A statistic is said to be pivotal if its sampling distribution does not depend on unknown parameters.
Pivotal statistics are well suitable for statistical tests – because this property allows you to control type I error, irrespective of any unknown parameters of the population the data came from.