A statistical procedure designed to distinguish real treatment difference from that due to chance (i.e., random variation). To distinguish differences due to sampling error from differences due to discrepancy between observations and hypothesis.