...
For more information, please read the Mail::SpamAssassin::Conf documentation.
...
In SpamAssassin 2.5 and 2.6, there were only three states for the autolearn result: ham, spam, and no. In SpamAssassin 3.0, the result was enhanced to have six states: ham, spam, no, disabled, failed, and unavailable.
ham and spam are the same. no is now explicitly that the message didn't achieve the proper threshold values.
disabled means that the configuration specifies bayes_auto_learn 0 and so no autolearning is attempted.
failed means that autolearning was attempted, but couldn't complete. This happens if SpamAssassin can't gain a lock on the Bayes database files, etc.
unavailable is now the catch-all for anything not covered above.