Normal Distribution Basics: Centre, Spread and Symmetry
Read probabilities as areas under a normal curve, state mean and standard deviation explicitly, and use symmetry without inventing table values.
A high curve is not a point probability
A normal distribution is a continuous model. The total area under its density curve is 1; an interval probability is the area over that interval. The vertical axis represents density, not the probability of exactly one specified value. A single point has probability zero in the continuous normal model because it has zero width. This is a statement about the model, not a claim that a measurement can never be recorded.
Name the parameters before reading notation
The mean sets the centre and standard deviation sets the scale. Some books write with variance as the second parameter; others use with standard deviation. Check the convention instead of copying a symbol without its definition.
Suppose X has a normal distribution with mean 40 and standard deviation 5. The values 35 and 45 are equally far from the centre:
By symmetry, .
Symmetry does not supply an invented area
Write the central probability as p. Each tail is then . That follows from symmetry and total area. The actual value of p needs a suitable table, tool or supplied information; it cannot simply be assigned 0.70.
The familiar statement that approximately 68% lies within one standard deviation is an approximate normal-model rule, not a guarantee that all observations lie there. Nor should every roughly bell-shaped dataset automatically be treated as exactly normal.
What changes when the curve moves or widens?
Changing the mean while fixing the standard deviation shifts the curve horizontally. Changing the standard deviation changes its width and requires the density height to adjust so total area remains 1. A taller peak does not imply a larger total probability: both complete density curves still have total probability 1.
A graph-reading checklist
- Does the graph show counts, relative frequencies or probability density?
- What are its centre and scale?
- Is the question asking for an interval, a tail or a standardised position?
- Did a numerical area come from information, symmetry or an actual calculation?
- Am I turning a sample observation into a model guarantee?
In a continuous model, replacing with at an endpoint does not change an interval probability. That is different from asking whether an endpoint belongs to a set. Keeping those questions separate prevents a mechanical transfer of interval-notation rules into probability.