10% condition

Sampling without replacement makes each draw depend a little on the ones before it, because removing a person changes what is left in the population. The 10% condition, written $n \le 0.10N$ (n at most one tenth of N, the population size), keeps that dependence small enough that standard deviation formulas built for independent draws still hold. In a school of 1,200 students, $0.10(1200) = 120$, so a sample of 80 satisfies the condition while a sample of 200 does not. You only need this check when sampling without replacement.