Welcome to WebmasterWorld Guest from 18.104.22.168
Google has traditionally considered the underscore as a true character and not a separator. They did this so people can directly search on technical keywords that contain the underscore character, such as _borders or mod_rewrite.
Google has done some fancy dancing in more recent times, in order to use the relevance signals that are present in underscore separated file names. However the hyphen still seems to have a slight edge. It's not enough of a difference to bother changing established urls, but if you're creating a new file, I see hyphens as the way to go.
By the way, don't be misled by the way words are presented in bold on the final SERP. That's a simple character-match routine, run after the SERP is generated for a query. It does not indicate that the bolded word was actually used in the scoring.
With underscores, Google's programmer roots are showing. Lots of computer programming languages have stuff like _MAXINT, which may be different than MAXINT. So if you have a url like word1_word2, Google will only return that page if the user searches for word1_word2 (which almost never happens). If you have a url like word1-word2, that page can be returned for the searches word1, word2, and even "word1 word2".
err... perhaps the content is better? Perhaps the better placed page is more semantically correct and optimised for long tail? Perhaps it is linked to from high pr sources? Perhaps it is because of more anchor phrase returns in links pointing to it?
The name of your file is virtually unimportant. Sure if you want to squeeze 100% out of your page, then consider looking at it.
But would you not be better off making the actual content better and getting links to this page (from both your own site and external)? Wouldnt it be better also to experiment with your title element and meta description?
I think your time is better served on the above rather than spend hours examining underscores etc.
Experience tells me they all work to some extent - so concentrate on your content.
I think that SE's crawl both now and the whole argument is some what historical. In todays SE's it simply doesn't matter which one you use.
By the way if you read what Matt wrote he does point out this is a historical argument. So no way do I think this is an outdated post.
[edited by: Pirates at 12:47 am (utc) on Jan. 10, 2007]
Do a search for the following ( enter them all at once ):
allinurl:widget allinurl:tv allinurl:series
Then do one for:
Use underscores if you'd like Google NOT to identify the strings as separate words in the filename. Which sometimes makes sense. But not for what most of you have been asking this for.