A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
The standard normal distribution is a normal distribution with mean 0 and standard deviation 1. Any normal random variable X with mean μ and standard deviation σ can be converted to standard normal via Z = (X - μ)/σ. This standardization allows all normal probabilities to be computed from a single table or calculator for the standard normal. Z-scores measure how many standard deviations a value lies from the mean, facilitating comparison across different scales.
Practice computing Z-scores and using standard normal tables. Solve problems about proportions below/above certain values. Understand that Z-scores linearize normal relationships.
Confusing (X - μ) with (X - σ) in the Z formula. Forgetting to divide by σ. Thinking Z-scores only apply to normal data.
You already know that a normal distribution is characterized by its mean μ and standard deviation σ — change either and you get a different bell curve shifted or stretched on the number line. The problem this creates is practical: computing probabilities requires integrating the normal density, which has no closed-form antiderivative. Historically, people printed probability tables for one particular normal distribution and then showed how to convert any other normal question into a question about that one table. The standard normal distribution — mean 0, standard deviation 1, denoted N(0,1) — is that universal reference.
The conversion formula is Z = (X − μ)/σ. Read this mechanically: first, subtracting μ centers the distribution at zero; then, dividing by σ rescales it so one standard deviation equals one unit. The result Z measures how many standard deviations X sits above (positive Z) or below (negative Z) the mean. A score of Z = 1.5 means the original value was 1.5 standard deviations above average. This is the essence of a Z-score: a standardized deviation that strips away the original units and scale, leaving only relative position.
The payoff is enormous. Suppose a student scores 680 on an exam with μ = 500 and σ = 100, and you want the probability that a randomly chosen student scores higher. You compute Z = (680 − 500)/100 = 1.8, then look up P(Z > 1.8) in the standard normal table (or compute it directly). You never have to know the original scale again; the whole question lives in Z-space. The same logic applies to any normal variable — heights, reaction times, measurement errors — and the single table handles all of them. This universality is what makes Z-scores so foundational for the confidence intervals and hypothesis tests you'll encounter next.
One subtlety worth internalizing: standardization preserves shape but changes location and scale. After the transformation, the distribution is still bell-shaped and symmetric; it has just been shifted and rescaled. Every normal distribution, no matter its original parameters, becomes the same N(0,1) after standardization. That's not a coincidence — it's a consequence of the linearity of expectation and the behavior of variance under scaling. The standard normal is simply the "canonical" version of every normal distribution.