A single piece of audio is used to build a ring of virtual sources around the listener. Each source is made individually too diffuse to localize, and the ring is then rotated. The question is whether a field whose contents cannot be located is still heard to move.
The variants below are rendered from the same twenty seconds of the same signal. Each one is aligned to the sample and matched in loudness against the untreated reference, so that moving between them varies the treatment while the material is held constant. The monitor shows source positions as recorded by the renderer during the render itself.
Variants (click, or press 1 to 8)
Monitor
The field is shown from above with the listener at the centre facing front. Filled dots indicate coherent sources and hollow dots indicate decorrelated ones. Dot size and fill track coherence, and opacity tracks level.
Scope
The head model is a rigid sphere, so left and right are carried by both interaural cues while front and back are rendered identically. Front-back information is available only in the variant using the measured HRTF. Distance is carried by level alone, without propagation delay, Doppler shift, air absorption or reverberation.
The IACC figure in the readout is interaural cross-correlation, where 1.0 corresponds to a single coherent image and values near zero to full decorrelation. It describes how diffuse a render is rather than whether the field moves. Blind listening on this instrument found the sensation of movement surviving a sum to mono, which removes every interaural difference and indicates that the effect is substantially spectral. The results so far come from a single listener who is aware of the hypothesis. Further detail is in the repository README.