For a small site, doing anything out of the ordinary is enough to stop the vast majority of spam.

For years, I had a non-captcha on my blog that simply read:

    Enter the word "elbow": ________

That was it. The word was hardcoded, and the server simply checked to see if that input field did, indeed, contain the word "elbow". Spam completely halted for a long time. Eventually, it started showing up again, and once I got tired of deleting a couple of spams per week, I changed it to:

    Enter the word "humour", but with American spelling: ________

Once again, spam stopped for a long time. As a bonus, I had a couple of commenters complaining that not everybody was a native English speaker and this question was too hard. I say "bonus", because I like keeping away commenters who can't be bothered to spend five seconds on Google.

This started to break down eventually as well. My latest effort, done more for fun than effectiveness, is to implement a hashcash proof-of-work system in Javascript. In short, the server sends a random salt to the client, and the client must then come up with a string which, when concatenated with the salt, has an SHA-1 hash with a certain number of leading zero bits. This is easy to verify (the server just takes the client's result, performs a hash, counts the leading zero bits) but hard to produce (the client has to brute-force it).

The relative slowness of JavaScript crypto makes this not actually very useful in a theoretical sense. I've calibrated the difficulty to take about 20 seconds of work in the browser, but a decent native code implementation can do the same amount of work in maybe a tenth of a second. So, against a determined attacker, it won't be much of an obstacle.

But as a relatively small site with a completely custom anti-spam solution, I don't get determined attackers. I get the occasional spam message with this in place, but from the server logs it's obvious that it's actual real human beings firing up the comment form and waiting the ~20 seconds for the hashcash to complete before posting their comment. Terribly inefficient for them, and it keeps stuff down to a manageable level.

In summary, if you have a small site, you can throw up nearly any custom anti-spam measure, no matter how silly or easy, and have it be effective. A standard captcha may be easier to add than a non-standard anything, but it's definitely not an "absolute requirement". Big sites are completely different, since they'll attract enough spammer attention for site-specific attacks.