Reading level – accessibility for web writers

The accessibility guidelines suggest we write content that requires only a lower secondary level education to read. Where that’s not possible, we should provide supplementary content (visuals, for example), or an easy-to-read summary.

Benefits of more readable content

Unfortunately a lot of web content is unnecessarily dense and complex. This guideline aims to help people with learning disabilities like dyslexia. However, there is evidence that writing in a more readable style benefits everyone, including the organisation that publishes the content.

Readability testing tools

The quickest way to measure readability is with readability testing tools. They were first developed in the 1920s and claim to measure readability, but their use is controversial. They only measure two, or sometimes three aspects of content: word length (in syllables), sentence length, and some include a list of uncommon words.

While you may need to do readability testing with these tools during an accessibility audit, you should understand the limitations. See ‘problems with readability scores’ below.

Two readability testing tools are built into Microsoft Word: Flesch Reading Ease and Flesch-Kincaid Grade Level. You’ll need to enable them and then run the spell-checker. When the spell-checker has finished, Word shows the readability statistics.

Example readability test

To meet the accessibility guidelines, content needs a Flesch Reading Ease score over 50, or a Flesch-Kincaid Grade Level below grade 10. Here’s an example, using content from the Centrelink (an Australian government agency) website.

Interpreter and translation services

To help customers understand Centrelink services, Centrelink provides interpreters at no cost to customers.

Where necessary to support a claim, Centrelink also provides a free translation service for customer documents.

Interpreters contracted by Centrelink are covered by confidentiality provisions and a Code of Ethics, which means customers can be reassured that any information learned through an interview conducted by an interpreter will remain confidential.

Bilingual staff may be available in some Centrelink Customer Service Centres to help with brief customer enquiries. If an interpreter is not immediately available, Centrelink staff may use a telephone interpreter service to assist customers.

I copied this text into Word and ran its readability tests. This showed a Flesch Reading Ease score of 16.6 (too low) and a Flesch-Kincaid Grade Level of 15.8 (too high).

Screenshot showing readability statistics from MS Word, discussed above
Readability test results

Retest after removing proper nouns

The guidelines let you remove proper nouns from content before testing because it can be hard to find shorter words to replace them. So I retested the content, using ‘X’ to replace the proper nouns and titles (so I was still testing sentences of the same length). The results: a Flesch Reading Ease score of 40.6 and a Flesch-Kincaid Grade Level of 12.5. The reading level is still too high.

Screenshot showing readability statistics for retested content, discussed above
Readability test results: retested content

Rewrite with shorter words and sentences

To lower the reading level, we would need to rewrite the content using shorter words and sentences. The example below does this and has a Flesch Reading Ease score of 58.8 and a Flesch-Kincaid Grade Level of 8.3.

Free interpreter and translation services

To help you understand our services, we provide free interpreters. We can also translate your documents if you need to include them with a claim.

If our interpreters are not available we may use telephone interpreters. Some of our staff speak other languages and may also be able to help with short enquiries.

Our interpreters follow a Code of Ethics and must keep your information private.

Readability statistics, discussed above
Readability test results: rewritten content

Numerous problems with readability tools

Readability testing is easy to do, but has significant weaknesses:

  • Scores for the same text differ when using different formulas
  • All words of the same length are treated equally, yet ‘agree’ is probably less difficult than ‘concur’
  • Shorter words are treated as easier words, but ‘abide’ is probably more difficult than ‘tolerate’
  • Shorter sentences are always considered easier to read. However, a sentence of 20 words is not necessarily easier to understand than one of 25 words
  • Sentence structure and style are not considered. Use of passive voice, double negatives, nominalisations, noun strings, idioms and other writing problems are not factored into the formulas
  • The length, structure and layout of the content are ignored. Long, poorly organised content with rambling paragraphs and few headings is likely to be less readable than well designed content
  • The use of graphics to support or present content cannot be measured by readability formulas
  • The degree of difficulty of certain concepts or topics is not given any weight
  • Readers’ interest and motivation are not considered, nor is their existing knowledge of the topic
  • It’s difficult to get meaningful results from testing tools when you use a lot of dot points or tables, which is often the case on the web.

Despite these problems, some argue that readability testing helps them identify or get agreement that content needs to be rewritten or tested with users. It’s certainly true that a high grade-level means content won’t be easy to read. However, the reverse is not true.

If you use a readability testing tool, be aware of its weaknesses.

A better approach

Use plain language techniques

A better way of making content easier to read is to use plain language. Plain language is more than choosing short, simple words.  The guidelines (linked below) include a range of writing and design techniques, and also require you to think about your target audience. Research shows they make a huge difference.

Test content with users

The best way to check if your content is readable is to test it with users.


Next article in this series