The Palimpsest Method
On revision
I’m a believer that second drafts should generally be rewritten from scratch.
By “from scratch” I mean, more precisely, from memory—usually after an incubation period of at least a couple days. While not always feasible due to practical constraints, and not always worth the extra effort that comes with starting over, rearticulating your ideas anew offers two major advantages.
First, rewriting from scratch serves as a kind of compression algorithm—one which helps crystallize what is actually essential and what is in fact extraneous material (however instrumentally useful it may have been in the course of the piece’s development). If you rewrite a story, speech, or essay from memory and find that you left out a “key detail” you’d had in your original draft, it’s worth asking yourself how key that detail truly was.
Second, rewriting from scratch can help shake loose structural assumptions that are difficult to spot when you’re working under the sway of your original version. For example, suppose that reading in reading over the two sentences below, you feel that the syntax in the second one is clunky:
Version I [original draft] — I was feeling worn down, struggling to get through my shifts at work. So I decided to request, after much deliberation, a week of unpaid leave.
In a standard revision process, you might move up the offending clause to create a more natural transition between the sentences:
Version II [standard revision] — I was feeling worn down, struggling to get through my shifts at work. So after much deliberation, I decided to request a week of unpaid leave.
This is, to my ear, an improvement on Version I. However, had you rewritten the sequence from scratch, you would almost certainly have produced a more radical departure in flow and word choice—for example:
Version III [re-creation] — I was so burnt out at work, I started thinking about taking a week of unpaid vacation. I turned the idea over and over in my head, floated it to a couple friends, and eventually asked my boss if I could do it.
The point here isn’t that Version III is dramatically better than Version II, or even very different; rather, it’s that Version III generates more new options than Version II and highlights more of the implicit choices in Version I.1
In a sense, what you’re doing here is trying to simulate a kind of internal “writer’s room,” in which two or more “different people” (i.e., you, at different points in time) provide different visions of the same underlying concept.2 As you move into later stages of editing and revision, you’ll then work to blend these visions into a coherent final product. You’ll typically have a bias toward the most recent version (which generally is better overall); still, you may find useful nuggets from earlier drafts you opt to reinsert or combine with your latest model—for example:
Version IV [hybrid of original + re-creation] — I was feeling worn down at work, struggling to get through my shifts. I was so burnt out, in fact, I started thinking about taking a week of unpaid leave. After turning the idea over and over in my head and floating it to a couple friends, I finally decided to ask my boss.
Since the examples above are intended to simply demonstrate a revision method, I’ve kept the sample passage short and focused on sentence-level changes. However, it’s worth noting that many of the most difficult choices to detect are actually operating at a much higher conceptual level in a piece—involving, for example, the underlying purpose of certain paragraphs or fundamental choices about pacing and emphasis. (As hard as it can be to notice that, say, the last sentence of your introduction might benefit from rewording, it’s even harder to notice that you don’t actually need the introduction at all; or that the introduction would serve you better if it were split in half and sandwiched on either side of paragraph seven.) You’re much more likely to stumble onto these kinds of conceptual revisions organically if you recreate the piece from scratch.3
Implicit in this from-scratch approach is that while you may apply it multiple times, you’ll eventually get to a point where you do stop rewriting and instead simply refine the writing that is already on the page; the longer the piece is, the earlier this point is likely to arrive.4 After all, there are only so many pages of content a person can recall from memory.
Still, regardless of the length of the text or how far along it is, there is often value in selectively employing this kind of strategy. If a particular paragraph seems like it isn’t working, you might try eliminating it and rewriting cold. If you’re struggling with how to articulate an idea without using complicated syntax, you might try starting with a novel sentence stem, just to generate some alternate version. Even if that version is worse, it may still reveal some options or implicit choices you’ve made.
While writing is often analogized to shaping a lump of clay or finding the statue in a block of marble, I think the more apt metaphor—especially in the early stages of a piece’s development—is overpainting layers on a canvas. In a literary context, the best term for this is probably palimpsest: an old manuscript which has been effaced and written over. (In the modern context, “palimpsest” refers more broadly to any work in which traces of the original remain partially visible, covered up by the final version.) Whatever you liken it to, the key point is this: A layer-atop-layer approach is effective not simply as a means of concealing or correcting mistakes; rather, it is a highly generative process in its own right.
In fact, in drafting this very post, I realized I’d published an earlier version of the idea two years ago, as part of a longer piece. Looking back on that earlier post, I find it less compelling than this one. Still, I did find one or two useful nuggets I was able to reincorporate. And however faint the traces of that first attempt, there is little doubt the piece you’re reading now would not exist without it.
If we break down the flow, we’ll see that versions one and two take the form of [sent. 1: assertion of burnout]; [sent. 2: unpaid vacation? → decision] whereas version three takes the form [sent 1: burnout → unpaid vacation?]; [sent 2: deliberation + decision]. Whatever the merits of each option, these are subtly different emphases.
I reference this study often because I find it fascinating, but it’s worth reiterating that asking yourself the same question twice—a.k.a. polling the “crowd within”—is an approach whose efficacy is supported by research.
Of course a good editor is also invaluable for spotting these kinds of big structural changes.
The method also implies a kind of goldilocks level of forgetting: Wait too long to rewrite, and you may find yourself at a loss for what was insightful or exciting about the piece in the first place; jump right into the rewrite, and your “re-creation” may be close to a verbatim duplicate—which more or less defeats the purpose.