the sample range
The simplest single number describing how spread out a sample is just the distance from its smallest value to its largest: the sample range. If today's temperatures ran from 12 to 27 degrees, the range is 15. It is quick to compute and easy to explain, which is why quality-control charts and weather reports lean on it.
Formally the range is R = X_(n) - X_(1), the largest order statistic minus the smallest. Because it is built from two order statistics that are themselves dependent, finding its distribution takes a little care: you work from the joint distribution of the minimum and maximum (which for an iid sample has a known form) and then find the distribution of their difference. The result depends on the underlying distribution F and grows, on average, as the sample size n increases — more draws give more chances for an extreme high and an extreme low, widening the gap.
Two honest cautions. The range uses only two numbers and throws away everything in between, so it is a crude and fragile measure of spread — a single outlier can blow it up, and it almost always understates the true spread for small samples while drifting upward for large ones. For careful work the standard deviation or the interquartile range is usually preferred; the range earns its keep mainly when speed and simplicity matter.
For five independent uniforms on (0, 1), the expected range is (n - 1)/(n + 1) = 4/6 = 2/3. Notice it is less than the full interval length 1: with only five draws you rarely hit values right at 0 and right at 1, so the observed spread typically undershoots the true span.
The range understates the true spread for small samples, since the extremes are rarely reached.
The range depends only on the two extremes, so it is very sensitive to outliers and ignores the bulk of the data. For most purposes the standard deviation or interquartile range is a more reliable measure of spread.