The second experiment investigated the possibility that the ΔF0 effects observed in classic "double vowel" studies might be conditional on the phase patterns of the vowels employed, as recent theories of vowel segregation based on temporal beat patterns and pitch period asynchrony might lead us to believe. We found little evidence that intra-vowel or inter-vowel phase patterns determine segregation, whether at unison or at a ΔF0 of 6%.
The third experiment investigated more precisely whether phase effects could have caused an artifact in a previous experiment on harmonicity. We found no evidence of such an artifact.
The fourth experiment reinforced this conclusion by replicating three main conditions of the harmonicity experiment with stimuli designed to minimize eventual phase or beating effects. We found, as previously, a strong dependency of identification on the harmonicity of the ground (interfering) vowel, consistent with the hypothesis of harmonic cancellation. However we no longer observed the paradoxical effect of target harmonicity (opposite to that predicted by the harmonic enhancement hypothesis) that we had found previously. Target harmonicity had no measurable effect.
The fifth experiment replicated several conditions of the previous experiments using a more classic task. Our one-or-two response task is sensitive to cues that signal the multiplicity of sources within a stimulus, whereas the classic two vowel forced response task ignores these cues, and is mainly sensitive to cues that determine mutual masking between vowels. Replication with the same subjects and conditions allowed us to assess the impact of the new task, and to establish a basis of comparison with prior results. As expected, we found smaller effects with the classic task, but overall patterns were similar.
The sixth experiment was a full replication our previous harmonicity experiment, using the classic two-response task. We found as before a strong effect of background harmonicity, but no effect of target harmonicity. The results once more support the cancellation hypothesis but not the enhancement hypothesis.
The seventh experiment investigated two conditions in which the target and background were both harmonic and had the same nominal F0. In one condition the inharmonic patterns were identical, in the other they were different. Identification was poor in the first case, as for harmonic stimuli at unison. It was better in the second case. No conclusion of interest is drawn from this result.