Writing Readably Doesn’t Mean You’re Stupid

Yesterday, I celebrated long writing. Today, I’m going to demonize it.

Fog Index Algorithm

The Fog Index Algorithm

My point yesterday was that long-form writing, such as a book, often engages readers effectively. But when it comes to words and sentences, length can be a reader’s enemy.

I hadn’t planned on this follow-up today. What spurred me on was a column by Meghan Daum in this morning’s Los Angeles Times. Headlined “Speaking down to Americans,” it discusses a recent study by the Sunlight Foundation analyzing the grade level of speeches by members of Congress. The analytical tool used by the study is the Flesch-Kincaid readability test, which, as Daum puts it, associates long words and sentences with higher grade levels.

The misleading implication of those grade levels is that the higher your writing scores, the smarter you are. As a result, much of the press coverage of this study has interpreted it to mean that by using shorter words and sentences, politicians are either becoming dumber or speaking down to Americans.

And that, I first thought, was the gist of Daum’s column, which a possibly overworked copy editor had subtitled as follows:

“Researchers have found that politicians’ rhetorical skills have taken a dive since 2005, when they were at an 11th-grade level.”

It turns out, though, that Daum doesn’t really think that. In fact, she argues that the foundation’s data are meaningless: such studies, she says, “measure not meaning or depth of thought but characters and syllables and average number of words per sentence.” And she criticizes both those who fault legislators for speaking plainly and the foundation for doing the analysis:

“It seems disingenuous to mourn the passing of a time when you needed a 17th-grade education to understand what your president was saying. Not as disingenuous, however, as using a hopelessly reductive study to assail one’s political enemies in the most predictable way.”

So Daum seems to be in favor of readability, but blithely tosses aside the Flesch-Kincaid test as useless. Similarly, The Atlantic’s Eric Randall, while in favor of short words and sentences, suggests that the test promotes verbosity: “Flesch-Kincaid rewards long words and winding sentences, but clarity rewards the opposite.”

Well, no. Flesch-Kincaid doesn’t reward long words. It penalizes them. The entire point of using the test is not to raise the grade level of your writing, but to lower it.

The confusion is the inevitable result of the decision to tie writing skills to reading comprehension grade levels. I’m not quite sure whom to blame for this fatal mistake. It doesn’t seem to have been readability maven Rudolf Flesch, whose original readability index used a 100-point scale (the higher the score, the more readable the prose). The error may have been introduced by Robert Gunning, inventor (in 1952) of the Fog Index. The Flesch-Kincaid test, developed in 1975, followed Gunning in using grade levels to assess writing.

Whatever the source of the grade-level equivalence, it’s a problem. When I used to lead in-house seminars on readability for editors of technical magazines, descriptions of the grading system always backfired at first.

Our readers have PhD’s, they’d think. We’re smart too. We can write at grade 20! Bring on the 50-word sentences and sesquipedalian locutions!

The real goal, of course, is just the opposite. If you can write at grade-level 5, you should—not because you want to reach fifth-graders, but because even your doctoral readers will likely find such prose lucid, appealing, and memorable.

The same caveat applies today as yesterday, however: It all depends. Talented writers can reel off long, polysyllabic sentences to brilliant effect; inept writers can produce wretched short ones.

The smartest writers combine both approaches. It isn’t the length of any one sentence that matters—it’s the average over the course of several paragraphs or pages. Long words and sentences are not proscribed by Flesch-Kincaid. They just need  balancing with short ones.

The Flesch-Kincaid test or the Fog Index, used circumspectly, can help improve your writing. But remember this crucial point: you’re testing for readability—not for your intelligence.

Journalism, Professionalism, and the Turing Test

What’s the way forward for journalists? Doubling down on the traditional ideals of objectivity and impartiality? Embracing the subjective, personality-driven approach of social media? Or is there some uncertain, ill-defined middle way?

Turing Test By Bilby (Own work) [Public domain], via Wikimedia CommonsThose are some of the questions being raised recently by a number of new-media observers, most notably GigaOm’s Mathew Ingram, who’s lately been rolling out one must-read blog post after another.

The problem with traditional news is that traditional journalists are increasingly unnecessary to produce it. Robot reporters are cornering the market on facts, as companies like Narrative Science and Automated Insights perfect the science of teaching software to turn data into news stories.

If basic news becomes a fungible commodity, one obvious alternative for journalists is what Ben Huh says great reporters already do: convey not simply the facts, but their subjective emotions about those facts.  But this, he says, is a “very, very dangerous” approach.

That’s one lesson that could be drawn from reporter Anne Sutherland’s recent suspension from the Montreal Gazette for remarks she made on Twitter. Covering a “nearly naked” protest by university students, she tweeted a number of photos of the protestors accompanied by “snarky” comments about their physiques. Neither her Twitter followers nor her employers found it amusing.

Writing about Ben Huh, GigaOm’s Ingram says that “in order to be effective, journalism needs to be personal.” But doesn’t Sutherland’s seemingly personal reaction to the protestors prove the opposite, and that the dangers of being personal outweigh the benefits?

I think not. I don’t know her, of course, but I’d guess the problem isn’t that she was being human or that she was being too personal. Rather, she was responding to the wrong instincts and emotions.  She was there as a journalist, but reacting as an average, and thoughtless, bystander.

In a post written before Sutherland’s misstep, Steve Buttry addressed a similar issue in explaining “how to respond to staff members who were using crude language and behaving unprofessionally on Twitter.” On social media, he says, journalists must be personable, yes, but also professional:

“A professional journalist using Twitter should behave professionally. Your profile should identify you as a journalist with your news organization. You should behave accordingly.”

I don’t disagree. But I wonder if professionalism is sufficient. The problem for me is that professionalism is more shield than guiding light. Too often, it is just a way of doing what won’t get you fired.

To succeed in a personal medium, you ultimately need a personal standard. The preeminent question to ask yourself now may not be Is this a professional and objective statement of the facts? but rather Is this my best, most honest, and most personally true assessment of those facts?

This might not seem like the appropriate corrective to the all-too-personal Sutherland. But I suspect her reactions were not truly personal. They sound, rather, like received views, the trite and formulaic reactions not of a person, but of a type of person. It is a behavioral response that could be easily programmed into a Narrative Science algorithm: If see hairy body, then tweet “Ewww.”

In gauging how to handle social media, maybe what journalists need is not so much a standard of professionalism as a kind of Turing test. That is, could what you’re writing be produced by a computer imitating a human reporter?

The test is not whether the content is dryly factual or snarkily silly, superbly impartial or grossly biased. Those traits are easy to replicate. Instead, the test should be whether the prose is truly personal. Does it reflect a real consciousness struggling to find the truth, or an automaton juggling ones and zeroes?

Such a test can never be very precise. But journalism, whether conducted in traditional or social media, would be the better for it.

Swabbing the Decks of the Titanic: Why You Should Learn Programming

Image of the Titanic sinkingLast week journalism professor Matt Waite wrote a blog post worrying about the typical defeatist reaction of journalism students when faced with a coding challenge, whether in HTML, JavaScript, or other language: “I can’t do this,” they tell him. “This is impossible. I’ll never get this.” When I tweeted a link to the article, I wrote “”Journos: If you fear coding, you fear the future.”

That prompted a response from a practicing trade journalist and former colleague, who asked “I can see why knowing things like HTML and CSS can be helpful but do most journos need more than that?”

His question wasn’t one I could answer easily on Twitter, because for me, at least, there’s no clear and simple answer. Does a typical mid-career editor on a print publication today need to learn software programming? From that perspective, it’s hard to come up with a compelling argument for it, though I’ve certainly tried.

Waite’s blog post, however, wasn’t about veteran editors but about the journalists of the future. Those journalists, he says, must be able to “construct, manipulate, and advance digital distribution of content and information.” If they don’t have a positive, can-do attitude towards programming, they won’t succeed.

Does this mean that most journalists will need to be experts in one or more specific programming languages? I don’t think so. My guess is that while the ranks of programmer-journalists like Jonathan Stray, Michelle Minkoff, and Lisa Williams will continue to swell, most journalists won’t become similarly hyphenated. There will always be some degree of specialization in journalism. But in the new-media era, to be a good journalist, to master your craft, you must at the very least learn enough about programming to understand it.

As my former colleague implied, even for veteran journalists there’s a benefit to understanding code like HTML and CSS if they do any work online. There’s nothing new about needing to comprehend the means of your production in order to perfect your message.

As an analog example, consider how easily in the traditional print world you can lose control of your editorial content if you don’t understand at least the basics of what your art director and your production manager do. The decisions they make can strongly influence your content, and if you don’t know what to ask for and to explain why you’re asking, your content will suffer.

Likewise, in the digital medium, studying what’s under the hood gives you greater flexibility in presenting and distributing your content. If you work with web developers and programmers, you’ll have a better idea of what to ask for, and better chances of getting it. And if it’s just you and WordPress, you’ll be better able to customize the code yourself to get the result you want.

But there’s another reason that journalists of the future should want to get their hands dirty in code. The value of learning how to program is not just in better understanding their jobs, but also in better understanding the world they write about. As Roland Legrand puts it,

“Every year, the digital universe around us becomes deeper and more complex. Companies, governments, organizations and individuals are constantly putting more data online: Text, videos, audio files, animations, statistics, news reports, chatter on social networks. . . . Can professional communicators such as journalists really do their job without learning how the digital world works?”

This trend toward digitization in all human endeavors has given rise to another journalistic specialty, computer-assisted reporting or data journalism. Though it may never account for the bulk of what most journalists do, knowing how to extract, manipulate, and present data will be an increasingly valuable skill. Even today, it’s possible that you’re sitting on a rich lode of data that, if you just knew a little programming, you could help mine.

If you are well advanced in your career as a journalist, maybe you don’t need to learn anything about programming. You’re set, right? But that’s probably what the crew thought as they swabbed the decks and polished the brightwork of the Titanic.

Why not play it safe? Your job as a journalist may not require you to have any familiarity with programming today. But one day, perhaps sooner than you think, it will. Why not prepare yourself by finding out more about data journalism,  by learning some programming basics at as site like Codecademy, or by joining a cross-disciplinary group like Hacks/Hackers?

As I’ve noted recently on this blog, some journalists are worried that their role will one day be eclipsed by software. If you don’t want to become an algorithm’s slave, you have only one choice. You must become its master.