Sound and visible pattern, each from the other
In May 1951 Franklin S. Cooper, Alvin M. Liberman and John M. Borst of Haskins Laboratories published, in the Proceedings of the National Academy of Sciences, a paper on the interconversion of audible and visible patterns as a basis for research in the perception of speech. It had been read before the Academy on 10 October 1950. In November 1952 the same group reported what came of it: synthetic syllables allowed a systematic exploration of the acoustic cues by which people perceive consonants.
Why it matters
Speech synthesis stopped being a demonstration and became a method. If sound can be built from any pattern you care to draw, then any guess about which part of the spectrum carries the identity of a sound can be tested by ear: change the pattern, listen to the result. That is how the cues for consonants were found, which spectrograms alone do not show.
The full text of the 1951 paper is a 744 KB scan and could not be read: the PubMed Central PDF path answers with a reCAPTCHA, the Europe PMC mirror returns a 520, and the Max Planck Society copy is marked private. So the title, the authors, the laboratory, the two dates and the pagination come from that source and no figure from the paper's body does. The 1952 acoustics paper is paywalled as well, and only its abstract was read. What the record does not claim: anything about how the apparatus was built - not the number of harmonics, the fundamental frequency, the upper limit of its range or the material of its belt. The research report behind this batch attributed to the 1952 paper the conclusion that the two lowest formants suffice for twelve English vowels; that paper's abstract is about consonants and the word vowel does not occur in it, so the claim did not enter. The name by which the apparatus is known is used by none of the sources that could be read.