Have there been any studies done into the similarity between individual latin letters? I'm hoping for something (preferably with letters in Arial font) where a letter has been given and the chance of other letters being confused with it is the result. Absolutely ideally the result would be something like this, where columns are the letter presented and rows are the letter returned (numbers completely made up):
<> A <> B <> C
A -.-- 0.43 0.32
B 0.43 -.-- 0.44
C 0.33 0.23 -.--
I've tried looking myself and am wondering if I'm looking for the wrong thing. Any help would be appreciated, thanks!