Publishers have one final recourse if the author refuses to make a change we think is critical: we can refuse to publish the book. Boilerplate contract language requires the author to deliver a manuscript that is "satisfactory" to the house. So if the author digs in her heels, and refuses to revise something that the editor feels will make the book unmarketable--or perhaps, may make it libelous or otherwise legally questionable--or again, let's say, untruthful--the author and publisher may simply part ways.
the book belongs to the author, but if the author wants a *published* book, he/she better listen to the publisher, at least that's what I gather. Am I mistaken in this view?Good question. My answer is, you're not mistaken, but the threat of cancellation is almost too blunt an instrument to use as a tool in editorial negotiations. canceling a book amounts to using the "nuclear option," laying waste to the whole project and vaporizing the time everyone has invested in it. (Quite likely vaporizing the publisher's money, too, as on-signing advances can be difficult to recoup.) I'm happy to say this situation has arisen very few times in own experience, but it has happened--it's very unpleasant for all parties involved.
In general, editors will sigh and let an author get away with some pretty misguided writing before they try reject a manuscript--threatening termination is so traumatic to an author that I at least rarely mention it unless it's a real possibility. In fact, with nonfiction, one would rarely cancel a book for mere differences of editorial opinion: it would most likely be over issues of veracity, accuracy, or legal liability.
(National Lampoon cover from January 1973,
photograph by Ronald G. Harris from a concept by Ed Bluestone)