When I receive galley proofs of a paper, I look at possible errors introduced by the copy-editing team. But while I proofread the article, there are sometimes small mistakes I'd like to fix, which were not introduced by them (i.e. they were already present in the accepted version of the manuscript).

Usually, the proofs are accompanied by instructions saying that extensive changes should not be introduced at that time, and any such changes would have to be approved by the editor (hence, I suppose, delaying publication). However, the limit is not very clear to me. What is considered extensive changes? In particular, what do you think of the following items (from my experience):

• Slight changes in wording, to improve clarity
• Updating a citation, because an “in press” or “ASAP” article now has page numbers
• Adding an important (but not crucial) citation one had missed, in the introduction
• Adding a citation to a paper that has been published since the manuscript was submitted; possibly adding a short sentence to the text

What I have done so far is change everything that I think should be changed to improve the paper (including all the above items), and let the typesetter decide whether he wanted to send it back to the editor. I never received any complaint or comment on my changes, which could indicate that it was the correct course of action.