Abstract. Gauss-Legendre quadrature formulas have excellent convergence properties when applied to integrals 1 0 f (x) dx with f ∈ C ∞ [0, 1]. However, their performance deteriorates when the integrands f (x) are in C ∞ (0, 1) but are singular at x = 0 and/or x = 1. One way of improving the performance of Gauss-Legendre quadrature in such cases is by combining it with a suitable variable transformation such that the transformed integrand has weaker singularities than those of f (x). Thus, if x = ψ(t) is a variable transformation that maps [0, 1] onto itself, we apply Gauss-Legendre quadrature to the transformed integral 1 0 f (ψ(t))ψ (t) dt, whose singularities at t = 0 and/or t = 1 are weaker than those of f (x) at x = 0 and/or x = 1. In this work, we first define a new class of variable transformations we denote S p,q , where p and q are two positive parameters that characterize it. We also give a simple and easily computable representative of this class. Next, by invoking some recent results by the author concerning asymptotic expansions of Gauss-Legendre quadrature approximations as the number of abscissas tends to infinity, we present a thorough study of convergence of the combined approximation procedure, with variable transformations from S p,q . We show how optimal results can be obtained by adjusting the parameters p and q of the variable transformation in an appropriate fashion. We also give numerical examples that confirm the theoretical results.