The normality assumption also ensures that these variances have a chi-squared distribution.
In other words, the explained and unexplained variations from a regression model have a chi-squared distribution under the assumption that the conditional distribution of Y is normal