According to the master's thesis there are no significant main effects:
http://www.ilo.gw.utwente.nl/ilo/attachments/032_Masterthesi...
I could imagine that this works well for specific kinds of dyslexia.
I could imagine that this works well for specific kinds of dyslexia.
I read the tables in the thesis. There are 12 comparations between the Arial and the Dyslexie fonts, but only 2 of then have p<.1 and none of them has p<.05. So it is very difficult to know if the differences between the fonts are real or come from random noise.
They made a font that they think is better for people with dyslexia, they tried it, but they found almost no difference. This work is only a master thesis, so perhaps they are planning to continue working on this. But for now it is not a "cure" for dyslexia.
---
I think that the problem is that they didn’t do enough "beta"-testing. (Perhaps they did it, but it is not in the thesis.) With infinity time, money and resources, they should:
* Find a small group people with dyslexia.
* Do some A/B testing on them for one year. Compare two fonts with very little difference. For example in one font the dot of the i has the normal size and in the other font the dot is bigger.
* Perhaps, from time to time, change the group of people, to be sure that the font is good for "normal" people with dyslexia, not only for the trained group that is in the laboratory.
* Perhaps, from time to time, try some bigger changes, but measure the improvement.
* Find a new bigger group people with dyslexia.
* Compare with this people the difference between the original and the final font.
"many of the letters look similar – such as v/w, i/j and m/n – thus people with dyslexia often confuse these letters"
Dyslexic people usually don't confuse letters M and N, they confuse M and P. Notice how they don't relate in any sort of way. This shows how much of a brainfuck (no offense) it is to live with dyslexia, and how hard it is for non-dyslexic to understand the disability.
To make an object-oriented parallel, dyslexic people know very well the difference between two classes (concepts), but when seeing or creating instances (realisation), they can't seem to make heads or tails of which one belongs to which class. So the instances m of class M and p of class P will get processed as some mashup M/P. The only way out is to work around the mashup by tying M class and P class each to another unrelated concept, but related to the instances.
So as an example, the letter 'm' is linked to the word 'marmalade' (which is an instance of the concept of 'Marmalade') and p to the word peanut (itself an instance of Peanut). 'Marmalade' and 'Peanut' being hopefully[0] unrelated, the person can find a solution by linking concepts (not instances) 'M' to 'Marmalade' and 'P' to 'Peanut', so that when the person tries to process the word 'motor', instead of being faced with a dreadful (M/P).O.T.O.R. they will be able work around the issue by thinking (Marmalade).O.T.O.R.
I took the 'm' shape which is present in both 'marmalade' and 'motor' as a link vector, but really any vector will do, and people use an extremely diverse array of vectors and associations. One can end up linking 'm' to 'Green' just because for some reason seeing a 'm' makes him thing of the color (hence concept) 'Green' whereas 'p' will be tied to 'Red'. Being able to break the M/P mashup is the key, and whatever association will do, so it just happens that M/P+Green == M for someone.
[0] I said 'hopefully', because concepts of any level can result in a mashup. If for some reason the mashup Marmalade/Peanut exists, it will be of no help to map m to Marmalade.
SLTs actually use those associations to quickly and accurately diagnose dyslexia, so sorry, I have no reference to give to you. It may look like anecdotal evidence, but I get that straight from the horse's mouth.