The True Story of How Laurel Became Yanni

At the beginning of the video with a picture of laurel or yanni, she talked about laurel and nothing but laurel. Here is the original version of the video , recorded by an opera singer working for Vocabulary.com. You will most likely hear it as laurel as well. Let’s compare it to a viral clip and we’ll see what exactly has changed and why half of us hear a viral clip like yannie.

The original recording was made by Vocabulary.com in 2007 to include the pronunciation of the site’s most commonly used words. It was a transcript of the word laurel . It was spoken by Jay Aubrey Jones , one of eight singers commissioned by the company to read the words from home using the provided laptop, microphone and portable sound booth. (Opera singers are trained to read the IPA , a pronunciation code that dictionaries also use.) Jones personally read about 36,000 words for the site.

According to Vocabulary.com co-founder and CTO Mark Tinkler, there were so many audio clips to be processed that the program did everything automatically: cut the beginning and end of each file, applied a noise filter if necessary, and converted to mp3 format, which saves space and bandwidth.

The beauty of the mp3 format is that it can compress audio files into small enough space so that they can be easily distributed over the Internet. The downside to the mp3 format is that it does this by removing sounds from the recording and creating glitchy, harsh noises. If you are comfortable with a file that is only slightly compressed, you can keep most of the original audio information. But, according to Tinkler, “it was 2007 or 2008, so we used downsampling quite a lot.”

I asked Tinkler if he had a chance to compare with the original file. He found it on a DVD in a closet and uploaded it to Soundcloud for everyone to hear. Technically, this is not an exact original because Soundcloud provides it as mp3, but it was originally recorded in an even better format. But this, so to speak, is more original than the version that you and your friends were arguing about. And it’s clear enough that we can now tell exactly what happened when the file was compressed.

What’s lost

Compression of the laurel clamp resulted in serious injury: the second speech formant . Without this part of the audio spectrum, L might sound like “ee,” and R might sound a little like N.

That’s why. Each sound wave has its own frequency. The higher the frequency, the higher the pitch we hear. But the sound of speech includes many frequencies at the same time. Our ears and brain pick the strongest frequencies from this jumble of sound. Speech scientists call these formants . The two lowest pitch formants give us enough information to distinguish one vowel from another.

If you open a sound file in an analysis program, you will see these formants. This is exactly what my father, a speech therapist at the University of California, Pennsylvania , did when I showed him the viral clip. He opened mp3 in a program called Wavesurfer and started showing me on his laptop at the kitchen table how wrong I was when I heard laurel, because the formants match the sounds “uh”, “a”, something ambiguous, which could be ” n “and the final” her “. Yanni.

But, he admitted, there is enough ambiguity to hear the laurel if you really wanted to. (At first he suspected that the file had been carefully crafted to be ambiguous, a deliberate illusion of sound .) For the record, I’ve never heard anything other than laurel from this file.

My dad can only read formants from a black and white spectrogram, but Wavesurfer helpfully detects them and codes them in color. The wave surfer agrees with my father and reads this file as “yannie”. But if you give it the original file, it will highlight a different set of formants: the ones that very clearly say laurel.

To understand what happened, take a look at the red and green lines. These are the first two formants, which we will call F1 (red) and F2 (green). They both hover at the bottom of the spectrogram. The third formant, F3, hovers high above them in blue, dropping very close to F2 during the R sound in laurel.

But then look at the rendered clip you shared with your friends. (Click the arrow in the slideshow above.) It is generally noisier thanks to the mp3 compression. This noise erases the difference between F1 and F2, leaving both the wave surfer and our brain thinking that there is probably only one formant, F1. This means that the top line, the one that goes down in the middle, looks like it should be F2.

At the beginning of the clip, F1 and F2, next to each other, are compatible with the L sound. But if F1 is on the bottom and F2 is much higher than 2000 hertz, it is an “her” or “Y” sound. Removing the second formant changes how we interpret sound.

But there is enough room to disagree

The viral processed clip does not completely remove the second formant, it is simply harder to distinguish among the noise. If you’re expecting to hear laurel, and if your ears and brain can find two formants at low frequencies, you can still hear it as laurel. But for another person, the yanni sounds might have stood out more, as Gizmodo reported earlier this week .

So what about those videos and slider tools that show you can change what you hear by listening to only high or low frequencies? It turns out that this phenomenon is also reduced to speech formants.

If you turn off the higher frequencies, you are only left with the lower end of the spectrum, where the original F1 and F2 were. Your brain will work a little harder to distinguish between the two formants instead of clinging to what it thinks is F2 above. In contrast, if you only listen to the higher frequencies of the clip, your brain picks up the real F3 as a possible F2 and suggests that there should be a single F1 at the bottom of the spectrum.

Thus, this trick changes your perception not because yanni is “in” one part of the spectrum, but laurel “in” another, but because in any case your brain hears only some of the formants and guesses what others might be.

More…

Leave a Reply