How to: Write a response to referees

I think I'm going to start a periodic series of "how to" posts.  First up, how to write a decent "response to referees" document.  While this is pretty much common sense, it's not bad to think about it a bit in the abstract, rather than in the heat of the moment of having just received some kind of (perceived) searing blast of criticism.  In brief, assuming you get some collection of referee reports, at least one or two of which are not particularly positive, and you intend to revise and resubmit:

• Read the reports, and then put them aside for a day, as your white-hot rage over the terrible injustice that has befallen you fades, and in the cold light of reflection you realize that perhaps the manuscript you'd sent in is not, in fact, the greatest non-fiction prose writing since Churchill's six volume history of the Second World War.
• Now that you're in a less annoyed frame of mind, read through the reviews again, carefully, trying to understand (a) what the reviewers are actually saying, and (b) what the reviewers want you to do (assuming that's not "dry up and blow away").  Often the answers to (a) will reveal either that the reviewers did not properly understand the main point or some subsidiary point of the paper.  Much as we like to grumble about referees, you may have to admit that the fault could lie in your presentation.  Were your figures unclear?  Did the abstract and the intro make your main point explicit, or did you bury the lede somewhere down in the conclusions?  Remember, scientific papers are not mystery stories.  Springing the cool observation on the reader after a lot of setup risks the reader not realizing that the observation is cool.  Moreover, often the answers to (a) will reveal that the reviewer has thought of a possible concern or objection that you either didn't consider, or you did consider but dismissed without pointing it out and explaining your reasoning.  An extremely important part of the response process is figuring out what the main point of the referee is, and realizing that frequently it's worthy of consideration.
• Regarding (b) above, write down and make a list of what you think the referees want you to do, or what you think it would take to address the points that they raise.    Then consider whether you want to or should do all of those things.  Sometimes the referees can be very demanding.  (We've all seen this.)  You have to use your judgment, and remember that referees are not generally gratuitously mean.  I'd say the default position should be to do what they want, unless what they want is really considered unreasonable by you and your coauthors.  This list, by the way, is a headstart on the eventual "list of changes" that you'll need to provide when you resubmit.
• When you sit down to write your response, have the referee remarks right there.  In fact, it's a good idea to use copy/paste to intersperse your point-by-point responses.  That way you can be sure you didn't miss anything, and you are forced to write your response in an order that will seem logical to the referee.  
• Always (always) thank the referees for your time.  Seriously.  You know what refereeing is like, and you'd like to be thanked, admit it.  
• Point out that after this process you believe the paper is much improved (it will be, too, assuming the referees were really on point and not just asking you to cite their seminal work on the topic at hand), and if possible explain why.  (e.g., we believe that our main point is now much clearer)
• Always be polite and professional.  If you fly off the handle in your response, even if the referee is overtly hostile, it won't do you any favors with other referees or the editor.  Similarly, just as tone is difficult to convey in email, I suggest avoiding attempted jokes or sarcasm.  This is a professional communication - keep it that way.
• Try to be timely about revisions.  It's much better to get revisions done while everything is fresh in your mind, rather than letting things linger.  (Don't write them in the heat of the moment, though.)

In your accompanying cover letter when you resubmit, make sure that you emphasize the changes you made in response to the referees.  Also, it doesn't hurt to point out to the editor if you think a referee either missed the mark or seems not to be objective, but it would be best to do so in a very professional way.  Calling the referee an idiot won't win you any friends, particularly since the editor likely chose the referee.  Still, if you really think the referee made serious mistakes, or was not competent, or didn't read the paper, you should bring that to the editor's attention in a professional way.  Again, this kind of response is an important part of your repertoire of professional communications, so it's best to get in the habit of writing them well.

That's it for now.  I'm sure I've left out points - please feel free to bring them up in the comments.