Jonathan presents a couple of examples that he thinks clearly contradict this claim. He says of submitexpress.com, for instance,

“If you go to Google and take a look at the backlinks pointing to submitexpress.com (using the link: command), you’ll notice that very, very few of those links have anything to do with search engine optimization or marketing at all.”

Jonathan’s point is that you don’t need links from pages with a similar theme to yours. You just need a lot of links.

This seems to square with my own observations – although I admit I have not done an exhaustive study of the situation. It should be pointed out however, that Jonathan’s study is not very exhaustive either. If, as many SEO specialists point out, a link from a “relevant” high PR site is worth significantly more than a link from a low PR “irrelevant” page, than having such a link (one from a relevant high PR site) will skew the result significantly. It is not really satisfactory to draw hard and fast conclusions until such an analysis is undertaken.

What would be most instructive is if it could be shown that there is little obvious difference in link value between a high PR link from a “relevant” site and one from an “irrelevant” site.

My own conjecture is that you would find there is very little difference (and thus I am agreeing with Jonathan’s thesis). But, apart from what I have more or less casually observed, my reason is quite speculative. I don’t think Google (or anybody else) has the ability to make the necessary evaluations about “relevance” to make such evaluations meaningful, and therefore they (probably) don’t even try to do it.

That isn’t to say they won’t have that ability in the future. And if they did, links from “relevant” sources would have more value in the future as a result.

In any event, links from “relevant” sources have other things going for them – the most important of which is to attract traffic from those sources.