One of the things that seems to work well is forcing human interaction by using a variable graphic with randomly generated text that has to be manually entered. This technique helps prevent bot scripts and other automated means of hijacking things as the graphics are very hard, if not impossible (currently) to interpret without the aid of Computer #1 (the human brain).
On my website the forms actually submit to a repository that I then scan with my own program. It filters off all but the legit looking posts, then passes them to me for approval where I can then either manually, or automatically have it posted up on the site under the actual approved file name with a digital checksum for verification (that way if the form is hacked, I get notified and the hacked form is dumped and replaced automatically).
Of course, my website gets relatively low volume due to it's rather niche appeal (Cyber Forensics). Even the email addresses on mine use an encryption to prevent them from being scooped up.
If you are interested in how I managed this, drop me a private email. I'd rather not post the specifics here for obvious reasons.
