We believe that the illusion reveals potent and ubiquitous contextual processing in audition, perhaps explaining why people are so incredibly good at dealing with the complex auditory scenes we navigate everyday (a conversation in a noisy place, a music concert...).

In a few words, everytime you hear two successive sounds, like two speech sounds in a sentence, or two musical notes in a melody, you need somehow to bind them over time to realise that e.g. pitch went up. In our examples, we designed the final two tones as a carefully-balanced situation: there were as many "up" transitions than "down" transitions between the frequencies that make up the tones. What the context did was to favor some of these transitions, by biasing you towards one possible binding between tones, the binding that minimized the overall frequency distance (after all, real sounds do tend to evolve smoothly over time).

This is only part of the story! For a lot of behavioral data, some brain imaging, and a probabilistic model explaining why this may be important for everyday sounds, please see: Prior context in audition informs binding and shapes simple features', by Claire Chambers, Sahar Akram, Vincent Adam, Claire Pelofi, Maneesh Sahani, Shihab Shamma, and Daniel Pressnitzer, Nature Communications, 2017.

This work has been done at the Laboratoire des Systèmes Perceptifs (CNRS & Ecole normale supérieure, PSL Research University, Paris), in collaboration with U. Maryland (USA) and the Gatsby Unit in London.