Text Grading

The key thing with reading is to have things that you can manage easily. The target is being able to correctly recognise at least 96% of the words, and preferably 98%. Anything lower than this, and you will have to keep looking up words – although with aids such as IT dictionaries, this can be relatively quick and painless.

This is relatively simple for you to check. The best way is just to read a few pages and keep track of how many words you don’t know. Anything greater than 1 in 20 will be a poor match.

Sometimes you may want to try something that is (just) a bit harder, particularly if you are really interested in it. This is ok, because you’ll find that authors tend to use the same words and after the first chapter or so, a book will get easier. By the last chapter, you might wonder why you found it so hard at the beginning!

Another approach is to analyse the text for readability, so that you can match it with your level of attainments. There are various formulae for readability, which are typically based on the number of syllables in words and the length of sentences. I’ve used the English versions as they strike me as having reasonable validity/accuracy and are easily available (see below).

One of these is built into Microsoft WORD. You activate it by going to ‘File’, then ‘Options’, then ‘Proofing’, then tick the ‘Show Readability Statistics’. Now, when you select ‘Review’ and ‘Spelling and Grammar’, you will also get a Flesch-Kincaid Grade Level at the end of the mini report. You can convert this to an approximate Reading Age, using the table below:

This does however, depend on being able to get some text into WORD. This can be difficult if you are for instance using the ‘Look Inside’ function in Amazon as you can’t use this to cut and paste. This means you have to type in a substantial amount of test.

An alternative is to use the Fry Readability Graph (Link). This involves counting syllables and words, then looking the grade level up on a graph.

To make this easier, I have created a little program to do the calculations for you (download below). You read the text to yourself, and as you do this, tap the ‘N’ key on the keyboard for each syllable, and the ‘M’ key for the end of each word (you’ll see the totals go up as you do this). You tap ‘.’ at the end of each sentence, and the program will display the reading age.

As an example, doing this with a French reading book Folie d'Ours, gives a reading age of 9 yrs. Having done a lot of work with reading over the years, this seems about right to me. If you check the CEF, this means you’d need to be making good progress above level B1 to manage this book at a high success level.

Remember that the reading age will be for a French student. This particular book is part of a series for French adolescents and upwards, who are behind with their literacy. It is one of the easiest books at Niveau (level) 1. The series aims to be at a reasonable interest level to keep the attention of the target French readers.

Using English Readability Measures with French Text

The Flesch formula has been modified to work with the French language (Kandel and Moles, 1958) simply by reducing the weighting which is given to the syllabic word length. This adaptation is not that large however, and using the English formula with French text seems unlikely to alter absolute figures that much (at most, it will inflate them somewhat). There is also some evidence that the formula in Microsoft WORD tends to score lower than other formulas, so there could be a partial cancelling-out of any inaccuracy. It also seems very likely that as both the English and French formulae are so similar, scores will retain their relative values.

Despite this, using simple formulas with any reading material (and particularly with foreign language materials), is always bound to be a somewhat crude approach (Heydari & Riazi, 2012). It is therefore probably best suited for a general, initial indication, which should be refined by actual use. There are some interesting developments (Francoise & Fairon, 2012) which use a larger number of more sophisticated variables and are the result of AI training on samples with established levels of difficulty. These give significantly greater accuracy, but unfortunately, are not yet freely available.

For what it’s worth, when I’ve compared the Fry and the Microsoft WORD (Flesch-Kincaid) formulas on the same samples of French text, they have given quite similar values. They also seem reasonably valid as they correlate well with reading comprehension - at up to about .9 (Link).

Text Grading

Matching Reading to your level