I’d be interested to know if you’ve continued to track data and what the current numbers tell you about your hypothesis?

What’s the relationship between the structure of the page and the way the browser renders it, and the reading time? How, if at all, do you account for the delay in rendering a page if the network speed is low and none of the content appears until most of the page HTML has been delivered? This seems like something which will vary between browsers, sites, and even pages, so the effect may average out, but we don’t know that.

There is a reasonably well documented knowledge and research, dating back from long before the web, of the relationship between user response and system response in an interactive graphics (I think electronic circuit design) application. As the speed with which the system responded to a user action decreased (i.e., longer response time), the time for the user to make his next request also got longer. Intuitively, you can think of this as the effect of the user’s attention wandering while he waited for the system to give him the go-ahead for the next input, and I think that’s a good part of what’s going on. There may also be an effect of the user loosing his rhythm when the system can’t keep up. I think it’s likely that at least the first of these two effects is what you’re observing: the user starts to think about other things while waiting for the page to appear. That’s why it’s so disproportionate. I would expect the effect to be much more severe here because the impatient user can switch to another brower tab or window, or even a different application How does your data take that into account?