I just encountered a very strange problem with tracking changes on microsoft word, maybe someone here knows what the source is. My editor sent me back an edited manuscript (no tracked changes on the first round). I took that document, turned on track changes, and revised it. I added about 3,000 words of text. I sent it to my editor with the changes still tracked. He then made his edits, accepted all the changes and sent it back to me. When I got it, about 2,000 words were missing. Maybe 70% of the stuff I had added last time around (including an entire section). I mentioned it to him and we compared various versions of the manuscript. It looks like when I sent him my version with tracked changes, not all my stuff was there. So it seems, on his end, for some reason, some of my tracked changes vanished, but not all. We are both using word 07, but I have Vista and he has XP. That is the only difference we've been able to identify.

Huh, that seems like an exotic problem. By "exotic" I mean that it's likely better for you to ignore the reason for the issue and work around it. If you'll use some other method aside from Word comments, or switch programs, you're unlikely to run into the same issue again. As you're not going to be doing this back-and-forth indefinitely anyway, it's likely not worth the hassle to try to fix it, unless a simple Internet search gives you an immediate explanation and solution. With a more persistent bug you might want to work on it, but in this case I don't see why you should.

Probably kind of late advice, but I'd suggest you not use multiple Track Changes "layers" in Word. From your description, you've got a draft in which you've accepted changes more than once. This is sort of Not The Way Things Are Done--Word isn't a content management system; it assume "single-pass" reviews.

Our tracking best practices at my job are as follows:
1) Author creates a new draft and turns on track changes.
2) Author sends out for serial review.
3) Each reviewer/editor gets it for X time, and there's a list for who to send it to next.
4) The last person on the serial list is the author.
5) Author saves a copy of the draft with ALL tracked changes still in it (for the record).
6) In a new copy, author reviews the changes, Accepting or Rejecting as appropriate.
7) Author sends that new copy out for serial review (again, if new content) or for parallel verification (someone wanting more changes replies with a copy with track changes on).

Steps 1 and 5 make sure you have a record of change control--who wrote or suggested changing what. They also allow you to roll-back, if you need to do so (e.g. corrupted review copy).
Step 3 (and 7) allows the reviewers to see what others have said before they reviewed it, to either save them the time making a correction someone's already caught OR to have an argument with that person on what should actually change WITHOUT having the author middleman the argument. [Parallel verification in Step 7--the final "review" stage--assumes that the reviewers and editors don't have any ALL-NEW content to add themselves or author additions to edit.]
Steps 4, 5, and 6 makes sure the author sees all the input AND saves it for posterity AND can manually accept or reject based on his or her knowledge of the content or style.

Yes, you'll have to use some kind of file-naming and -versioning scheme, to keep from overwriting copies. But that can be as simple as putting the date in YYYY-MM-DD format at the end of a file name (if you don't want to use version numbering like "1.2.6" or software terms like "alpha, beta, and gold master").

HTH;
David

Thanks Eero and Artman. In response to this issue, I have worked out a procedure similar to the one artman lays out (I worked in an editorial office some time ago and adopted the one we used to use when using word--but that was before vista and microsoft word 07). For some reason, I never encountered a problem before with sending a tracked changes document back and forth multiple times (it usually color coded our respective changes, so it was easy enough to handle). But I think in future, doing in it stages as you suggest is the way to go. It is a much clearner looking document to work with that way as well. I spoke with a microsoft certified buddy of mine, and he seems to think it is an issue of the settings on our programs somehow causing them not to communicate properly. For now, have new procedure, an increased sense of paranoia and an additional layer of editing. We are also keeping a much closer eye on our word count, to make sure nothing vanishes.