You know, like the kid who thought the U.S. national anthem began "José, can you see?"
These manglings are called mondegreens, and a scientist explains part of the problem:
In the small study, Ma had 33 volunteers watch videos of people saying words with different levels of background noise, then had participants report what they thought they heard. He found that, depending on the noise level, participants got the words right a mere 10 percent of the time when there were only sound cues. Seeing a person's lips move improved understanding of the spoken words up to 60 percent.When you can't read their lips,
the meaning may slip
away from the script.