“A poem is a machine made of words.” (William Carlos Williams)

Think of a machine as a structure that is “geared” for action, an arrangement of parts that does something once it has been set in motion. Indeed, Williams thought of a poem as a “field of action”, just as Hemingway sought a “dignity of movement” in his prose. Good writing is effective writing — writing that has a series of intended effects on the reader. A paragraph is a little machine that makes a claim easier to believe, understand or agree with. An essay is constructed by arranging paragraphs in a series, one effect after another. In my weekly Wednesday talk this afternoon I’m going to try to leverage, if you will, these metaphors in thinking about the structure of a research paper. I want to help you distinguish the paragraphs in each section of a paper in terms of their means and ends, their causes (in Aristotle’s “material” sense) and their effects, or, as I have been calling them in my previous talks, their bases and their aims. The trick is to bring them all together for an overarching purpose, to get them working.

What, then, is your research paper based on? The obvious answer is that it is based on your research, but we can be much more specific about this when we consider each section in isolation from the others. The background section, for example, is based on publicly available sources of information, i.e., newspapers, books, company reports, government whitepapers, official statistics, and so forth; the theory section, by contrast, is based on the scientific literature, which you explored when doing the literature review. The crucial difference here is that your reader needs no special qualifications to access and understand your background sources, while your theory section is really only going to make sense to a trained specialist with access to the relevant journals. Your analysis is, of course, based on your data, to which you have privileged access as a writer. (No matter how open you are about your data, you should write about it as though the reader hasn’t seen it.) Your methods and discussion use sources of a different kind: experience and reason, respectively — your doing and your thinking. Think of your basis in each case as what qualifies or author-izes you to write it. Your introduction and conclusion, for example, are based on what is in the rest of the paper, which makes you, the author, the ideal writer of these sections.

As you can see, we can easily distinguish the sections of your paper on the basis of their sources. But we can also look at their aims — what each section is trying to accomplish. The introduction is, obviously, going to introduce your paper, which is to say, it’s going to open a dialogue with a knowledgeable peer about a subject that interests you. Your background section will inform the reader about facts you won’t presume the reader knows. The theory section will activate a set of expectations of your object in your reader. Since the analysis is going to try to bring about an “artful disappointment” of those expectations, your methods section will build trust in your data, so that your reader won’t just dismiss your results. The discussion section will then identify the implications of whatever tension exists between the theory you have used and the practice you have studied. Finally, the conclusion will (just as obviously as the introduction introduces) conclude the paper, bringing the conversation to a close, and bidding the reader farewell.

This smooth sequence of aims, with one task leading to another after it has been accomplished, should remind us that reading is a linear experience that moves forward in time. We have constructed a series of one-minute (i.e., one-paragraph) experiences that will ideally (though not always really) be lived by the reader from beginning to end, lasting about forty minutes altogether. Reading, like writing, is a process. But don’t forget that, at the end of the day, a paper is also a structure; it remains “standing” after the reading is done. Or, at least, it should remain standing. We might say you’ve walked the reader through a building that you have constructed and you don’t want them to come away with the feeling that it’s about to come down (that you intend to demolish it tomorrow). The discussion section should feel like it was “set up” by the background and theory sections. The analysis should challenge the theory but not overwhelm it. The methods section should establish limits that your conclusion respects. And your introduction should promise no more and no less than what your paper delivers. When the reader puts down your paper, there should be a clear image in their mind of a place they could (and hopefully will) revisit.

Williams’s friend, Ezra Pound, encouraged us to remember that not all images are still. There are moving images, imaginary films. Likewise, we must remember that not all structures are static. To say that something has structure is not to say that it doesn’t move, only that it moves, when it does, in a particular way. Even wholly imaginary dragons are constrained in their movements by their imaginary skeletons. A machine is a structure that repeats a series of motions over and over, belaboring a set of materials to produce an effect, a product, and even post-structuralists have this kind of structure. As I said a couple of years ago, 1968 marks a kind of “epochal shift” in our thinking about society, a movement, we might say, from “structure” to “machine”. It can be found in the famous opening lines of Deleuze and Guattari’s Anti-Oedipus:

It is at work everywhere, functioning smoothly at times, at other times in fits and starts. It breathes, it heats, it eats. It shits and fucks. … Everywhere it is machines — real ones, not figurative ones: machines driving other machines, machines being driven by other machines, with all the necessary couplings and connections.

These “desiring-machines” may just be the “poetic” counterpart of the “cognitive frames” of our prose. I tend to agree with Deleuze and Guattari that these tensions are not merely metaphorical, but I’m less inclined than some to abandon the prose of the world. And let’s watch our language, friends; let’s keep it clean out there!

Or don’t. Fu…

“Thus your data shimmers.” (Lisa Robertson)

I’m really enjoying preparing our weekly Wednesday talks. I’ve now had a chance to cover the theory and methods sections in some detail. This week I’ll be talking about writing the analysis. Because I’m trying to keep these talks applicable to the different levels that students are working at, as well as the full range of CBS degree programs, I’ve found myself occasionally waxing philosophical. I think this week’s talk will be a little more practical, but still general enough, I hope, to be of use to everyone. The overarching theme will be that of using your data to support factual claims about the object you have studied. That is, in our analysis we’re always moving from our direct observation of reality to our interpretation of that reality. It will be useful to think of each paragraph as including both an interpretation, which will be expressed in the key sentence, and some observations, which will support the factual claim it makes. That is, each paragraph in your analysis will assert a fact on the basis of your data.

Let’s begin with the data, which we have talked about before. It consists of what you have directly observed. In ethnography, it’s your record of what what people have done and said. In survey research, it consists of how they filled out your questionnaire. In financial market analysis, it consists of the stock prices you have exported from a relevant financial database. In discourse analysis, it’s the archive of documents you have collected. However you have gathered it, you deploy it in your analysis section by quoting (words or figures) as they appear in your data set or by summarizing aggregates. Your statements about your data are true or false in a highly objective and unambiguous way. People either said what you quote them for or they didn’t. A certain number answered “yes” to a question and another just as certain number answered “no”. You just have to count them. The documents either invoke the codes you’re looking for or they don’t.

But an analysis is not just a summary of your data. You have collected the data in order to represent the facts as they are, independent of your data and your analysis, and making your data represent facts always requires an interpretation. The amount of days employees are off on sick leave in a particular company is a data point. Whether the company has a stressful working environment is a fact to be determined by your analysis. You gathered the data in order to determine the fact but, interestingly, if your readers want to observe the same fact, they don’t have to use the same data. Facts are not made of data, we might say, they just “give off” data. Like an astronomer gathers the light from a star, you design your instruments to be sensitive to data about the people you study. To borrow Lisa Robertson’s image, data is a “shimmer” on the surface of your facts. The data are ultimately ephemeral (which is why you have to keep a good record of them); the facts are made of sterner stuff.

Again, your analysis doesn’t just describe your data; it doesn’t just make claims about your sample. It makes claims about the world in which we live out of interpretations of your data. It tells us what your data has shown you, what it has taught you about your object. As I have said before on this blog, this lets you think of each paragraph in your analysis as repeating a simple pattern: the key sentence tells us what you mean and the rest of the paragraph tells us how you know. The key sentence may tell us what the people you have studied believe or desire, but the rest of the paragraph will tell us what they said or what they did to make you think so. Present your interpretation in the key sentence and build the rest of your paragraph around your observations. Obviously, you should make sure your observations support your interpretations.

It is tempting to see the analysis as a “write up” of your data. If we’re working with qualitative data, we’ll often start with memorable quotes from our interviews or striking observations from the fieldwork. Quantitative researchers might start with the “significant” results in their contingency tables. Either way, the writer thinks of their prose as a way tying these data points together, connecting the dots. But it is much better to organize your analysis around a set of claims about the world — statements of actual, ordinary fact. You will ultimately be composing a finite series of paragraphs, each of which says one thing, and supports that claim with your data. So plan out your analysis section as a series of claims that you are able to support, not just a number of themes inspired by your data. After all, your readers don’t just want to know about your data; they want to know what your data shows us about the world in which we live. They want your observations and your interpretations of them.