From a theoretical standpoint, a PhD is obviously not a magical "get good at science" button: there are plenty of other ways to gain competence. An example is Rick Stevens at Argonne National Labs, [1], who I'm pretty sure doesn't have a PhD (my friend worked for him once, and his bio doesn't mention it, but I don't know him personally). People like that show that a PhD is obviously not necessary to do first-class work.
On the other hand, if you want to be a principal investigator on a grant of any decent size, almost no one will believe you have that competence without a PhD. Part of this is the ridiculously rigid US scientific job market: all the arbitrary structure of a resume, even less of the rationale. You can't get a postdoc, almost trivially, without a doctorate. Even if someone wanted to give you a job as a postdoc, the university wouldn't like it, and why choose the guy without the PhD when you have other talented people who also have those three letters? And it's really hard to get a faculty position without a postdoc, so there you go.
So if I could go about becoming a principal investigator or having a nice faculty position without getting a PhD, I'd seriously consider it. But practically speaking, that's not a real possibility yet. Hopefully someday it will be.