FizzBuzz is a test, not the results. If you can point to a study that has administered it to a well selected group representative of professional software engineers that displays a bi-modal distribution, then I'll accept that as evidence. Until then, it's just anecdotal data about the quality of a self-selected group attempting to get a well paid job.
Fizz Buzz is a pass/fail test that wouldn't even exist if the pool being tested weren't bimodal. Why would anyone administer the Fizz Buzz test to a normal distribution of programming skills, for example? (I mean https://www.google.com/search?q=normal+distribution)
It would make no sense - below what part of the normal curve of programmers would you put someone who failed at FizzBuzz (cannot code it)?
In my interpretation it's because programming skill doesn't follow a normal (bell-shaped) distribution, but is bimodal. Which is why Fizz Buzz exists, to tell you quickly which pool (left or right) your candidate is from.
Either that or we have a completely different interpretation of why Fizz Buzz exists or what it means. Why do you think it exists?
