Neural studies of combination-tone (CT) responses in the eighth nerve and antroventral cochlear nucleus of anesthetized cats give evidence of stimuluslike intracochlear CT's with a similar amplitude spectrum as inferred from human psychophysics. An unsolved problem raised in these studies is the gross discrepency between the phase of CT's measured psychophysically and neurally. It is unknown whether incompatibility lies in the cochleas or the criteria for measuring CT's in the different experiments. New psychophysical experiments were developed to clarify this issue. Secondary CT's (SCT) were generated by the interaction between a primary cubic CT (CTT) and a third tone in the stimulus. The SCT measured by cancellation exhibits similar properties to those of a CT generated by the comparable two-tone stimulus. The SCT is eliminated when the CCT is cancelled. These findings support the view that CT's exist in the cochlea as spatially distinct, stimuluslike excitations, and that CT excitations are eliminated by psychophysical cancellation. The SCT phase provides a measure of the CCT phase without requiring direct cancellation of the CCT. Phase measurements by the new indirect and older direct methods imply that the CT phase may be constant with changing sound level of the primary stimulus as found in the neural studies; these measurements also reveal nonlinear phase effects not found in neural studies. The new data suggest that the CT phase discrepancy may be caused by a real difference between the nonlinear mechanisms in the alert humans and anesthetized cats, and provide new constraints for clarifying this issue through further study.