A human attacker can break a math captcha but he can break an image captcha too. But if there is no semantic interconnection (HTML-wise) between the question and the form field for the answer, I would doubt, that a machine/bot can break a math captcha, especially if the mathematical formula is "hidden" in a sentence ("What is 1+1?" vs. "1+1").

Tschö, Auge

My main concern is the automatic bots. Humans can always brake the security, unless their is some kind of IP origin verification... if you are from some country that is always causing problems or at least being used for criminal activities... you can possible block them, but is currently not easy because you can block IP's, but you can not just drop down an list of country's and block them, or just allowed some country... it could one more option, if it's possible to integrate that too. They can use proxys, is true, but proxies have to much problems, and are much slower in most cases... and it's possible to add protection against (known) proxies.

About mathematical formula, I'm not a programmer my self, but I can imagine if they can brake most of images, they can much easier brake some question asking "What is 1+1?" or "What is 1 + 1?" or "Say to me, 1 more 1 is how much?", they have just to detect where is the numbers an what is the type of calculations.