electronicwoft
commented
3 weeks ago this is just swapping one for the other, isn't it? Should the W3 be tacitly endorsing any particular test for readability? Some links to contemporary research on evaluating readability for different contexts rather than a particular test might be preferrable to linking to an actual test?
Hi there,
In the Examples section of the Understanding document for SC 3.1.5, there is a reference to the Flesch reading ease score and another to the Flesch-Kincaid grade level. There's also a link at the bottom of the page, albeit with a disclaimer, to the Wikipedia pages on the Flesch–Kincaid readability tests.
In context, even with the disclaimer, it sounds like W3C is throwing its weight behind these formulas as being accurate depictions of the reading ease or grade level of a text. However, it's not clear that they are. If you look at the formulas (available at Wikipedia), they are very mechanistic, taking into consideration only the average number of syllables per word and the average number of words per sentence. These are certainly correlated with reading ease and grade level, but one important factor they don't take into account is how familiar the vocabulary is to the target audience.
The linguistics blog Language Log has looked at the Flesch and Flesch-Kincaid metrics on several occasions and consistently finds them wanting. At one sample post, a commenter concisely notes that 'Flesch-Kincaid is the test that treats "tort" as a simple word and "hippopotamus" as a difficult one.'
Dr. Janice ("Ginny") Redish has similar criticisms. Her post Readability Formulas: 7 Reasons to Avoid Them and What to Do Instead effectively lays out the issues with readability formulas, and, yes, what to do instead.
So I'd like to propose that the Understanding document be updated in either or both of the following ways:
If you'd like to see what that might look like, I could put together a pull request for review.