Search engine optimization, as an industry, is about as consistent as fortune telling and has a lot of the same skepticism built in. The problem with SEO is that there isn't any Periodic Table of the Elements where you can go and lookup what is the truth is. SEO is based mostly on bits and pieces SE reps let drop, personal observation, and a few true experiments. Since there is no standard, no measure, no factual reference, a lot of people draw incorrect conclusions as to how search engines really work. For instance they may claim until they're blue in the face that meta tags will help ranking or that to get listed highly its outgoing links, not incoming links that matter.
I've faced the same issues in my day to day life of writing about search engines and advising people. There are those who swear to high heaven that I'm wrong and there isn't anything I can do to show them otherwise. It was with the goal of providing factual examples that I could refer to that I started out with the endeavor to run controlled experiments on search engine algorithms.
I have created a website (http://www.searchenginelabs.com) in which I can run experiments on search engines in a controlled environment. It is a controlled environment because I control every variable, from incoming links, to anchor text, to keyword density. I have used this site to run a series of experiments on search engines which should serve as final proof on the issues they address. Well, the results (for Google atleast), are in and I'd like to share them with you.
There are two issues that you must address when trying to figure out if something helps with search engines. You must test to see whether or not something helps you get listed, that is ranked at all, and whether or not something helps your ranking once you are listed.
For meta tags it has always been very easy to test the first situation. All you need to do is a do an advanced search on a site that has a word in it's meta tags that does not exist anywhere else, including in the anchor text of incoming links. Limit your search results to the domain of the site you are testing and then if any results come up the search engine uses meta tags to list your site, if they do not come up then the search engine does not use meta tags to list your site.
For most people this test is enough, after all if a search engine does not use meta tags to list a site then why would it use them to rank a site? Alas some people believe that even though search engines like Google do not use meta tags to list a site (and by list I mean insert in SERPs - everyone knows that Google can sometimes use your meta description tag to create your site's abstract) that they still use them to rank a site if the words exist elsewhere on the page.
This test was a little more difficult, but still doable. I created 10 pages with an identical story about a dog, the only difference is I changed the dog's color in each individual page. The end result is I had 10 pages with equal keyword density for the word "dog." I then gave 5 of the pages meta tags.
If a search engine uses meta tags to rank sites then the 5 pages with meta tags will rank higher than the 5 pages without meta tags when you search for "dog" while limiting results to the site's domain name.
The result of this experiment was that Google did not list the pages in any visible order, they were listed rather randomly. Thus we can conclude that Google does not at all use meta tags to rank your site. As other search engines pick up the test site it will be possible to test them as well.