Colette Marchand

A readability score is a prompt to look at a paragraph, and nothing more. Some count syllables because syllable count correlates with difficulty. Some count sentence length, which is the easiest thing to measure and the least reliable. Some count the proportion of words that are not common. That last one depends entirely on which word list is used, so two tools with the same formula and different lists give different numbers on identical text. Syllable counting is the weak link in the widely used formulas, because counting syllables in English is ambiguous at the end of a word and in loanwords. Every formula that uses it carries that error forward. Formula choice changes the grade by several years on the same paragraph, so a score means nothing without the formula named. I put that on every page. What the score cannot see is the main limitation. It has no notion of whether the sentence is a heading, a caption, a table cell or a paragraph of explanation, and it treats all of them alike. A technical page scores badly and reads well to the audience that needs it. What I recommend instead is reading the paragraph aloud and checking whether anything needs unpicking, which is slower and considerably more accurate. Where a score has to be reported, I report the formula, the word list and the sentence it scored, so the number can be reproduced or dismissed rather than taken on trust. A score also rewards short sentences mechanically, which pushes writing towards choppy prose that scores well and reads badly. That is the clearest demonstration that the number is a rough instrument.