The singular-field-based mode-mix decomposition procedure is assessed for its ability to accurately predict delamination growth in various test geometries. Initially, toughness versus mode mix is obtained from laminates with delaminations at their midplanes. These results are then used to predict delamination growth in other geometries, and the associated tests of these geometries are performed. A typical graphite/epoxy material system is considered. For this material, it is shown that predictions of delamination growth using a singular-field-based mode-mix decomposition will result in significant errors. That is, when a classical, singular-field-based definition of mode mix is used, tests of different specimen geometries predicted to be at the same mode mix do not display the same toughness. All the test data is then reinterpreted with the aid of a previously developed crack-tip-element analysis. This approach allows for various definitions of mode mix to be evaluated, of which the singular-field-based result is one special case. It is shown that defining mode mix in terms of non-classical parameters results in accurate predictions of delamination growth, and test geometries predicted to be at the same mode mix now display the same toughness. This new definition of mode mix is based on parameters that uniquely characterize the conditions at the crack tip, but which are insensitive to the scale of the near-tip damage. A systematic method for determining the “appropriate definition” of mode mix for arbitrary material systems is described.