This is blindingly obvious when someone points it out to you, but not so obvious when you are actually working on multi-threaded code that has shared data. Whenever you have performance issues, the issue more likely than not is shared data protected by locks that causes all threads to serialize once in a while and minimizing this leads to performance improvement.
I've never thought about the fact that 'serialize' as a verb is opposite and antitetical to 'parallelize'.