In this paper we present methods for calculating both the average and worst-case quadratic performances for a family of linear time-invariant systems with real constant parametric uncertainty. The linear systems are described by state-space models with affine linear dependence on the uncertain parameters. A closed-form expression for the average performance is given, and it is shown that the worst-case calculation is similar to that of the H-infinity norm of a certain augmented system. Both continuous-time and discrete-time cases are considered.