All about that bass: exploring auditory perception using a meme
By: Israel Nir, Caroline Lo
This piece is based on public posts made in five English speaking countries (US, CA, AU, NZ, GB) from May 15th to 18th, 2018. All data has been aggregated and de-identified.
Back in 2015, a single dress “broke the internet,” showing that we humans might not all experience color the same way. Our fellow researchers at Facebook even wrote a blog post about it. We had to wait three years for the auditory equivalent of the dress to appear, but in some respects it was no less glorious. One sound clip divided the world—some people heard Yanny, others Laurel. By now, manyexperts have tried shedding some light on what might make us perceive that clip differently. When we started analyzing the public data from the Facebook side, we thought we might validate some of those theories. We actually did, but more interestingly also found evidence to a phenomenon that we weren’t aware of before.
We de-identified and aggregated public comment threads from people in English speaking countries from May 15th to the 18th that mentioned both “Laurel” and “Yanny.” The assumption is that if a thread includes both words, it’s probably discussing that audio file. We then looked at the percent of comments mentioning either “Laurel” or “Yanny,” but not both. Overall, most comments mention “Laurel” (which turns out to be the original word in that clip). But the story gets even more interesting when we segment by age and gender.
Gender Only
Breaking responses down by gender, we found that women were more likely to hear “Yanny” compared with men. To be specific, the odds that a given woman hears “Yanny” are 11% higher than a given man. This jives with research that suggests that men have less acute hearing than women, especially at high frequencies, aka the “Yanny” frequency. Essentially, hearing “Yanny” or “Laurel” depends on how well one perceives high pitch tones vs. low pitch tones [1]. And being more sensitive to high pitch tones makes one more likely to hear “Yanny” [2].
Gender + Age
If hearing loss is correlated with age, are people less likely to be able to hear “Yanny” as they age? Indeed, we found that across both genders, older folks became less and less likely to hear “Yanny” all the way up to their 40s.
But something mysterious happens past that age. While “Yanny” rates remain relatively constant for older men, older women actually became more likely to hear Yanny, not less. What could be happening? Does women’s hearing magically improve after a certain point in life?
Amazing as that would be, as far as we know, auditory regeneration isn’t a thing (yet). We were baffled by our findings until we came across research that shows that past a certain age, women experience greater hearing loss in lower frequencies than men [3]. Thus, that awesome baritone recording of “Laurel” actually gets harder to hear. So, while women past the age of 50 still hear low-pitch voices better than high-pitch voices, at that age for some women, the pendulum swings back towards favoring high pitch sounds, or in our case, “Yanny.”
What other words people are hearing
We’ve been focusing on the Yanny vs Laurel debate this whole time, but is that all that people heard? Taking comment threads that mentioned both “Yanny” and “Laurel,” we identified comments that matched the approximate phrasing of “I hear <word>,” and took a look at which words showed up. You can see a word cloud of the creative variety of alternatives, based on how often users mentioned that word, below. And if you’ve been wondering why you haven’t been hearing either Yanny or Laurel, don’t worry, you’re in good company: by our count, at least 10% of commenters hear something else.
Conclusion
When we started looking into the “Yanny” vs. “Laurel” debate, we wanted to see how the theories so many people talked about in the last few weeks were reflected, and hopefully validated by the actual data. We think we achieved that, actually observed a surprising phenomenon that we haven’t heard about before (some pun intended), and learned something new.
Many thanks to Shrivats Iyer for helping us better understand the data we were seeing.
