There's definitely some research out there about the ways that people navigate the web, and whether different ways of structuring a site might have different effects on people's ability to navigate and engage with its content. Below are some of the sources I've come across, but there are almost certainly others.
People's existing knowledge interacts with information structure. Calisir & Gurel (2003) suggest that having pages in a traditional hierarchical structure may be more useful for people with low knowledge of the relevant topic area, because it gives them some kind of structure to hang the information on (if you're an expert, you're less reliant on structure, because you've already built your own).
In a similar vein, Chambrot's 2008 Masters thesis found that people remembered parent-page content better than linked-to content. It also found, interestingly, that the content of unrelated linked-to pages was recalled better than information from related linked-to pages. However, the author does point out that the study lacks ecological validity — i.e. that some of the results may relate to the non-real-world nature of the experimental task, which involved many visits back to the parent pages. (I'd also guess that the first few pages you encounter on your web wanderings are remembered better because of the serial position effect)
Information structure affects refindability:
There's a really interesting paper by McDonald & Stevenson (1996?)
about people surfing between hypertext documents and then trying to navigate back via hypertext links to a particular piece of information in a hypertext document (cached version, sorry; the original seems to have disappeared). People seemed to find it easier to find something again in a linearly-structured document (think conventional Word doc or similar) than in a hierarchically-structured document (think traditional website with parent and child pages). In turn, it was easier to find something again in a hierarchical document than it was in a non-linear document (think wiki). This makes intuitive sense; we often have a sense of how long ago/how far through a book or magazine something was).
Is wandering the web intrinsically rewarding?
The instinct to seek, and how clicking links on the Internet supposedly turns us into rats pressing buttons in search of the 'reward' of finding new things, has been widely written about, such as this article from Slate about how we are predisposed to find the web addictive.
Should we be seeking to minimise wandering, or are we just evolving ways of coping with our desire to do so?
There's also been some blog commentary about this in the last year or two, revisiting the idea that instead of putting in distracting links mid-post, it might be better to put all the links at the end, more in the style of a bibliography - concise explanation and example here. However, according to research by Mozilla, it seems that people often use browser tabs as placeholders for linked content while they're reading a page — that is, they open the links as they encounter them, but will often finish reading the page before they go and explore those links.
This feels like an inadequate summary, but I hope it gives you some useful starting-points :)