Two approaches for extending relational database tables to allow storing uncertain and incomplete information have been proposed in the past. Grahne in 1984 introduced constraint tables that allow constraints to be attached to tuples and tables. Independently, Barbara et al. introduced in 1992 probabilistic tables that can contain random variables for some of the fields. In this paper, we combine the two approaches by introducing the concept of a mixed table: a table that allows storing both random variables and linear constraints on them.