A general method is given for finding optimal values of the constants occurring in the numerical inversion of Laplace transforms. It is then shown that one particular set of constants obtained by this method ensures that the process of inversion is well adapted to transforms which arise frequently in the analysis of dynamical systems and in other applications.