This perceptual transformation required that the intervening repetitions be exact; it did not occur when they were transposed slightly, or presented with the syllables in jumbled orderings.
For this reason, theories of the brain substrates of speech and song perception have invoked explanations in terms of the acoustic features involved.
[7] Yet in the speech-to-song illusion a phrase is repeated exactly, with no change in its features; however, it can be heard either as speech or as song.
[10][11] Phrases that are marked by syllables with stable pitches and that favor a metrical interpretation tend to be conducive to the illusion.
[4][3] Margulis and Simchi-Gross have reported related illusions in which different types of sound are transformed into music by repetition.
It has been suggested that in speech the neural circuitry underlying pitch perception is somewhat inhibited, enabling the listener to focus attention on consonants and vowels, which are important to verbal meaning.
Exact repetition of spoken words may cause this circuitry to become disinhibited, so that pitches are heard more saliently, and so as sung.
He presented spoken phrases in stereo and looped them, gradually offsetting the sounds from the two sources so as to create musical effects, and these were enhanced as the discrepancy widened.