How to Read the Z-Score Table
The part that trips most people up is the grid logic itself: rows represent your z-score's integer and first decimal digit, while columns carry the second decimal place. Once you see that structure, every lookup takes about five seconds.
To find P(Z ≤ 1.23), locate row 1.2 and follow it to column .03, where the intersection reads 0.8907, meaning roughly 89% of the distribution falls below that point on the standard normal curve. The old-school approach of literally sliding your finger across the row still works well for building intuition, because tracing through the table reinforces what the cumulative probability actually represents.
Standard Normal Distribution Table — Cumulative Format
The single biggest source of wrong answers in z-table work is not arithmetic — it is grabbing a table in the wrong format. The NIST/SEMATECH e-Handbook of Statistical Methods (section 1.3.6.7.1) lists the 0-to-z area, so z = 1.00 reads 0.3413 there.
Where NIST lists 0.3413 at z = 1.00 (0-to-z format), this table shows 0.8413 — cumulative from the far left.
That distinction matters in practice. Exam questions almost always ask for P(Z ≤ z) or P(Z > z) — one lookup, one possible subtraction. If your textbook answers run exactly 0.5 lower than ours, the book uses the NIST 0-to-z convention; add 0.5 and you are reconciled.
Negative Z-Score Table
Most people figure out quickly that the negative side is just a mirror, and they are right — P(Z ≤ −z) = 1 − P(Z ≤ z) means every negative entry is already implied by the positive table. We built the toggle anyway because doing that subtraction under exam conditions, when you are also tracking sign conventions and degrees of freedom, is where mistakes creep in. Flip to the negative side and the value is already there.
| z | P(Z ≤ z) | Where it appears |
|---|---|---|
| −1.000 | 0.1587 | One SD below the mean — bottom 16% cutoff |
| −1.500 | 0.0668 | Common textbook exercise value |
| −1.645 | 0.0500 | One-tailed test at α = 0.05 (left tail) |
| −1.960 | 0.0250 | Two-tailed test at α = 0.05 (lower bound) |
| −2.326 | 0.0100 | One-tailed test at α = 0.01 |
| −2.576 | 0.0050 | Two-tailed test at α = 0.01 (lower bound) |
| −3.000 | 0.0013 | Three SD below the mean — outlier territory |
Common Z-Score Values
| Confidence Level | Z-Score | Use Case |
|---|---|---|
| 90% | ±1.645 | Surveys, preliminary research |
| 95% | ±1.960 | Most common in academic research |
| 99% | ±2.576 | Medical research, high stakes |
| 99.9% | ±3.291 | Particle physics, rare events |
Critical Z-Values for Hypothesis Testing
Confidence intervals and hypothesis tests use the same z-values but frame them differently. The tail you care about decides which column to read.
| α Level | One-Tailed z | Two-Tailed z | Typical Use |
|---|---|---|---|
| 0.10 | 1.282 | ±1.645 | Exploratory, marketing A/B tests |
| 0.05 | 1.645 | ±1.960 | Standard academic threshold |
| 0.025 | 1.960 | ±2.241 | Bonferroni-adjusted for m=2 tests |
| 0.01 | 2.326 | ±2.576 | High-confidence decisions |
| 0.005 | 2.576 | ±2.807 | Clinical trials, safety studies |
| 0.001 | 3.090 | ±3.291 | Fraud detection, particle physics |
One pattern worth memorizing: the two-tailed z at α equals the one-tailed z at α/2. So ±1.960 shows up for two-tailed 0.05 and one-tailed 0.025 — same number, different framing. The particle-physics community's 5-sigma rule corresponds to a one-tailed α of roughly 3×10⁻⁷, which is why discovery thresholds feel worlds apart from everyday p-values.
Frequently Asked Questions
What does the z-table value represent?
The table value P(Z ≤ z) gives the cumulative probability that a random observation from a standard normal distribution falls below your chosen z-score. It represents the shaded left-tail area under the curve — picture a bell curve with everything to the left of your z-score filled in, and the table tells you how much area that shading covers.
How do I find the area to the right of z?
To find P(Z > 1.23), look up 1.23 in the table to get 0.8907, then compute 1 minus 0.8907, which gives 0.1093. This complement rule catches people off guard because most tables only show left-tail values, but it is the same subtraction step every time.
How do I use the table for negative z-scores?
The standard normal curve is perfectly symmetric, which means negative z-scores are just the mirror image of positive ones — and that symmetry is the shortcut that eliminates the need for a separate negative z-table entirely. For negative z-values, P(Z ≤ -z) equals 1 minus P(Z ≤ z). So for z = -1.23, subtract 0.8907 from 1 to get 0.1093. The NIST Engineering Statistics Handbook confirms that this symmetry property holds exactly for the standard normal distribution, which means one table genuinely covers every possible lookup.
Why does z = 1.96 show up everywhere in statistics?
1.96 is the cutoff where exactly 2.5% of the standard normal distribution sits in each tail, so it is the critical value for a two-tailed 95% confidence interval — the default choice across most academic research. The NIST Engineering Statistics Handbook lists it as the common z-value for that reason, and it surfaces in everything from political polling margins to clinical trial reports. Rough back-of-envelope work rounds 1.96 up to 2, which stays close enough when the audience does not need four-decimal precision.
What is the difference between a one-tailed and two-tailed critical value?
A two-tailed test splits the α budget between both tails of the distribution, so α = 0.05 puts 2.5% in each tail, which lands on ±1.960. A one-tailed test keeps the full 5% in a single tail, giving a smaller critical value of 1.645. The upshot: one-tailed tests reject the null more easily, which is why reviewers push back when researchers switch from two-tailed to one-tailed without pre-registering that choice.
What does 1 − 0.8907 mean when using the z-table?
We see this question often enough that it is worth saying plainly: the z-table has no right-tail column. Every number in the body is a left-tail probability, so P(Z > 1.23) is not in there directly. You pull 0.8907 for z = 1.23, which covers the left side, and 1 − 0.8907 = 0.1093 covers the right. That is the whole answer.
How do I find the z-score for a given probability (reverse lookup)?
Confidence intervals are the usual reason people need this in reverse: you start from a known probability (0.9750, say) and work back to the z-score that produced it. The way to do that in the table is to find 0.9750 in the body and read off the margins — row 1.9, column .06 — giving z = 1.96. That number (±1.96) is the one in every 95% confidence interval formula you have ever seen, which is a useful sanity check that you found the right cell.