I have this system whereby I break down letters into numerical components that are sized according to the 'length' of each letter.
Basically, I started with the letter l. I took l to be 1. I then based each letter in relation to l. I went through alphabetically, starting with consonants. So b, for example is roughly 2 and one-quarter times the length of l. C is about one equal to 1 and one-quarter of l, so that became the value for c. Note, I only use lower case letters.
So using that system for each letter I came up with the following results;
b,d and p = 2 and 1/4
c = 1 and 1/4
f = 3
g = 2 and 3/4
h = 1 and 7/8
j = 1 and 5/8
k = 2 and 3/8
l,t and v = 1
m and w =2
n = 1 and 3/8
q = 2 and 1/2
r = 3/4
s and x = 1 and 1/2
z = 1 and 3/4
Next I did the same system for vowels with the following results;
a = 1 and 3/4
e = 1 and 7/8
i = 1/2
o = 1 and 1/2
u = 1 and 1/4
Now that I have the value for each letter, I can apply the values to words. Say I see the word 'next'. When I see 'next' I don't see the letters, I see their numerical value. Once I've got the value for each letter in the word I try to make each of the values as similar as possible. For 'next', the values are 1 and 3/8 (n) 1 and 1/2 (x) and 1 (t). At this point the vowels are ignored. For these values, to get them as even as possible I change the 'x' and the 't' to both become 'c' (1 and 1/4). Now the new value for the word is 1 and 3/8 and two lots of 1 and 1/4. This is as even as I can get the values whilst still using letters.
That is the first thing I do with each word. The next thing I do involves using the vowel present in each word. Using the example of 'next' again, you can see it has one vowel, 'e'. The value for 'e' is 1 and 7/8. Because the 'e' in this word is in between two consonants I halve the value of it (for 'e' I round the value to 1 for simplicities sake) and I apply that new value to the 't'. Therefore, the 't' becomes equal to 2.
In the case of words that have more than one syllable, such as Toribash, I apply the highest vowel value, after any appropriate halving, to each letter, minus the value of the vowel that is already next to the consonant. So for 'Toribash', the highest vowel value is the combination of the 'a' and the 'i' next to the 'b'. The combined value of this, after halving is 1 and 1/4. I then apply this to each consonant in the word. The final result is; the 'T' becomes 'n', the 'r' becomes 7/8 (yes this is not a letter but that doesn't matter with this operation), the 'b' remains the same, the 's' becomes 'j', and the 'h' becomes 'g'.
Once I've done this operation to each letter, I perform the first operation on the word.
I apply this entire process to each word I read, and each word that I hear. So whenever I'm having a conversation I'm applying this entire process to each word that is said. I've spoken about this to a psychiatrist and apparently it is a form of OCD.
So yeah, that's my quirk.