Response: Why Readability Testing is not Enough

In September 2004 I had an essay published on usabilitynews.com, on why readability testing shouldn't solely be used to gauge how understandable your site's content is.

The recent press coverage of the Bath University research paper "Readability Assessment of British Internet Information Resources on Diabetes Mellitus Targeting Laypersons" (including UN's own story: NHS Direct Website too complicated for Diabetes Sufferers) has raised interesting questions about some of the methodologies used to measure users' experience on the Web.

The research assessed the readability of 20 information resources on diabetes using Microsoft Word's automated testing feature, which gives a Flesch-Kincaid Grade (rates text on a US grade-school level; the lower the score, the easier it is to understand the document.) and Flesch Reading Ease level (rates text on a 100-point scale; the higher the score, the easier it is to understand the document) score. This was then compared against the 'average' reading age of the population giving the conclusion that a majority of the sites tested were 'too complicated' for their readers.

On the face of it, this conclusion and the methodology used is fine, but due to the indiscriminate nature of automated testing tools, it doesn't present the entire picture and, at worst, can give the impression that the users of these websites can't understand the content at all, which may not be the case.

• A health website which uses (and in most cases defines) medical terminology would score high when evaluated using 'Readability formulae that estimate the reading level of a document based on the words that are used and the lengths of sentences'.
• The readers of such information would most likely have a good understanding of the medical environment surrounding their condition, which would make the articles easier to understand.

It's this lack of context in the research that gives the biggest cause for concern, as some of the sites tested may automatically try to lower the reading age of their articles. They would then run the risk of alienating their actual users, who might not need a simpler article!

I would forgive anyone reading this for thinking that I'm dismissing the research out of hand, but in fact I welcome any research that can help us to provide a better service to our users. This research has done that because it’s made the people in the healthcare community look at how readable their websites are. But the key message is, and should always be, know your users and test against them not just by using automated tools.

(In case you're wondering, the Flesch-Kincaid Grade and Flesch Reading Ease level of this article is 60.5 and 11.1. I hope you've managed to understand it okay!)