I recently became involved in a discussion on how to write a scientific paper, and the first thing I had to concede was that on the whole we do not do this well, and sometimes it is written almost as if the author said to him/herself, "Nobody will read it anyway." In many cases it may not matter. Many papers are written to archive an observation, or a procedure, such as how to synthesize something. These involve putting things down in the order that they were done, and making sure all terms are defined. The writing style probably does not matter much, because the only people who will read this are those who wish to either use the observation or to follow the synthesis. The first group will accept the statement and the second will have to work through it, no matter what.
More difficult is when you have to interpret what you found. An obvious example is the structural determination. The problems include the fact that there will be several interpretations of any given observation. The usual approach is to eliminate them one by one, usually in a sequence of experiments, and if that is what you did, so must you report it. The major problems include a failure to eliminate all possible alternatives, in which case the report is unconvincing, or alternatively, the alternatives are eliminated, but the eliminations are scattered, and it is too difficult to keep them all in mind, in which case the argument runs the danger of being confusing. A common problem is the presentation of data that supports your hypothesis. It may, but equally it may support something else.
Time for a confession! I once wrote a series of papers on the substitution patterns of red algal galactans. Prior to writing these, structural elucidation was very difficult. These initial procedures involved the molecules or substituted derivatives being broken down into fragments, following which a sequence of fragments were further fragmented, and from the resultant structures, the overall structure was inferred. Because there were so many different aspects that had to be kept in mind, in one paper I wrote with two others, in parts the sentences became so complicated that later even I had trouble working out what they meant. So I came up with an answer: represent everything mathematically. Rather than get to the structure linearly, I carried out a number of different operations on the parent molecule, and from nmr spectral data, each operation was consistent with a set of structures. The real structure was given when the intersection of all sets gave one element. I then wrote papers representing structural units as matrices and data as ordered sets. Mathematical manipulation was unambiguous. The problem was, the rather restricted audience was not very happy with discrete mathematics, and eventually an editor told me to stop doing that or else the papers would be rejected. As it happened, I did not care so I stopped publishing. I was self-employed, and this activity was to bring no income, so the decision to stop was not that difficult. The real shame was that the methodology was just becoming productive
Nevertheless, this raises the problem, what concessions should be made to the reader? My view at the time was, the statement of what I believed the structure to be should be put down in the simplest form possible. However, while how I deduced them should be as clear as possible, I thought it is not unreasonable to expect some effort to be made by those who wished to question the structure. My view was that to put down mathematically the arguments leading to the structures was optimal because all logic steps are explicit and unambiguous. There is no question of acceptance; to disagree you must show some step does not follow. However, many scientists did not agree with such an approach, and prefer comfortable sentences, which will generally be read without questioning them. What do you think? Be unambiguous, but have few readers, or be comfortable but with questionable ability to convince?
