This paper presents an approach that extends the classical error bounding techniques to parametric models. It departs from appropriate pairs of complementary solutions of a linear elastic problem, obtained using a Proper Generalised Decomposition methodology, to determine approximations of selected local outputs and strict bounds of the error of these approximations. The paper starts by presenting the procedures used to obtain the complementary solutions. The properties, the convergence characteristics of the global error and the determination of an indicator of the error distribution are illustrated for a very simple example. The demonstration of the procedure used for determining local outputs and their bounds, also accompanied by illustrative examples, completes the paper. This integral, when computed using a numerical quadrature formula, is in general approximate, because the solutions in space are not constant in j . Nevertheless, preliminary tests indicate that influence of the number of Gauss points on the quality of the PGD approximations is not significant.There is minimum number of points that must be used, those that are necessary for the integration of the explicit polynomial of degree 2 p j , that is, p j C 1 Gauss points. Using a higher quadrature rule increases the quality of the solution, but the change is not significant. Using fewer integration points results in singular Galerkin forms of the equations used to determine the parameters in E 1 and in E 2 . , to whom I extend my gratitude. Preparing these presentations was crucial to have the concepts fresh and organised, ready to respond to a new challenge.