This is a bug in the way perldoc interacts with groff (it first appeared in Red Hat Linux 8.0, becuase this is the first version with UTF-8 support), it is fixed in the latest rawhide version of Red Hat Linux (which means the fix will be in the next release), but in the meantime the easiest way to work around it is to set the environment variable LESS to 'isr'.

There are several ways you can do this:

For bash, add to your .bashrc: export LESS='isr'

For csh or tcsh, add your .cshrc or .tcshrc: setenv LESS isr

Or you can just specify it on the command line when you run perldoc: LESS=isr perldoc perlop

Update: I should mention that only the -r option is really necessary to fix this problem, I just used -isr because that's what I have in my own settings, -i tells less to be case-insensitive when searching, and -s says to compress runs of multiple blank lines to a single blank line. The -r option causes the formatting escape codes to be passed through directly to the terminal.

Update: Oops, has nothing to do with UTF-8, it's caused by new options in groff... The bugzilla.redhat.com entry has some more information.