I’m still thinking about browsing …
Browsing for research (rather than pleasure) is usually a very purposeful activity, and the browser usually has a pretty good idea of what she is looking for. I think that research browsing falls between “known item searching” and “keyword searching”. When I am browsing for research purposes, I usually start with a known item, and I am looking for items that are similar or related on some dimension. I’m usually looking for other items by the same author, or items that cite or are cited by the original item, or items that use the same methodology, or theoretical perspective. I’m rarely interested in seeing all the other items on the same topic/subject — academic browsing is so much more selective.
Because academic browsing is so selective, browsing the stacks of a large research library strikes me as a pretty inefficient way of finding items of interest. LC call # order is a lousy approximation of the kinds of similarities or relatedness that I am looking for when I browse. It seems unlikely to me that scholars find much useful material by browsing the stacks.
This is why I am optimistic that online browsing, faceted searching, citation linking, tag clouds, and lots of other developing discovery tools will make browsing easier, better, faster … not harder.