Structure

The basic structure of a paper is to make a claim that will be proven, to provide the evidence, and to conclude by bringing all the evidence together to show that it proves the claim and by suggesting the implications of the claim -- how it fits into larger issues, what further work might be appropriate, and so on.

This structure repeats at larger and larger scales. A paragraph makes a claim, provides evidence, and concludes. But paragraphs occur in sections, and each section has an argument, which is supported by the separate paragraphs, which fit together to provide the conclusion. Sections occur in papers, and papers have an argument (the thesis) which is supported by the sections, and ends with a triumphant conclusion. In a six-page paper, there may be only one section, and at most two or three; in a twenty-page paper, there usually need to be several well-defined sections. A book has a large argument that depends on the claim of each chapter; each chapter has a claim that depends on points established in each section; each section has a point that is established by the small arguments of individual paragraphs.

To make this structure work, one needs to pay attention to transitions. In general terms, transitions not only connect what has gone before to what comes after, but, when appropriate, they are the parts of the paper that remind readers how the specific points being made connect to the overall thesis. Consider this transition from a paper arguing that Malory presents different styles of chivalry: "While Launcelot's form of chivalry is religious, it emphasizes good deeds done in the world; Galahad's chivalry is more concerned with seeking communion with God." this would the first sentence of a new paragraph, and it sums up the section that went before (Launcelot's chivalry) and announces the claim that will be proven in the new section (Galahad's chivalry is uniquely religious), and its emphasis on different styles of chivalry is a reminder of the general argument being proven. Because it announces the new topic, it belongs in the new paragraph and the new section. While concluding sentences may foreshadow what is to come, avoid the mistake some writers make of foreshadowing so strongly that the topic sentence of the next paragraph is actually placed as the last sentence of the preceding one.

A number of students are taught a "five-paragraph theme" structure: an introduction announcing an argument, three paragraphs each making a point that supports the argument, and a conclusion that repeats the argument. This is a good for teaching the basic idea that papers must be structured to support an argument, but this structure does not work as well for more advanced arguments. (It is most pernicious when the body paragraphs are described as providing "examples" -- the body paragraphs do provide evidence from which one can argue for the conclusion, but if the thesis is so simple that there are three self-evident examples of it, the paper's not going to be that interesting). Very often there are not three separate points that support a thesis, but a chain of reasoning. Point A proves point B, point B proves point C, point C proves point D, and point D can be used to prove the thesis. Points A, B, and C by themselves do not prove the thesis, so one needs to connect them in a chain, linked to each other and only indirectly to the main argument.

One simple way to test if the structure of a paper is working is to take an "x-ray": outline the paper after it is written by writing down in order one sentence per paragraph encapsulating the main point. If you can't write down the point of a paragraph, the paragraph is broken -- it either doesn't have a point or it has more than one and needs to be split into separate paragraphs. Once you have this list, read through and make sure the ideas progress smoothly. If there is a gap -- the topic of one paragraph doesn't seem connected to the one before it -- or if the paragraphs seem out of order, go back and fix the structure. If you could rearrange the paragraphs without much difference to the sense of the paper, you know you have a very weakly structured paper.