While translating some code from Perl to a toy Lisp, I noticed I was treating something differently in the two languages: variables with one-letter names look much worse in Lisp than in Perl.
$s obviously doesn't collide with anything predefined, but
s looks like it might collide with something (perhaps the
s combinator?), so I felt obliged to give it a longer name in Lisp, even though this didn't really help readability. But I was perfectly comfortable with the one-letter name in the original Perl.
There are some good reasons for worrying less about name collisions in Perl. Variable sigils reduce the risk of collisions by putting functions and scalars in separate namespaces. Perl doesn't have many predefined names, and entirely avoids one-letter names, so there's less need to worry about collisions with them. But I suspect my main reason was just that
$s doesn't look like a one-letter name.