution’s probabilities still have to obey.

Sources

Sources

  1. NIST/SEMATECH e-Handbook of Statistical Methods — 1.3.6.6.1. Normal Distribution NIST
  2. NIST/SEMATECH e-Handbook of Statistical Methods — 1.3.6.7.1. Cumulative Distribution Function of the Standard Normal Distribution NIST
  3. NIST/SEMATECH e-Handbook of Statistical Methods — 6.3.2. What are Variables Control Charts? NIST
  4. OpenStax — 6.1 The Standard Normal Distribution, Introductory Statistics 2e OpenStax
  5. SciPy — scipy.stats.norm reference SciPy
  6. The R Project — stats::Normal (dnorm, pnorm, qnorm) documentation The R Project
  7. Microsoft Support — NORM.DIST function Microsoft
  8. Microsoft Support — NORM.S.DIST function Microsoft
  9. MacTutor History of Mathematics, University of St Andrews — Johann Carl Friedrich Gauss biography University of St Andrews

FAQ

Frequently asked questions

What Is a Normal Distribution?
A normal distribution is a symmetric, bell-shaped continuous probability distribution where the mean, median, and mode all coincide at the center. It is fully described by two parameters — the mean (μ) and the standard deviation (σ) — and the probability of any interval is the area under the curve between those two points, not the height of the curve at any single value.
Why Is It Called the Bell Curve Normal Distribution?
The name comes from its visual shape: plotting probability density against value produces a curve that rises to a single peak at the mean and falls off symmetrically on both sides, resembling a bell. "Normal distribution," "bell curve normal distribution," and "Gaussian distribution" all refer to the exact same mathematical object, named after Carl Friedrich Gauss.
What Is the Difference Between a Normal Distribution and the Standard Gaussian Distribution?
A normal distribution has two free parameters, μ and σ, which can take any values. The standard Gaussian distribution is the specific case where μ = 0 and σ = 1. Any normal distribution converts to the standard Gaussian by computing z = (x − μ) / σ, which is why probability tables and most software functions are calibrated to that one standardized version.
What Does the 68-95-99.7 Rule Mean?
It means that for any normal distribution, about 68% of values fall within one standard deviation of the mean, about 95% fall within two standard deviations, and about 99.7% fall within three. These percentages come from integrating the normal distribution’s density function over those ranges and hold for every normal distribution regardless of its specific mean and standard deviation.
How Do You Find a Probability From a Normal Distribution Chart?
Convert the value to a z-score with z = (x − μ) / σ, then read the cumulative probability Φ(z) from a table, software function, or calculator. For an interval between a and b, compute Φ(z_b) − Φ(z_a). On the chart itself, that probability is the shaded area between a and b, never the height of the curve at either point.
Can Any Dataset Be Modeled as a Normal Distribution?
No. A normal distribution model fits data that is symmetric, unimodal, and bell-shaped. Datasets with strong skew, heavy tails, multiple peaks, or a hard lower bound — counts, wait times, income — usually need a different distribution, such as the lognormal, Poisson, or exponential, or a transformation before normal-distribution methods apply. Plot the data and run a formal normality test before assuming it fits.