Rate-of-language: The amount of underlying information added by each successive letter of a message. English prose, for example, turns out to contain something like 1.3 bits of entropy (information) per letter. This might seem outrageous to claim -- after all there are more than 2^(1.3) letters in English! But the problem is that some letters occur a lot more than others, and pairs (digraphs) and triplets (trigraphs) of letters cluster together also. The rate of English doesn't depend just on the alphabet, but on the patterns in the whole text. The low rate of English prose is what makes it such a good compression candidate.