One recent feature added to Google Webmaster tools was the concept of keyword variance. This new feature shows how weak Google's Latent Semantic Indexing probably is. The most variance I've seen for a keyword is past versus present tense. I've never seen Google give credit for a synonym.
In the field of science lots of things are related. How is Google supposed to give credit for something being related or not when so many things have the potential for relevancy? So many words have subtle different meanings and some words take on new meanings like lose weight becoming synonymous with lose fat but weight having a very different meaning than fat. The answer is that Google only uses the exact words when determining relevancy.
What does that have to do with how to boost relevancy? Earlier I wrote an article about The Keyword Academy. I was a member for the one month trial period. When I was a member and submitted articles I noticed that the website developers used the following tags: One tag for the title and two with the keywords of the links. Now the article hosts have incentive to be helpful because they want you to keep submitting articles to them rather than another member on The Keyword Academy spectrum. They want your links to be as strong as possible.
I realized why they do this. I think Google's application of Latent Semantic Indexing is so narrow that all that matters is the exact text when determining relevancy. You can apply this principle to other article sites as well. Ezine Articles lets you choose labels so choose labels with the same anchor as your links. It should be easy to perform a test to see if going back and doing this dramatically improves ranking.