(from an interview conducted by Thomas Kuhn on Jan 28, 1964. http://www.aip.org/history/ohilist/4570_2.html
CONTEXT: As a graduate student, Dennison had requested to do his thesis research in theoretical physics (this would be the first theory Ph.D. granted by Michigan). Guided by Walter Colby and Oskar Klein, Dennison developed the theory to understand the infrared spectrum of methane, a project that fit well with the experimental work being done by Randall, Barker, Sleator, and others in the Michigan physics laboratories. During the interview, Dennison told Kuhn:
At this time (1923) Oskar Klein suggested that I write a preliminary version [of the methane paper] which I would send on to Copenhagen. . . . I was sure that I was pretty good about writing. I was positive that I knew how to write. So I wrote a version of the thing up, and Klein sat down with me, and every single sentence was wrong, had to be revised and changed. ‘Exactly what do you mean by this, and what’s the shade of meaning here?’ and so on. Everything, piece by piece. I remember at one session I was beginning to be a little annoyed. I said finally, “Professor Klein, you would write it one way; I will write it another way. This is my way of writing it.” Klein said, “This is nonsense. There’s only one way to write a paper. If you have something to say, there is only one way; there are no two ways of writing it.” This actually, to a certain extent, was so banged into my head that I tend to believe it. I tend to believe it now that a good paper is one in which — in a sense it’s a little bit like an analytic function — each sentence has a little bit of what was past and a little bit of what is coming, so that from a portion of it you almost can reconstruct the rest. This is an exaggeration, of course. But, nevertheless, there is really only one way of writing well.