In May 2018, a Twitter meme containing an audio file from an online dictionary asked readers to decide if they heard the name “Laurel” or “Yanny” (the original audio was linked to the word “laurel”).The meme went viral and responses showed nearly an even split between listeners for each name.
In order to rule out any bizarre or particularly special quality of the original Twitter audio signal in the meme, and to allow for flexible control of voice source and vocal tract parameter, a version of “Laurel” was generated with TubeTalker*, a computational model of speech production.
The powerpoint file contains a demonstration of the Laurel vs Yanny effect based on the TubeTalker version of “Laurel”. A series of “Laurel” samples is presented with a progressive downshift of frequency. It may be useful as a demo for courses in Speech Science, Speech Perception, Cognitive Science, etc.
The audiofiles are embedded in the powerpoint presentation, but are available as separate downloads. The files labeled “normal” are based on the raw simulation of “Laurel”, whereas those labeled “preemp” have been pre-emphasized to generate a +6 dB/octave spectral tilt. The bottom two files are the same at the upper two but have been reversed in time. They produce some interesting effects too.
Brad Story
May 22, 2018