i've recently taken a look at a year's worth of search log data from the Guggenheim Collection on-line -- a pilot study for some work within the steve.museum project. I've attached a draft paper to this post -- comments are welcome! It's still rough in spots, but I need to step back.

One of our premises in discussing folksonomy in the museum is that allowing users to tag collections will improve their retrivability... but surprisingly, we know almost nothing about what searchers of museum collections really do. i couldn't find a single serious IR study in the museum domain. There's lots of literature about what we 'should' do, how standards will help and why controlled vocabularly is really important, with almost no evidence to support those claims. We need to look hard at the data.

Notable findings in the Guggenheim data: