It seems to me that the main reason most hypertext sources seem to produce shallower reading is not the fact that it contains hypertext itself, but that the barriers of publication are so low that the quality of most written work online is usually much lower than printed material. For example, this post is something that I might have spent 3 minutes thinking about before posting, whereas a printed publication would have much more time to mature and also many more filters such as publishers to take out the noise.
It is more likely that book reading seems more deep because the quality is better.
Also, it wouldn't be difficult to test this hypothesis with print and online newspaper since they both contain the same material.
It seems to me like "books are slower to produce than online material, so they're higher quality" would belong to the class of statements that are true on average but close to meaningless in practice. There's enormous variance in the quality of both digital and printed texts, and whether you absorb more good or bad material depends more on which digital/print sources you seek out than on whether you prefer digital or print sources overall.
This thread is for the discussion of Less Wrong topics that have not appeared in recent posts. If a discussion gets unwieldy, celebrate by turning it into a top-level post.