The McGurk effect: the eyes change what is heard
On 23 December 1976 Nature printed a paper by Harry McGurk and John MacDonald of the University of Surrey. When the sound [ba] was dubbed onto film of a woman saying [ga], adults heard [da]; with the reverse dubbing most heard [bagba] or [gaba]. Sound alone or the untreated film gave the syllables correctly.
Why it matters
Speech perception had been treated as purely auditory; the paper showed that people fuse what they hear and what they see into one sound. By MacDonald's account the illusion was noticed by chance, while preparing recordings for a study of infants. On this rests the idea that a machine that sees the lips can recognise speech better than one that only hears, which machine lipreading later tested.
Volume 264, pages 746-748. The full text is paywalled, and the record rests on the publisher's abstract and the co-author's account. John MacDonald wrote in 2018 that the paper went to Nature on 14 July 1976, and that in the first study the illusion ranged from 98% of adults to about 50% of children, depending on the combination of sound and image; the same 98% of adults hearing [da] is given by the 2017 review of Alsius, Paré and Munhall, together with 64% in the same authors' 1978 replication and a range of roughly 20 to 98% across later studies. The original recordings, MacDonald says, were lost. The record does not claim the number or ages of participants: the widely repeated 21 preschool children, 28 schoolchildren and 54 adults were not found in any document read.