The Herington test for thermodynamic consistency is examined in detail and shown to be incorrect. New expressions are proposed for the cases where information on heats of mixing is or is not available.