Good question. By hand, and I don't know how my results match up against more "official" methods, but here's how I did it (a few months ago, so I remember it well):

1) I time myself reading 100 pages, and randomly sample from those 100 pages until I have a tight confidence interval for the word count of those 100 pages. This gives me my wpm.

2) The following day, I generate 100 random numbers between 1 and the number of lines per page in that book. For each page, I check 3 lines starting at that number (may bleed onto the next page). I count it correct if I remember the content of those three lines, and incorrect if I don't. It's subjective, but I'm strict. If the lines are

how childrens minds work. One of the Nickelodeon / producers, Todd Kessler, had actually worked on Sesame / Street and left the show dissatisfied. He didn't like the

I count that as the facts: 1) Todd Kessler was a Nickelodeon producer 2) He worked on Sesame Street 3) but left dissatisfied

If I get any wrong (including failing to remember Todd Kessler's name), that counts as a strike against the whole phrase, not just 1/3 of it.

Thus, my 90% comprehension rate means out of 100 samples, I got 7-13 wrong (most of those were in fact forgetting a person's name, location, or organization).

3) Repeat for at least two more books to see how different writing styles affect my reading rate.