Here’s a post from Anne R Allen, which I saw via The Passive Voice Blog. The post is about WordPress, but here’s the part that caught my attention:


With the Yoast plug-in, you don’t get a list of rules. You discover each one when the elves give you a red, amber or green light on your copy. If you get a red or amber light, you must scroll down and find out what you’ve done “wrong” according to the Yoast rules.

Here are the things the readability elves will ding you on:


They give you an automatic red light if you start three sentences in a row with the same word. So never quote Charles Dickens “It was the best of times. It was the worst of times. It was the age of wisdom…”

Ah, I said. One of THOSE “readability” algorithms. Yes, indeed. How very helpful. Other things that the “readability elves” dislike: long sentences, long paragraphs, more than 300 words after a subheading, the passive voice, words with more than four syllables.

Sometimes I’d like to go through the world of words and find examples of things that absolutely meet “readability” standards in every way, without actually being readable. I actually have a specific example in mind. Long, long ago, when I was TAing a basic Bio class of some sort, the class switched one semester from the pretty good textbook we’d been using for years — I’ll call that Textbook A — to a much less good textbook, Textbook B. So I went to the TA coordinator for this class and asked about that, and without a word she handed me the newest edition of Textbook A.

It had been revised according to current readability standards. This was a Biology textbook that now had very few words over four syllables and was trying to get by at a seventh grade reading level and whatever else was mandated by the readability standards of the time. It was therefore utterly useless. A complete waste of paper and ink. No one could have learned anything from it.

“Ah,” I said. “Textbook B looks fine, then.” And that was it for Textbook A. I don’t know if a future edition ever repaired the damage.

Even a couple of decades later, this incident is what comes to mind when I hear about readability.

Anyway, the rest of the post is about getting Google to move your blog higher in search results and I’m sure that’s interesting, but it’s not what caught my eye.

Here, if you are interested, is a Readability tester. I imagine it’s somewhat different from the one Allen refers to, but it pops right up in a Google search, so I imagine the Google elves like it. I plugged a thousand words from TUYO into this tester and here’s what I got:

Flesch-Kincade Grade Level: 4.8 — I suspect this is because Ryo usually thinks and speaks in short, relatively simple sentences.

Gunning Fog Grade Level: 7.1 — I wonder if this is because the vocabulary is more advanced than fifth grade? I don’t know how the two grade level things make their decisions.

One hundred and two “issues” that should be addressed. A hundred and two! In a thousand words! Wow. Let me see. Lots of spelling queries, which yes, that’ll happen in a fantasy novel. This readability checker wants me to break up all long sentences. It appears to think anything over 25 words is “really long” and anything over 15 words is “long.” I’m tempted to paste in some sentences that are actually really long and see what it says.

Unsurprisingly, perhaps, this checker would like me to remove all adverbs. I do get weary of the constant adverbs-are-bad drumbeat. I wonder if any of you noticed that Ugaro basically have two intensifiers: very and extraordinarily. Ryo seldom if ever uses other intensifiers. The Lau have a more varied vocabulary and generally speak with longer sentences and so on.

I note that the checker thinks I did a good job avoiding cliches. Well, glad it’s happy about something.

Out of curiosity … I have Door Into Light here … let’s try 1000 words or so of that one … ah, the very different style certainly comes through.

Flesch-Kincade Grade Level: 7.3

Gunning Fog Grade Level: 9.5

Interesting, isn’t it? I knew the style was different and I specifically knew I was giving Ryo a distinctive voice with short, relatively sentences and more than usual repetitiveness of vocabulary, but wow, this is sure very different.

On the first page, this readability checker thinks that ALL BUT THREE sentences are “long” or “really long.” It identifies the word “unpredictable” as a “hard word.” (!) It does not recognize somewhat obscure words like “hewn” and “skirl” and tells me those are spelling mistakes. Good heavens, it thinks that if you start a sentence with “when,” the sentence is probably a fragment. I wonder what other dependent clauses it fails to recognize.

It’s moments like this that make me think we really should just drive a stake through the heart of all readability algorithms. To the extent anyone takes them seriously, these sorts of scores have GOT to drive reading ability and general text comprehension downward, while potentially producing textbooks that are completely useless.

Please Feel Free to Share:


10 thoughts on ““Readability””

  1. I am hard pressed to think of a disadvantage to axing all readability algorithms. I don’t think this is just my (admittedly highly literate) perspective.

  2. The more basic the language, the wider the audience. That’s the theory.
    When writing technical manuals for a worldwide audience, my team relied heavily on algorithms like these. Supposedly, non-native English readers cannot process them big words.*
    Many tech companies are moving toward publishing most of their manuals in English only, since most tech people have some familiarity with the language, albeit rudimentary.**

    * And yet, I have worked with many a non-native speaker whose written comprehension was quite advanced.
    **i.e. I would have been slapped down for using i.e., albeit or rudimentary

  3. Allan Shampine

    I agree with Craig. Splitting the world of writing roughly into three groups (leisure reading; technical writing for a technical audience; and technical writing for a lay audience), I can think of a possible use for algorithms in only the third situation – technical writing for a lay audience. Particularly an audience who may not want to be reading the material (e.g., judges, juries), in which case using words they don’t immediately know or getting them bogged down in complex language may result in a DNF on their part. But given the stakes in the third situation, I might be willing to look at the advice of an algorithm but I would never just apply it without exercising my own judgment.

  4. Allan, I think you’re right. Situations like producing technical manuals for a worldwide audience, like Evelyn, or as you say written materials for a captive audience like a jury, are the sole reasonable use of readability algorithms — and if I were a lawyer, I would insist on going over materials myself to make sure they were REALLY readable.

    Any algorithm is unbelievably and imo irretrievably flawed when it says words like “unpredictable” are “too hard.” Any algorithm that just counts syllables is never, ever going to be able to decide whether words are hard or not.

  5. What you want, even in the limited situations, is someone who actually knows what they’re doing. If you have that person going over the material anyway, I strongly doubt that an algorithm is even a useful tool.

    So my cynical side suspects it comes down to a CYA: the algorithm seems more objective and it leaves you without any person who has to take the blame for errors — especially if a committee decided to use it in the first place.

    I think I’m sticking with my initial “axe the whole thing” reaction.

  6. I expect you’re right. If a knowledgeable person has to supervise that closely, which seems inevitable, then what is even the point? Other than, yes, a pseudo-objectivity you can point to if that seems helpful for some reason.

  7. I find that things like shorter sentences and shorter paragraphs often make work emails more readable (and also easier to skim to determine if said email is relevant or urgent). Whenever I get to check through colleagues’ emails, I want to make sentences shorter, because I suspect that people are more likely to read, respond and remember if it’s clear and easy to read. So I can see how an algorithm could maybe make some — some! — helpful suggestions in this sort of situation.

    But I would hate to live in a world where everything was written like a practical work email. Ugh.

    And I don’t think algorithms are smart enough to properly work out whether words are “too hard” or not, because that depends so much on context. Is the possibly- unfamiliar word explained or does the text just assume the reader knows it? (A tendency to define unfamiliar terms is perhaps what fantasy novels and science textbooks have in common.)

    I’ve noticed that, when it comes to books for kids learning to read, the easiest books sometimes have longer words than the books in the next level up, because the simple, repetitive sentences and pictures mean those long words are very easy to predict. Books with longer sentences, less repetition and more complex concepts might use shorter nouns (eg. “cat” rather than “elephant”) but the text is still more challenging because the reader can’t predict, or rely on the pictures, as much.

  8. Herenya, shorter paragraphs are the one thing I think is most important for readability online, on a computer monitor, not just for emails but for, eg, blog posts.

    Those are good observations about early books for kids learning to read. I’m not familiar with the guided practice offered to those readers, but it makes perfect sense that longer words with explanatory pictures are easier than shorter words embedded in longer sentences with fewer pictures.

Leave a Comment

Your email address will not be published.

Scroll to Top