random error
/ RAN-dum EHR-ur /
Try to read the exact level of liquid in a measuring cylinder ten times in a row. Your numbers will wobble a little — 25.31, 25.29, 25.32, 25.30 — not because anything is broken, but because tiny, uncontrollable wisps of variation creep into every reading: a slight tremor in your hand, a flicker of light, a touch of warmth. That irreducible jitter, sometimes up and sometimes down, is random error.
Random error is the unpredictable scatter in repeated measurements, with no fixed direction: it is just as likely to push a reading high as low. Because of this even-handedness, it tends to cancel out when you average many measurements, which is exactly why we take replicates. It is the source of imprecision, and it is described statistically — by the standard deviation and the bell-shaped Gaussian distribution that repeated readings usually trace out.
Random error matters because it sets a floor on how precise a measurement can be, and unlike systematic error it can never be fully eliminated, only reduced — by averaging more replicates, using a steadier instrument, or controlling the environment. The honest caveat: averaging tames random error but does nothing about systematic error, so a tight, low-scatter result is reassuring about precision yet says nothing about whether the answer is true.
Weighing the same coin five times on a sensitive balance gives 3.2741, 3.2739, 3.2742, 3.2740, 3.2738 g. The last digit dances around because of random error; averaging the five readings gives a more precise best estimate.
The wandering last digit is random error at work.
Averaging reduces random error in the mean roughly in proportion to the square root of the number of measurements — so four times as many readings only halves the scatter of the average.