For several years I flew search and rescue missions with the Civil Air Patrol (CAP) in Colorado. One of our jobs was searching for lost hikers and missing airplanes. As part of our training, we learned that when you search an area looking for a lost hiker or a downed airplane, you may not see your target, even if you fly directly over it. Therefore, we would never search an area just once. Instead, we would send several flights over each area with different crews and in different conditions, and after each flight we would calculate a probability of detection (POD), a measure of how well we had covered each search area. Often, it took several flights over the same area before we could say with a high probability (but never 100%) that the target was not in that area.
I just spent the last week doing final production and proofreading on a new book (The Language of Content Strategy, by Scott Abel and Rahel Bailie), and the effort reminded me of how we did searches. The copy-edit and proofreading passes were done in three different media (PDF, print, and the development wiki we used to author the book) and with five different participants (three content editors, one copy editor, and one indexer).
I was one of those participants, and when I saw the printed proof for the first time, I was pretty sure that the book was 99% ready. I took what I thought would be a quick production pass through the book and discovered, to my surprise, that I ended up making changes to 90% of the pages. In fairness, some of the changes were nitpicks that are normally only picked up at the very end (for example, two consecutive lines that end with hyphenated words), but some were clear errors that all five of us had missed.
Looking back at what happened, I realized that the same principle that applies to search and rescue also applies to proofreading. You need to have different people looking at the book from different perspectives (e.g., indexer, copy-editor, proofreader) and in different media in order to have the best chance of catching as many errors as possible. I also confirmed what I had suspected for a long time; the most productive medium for proofreading a printed book is a printed copy. I'd even argue that print is the most effective medium for an ebook. There is something about print that, for me at least, makes errors easier to spot.
Next in line in effectiveness, possibly surprisingly, is reading the book out loud. We received confirmation of this just this morning. We are doing an audiobook of The Language of Content Strategy, and the manager of that activity just sent us email pointing out three previously undetected problems (fortunately minor).
With search and rescue, we don't stop until we find our target or reach a point where the odds of survival are effectively zero. With proofreading, the choice isn't life or death, but you also don't have a single target; in fact, you don't know how many targets might be out there. For any non-trivial book, there will be errors that pass through whatever proofreading defense you mount. And in the end you need to decide when you've hit the point of diminishing returns. But I think that if you vary the people, perspectives, and media and take at least one pass through using a print copy of your content, you should be able to catch the worst problems.
Richard L. Hamilton is the founder of XML Press, which is dedicated to producing high quality, practical publications for technical communicators, managers, content strategists, and marketers and the engineers who support their work. Richard is the author of Managing Writers: A Real-World Guide to Managing Technical Documentation, and editor of the 2nd edition of Norm Walsh's DocBook: The Definitive Guide, published in collaboration with O'Reilly Media.