It is shown that two tests for the adequacy of a fitted ARMA (p, q,) model, based respectively on comparison through the Lagrange multiplier test with an ARMA (p + k, q) model and on analysis of the first k residual autocorrelations from the estimated model, are equivalent.