Which English letters and numbers are most commonly confused when read
out loud over the phone?

For example, if you read the digit "five" to someone over the phone,
they sometimes hear "nine." When you say "bee" they sometimes hear
"pee" and so on.

On the other hand, W ("double you"), it seems, is very distinctive and
would very rarely be confused.

I need to create an id number which will be read by novices over the
phone, and I am trying to choose the subset of letters and digits
which are LEAST
likely to be misinterpreted. I can leave out letters and numbers which
are likely to be misheard from my id numbers. For example, since 5 and
9 are frequently mistaken, my id numbers will never have 9 in them.

I'm sure someone, somewhere must have done some real research on this,
which would be very useful to me.

I am looking for ACTUAL RESEARCH RESULTS on the topic, not random
conjecture. This is the kind of thing that Bell Labs used to research,
I think, because they used to have policies for operators to read the
number 5 as "fie-uv" to avoid confusion with nine.

Thank you for accepting my findings as your answer. I am glad to have
been able to help you locate a study that meets your needs.

This entire paper may be of interest, particularly Table 1,
"Classified performance for individual letters" on Page 3:
 
University of Pennsylvania: Spoken Letter Recognition
http://acl.ldc.upenn.edu/H/H90/H90-1075.pdf

My Google search strategy:

Google Web Search: "letter OR alphabet OR number OR digit OR numeral
recognition" spoken
://www.google.com/search?hl=en&q=%22letter+OR+alphabet+OR+number+OR+digit+OR+numeral+recognition%22+spoken

If anything is unclear or incomplete, please request clarification;
I'll gladly offer further assistance before you rate my answer.

Best regards,
pinkfreud

Very quick answer to a difficult question! Thank you!

The usual solution for such matters is not to avoid "bees" and "pees"
(or similar combinations, e.g. "em" and "en"), but to use phonetic
alphabet (the way might have seen done in the military: alpha, bravo,
charlie, delta etc).

Thanks! I'm aware of the phonetic alphabet, but my users won't know it.

As a part of a software/internet product I'm developing, I need to
tell complete novices (mostly Americans) to call someone up and recite
a long ID code. For example we will tell person A to call up person B
and recite the ID code 8735-JHUI-8375893-JJHGK-6716, and person B has
to type this into their computer without making mistakes.

Very few of them will know the phonetic alphabet. I'm just trying to
pick a subset of letters and digits that will minimize the number of
mistakes.

Table 1, "Classified performance for individual letters" on Page 3 of
this document may be of interest to you:

http://acl.ldc.upenn.edu/H/H90/H90-1075.pdf

That's close enough! I'll accept this as an answer

I wanted to add that obviously many many English speakers have had
this same problem, and the most common solution has become the
phonetic alphabet.

http://en.wikipedia.org/wiki/NATO_phonetic_alphabet

I would just forget about making up codes and tape a printout of that
list to every desk.