Normal Distribution — Bell Curve, Z-Score & Formula

Plot enough exam scores, repeated measurements, or quality-control readings and they often pile up into the same bell shape: most values cluster near the center, and the rare ones trail off symmetrically on both sides. That is a normal distribution, and reading it comes down to a short, repeatable procedure.

When this procedure applies

The normal distribution is the right model when measurements cluster around a central value and extreme values are relatively rare — measurement error, test-score interpretation, quality control, and the behavior of sample averages. The condition that licenses everything below is fit: the normal shape must be a reasonable match for the data. Not all real data are normal; the model is a useful approximation only when the shape, context, and assumptions support it. You will often see the notation

meaning $X$ is modeled as normal with mean $\mu$ and variance $\sigma^2$, so the standard deviation is $\sigma$ with $\sigma > 0$.

The four-step reading

1. Find the center. Identify the mean $\mu$ — it marks the middle of the bell curve and moves the whole curve left or right.

2. Check the spread. The standard deviation $\sigma$ sets how wide or narrow the curve is. A small $\sigma$ packs values tightly around the mean; a larger $\sigma$ spreads them out. (The full density formula is $f(x) = \frac{1}{\sigma \sqrt{2\pi}} e^{-(x-\mu)^2/(2\sigma^2)}$; you do not need to memorize every part — what matters is that $\mu$ shifts it and $\sigma$ widens or narrows it. Note this describes density, not the probability of one exact value, so probabilities come from intervals like $P(X < 80)$.)

3. Standardize the value. Compute the z-score to place a value relative to the rest:

If $z = 1.5$, the value is $1.5$ standard deviations above the mean; if $z = -2$, it is $2$ below.

4. Read probabilities carefully. When the normal model is reasonable, use the empirical rule:

It is an approximation, not a guarantee for every data set.

Full worked example, step by step

Suppose exam scores are modeled by

so the mean is $70$ and the standard deviation is $10$.

Steps 1–2 (center and spread): $\mu = 70$, $\sigma = 10$.

Step 4 (empirical rule): about $68\%$ of scores fall within one standard deviation:

About $95\%$ fall within two:

Step 3 (standardize): a student scored $85$, so

The score is $1.5$ standard deviations above the mean — clearly above average, but not deep into the tail.

Where each step goes wrong, and how to verify

• Step 1, treating every bell-shaped graph as normal. Some data are skewed, heavy-tailed, or multi-peaked; check the shape before trusting the model.
• Step 2, confusing density with probability. $f(x)$ is not the probability $X$ equals one exact number — for continuous models, exact-point probability is $0$, so work with intervals.
• Step 3, mixing up variance and standard deviation. Variance is $\sigma^2$; the z-score uses $\sigma$, not $\sigma^2$. Self-check: did you divide by the standard deviation, not the variance?
• Step 4, using the empirical rule without checking the model. The $68$–$95$–$99.7$ rule belongs to the normal distribution and should not be applied automatically.

Try the procedure yourself

Change the example to $X \sim N(100, 15^2)$. Run the steps: confirm the center and spread, compute the z-score of $130$, then find the interval that covers about $95\%$ of values. Watching the bell curve shift as you change the mean or standard deviation is the fastest way to internalize the procedure.