I voted on "spoken words", but in a few rare cases, "other sound" applies too. Specifically, when I have to count some event repeating fast, I don't have time to think of the numbers as I would normally do, I just think of it as beats coming fast like "TUM tum tum tum Tum tum tum tum TUM tum tum" (in base 2, yeah), and then replay them more slowly, counting the beats. This works up to about twenty.