After 3 months the CAPTCHA was switched 
  to the other CAPTCHA setting.

Wouldn't it be a better study if instead of doing the switch once after 3 months, captcha went on and off randomly per visitor throughout the 6 months? This would take away unintended effects on the study due to changes in traffic pattern over time.

Interesting read up, but graphs are seriously misleading. The first one being from 0-900 scale, and the next one a zoomed in 600-820. This really throws your off... took me a while to realize that the failures are actually a VERY SMALL fraction... (even after reading the percentages explained in text).

Yeah, the guy was obviously looking to do nothing more than confirm (and push) his bias.

The quality of his captcha is also questionable, since it was apparently not recaptcha.

Exactly what I thought! Talk about statistical manipulation...

It would be interesting to see the effect of captcha on signups (and not just comments). Also it would be good to know if re-captcha works out to be better (since it has audio support).

> It would be interesting to see the effect of captcha on signups (and not just comments).

Yes, I think that author fails to take into account that on certain sites a lot of spam would cause potential users to shun from even trying to signup.

I saw a comment form recently where there was a field labeled 'Are you human?', if you enter 'yes' your comment is not marked as spam. I wonder how well silly but obvious questions, to humans at least, can stop spam? From a UX perspective it's slightly less effort than deciphering a CAPTCHA box. It's not as thorough a solution as CAPTCHA but is super quick to implement.

Finally someone figured it out.

I guess it's even worse depending on he kind of captcha you use. Some captchas are so awful that they probably make more people abandon the form.