New to Typophile? Accounts are free, and easy to set up.
So after a long delay, I've finally compiled something that I feel could really help type designers: a list of the most frequent letter-pairs according to actual language, so that designers can better figure out what pairs need kerning. The following list is for English, and it's derived from the venerable Brown corpus (by Kucera & Francis). It's not a huge corpus by today's standards ("only" about a million words), but it's nicely generic (not limited to Shakespear, newspapers, or the Bible like some other corpora) and there's been a lot of derivative research based on it - which is in fact what actually allowed me to do this, since I don't yet have the Brown corpus as a digital file; I compiled the files from a linguistic analysis book by Dominic Massaro, where all pairs in all 3-7 letter words from the Brown corpus were quantified (in fact in all word positions). For the 2-letter words I went through the Brown corpus myself: there were about 25 of them - although they account for ~20% of English text!
The list is broken up into three frequency groups. The cutoffs are pretty arbitrary, but what I tried to do is group them such that there's a triple difference: group 2 is three times the size of group 1, and group 3 is triple group 2. In this way a designer can decide how "deep" he wants to go.