Welcome to WebmasterWorld Guest from 220.127.116.11
My information could be obsolete. Let me know if you know otherwise.
I have some "unknown" in my stats (not Google's - haven't signed up yet) and I think they're tentatively identified as bots because their behavior matches typical bot behavior, though they don't identify themselves as robotic user agents and they don't hit the robots.txt file.
Consider this example, someone creates a very simple plugin for FF that all it does it inserts a meta-refresh tag for a page the browser downloads. Now this meta tag will trigger a periodic refresh of the page and will be recorded as a human action. The browser could allow jscripts so it will pass through the GA filters but in reality this represents automation or what I could call a bot. Of course there are lots of other implementations but you get the idea.
Motives for it? Lots, say for instance you have a marketing campaign in place like ppc.
Some bots do request image file types, of course. It's a matter of how they are programmed, for example whether the programmer wants it to avoid being detectable as a bot because it doesn't request images. The analytics world had a major incident with one of these last year. Or, the bot might be an image collector, period.