It's most likely an attempt to poison Bayesian spam filters.
Bayesian filters work by examining the words in a message and determining how likely they are to be related to spam. The filtering program will keep a database of these statistics based on messages that the user has received and classified as spam or not spam.
These nonsense messages are meant to throw off those statistics, thus reducing the usefulness of Bayesian filters. The idea is that sending spam with the kind of words you'd find in a normal e-mail will cause the filter to associate those words more strongly with spam, causing more false positives and ultimately forcing people away form Bayesian filters.
How effective this is in practice, I don't know. If you're interested the Wikipedia entry on the subject
is a good read.