You calculate a z-score, open a normal-distribution table and suddenly wonder whether the sign should change. The table shows positive numbers, the question asks for an upper tail, and several different rules seem to compete.
Separate two decisions. The sign of the z-score tells you where the value lies relative to the mean. The probability calculation depends on which area the question asks for and what your table reports. A right-tail question does not automatically require a positive z-score.
Let subtraction determine the sign
For a value x, population mean μ and positive population standard deviation σ, the standardised score is z = (x − μ)/σ. Values below the mean give negative scores, values above it give positive scores, and the mean gives zero. The standard normal distribution has mean zero and standard deviation one. NIST, Normal Distribution
Make a small number line before using a table. Mark the mean and the value. If your value is visibly left of the mean but your z-score is positive, check the subtraction order. Do not change the formula to μ − x just to match the portion of a table you happen to see.
Standardising a value does not make its distribution normal. The normal model must be stated or otherwise justified before you use normal probabilities. The arithmetic of a z-score and the probability model are related but distinct steps.
Worked example: a value below the mean
In an invented exercise, completion time is modelled as normally distributed with mean 50 minutes and standard deviation 8 minutes. What is the z-score for 42 minutes?
Calculate (42 − 50)/8 = −8/8 = −1. The value is one standard deviation below the mean. The negative sign does not mean the completion time itself is negative or that the result is poor; it describes position relative to the chosen distribution.
If the question asks for the probability of finishing within 42 minutes, it asks for the area to the left of −1. Under the stated normal model, that area is approximately 0.1587. If it asks for taking longer than 42 minutes, the answer is the complement, approximately 0.8413.
The z-score remains −1 in both questions. What changed was the requested area, not the location of 42 minutes.
Identify your table's convention
Some tables give cumulative area to the left of z. Others give the area between the mean and a positive z-value. Still others focus on tail probabilities. Read the heading and accompanying sketch before using an entry. NIST's standard-normal table illustrates the importance of knowing which area a tabulation represents. NIST, Standard Normal Distribution Table
For a cumulative-left table, Φ(1) is approximately 0.8413. For a mean-to-z table, the corresponding area from zero to one is approximately 0.3413. Both describe the same curve, but they are not answers to the same area question.
Write the probability expression before looking up a number. “P(Z > −1)” protects you from losing the tail direction while searching the table. A shaded sketch provides a second check: the requested area should be larger than one half in that example.
Use symmetry without losing the question
The standard normal curve is symmetric around zero. Therefore the left-tail area below −a matches the right-tail area above a. This lets you use positive entries when the table's convention requires it, but the conversion must preserve the area you actually want. NIST, Normal Distribution
For an original practice question, find P(−1 < Z < 1). Using cumulative areas gives 0.8413 − 0.1587 = 0.6826, approximately. Using a mean-to-z table gives 2 × 0.3413 = the same result.
Do not subtract a left-tail value from a mean-to-z value as though they shared a definition. Label each lookup before combining numbers. Most table errors become obvious when every value has a small shaded diagram beside it.
Reverse the calculation
Sometimes the question gives a z-score and asks for the original value. Rearrange to x = μ + zσ. In the completion-time example, z = 1.5 gives x = 50 + 1.5 × 8 = 62 minutes.
Check the direction before accepting the answer. A positive z-score should produce a value above 50; a negative one should produce a value below it. This simple check can catch an incorrect sign even when the multiplication is correct.
Keep units on x, μ and σ. The z-score has no time unit because the units cancel. Writing “z = 1.5 minutes” mixes the standardised position with the original measurement.
Self-test
1. Mean 100, standard deviation 12, value 82. What is z?
Answer: (82 − 100)/12 = −1.5. The value is below the mean.
2. Does asking for an upper-tail probability change that score to +1.5?
Answer: No. The requested area changes, not the standardised location. Use a complement or symmetry carefully if needed.
3. Mean 30, standard deviation 4, z = −2. What is x?
Answer: 30 − 8 = 22.
4. Can any dataset be treated as normal after converting its values to z-scores?
Answer: No. Standardisation changes location and scale, not the underlying shape into a normal distribution.
Practise the sketch, sign and shaded area before touching a table. Then solve the same area using the table convention supplied by your course. Our active recall guide can help make short prompts that distinguish a sign error from a tail-selection error.
Sources and further reading
- NIST/SEMATECH (online handbook). Normal Distribution. Definition and properties.
- NIST/SEMATECH (online handbook). Cumulative Distribution Function of the Standard Normal Distribution. Table and convention.
