Microsoft Research wrote a paper about the intentional use of bad grammar/spelling in 419 scams. Keeping the ostensible source as the email Nigerian is another tactic along these lines, since anyone even slightly savvy will see that as an immediate red flag after all these years.
I skimmed the actual paper and wasn't very impressed. It uses theoretical models without any actual data to support its claim. The claim the paper actually proves is:
If there is a technique that successfully weeds out false positives, then such a technique can be used profitably.
Which is a far cry from their supposed claims, that the spammers are using bad grammar and "Nigeria" to weed out false positives. They basically proved a molehill and claimed a mountain.
http://blogs.msdn.com/b/tzink/archive/2012/06/24/bad-spellin...