What is Classic Style? A Primer for Social Scientists

This quarter, I'm teaching a course called Writing About Thinking. The course got a soft launch a couple of years back, when I taught it as an undergraduate seminar at The University of Miami. Now that I'm at UCSD, I am teaching a more advanced version of the course to a very nice group of our PhD students. The course is based on a simple premise: Writing about thinking, which every psychologist must do, is hard, but it's possible to get better at it by first thinking about thinking. The course, therefore, involves excursions into psychological research on communication, cooperation, memory, syntax, argumentation, and, of course, style.

One of the books we're reading is Francis-Noël Thomas and Mark Turner's little book Clear and Simple as the Truth, in which they explicate a style of writing they call Classic Style. It's an intentionally coy, playful little book that teaches as much about Classic Style by showing what Classic Style is as by telling what Class Style is.

The Classic Style, as Thomas and Turner lay it out, involves several guiding principles. Here are eight principles that I think are among the most important. 

(1) It is based on the conceit that it is possible to say things about the world that are true, and that it is the writer's job to point to these things. 

(2) It assumes a writer that "takes the pose of full knowledge," and is competent to explain everything the reader needs to know to understand the subject.

(3) It rests on the gambit that the reader is no less intelligent than the writer. The only difference between the writer and the reader is that the writer happens to know something that the reader doesn't. The reader is perfectly competent to acquire this truth.

(4) It relies on a writer who is confident in her own abilities. She resists the temptation to argue for the importance of her subject matter, she abstains from complaining about how hard writing is or how hard-won her insights are, and she avoids self-reflection and rumination. The classic-style writer hides her effort, but because she exerted herself so mightily in advance, the end product of her effort appears effortless, as if it could have been written in no other way.

Here, Thomas and Turner convey this idea in what I regard as a triumph of Classic Style:

The classic writer is not like a television cook showing you how to mix mustard and balsamic vinegar. He is like a chef whose work is presented to you at table but whose labor you are never allowed to see, a labor the chef certainly does not expect you to share. There are no salt and pepper shakers on your table.

(5) Because the writer and the reader are intellectual equals, and because the writer is pointing at true things in the world, the two of them can have a conversation. Classic-style writing, when read aloud, sounds like one person talking to another, like a really good tour guide when you're visiting a museum or a foreign city.

(6) Sentences and paragraphs go somewhere. Each unit of meaning, Thomas and Turner write, "has a clear direction and goal." The payoff comes at the end of the sentence or passage, but to get to that payoff, the reader must follow a path, made of several steps, along which the writer is leading him.

(7) With all of its reality and pointing and seeing and touring , the Classic Style relies on the same image schema we use to interact with the physical world. Ideas have weight; they develop. Arguments go somewhere. We follow lines of reasoning. By relying on physical imagery, Classic Style is able to depend on some of the cognitive processes that use so successfully to navigate the real world.

(8) No topic is so complex that it cannot be explained.

The first part of Clear and Simple as the Truth is the exposition. The second part is "The Museum," consisting of a variety of classic-style passages, along with Thomas and Turner's analyses of them. The Museum is well worth a visit, but its examples are not as helpful for social scientists as examples from actual social science might be. I was therefore very pleased to discover yeseterday that one of my favorite articles in Psychology--Denny Borsboom, Gideon Mellenbergh, and Jaap van Heerden's The Concept of Validity (which I am currently re-reading for a paper I'm working on, and which I blogged about earlier here)--is an exemplar of classic style.

At the opening of the paper, you find this marvel:

Please take a slip of paper and write down your definition of the term construct
validity. Now, take the classic article of Cronbach and Meehl (1955), who invented the concept, and a more recent authoritative article on validity, for instance that of Messick (1989), and check whether you recognize your definition in these works. You are likely to fail. The odds are that you have written down something like “construct validity is about the question of whether a test measures what it should measure.” If you have read the articles in question carefully, you have realized that they do not conceptualize validity like you do. They are not about a property of tests but about a property of test score interpretations. They are not about the simple, factual question of whether a test measures an attribute but about the complex question of whether test score interpretations are consistent with a nomological network involving theoretical and observational terms (Cronbach & Meehl, 1955) or with an even more complicated system of theoretical rationales, empirical data, and social consequences of testing (Messick, 1989).

Who in psychology opens a paper like that? Too few of us.

A little further along, there's this:

The argument to be presented is exceedingly simple; so simple, in fact, that it articulates an account of validity that may seem almost trivial. It is as follows. If something does not exist, then one cannot measure it. If it exists but does not causally produce variations in the outcomes of the measurement procedure, then one is either measuring nothing at all or something different altogether. Thus, a test is valid for measuring an attribute if and only if (a) the attribute exists and (b) variations in the attribute causally produce variations in the outcomes of the measurement procedure. The general idea is based on the causal theory of measurement (e.g., Trout, 1999).

And then this: 

That the position taken here is so at variance with the existing conception in the literature is largely because in defining validity, we have reversed the order of reasoning. Instead of focusing on accepted epistemological processes and trying to fit in existing test practices, we start with the ontological claim and derive the adequacy of epistemological practices only in virtue of its truth. This means that the central point in validity is one of reference: The attribute to which the psychologist refers must exist in reality; otherwise, the test cannot possibly be valid for measuring that attribute. This does not imply that the attribute cannot change over time or that that psychological attributes are unchanging essences (cf. Kagan, 1988). It does imply that to construe theoretical terms as referential requires a realist position about the phenomena to which such terms refer. Thus, measurement is considered to involve realism about the measured attribute. This is because we cannot see how the sentences Test X measures the attitude toward nuclear energy and Attitudes do not exist can both be true. If you agree with us in this, then you are in disagreement with some very powerful philosophical movements that have shaped validity theory to a large extent.

In spite of their scholarly apparatus (such as citations in parentheses, maybe slightly too much meta-discourse), these passages bear all of the marks of Classic Style. No hedging, no apologizing, no showing off, plenty of grounding in spatial imagery (with its taking of positions, reversings of causal orderings, and so on), and a confidence that even a very complicated idea can be expressed in plain English to any reader who is willing to take some time out to "talk" with an expert about it.

As a bonus, the paper itself pushes what I regard as a classic-style view of science, measurement, and validity. On Borsboom and colleagues' view of measurement, things either exist or they don't, and it's only the things that exist that can be measured. And a measure has validity as a measure of that invisible entity (intelligence, self-esteem, reading comprehension, or whatever) only if that invisible entity is real and if that entity is involved in the chain of causal processes that lead to the representations that we take to be "measurements." Reality is out there, validity is much simpler than you think, and when we do measurement, we take a sounding of real things. I love the fit here between the the writers' medium and their message: Borsboom and colleagues help their case along through clear, confident, conversational writing that asks the reader to no more than look where the writer is pointing.