Bessel's correction lies in dividing by $n-1$ instead of $n$ in computing the sample variance.

Bessel's correction corrects the bias in estimating the population variance, as well as some of the bias in estimating the population standard deviation. It often increases the mean squared error in such estimations.