natalie - not gonna argue with ya cause I agree that everyone is welcome to try what they like
However - in all of the actual controlled tests where the keywords were not on the page (that would introduce false positives) and in the links to the page (again a false positive) and only the url contained the words, in all cases the dash separated keywords ranked for the phrase typed in as a search of the two words, wheras the underscore did not show up for the phrase.(It did show up for the two words combined into one word)
I would also agree that Google's engine has not been changed to recognize the underscore yet - although I believe Y and MS have. Its just a difference of the way the programmers like things at Google.
The other problem is that in the last year, some variations have been added to the algorithm to separate combined words if(and they are getting better at it) the words are commonly used words recognized by Google. This adds the secondary effect that if you use an underscore, the algo sees the words combined as one, but then if they are common words they get separated again
Of course in all of the testing we've tried, the url has so little to do with the ranking of the page vs. using keywords and other on-page and link factors - they way outweigh anything in the URL - and that can be seen on any search done in Google for a two or more word phrase - assuming that you see pages with keyword URLs I would be willing to bet that they are doing just as well as pages without