What makes writing more readable?

Writing text that can be understood by as many people as possible seems like an obvious best practice. But from news media to legal guidance to academic research, the way we write often creates barriers to who can read it. Plain language—a style of writing that uses simplified sentences, everyday vocabulary, and clear structure—aims to remove those barriers.

Plain language is useful for everyone, but especially for those who are often denied the opportunity to engage with and comment on public writing. This includes the 20% of the population with learning disabilities, a number of the more than 7 million people in the US with intellectual disabilities (ID), readers for whom English is not a first language and people with limited access to education, among others.

These audiences are routinely excluded from public dialogues, including dialogues about themselves. People with disabilities are also often excluded from writing or sharing their own stories first-hand due to lower vocabulary skills, learning differences, and intellectual disabilities. For example, throughout much of US history, people with ID have had decisions made on their behalf based on the presumption that they do not and cannot understand. This, on top of discriminatory attitudes and stigma, has led to infantilization, institutionalization and eugenic sterilization.

Additionally, there is a tendency to censor content for these audiences rather than explain it, which can contribute to continued disparities, like the higher rate at which people with ID experience sexual violence than nondisabled people.

The benefits of plain language aren’t limited to universally challenging texts like legal documents and tax forms. Even everyday writing, like news articles, can still pose a barrier for some readers.

Let’s walk through how Rebecca, an expert in plain language, translates a text to be more readable. We'll use an excerpt from her translation of a ProPublica article by Amy Silverman in the following example.

As more people have recognized the practical value of plain language, researchers have sought to quantify the “plainness” of writing through readability formulas—mathematical models that assign numerical scores to text, indicating how understandable they are.

Though most readability formulas were designed to offer rough difficulty estimates for specific groups of readers, their usage varies greatly, with the Agency for Healthcare Research and Quality warning that “these formulas are often interpreted and used in ways that go well beyond what they measure.”

Moreover, the simplicity of readability checkers has enabled their widespread adoption. Military engineers use them to help write technical documents. Governments and doctors use them to guide communication for a general audience. Schools and textbook manufacturers use them to tailor reading assignments to particular grade levels and students.

To better understand how readability scores work—and how they can fail—let’s look at three representative examples.

The Flesch-Kincaid Grade level formula looks, in part, at syllable count, based on the idea that words with fewer syllables are easier to understand.

The Dale-Chall Readability Formula considers the proportion of difficult words, where it deems a word “difficult” if it is not on a list of 3,000 words familiar to fourth-grade students.

One risk in the use of vocabulary lists is that the vocabulary of the readers surveyed to create them may not match the vocabulary of the intended audience. The original Dale-Chall list of “familiar words” was compiled in 1948 through a survey of U.S. fourth-graders, and even the most recent update to the list in 1995 retains obsolete words like “Negro” and “homely” while omitting “computer.”

More recently, US schools and textbook manufacturers have standardized their curricula on the Lexile Framework for Reading, a set of readability algorithms and vocabulary lists designed to automatically match students with appropriately difficult books. Publishers apply Lexile to their texts to rate their difficulty. A Lexile score of 210 indicates an easy-to-read book, while a score of 1730 indicates a very challenging one. A reading comprehension test assigns a corresponding reading score to each student, after which teachers pair students with texts that have comparable Lexile scores.

The approach sounds simple enough, but critics have pointed out absurdities in Lexile's results. According to Lexile, The Grapes of Wrath (Lexile score: 680) is easier to understand than the Nancy Drew mystery Nancy's Mysterious Letter (score: 720), but neither of these is as challenging as The Library Mouse (score: 860), a 32-page illustrated children's book.

How exactly are Lexile scores calculated? Unfortunately for us, the Lexile Framework is the intellectual property of MetaMetrics, the private company that created it, so we can only guess at the secret recipe, but it's a pretty good bet that Lexile scores depend on a mixture of the same factors used in Flesch–Kincaid and other open-source readability measures.

Formulas based on surface level text properties and word frequency have clear limitations. None of them consider how well organized the information is, or whether the sentences and paragraphs are coherent. None consider the role of grammatical tense. None account for the explanation of acronyms and jargon. None would balk at Jack Torrance's rambling and meaningless draft in The Shining, endlessly repeating “All work and no play makes Jack a dull boy.”

But proprietary tech like Lexile has some of the most disconcerting aspects of both worlds. As flawed as Flesch-Kincaid or Dale-Chall, but opaque and unexplainable. The main benefit of the F-K and D-C formulas (and other simple algorithms like Gunning-Fog and SMOG) is their transparency. A broken system locked in a black box can't even offer this.