Is it possible to determine whether suspicious sign-ups are bots or humans?

[Sono]

I have 2 verification task added to the registration form of my forum:

1. Matching photos from www.peoplesign.com

2. The default Image Captcha of SMF.

In spite of this, I sometimes have suspicious sign-ups, though since I added both tasks, they are not as common as before. I wonder if these sign-ups are bots or manual registrations? I thought the Peoplesign image matching task cannot be handled by bots, but I may be wrong. What is your opinion? I am just interested, because in case some spammers sign up manually to my forum, I think they don't mean too big a risk, as they have to pass verification for the first 10 comment, and I doubt they are as devoted as to complete the tasks that many times. If they are bots, then I worry though. Recently I approve every new registrations manually, but in case these suspicious registrants are not bots, I could switch to automatic approval, to save time.

[Kindred]

Captcha is basically useless at this point.

Questions are actually the most effective means at this time...

Read the FAQ article on spam in our wiki/FAQ.

[Sono]

The subject of my forum is quite rare. In case I set up a security question based on my subject and there are still suspicious registrants appearing, it is correct to assume they were registered manually? Or some bots can mine out possible answers to questions from the internet?

[Kindred]

If you only have one question, it will get quickly documented into the spammer database.

You should have 10-20 questions and ask for the answer to 2 or 3

[Professor Sea]

Oh, is that what I did wrong? I will take note of that and try to improve participation!