Zimbardo discusses the members of the experiment in his book the Lucifer Principle. They were from very different backgrounds than criminals, though they seemed to be very countercultural as well, which still makes the end result surprising. Regarding Zimbardo's take on the whole thing (ethics & impact), I think he does cop to the experiment being unethical, and his behavior being unethical as well. I don't know if it's a case of being localized to Stanford, but I do completely agree that it's a case of guessing the password; in fact, that's pretty much how I read Zimbardo's take on it. He was creating an environment where the password was the increasingly abusive behavior. That's why he sees the experiment as being relevant to Abu Ghraib; guards were living up to an implicit password within the context of their roles on the nightshift. The next and more practical question would be whether or not the existence of such passwords is really universal.

There was an attempted replication that Zimbardo critiques in his book. That replication (which was a reality show, btw; so more confounds are present, though whether that's anymore real than a Stanford LARP is up for debate) had very different results than Zimbardo's. Prisoners there found some solidarity, which Zimbardo predicts would be broken by a more repressive prison structure. It's at that point where I'm fine with the arguments staying observational and theoretical and not moving into experimentation. It's not that I don't think the study of humanity's darker sides aren't important, it's just that I don't think it's acceptable to move into what would seem to be very unethical experimental setups.