Turus and I had been discussing how to handle the increasing sign-ups related to spambots on the forums for quite some time. I had suggested we try implementing reCAPTCHA which takes unknown text acquired from OCR software and uses the visual CAPTCHA to isolate the true identity of the unknown word. In theory, it solves two problems: identify words that OCR software can't during the process of digitizing old books and, if OCR software can't read the source CAPTCHA text, chances are the spammers can't either. Unfortunately, some of the CAPTCHA challenges are a bit difficult to read, so you might need to click on the little red refresh button (above the speaker) if you attempt to sign up and can't quite figure out what it's asking.
We will also very likely be changing user sign-up authorization back to user control if it looks like this method is going to work well. Chances are, I think it will.
(We also had to remove about 180 sign-ups that were generated by spammers )