How does Google judge subdomains?

That's a real tough nut fogo. It would take a little bit more analysis. 'en' would not (I don't think) be a heavily targeted keyword, so it is doubtful many marketers would be using 'en' as anchor text (in backlinks).
en.wikipedia.org would have high strength based on the page-rank funnel that is created by EVERY one of it's "interior" pages linking to en.wikipedia.org (upper left corner) and all those pages have TONS of pages linking to them from all over the web (as we all know) and since deep-links can transfer some percentage of their "link-juice" to any other page they link to (such as the root en.wikipedia.org) those multiple thousands of deep pages at en.wikipedia.org/whateverpage lend PR to en.wikipedia.org, with or without anchor text that says "en" exclusively.
So, it does NOT tell us that en.wikipedia.org is necessarily "better" than wikipedia.org/en (or the other way around).

Unfortunately, it's really impossible to infer a real "what is better?" answer, b/c "it is what it is" and wikipedia chose to setup in that way (using subdomains).

What it does tell us is that it CAN be something to help rank, but it is really impossible to know conclusively if subdomains are better than sub-folders.

I suppose if one really wanted to cross-test this, they could reg endomain.com and setup subdomain en.endomain.com/en/en.php and backlink the crap out of that with 'en' as an anchor... I don't see this being profitable, but maybe "I outranked en.wikipedia.org for the phrase 'en' " might make a nice plaque on your wall

Used to be G would classify them as a separate external link in GWT.
You could see 'em show up there.

Not any more. Now they are considered *internal* link structure in GWT.

Jury's still out on what affect that's gonna have; but the switch seems to
suggest the mighty G is on to it (and likely may change their algo?)

It still considers subdomains to be a separate site, at least in regard to Panda. For backlinks and EMD love? I'm not sure. It's hard to say. I'll need to run some tests on this.

@ dv8domainsDotCom, yes testing the wikipedia page probably isn't particularly conclusive.

Yeah that would be pretty cool haha. Tough competition though with Wikipedia, a couple of UN websites, and 17 billion others... maybe someday...

I would have thought this must be the case because of sites like blogspot.com and wordpress.com. If subdomains weren't treated separately by Google then you could easily make random.blogspot.com and leech off blogspot's high page rank to easily rank high (without actually having earned that page rank)... which I can't imagine Google allowing.

@MaxReferrals, But if GWT has changed that doesn't necessarily mean that the search algorithm has changed. (And I doubt it has for the above reason.)

Wow - didn't know that. That seems pointless. I know they changed that on GWT. In this case one would presume they also changed the algorithm because that's just foolish.

Thanks for your replies everyone

In the end I think I'm going to go for two separate domains of keyword1.com and keyword2.com rather than subdomaining keyword2. It makes a bit more sense in my particular position. But I do think the question is still an important one.