Good experimental practice therefore requires the use of specific tests to establish whether observed differences in performance are truly statistically significant. A popular method for this purpose is the two-tailed t-test, which gives an indication about the chance that the values came from the same underlying distribution.