(This article was originally published at Statistical Modeling, Causal Inference, and Social Science, and syndicated at StatsBlogs.)
Following up on my discussion of Steven Pinker’s writing advice, Pinker and I had an email exchange that cleared up some issues and raised some new ones.
In particular, Pinker made a connection between the difficulty of writing and some research findings in cognitive psychology. I think this connection is really cool—I’ve been thinking and writing about writing for awhile, now, but I’e never really seen the connection to psychology research. So I wanted to share this with you.
Pinker’s remarks came at the end of an email exchange. I’ll share the earlier messages to give the background, but by far the most interesting part is what Pinker said, so I’ll give that right away.
Here’s Pinker, discussing the difficulty of communicating complex ideas (in particular, in academic writing we typically aren’t just making the case for position B, we’re also arguing why previous position A, reasonable as it might sound, is not correct):
I address it in part in chapter 5 of The Sense of Style in discussing our comprehension of negation. The human mind cannot represent a proposition without a truth value – to think “X” is to think “X is true,” at least temporarily. Negation requires an extra mental step—which can easily fail when the person is overloaded or distracted. A number of systematic kinds of error and difficulty follow. My colleague Dan Gilbert has an insightful review of this literature in his 1991 article, “How Mental Systems Believe.”
On top of that people’s comprehension is often driven more by expectations than the literal content of the text (again, particularly when not paying close attention). When I see certain misinterpretations I often think of the puzzling neurospsychological syndrome called “deep dyslexia.” Surface dyslexia consists of misreadings based on alphabetic confusions – misreading “pear” as “bear,” for example. In deep dyslexia, the patient might misread “pear” as “apple.” The puzzle is that if the patient’s word-recognition system could parse the letters well enough to realize it referred to a fruit, it must have been because he matched the input with a stored template for “pear” – so why didn’t he successfully read it as “pear”? Presumably such patients’ semantic representations (the definitions in their mental dictionary) were so degraded that the word merely activated a coarse semantic ballpark, without enough precision to pinpoint the exact entry. From there sheer base-rate frequency determines the output. It’s a crude analogy to the way we non-brain-damaged people often parse a sentence coarsely enough to remind us of a semantic neighborhood and then we fill in the rest from base rates. Often a writer will have to anticipate this and explicitly disavow an expected confusion: “You probably think I mean this, but I really mean that.”
I’ve been trying to do this more often, for example in this 2011 article on the philosophy of Bayesian statistics, where I’m pretty explicit about what I’m disagreeing with (for example, the first section of this paper is entitled, “The Standard View of the Philosophy of Statistics, and Its Malign Influence on Statistical Practice”).
Still, I think the necessity of clearing-away-the-old creates an additional degree of difficulty in much of academic writing.
What I found exciting in Pinker’s note above was his connection of this vague idea, which I arrived at by introspection, to research in cognition.
Background
And here’s how we got there.
Our conversation started with Pinker reacting to my statement that it’s not so remarkable that academics don’t in general write well; after all, writing is hard.
Pinker wrote:
The paradox is that academics might be expected to do better than laypeople in their writing, since the communication of ideas is part of their job description. So if academics are no better than laypeople at writing, that raises a puzzle. And . . . academics are not just average writers – academic prose is often, notoriously, far worse than nonprofessional writing.
Pinker also pointed out a place where I’d misread what he’d written even though he’d taken “a great deal of care in crafting the sentence” (as he put it).
This was interesting and reflected experiences I’ve had, where I try so so hard to be clear, and people still come up with a misreading of my prose.
I reflected on this and replied to Pinker as follows:
My misreading of your article raises an interesting meta-issue about writing which might interest you, as I think it’s also relevant to the puzzle of why academic writing is often so bad, despite the fact that academics typically get a lot of practice at it.
What happens to me a lot in writing, whether it’s a scholarly article, a textbook, or a popular article, is that I am aware of a possible misunderstanding, and I carefully craft my sentences to avoid the possible pitfall—but people fall into it anyway! For example, when writing about significance testing, I am careful to avoid saying that the p-value is the probability the null hypothesis is true, so I use a very careful wording, but then people read my writing as if I’d said that wrong thing. I realize that in such cases it’s not enough to say it right and avoid the error; I really need to explicitly state that I’m not saying that other thing that people are expecting to hear. (Which may be what happened in the cases where I misread you; I read a sentence on a certain topic, was expecting to hear a certain thing (“professors have a foreign policy” or whatever), and then I heard it, misprocessing the incoming words. So in my more recent writing I’ve tried harder (not always with success) to not just avoid making an error but also to making clear to the reader where the error arises.
Anyway, back to academic writing. Perhaps academics are often in this situation: trying to explain an idea. while also explaining why a certain natural-seeming interpretation is not correct. Which means that, to be most effective, we have to both convey our idea and also convey the idea that we think is wrong. I think this is an inherently difficult task (and I suspect you’ll agree on this, as I know that a lot of your writing has this feature, that you explain the appeal of the wrong model in the context of explaining your own ideas). So . . . perhaps one contribution to the ugliness of academic writing is that academic writing, to be compelling, often has to do both these tasks, and that’s not easy.
And all this is, perhaps, especially true when academics are writing about topics of general interest (which is, of course, the time that outsiders are likely to encounter our work. A nonstatistician might well read my article on the statistical crisis in science; such an outsider is not so likely to read one of my articles that’s full of math): in our general-interest articles we’re often explaining how a certain common idea is actually wrong.
It was in response to this message that Pinker sent the note given above, with the connection to psychology research.
The post Discussion with Steven Pinker connecting cognitive psychology research to the difficulties of writing appeared first on Statistical Modeling, Causal Inference, and Social Science.
Please comment on the article here: Statistical Modeling, Causal Inference, and Social Science
