The sound spectrograph
In July 1946 W. Koenig, H. K. Dunn and L. Y. Lacy of Bell Labs published the design of the sound spectrograph, a wave analyzer that produces a permanent visual record of how a sound's energy is distributed in frequency and in time at once. The paper describes how the device works, gives the mechanical arrangement and the circuits of one particular model, and shows spectrograms of voice, animal and bird sounds, music and frequency modulations.
Why it matters
Until then speech could only be listened to and described in words. After it, the speech spectrum was a two-dimensional sheet on which transitions between sounds and resonances could be seen and measured with a ruler. Everything that came next in speech work, recognition and synthesis alike, rested on there now being something to look at.
The full text is behind a paywall, so the record claims only what the abstract on the publisher's page states. The paper also discusses the problems of this kind of analysis, particularly the need to handle and portray a wide range of component levels in a complex wave such as speech. What the record does not claim: the length of the sample the device took, the material and size of its recording medium, the widths of its analysing filter, the sweep time, the dynamic range or the high-frequency pre-emphasis. All of those figures were in the research report behind this batch, none of them could be read in any accessible source, and so none entered.