Update: Some more thoughts about that flawed research paper, plus a few added links.
Ruth just pointed me at benutzerfreun.de, a German usability site. I've only just started reading the site, but they have some great links (and yes, the links point to English content, too), even though their main goal seems to be making flow charts more popular among site designers.
Anyway, they also pointed me at Jakob Nielsen's site. This guy works with Bruce Tognazzini and also has an interesting article on web ads and usability. This one only shows how not to do it, but at the bottom is a link to a small article with at least some tips on how to do it right.
Nielsen also has other interesting shorts on How important it is to provide related links in context. The most important quote from this article, an executive summary, if you so wish, would be: (...) users typically return to a previous location to follow-up on related actions. Familiar places are easy to find and create less mental resistance than having to actively construct new, independent actions. He also has good examples in there.
Other stuff by him is about Global site navigation and its pitfalls and Site Map usability. Darn, I sense another web site rework coming...
Another article I found fits better in the "well, duh!" category. Basically, they're doing tests on arrangement and grouping of things with whitespace and boxes, and how that relates to users' search times. Surprisingly, the methods that group things more visibly come off worse than those without any grouping, which isn't that surprising once you realize that the text they're using is in a completely arbitrary order.
So, the worse search times probably occur when the users try to find a system in how the data is grouped, whereas the ungrouped text is instantly recognized as the unsorted mess it really is.
Even though that stuff is from Uni researchers and I'm not a psychologist, I'm inclined to go with my previous knowledge and past experiences, which contradict their results and warn me this test may not be kosher, since the article doesn't address this possibility at all. Not to mention that the example images they provide mostly are still what I'd consider "unreadable", so maybe they just picked images that really made no difference because they're still too similar to show the benefits of the different arrangements.
After some more talk with Ruth, I've come to the conclusion that the "bug" in this research paper is mainly that they tried to judge the usability of boxes, grouping and other structuring tools without an actual structure. Just arbitrarily segmenting text or lists isn't really helpful, that's all they proved. In the real world, you want to group items that are somehow related, so people can look at one or two items and immediately being able to deduce what the rest of the box contains. Grouping is intended to be a time-saver.
In their samples, people were first misled by the seeming groups, only to have to actually scan the complete data set a second time to find the needed information by brute force. And this fits quite nicely into Nielsen's findings on How users read on the web (they don't). That's it. I rest my case.