A good comment on one of my papers

An anonymous reviewer wrote:

I appreciate informal writing styles as a means of increasing accessibility. However, the informality here seems to decrease accessibility – partly because of the assumed knowledge of the reader for concepts and terms, and also for its wandering style. Many concepts are introduced without explanation and are not clearly and decisively linked in developing a narrative argument. I think the prose and argumentation would be much stronger if ideas were introduced and developed more deliberately and not assuming insider knowledge of the reader.

Good point. I have an informal writing style and that often works well, even for technical papers. But sometimes an informal paper is harder to follow for readers without the background knowledge. Paradoxically, a more stilted style with lots of notation and many stops to make precise definitions, can be more readable for the less-than-expert audience.

22 thoughts on “A good comment on one of my papers

  1. I can’t see that it matters whether a formal or informal style is used, provided the author gives sufficient information to enable the piece to be understood. I have spent 40 years in research establishments reading applied papers (in biology, ecology, agriculture,ecotoxicology and so on) in which there was often a significant statistical component. The vast majority were written in the stilted style and often the statistical content was incomprehensible, so I tend to associate stilted with unreadable. I always judged the the author(s) to be inarticulate, possibly deliberately. My favorite paper of all time is “Baby Bear’s Dilemma (Carmer and Walker, 1982)”, which, if I remember, is very funny and is about multiple comparison tests and porridge. https://www.soils.org/publications/aj/abstracts/74/1/AJ0740010122

  2. Perhaps it is probably a good idea to remind the reviewer (and many of us) about the “Plain Language” movement.

    “Plain language is clear, succinct writing designed to ensure the reader understands as quickly and completely as possible. Plain language strives to be easy to read, understand, and use. It avoids verbose, convoluted language and jargon.”
    (http://en.wikipedia.org/wiki/Plain_language)

    Note the last sentence: “It avoids verbose, convoluted language and jargon”. How many scientists and other professionals aren’t aware of this simple rule?

    • The point is not whether anyone would disagree but quite what this means in practice.

      Incidentally, words like “succinct”, “verbose” and “convoluted” wouldn’t qualify as plain or simple to many readers. Who is to judge the judges?

      The bigger issue is that what a reader may see as jargon the author probably sees as standard terminology, and vice versa. No technical paper can be written without some assumptions about level of knowledge. Moreover, a key goal of many technical papers is to show something rigorously, and that goal severely constrains acceptable style.

      To be clear, I’m on your side, and have spent a large fraction of my time as teacher, reviewer or editor cutting out or advising against obscure writing. It’s just not as simple as these manifestos imply.

    • It’s easy to vilify jargon but often it’s a useful shorthand of a field: If I use a word like heteroscedasticity / ergodic / anisotropic, is that jargon to be avoided?

      Also, the audience matters. A complex argument written in “Simple English” would probably be annoying. Clarity need not mean plainness. Sometimes the perfect word may be a difficult one but avoids verbosity or imprecision in the communication.

      • As to jargon, I couldn’t agree more. One person’s jargon is another person’s precise technical term. I don’t like the practice of calling something “jargon” because you don’t like the term or don’t undrstand it. Andrew’s always railing against “jargon” and then using terms like “MRP” Not only is “MRP” itself jargony, even if you know it stands for “multilevel regression and poststratification”, all three component nouns are themselves jargony in the sense that they’re not understandable from a basic understanding of English.

        As to simple English, I’d say it’s always a good idea when trying to communicate to be as simple as possible. The problem is that simple can be the enemy of precise.

        • I’m ok with jargon but I hate un-expanded acronyms. It’s easy to google up “heteroscedasticity” but harder to google MRP.

        • > I’m ok with jargon but I hate un-expanded acronyms.

          +1

          No, scratch that: +5. Acronyms are ubiquitous within DOD. For a long time I’d just nod along, pretend I understood, then go look up the ones I didn’t know after the meeting/telecon/etc. More recently, I’ve taken to asking. Probably at least 1/3 of the time the user doesn’t know the original phrase. The acronym has devoured it. Sort of like AARP.

        • > As to simple English, I’d say it’s always a good idea when trying to communicate to be as simple as possible. The problem is that simple can be the enemy of precise.

          One curious problem I ran into many years ago when I was working on quantum measurement stuff is that we would publish a paper in which we used the word “measurement” as jargon, with a very precise meaning, and we would have people outside the field try to argue with us, sometimes literally citing dictionary definitions of “measurement” without realizing that we were using the word in a very different way.

  3. The most irritating thing to me about statistical papers is the use of symbology that’s lightly explained in a spot that’s hard to find, e.g. finding theta_ijk on page 24 and eventually finding the reference to what theta means on page 7.

    Or, symbology that’s assumed because it’s common in that subfield, but perhaps that’s a subfield I’m not familiar with.

    My long term suggestion would be that every symbol such as theta_ijk be a hyperlink that points back to the point in the article where it is explained.

    Stat books can be worse. Unless you are taking/teaching a course, you are likely to be reading a later chapter — e.g. you might find VAR analysis discussed beginning on page 325, and don’t really want to do more than lightly skim the pages up to that point.

    • The most irritating thing to me is having to hunt through a long narrative to try to piece together the model. I think if everyone laid out the data first, then the parameters and probability model, it’d make everyone’s life a lot easier. Especially those of us trying to implement the models.

      • Yes! I very much agree with this!!

        Books can be annoying when *very important* assumptions are stated in chapter 3 and assumed for the rest of the book, but I think stats has more of a problem with clearly stating models (I still can’t work out what MRP is…) [or, maybe, how MR isn’t just a hierarchical model and what P is] than it does with “long-range” dependence.

        I’ve never really understood why this is so much more of a problem in stats than it is in other maths-y fields. Maybe it’s tradition (as Philip Larkin said “man hands on misery to man”), or the number of people who come from other backgrounds and don’t have the general training (rather learning stats methods in their application-specific context). The hardest questions for me to answer when giving short courses somewhere is “how can i do …”, when (unless you’re lucky) you have to extract what “…” is in terms of a model and work out an answer pretty much “on the fly”. It can go *very* wrong…

        • Mr is basically just a hierarchical (or multilevel) model. The P is poststratification, with the coefficients from each part of the model postratified to be a true representation of the population of interest.

          For instance, if you’re modeling something conditional on age, education and gender by state, then you’d fit a model with these included as categorical variables. You’d then postratify each of the categories for these variables onto census data so they’re weighted at their true proportions (actually each sub-category, so you’d have a census cell for 18-29 year old males in Maine with a college education, etc).

          Once you get down to it, the concept is pretty simple (unless I’m missing something, Andrew?).

  4. Andrew – don’t be discouraged. I am trying to get further and further towards your informal style (without sacrificing good old notational rigour), but it requires constant pushing against editors; actually, a “wandering style” is my personal tendency too (along with excessive [and sometimes nested] parentheses, and long sentences), but I’m never sure whether it is a good thing to be squeezed back to a dry, punchy style all the time, notwithstanding the desire to be widely read and understood (whatever the reader’s mother tongue). I am what I am.

    While we’re on zbicyclist’s point, it seems to me that everyone used to have a section to explain all the notation up front, and that seems to have gone out of fashion, though it’s really helpful. For example, I like the Hoffman-Gelman NUTS paper, but you don’t fully explain detailed balance until page 11 of 12 (unless I missed it). I haven’t known such a nail-biting ending since Who Shot JR.

  5. > … a more stilted style with lots of notation and many stops to make precise definitions, can be more readable for the less-than-expert audience.

    When I’m learning a new area I find that to be true.

  6. Andrew:

    You’re a good writer and do a lot of outreach. That said, a little explication can be very useful and make the medicine go down smoother. Don’t blather, but use a little text. For the expert that already knows it, he really won’t mind. for the semi-expert, to people from other fields, to newbies, to those less smart…it is all beneficial. Even the reader who already knows something, won’t mind a little bit of reading stuff he understands. ;)

    See also: http://courses.media.mit.edu/2010spring/mas111/NASA-64-sp7010.pdf (clear technical writing will get you laid and destroy your rivals!)

    • Nony:

      I don’t disagree with your general advice but in this case the request was not for more explication in the form of text. On the contrary, they wanted a more logical mathematical development. They wanted less words, more symbols. Which I can believe was good advice in this case.

  7. “a more stilted style with lots of notation and many stops to make precise definitions, can be more readable for the less-than-expert audience”

    even for the expert audiences, i think this can be quite useful. eg, as news ideas are developed, it may take some time before terms have precise meaning. and see bob’s comment about experts who are trying to implement models. experts also may attach different meaning to the same term, even terms that have been around forever.

    as you have pointed out many times, communication is hard.

Comments are closed.