Deep in a long discussion, Phil writes, in evident frustration:
I don’t like argument by innuendo. Say what you mean; how hard is it, for cryin’ out loud?
Actually, it is hard! I’ve spent years trying to write directly, and I’ve often noticed that others have difficulty doing so. I always tell students to simply write what they did, in simple declarative sentences. (It can be choppy, that’s fine: “I downloaded the data. I cleaned the data. I ran a regression” etc.) But it’s really hard to do. As George Orwell put it, good prose is like a windowpane, but sometimes it needs a bit of Windex and a clean rag to fully do its job.
P.S. I feel similarly about statistical graphics. (See also here.)
Agreed – it's difficult for people. The beginning of the semester is full of homeworks that spend two paragraphs describing a Wald test.
One of my heroes of science and science writing is physiologist Knut Schmidt-Nielsen. In about 1993 I bought his short book "How Animals Work" — which may be the best short data-based book ever written — from a used book sale, and thought I had discovered an unknown gem. I later found that Knut is relatively famous for both his science and his writing.
In 1998 Knut's autobiography (The Camel's Nose) came out, and I read it. He devotes four pages near the end to a discussion of writing for scientific journals…looking at those pages now, I'm reminded that for more than ten years he taught a course on scientific writing at Duke.
One thing Knut emphasizes is the importance of the title. He notes that a potential reader might glance at the title of 5,000 or 10,000 papers in a year — if you just read the titles of papers in Nature and Science, you're at 4,000 already — so if you want people to take the time even to read the abstract, the title had better interest them. He gives an example: "Further Comparative Studies of Food Competition among Grazing Herbivores." He says this would be better as "Food Competition among Grazing Animals."
And of course he gives advice that we all know but find hard to follow, like, you can almost always cut "It has been shown that"; and often a long phrase ("due to the fact that") can be eliminated or replaced by something simpler ("because"). He says people often try to sound sophisticated when they should be trying to simply be clear.
Around the time I read the book, Andrew and I were working on some papers together. I told him about Knut's principles, but couldn't remember any exact examples when Andrew asked for one. I made one up: "Instead of 'Even desert-adapted mammals show signs of distress when deprived of hydration,' Knut would recommend 'Even camels get thirsty.'" I don't know if Andrew remembers, but for several years, when Andrew and I were struggling with a sentence or paragraph that seemed convoluted, confusing, or simply too long, one or the other of us might say "even camels get thirsty," to remind ourselves of the ideal of clarity and simplicity we were trying to achieve. Amusingly, looking through Knut's autobiography now, I see that he gives no example as good as the one I made up for him.
I strive for Knut-like clarity in all of my papers, and the papers are better because of it. But I'm very far from perfect, perhaps especially with the titles. I think I did OK with the title of a paper a couple of years ago, "Trade-offs between moving and stationary particle collectors for detecting a bio-agent plume," but this is still nowhere near as good as the title of the paper Andrew and I were working on during the period I was reading Knut's autobiography. It's probably no coincidence that that title remains the favorite among all of my papers: "All maps of parameter estimates are misleading." The title truly says it all. You don't even have to read the abstract.
Phil:
Good points. I did list Knut in my list of best nonfiction books ever. And here's my advice for writing research articles. I forgot your point about the title.
One thing, though. You write that Knut says that people often try to sound sophisticated when they should be trying to simply be clear. Maybe so. For me, though, I'm always trying to be clear, not to sound sophisticated–but still I find it difficult to be direct. I think it's just hard, even without any sort of mixed motives.
Andrew — funny, you and I could just be exchanging emails, if it's just the two of us — yeah, I agree, even if one is just trying to be clear (independent of whether or not you're trying to sound sophisticated) it's not so easy. But for me, at least, I can often make headway at improving a sentence or a paragraph if I try to separate the content from the presentation. "What is it I'm trying to say" is different from "how can I get this point across." When you're just sitting there typing, as I am with this sentence, you're combining both of these into one act. To the extent that we think in words, especially when trying to write something, it's hard to separate them, and you, or at least I, find myself operating at way too detailed a level, focusing on individual words, when my problem is often just that I'm not clear on _exactly_ what I'm trying to convey.
Homework is a little different. You don't want to know what the students did – you already know what they did, and the students know that you do. What you want to know if they know what they did.
If they really would simply write what they did, they would say something to the effect of "We prepared the data and performed a Hochtrabend-Besserwisser test as described in the course textbook{ref], following the procedure in lab assignment[ref]." That's what you'd do in an actual paper or similar after all. But that would defeat the entire point of having an assignment, and they know it.
Instead they give you what you want – a detailed explanation intended to show you they got the point of the exercise. And since they can't read your mind, they'll err on the side of longwindedness and redundancy lest you think they try to gloss over something they didn't get.
The problem is partly that direct writing is hard. But WHY is it hard? Little kids do it. Look at how 2d and 3d graders write. Not grammatically (usually) but very directly. Then we train them out of it, and the training gets worse in grad school.
I have been reading about the history of the Royal Society; I found that it was founded (per what we moderns would call its "mission statement") to foster clear writing in the sciences.
Yet, in all fields, but certainly in the social sciences, we foster lack of clarity. Long sentences. Lots of clauses. Look at almost any journal article.
Some suggestions:
1) Any sentence with more than 10 words should almost certainly be two sentences
2) Any word with more than 10 letters should almost certainly be changed.
Peter.
I completely disagree with you. Little kids write with great difficulty. It might take a second grader twenty minutes to write a thank-you note to grandma. I don't think this is at all comparable with the sort of writing you need to do to explain something to somebody.
Also, I disagree with the idea that people are being trained to write in a muddy way. I think it's the opposite: year after year, students are told to write what they know, to be specific, etc etc. But it's hard to do. I suspect the teaching of writing would go better if the teachers recognized the difficulty of clear writing, rather than thinking it's something that just comes naturally.
As a teaching assistant in econometrics and a student myself, they main reason I have a difficult time being short and pertinent is simple. Sometimes, I am not entirely sure myself about the subject I am trying to learn/teach.
Being immersed in the field of applied econometrics, you get lost in the jargon and forget the main point of all these complicated tests and analysis. In my teaching, it's hard to be direct on the "how's" of econometric, that's just difficult all around. However, I always try to be clear and direct on the "Why's" so they understand our goals and understand why it is problematic in an intuitive way.
If I find myself not be able to express something directly, I take that as a clue that I lack some information, regarding the problem, issue, goal or the situation.
On a concluding note, I really enjoy reading Mr. Gelman's blog, as an R aficionado (making me rather an outsider in my department) and my budding interest in multilevel modeling!
Cheers.
I agree with Andrew Kids do find writing hard.
My favourite quote on this is from Dr Johnson :
"Easy reading is dammed hard writing"
and ?Verlaine who said "A poem is never finished, only abandoned".
I have also taught creative writing and the pursuit of clarity is just as hard there too.
Actually your camel example is not so good Tinbergen founded a whole science on the difference between that statement and "After 4 weeks in the desert our Camel had reduce buccal secretions and ambulatory oscillations".
The difference between observations and inferences.
Hi Andrew
I didn't say little kids write *easily* I said they write directly. They don't use obscure words (partly because they don't know any) they don't use elaborate sentence structures (although some seem to use "and" in place of a period). We get better at very basic writing through practice, true. There are schema for thank you notes, for instance.
And when older kids write very directly, we tell them their writing is boring. Further, the writing they are exposed to is more and more indirect; in fiction, people use metaphor and simile, for instance. In the sciences and mathematics, we learn more and more elaborate terms.
I had one professor in grad school who said that we go to grad school to learn complicated words like heteroskedastic. I am not sure how much he was joking.
I also recall one study where they handed in pairs of dissertations. One in each pair used simple declarative sentences and a minimum of jargon. The other had similar content, but just the opposite style. The latter got better grades!
Phil: If you and Andrew were exchanging emails, it wouldn't be here in archive form for the rest of the internet to benefit from. Even if it seems like no-one else is participating, you have to remember how many people could be silently reading your discussions. Not just now, but even years from now after they do some web search and hit the blog.
I was surprised recently to find out which of my blog posts were the most popular via web search.
Re: children't reading
I have my 11 yr old read the occasional obituary from The Economist. Often it's non-technical, giving some recent cultural history. For example from Nov 5th 2009 (subscription required):
I'd like to draw attention to Xantippa's comment that "If I [Xantippa] find myself not be able to express something directly, I take that as a clue that I lack some information, regarding the problem, issue, goal or the situation." This is an issue for me too. Often, if I can't find a way to express myself clearly, it's because I don't have a clear thought in the first place.
I halfway agree with Peter's Principles: use short words, and write short sentences. (Devoted readers of this blog will have seen me write, more than once, "I am a big fan of the simple declarative sentence.") But in practice I often use fairly long sentences. The ten-word limit that Peter proposes is clearly far too short. (For example, that last sentence exceeded it, and I don't think that's an unreasonably long sentence.) And although I agree with using short words where possible, I think a long word like "unreasonably" is fine, so again I'm not sure I'd agree with a rule of thumb to avoid words longer than 10 letters. I guess I'd say I agree with the spirit of the principles Peter espouses, but not with the letter of them.
Re Long sentences good or bad:
Strunk:
Hi Phil
I just sort of picked 10 as a nice round number. I've thought some more, thanks to your post. We can distinguish long words based on short ones from ones with long roots. "Unreasonably" is based on "reason", a word most will know. "Heteroskedastic" is not.
About 196 years ago, in may, some people were united in Eidsvoll writing a new constitution for Norway. (In four years time there will be some expensive celebrations!) One of their more famous declarations were "all power in this hall" (all makt i denne sal). But that was not what they really wanted to say, as they wrer not making a dictatorship. What they wanted to say were: "zero power in the king's castle". ¿So why did'nt they say so?