Cards for a dissertation chapter

A dissertation chapter is the longest piece of argued prose most people ever write, and it is written under a condition that no other writing has: a committee will ask you to move things. Not to fix sentences — to put the theoretical framing before the case rather than after, to bring the second study forward, to split the chapter in two. Every candidate has had the meeting. And every candidate has then spent three weeks doing what the meeting asked, because in a sixty-page document, moving a section means every transition breaks, every cross-reference is wrong, the introduction promises the old order, and the footnotes renumber.

Cards make the committee's request an afternoon. That is the practical reason to write a chapter from them. The deeper reason is the one every guide before this has made: the chapter has an argument, the argument has a structure, and at sixty pages the structure is impossible to see in prose and easy to see on a table. This guide is about the scale, the committee, and the parts of a chapter that the undergraduate paper does not have.

The scale

A chapter is two to four hundred cards. That is not a table; it is a room, and the method changes in three ways to fit.

Sections are the unit of arrangement. A chapter's order is an order of sections, and within each section an order of claims. On paper this means piles with a heading card on top, arranged in a row, each pile opened and laid out when you work on it. In software it means an outline with headings at two or three levels. Either way, the top-level order — the sections — is the thing the committee will ask you to change, and it should be a single row of six to twelve cards that you can rearrange in a minute.

Every section is a paper. Each section has its own claims, its own evidence under them, its own counterpoint beside, its own thesis-sentence at the top that says what the section establishes. Build each section as student guides 1 and 2 build a paper. A section that does not have a sentence at the top saying what it establishes is a section that will read as and then I also looked at, and the committee will say so.

The chapter's thesis is one level up. What do the sections' theses have in common? That is the chapter's claim, found last (student guide 4), and it goes in the introduction that you write last, and it is the sentence the abstract is built from.

The parts a chapter has that a paper does not

The literature. A chapter usually opens with, or contains, a review of what the field has said (graduate guide 1). Build it as that guide says — by question, with disagreement cards — and keep its cards in their own section pile. The review's gap is the joint between the literature section and the rest of the chapter: the sentence that says and so this chapter asks.

Method and data. In an empirical chapter, the method is a section whose cards are not claims but descriptions — what was done, to what, with what — and the results are Result cards: one statistic each, in full reporting form, with the analysis it came from and whether you have checked it against the output. The claims in the discussion section link supports to Result cards, which means that when a result changes — a re-analysis, a corrected coding, a reviewer's request — you can see every claim that rested on it and check each one. This is the single most valuable thing cards do for an empirical chapter, and it is the thing prose does worst: a changed number in a results table silently invalidates sentences forty pages away.

Figures, tables and equations. Each is a card — the data or the source on it, the caption with it — placed in the order where it belongs, numbered by that order. Moving a section renumbers them. A reference to Figure 3 is a link to the card, not a number typed into prose, so it survives the move.

Matter. The abstract, the acknowledgements, the appendices, the data-availability statement if the field asks for one. Each is a card of its own kind, kept apart from the argument, placed by the format's convention at the end. They are written last, from the chapter, and they are the parts a committee reads first.

The committee's requests, and what each costs

"Put the framing before the case." Move one heading card in the top row. Rewrite two bridge sentences and the paragraph of the introduction that previews the order. An afternoon.

"Bring the second study forward." The same, one level down, with the results cards moving with their section and the figures renumbering themselves. The discussion's claims still link to the same results; nothing about the argument changed, only where the reader meets it.

"Split this into two chapters." Two arrangements from the same cards: the first six section piles in one, the last five in the other, each given its own thesis sentence at the top. The cards are not copied and nothing is retyped. This is the request that costs a month in prose, and it costs a day.

"I don't see how section four follows from three." The bridge is missing or the order is wrong. Lay out the two sections' thesis cards side by side and write the sentence that gets from one to the other. If it cannot be written, try the other order (student guide 6); a join that will not hold in either direction is a section that belongs elsewhere.

"Where does this claim come from?" Point at the card, and at the evidence under it, and at the source the evidence came from. A committee member who can be shown the chain from a claim to a page number in a source is a committee member who stops asking. If the chain is not there, you have found the claim to cut before the defense, which is the point.

"This is too long." The cut pile (graduate guide 5). Every section has cards doing nothing; the table shows which.

Working with a supervisor from the cards

The traditional exchange is a draft sent, a draft returned with comments in the margin, a revision, three weeks per round. The card exchange is faster and better because the supervisor sees the structure instead of the prose: the row of sections, the claims under each, the evidence under the claims, and where the evidence is thin — before the prose exists, when a request to move something costs nothing.

On paper this is a meeting with the table between you. With software it can be a link the supervisor opens: the argument's structure, the map of where the evidence is thin, and a place to comment on a card, with nothing of yours handed over that you did not choose — and in particular nothing about when you worked, which is not the supervisor's business and which a supervisor who is also a gatekeeper should not be shown.

Do the structural conference before the prose conference. A supervisor who has agreed the order and the claims will read the draft for the writing, which is what drafts are for.

Writing the chapter

From the arrangement, section by section, claim by claim, as every guide says: the claim as the paragraph's first sentence, the evidence as its body, the counterpoint answered beside it, a bridge at every junction. At sixty pages the bridges are the difference between a chapter and a pile of sections, and they cannot be written until the order is settled, which is why the committee's request to move things is so expensive in prose and so cheap here: the bridges are the only prose that depends on the order, and there are twelve of them.

The introduction last. The abstract from the introduction. The acknowledgements whenever you like.

Keeping the chapter alive

A chapter is not finished when it is defended. It becomes an article, or two; it is cut to a conference paper; a reviewer asks for the study the committee asked you to bring forward to be moved back. Each of these is a new arrangement of the same cards — the article's order, with the literature section cut to two paragraphs and the method compressed; the talk's order, with the figures and the claims and nothing else. The cards were the work. The chapter was one order of them. There will be others, and the cards are ready.

Doing this in software

Software that works from cards holds a chapter's scale without the room: an outline with headings and levels, several arrangements of the same cards kept and compared side by side, figures and tables and equations numbered from the order at compile, Result cards whose change flags every claim that depends on them, matter placed by convention, and a set of dissertation templates that typeset the chapter with the front matter, approval page and pagination a scheme expects. It gives a supervisor a link to the structure, the thin-evidence map and a comment box, and nothing else. What it does not do — and what nothing should — is write a section's thesis sentence, decide whether a committee member's request is right, or tell you which order is better. It shows you both. You decide.