A few techniques I'd try...
1. Keep a blacklist with spammy terms. Upon post, if a spammy term is detected, pre-populate the form with the posted data and ask for captcha. * I know you said you're against captcha, but at least here, you're not forcing everyone to input it.
2. I guess this is also a captcha, but you could localise the question instead of using a complex, unreadable captcha service like re-captcha. This is obviously dependent on who the target users are, but you could ask questions like, "What are the last 4 letters of the company name?", "Who won the 1995 Rugby World Cup?", etc.
3. This is somewhat more complex, but based off CSRF protection/techniques. A bot doesn't usually load javascript or read images (unless it thinks it's a captcha image and tries to decipher it using a 3rd party service). Upon load of the contact form, you could do an ajax get request to a basic web service. The web service stores a token in a session and passes it back to the js. The js then populates a hidden form field with this token. When the form is posted, you compare the posted token to the token you have in the session. If it's a mismatch, return the user back to the form.
You may have minimal false positives with this last technique. (session issues - cookies, or if it's a really slow connection, the ajax request may not complete before the user posts). You could therefore counter this by returning the user to the pre-populated form and prompting for a captcha. This time, you ignore the CSRF check upon post and only do a captcha check.
