[meanguy]

I did some work here.

There's a lot of problems with legibility research. It's very, very hard to test for "legibility" or "readability" -- as, not surprisingly, the typeface is only a small part of what's going on when the human brain turns symbols into thoughts then plays them back later.

To the extent that we could get valid data, people did best when they were familiar with the typeface in question. And back in the age of newspapers, the closer to the local newspaper layout and typography, the better they scored with news-like information.

melloclello's comment hints at this part: serifs are high-frequency data. So there's more "information" in a serif face than a sans-serif face. In fact, simple filtering of a serif face makes something that looks very much like a sans-serif face. But whether this extra information is helpful or distracting to readability or legibility depends on what the reader is used to. (But even that didn't matter much in our tests!)