tedster - 8:13 pm on Sep 29, 2010 (gmt 0)
I don't know of another study just off-hand - it's hard to keep enough of the variables fixed so you can isolate just what you want to focus on.
I can tell you about some data I have from an earlier project in online medical writing. We found that 5 screens worth of scrolling was the sweet spot for pagination. It generated more read-to-the-end visits than any other page size. That was without some of the usability enhancements I tried in this more recent case.
You can't use word count or character count to decide on page breaks, since the width of the display is an essential variable that changes the read-through data.
When it comes to bounce rates - the big challenge is how well qualified is the traffic to begin with. That's one reason I really like read-through rates for paginated articles. If someone starts an article, and then goes to the next page, and then the next - now you've got some data you can sink your teeth into. Taken in aggregate, you've got a much better metric than a single page article where who-only-knows leaves the page whenever.
Each new page in a paginated article is a kind of "mini-conversion". And when it comes to bounce rates, I would never let a client back away from serving their MOST INTERESTED target just because others bounce away. As I'm quite fond of saying, you can deposit a percentage point in the bank. You only deposit revenue, and that comes from your most interested visitors, not some generic, random person (or bot!) who happens onto a page.