In a linear model under normality it is shown that the error sum of squares is characterized by its distribution. Two proofs are presented, one using the almost-sure ...