I first began doing this kind of work from the problems set forth by Ruhlen in "The Origin of Language" (New York 1994). In those problems we are usually presented with about 13 words from each of about 12 different and remotely related languages. In that study we are to analize the lexical items to see their correspondances so that we can classify the languages into families. Then we are to notice that the evidence of some correspondance would lead us to conclude that the families are in turn remotely related.
In this study we start with two closely related languages, Gera and Bole of the Plateau Sahel branch of the Chadic languages which are of Hamito -Semitic (Afro-Asiatic). The discussion will concern the basic vocabulary items only, but which are more important in considering proto-languages. I looked at lexical items from nearly 60 Chadic languages of the Biu-Mandara branch and the Plateau Sahel branch. Lexical items were obtained from "Chadic Wordlists" Charles H. Kraft, Berlin 1981. I found that any given Biu-Mandara language is aproximately 50% related to any given Biu-Mandara language and very aproximately 20% related to any given Plateau Sahel language, while any given Plateau Sahel language is about 50% related to any other Plateau Sahel language. To arive at that conclusion about 12 lexical items where analized for numerous languages, so the results are very approximate. And we are only concluding that the relation is of the basic vocabulary. Basic vocabulary is most resistant to word borrowing and other types of linguistic change, so analyzing basic vocabulary helps us to say weather two languages are related or not according to their more distant history or origin.
I set out to compare more lexical items for two languages and then compare the percentage of relation of the two languages, thinking that the larger lexical inventory should give me a more accurate indication of the percentage of relationship between the two languages. I selected 35 of the more basic lexical items out of about 400 basic words. Then I selected a subset of 13 items to compare the percentages between an accurate and less accurate assesment. Then I noticed that the first 21 items tended to be more resistant to linguistic change than the next 14 items in the list, so I then computed percentages in connection to the first 21 items between Gera and Bole.
Of the 35 items, basic Gera and basic Bole have 19 items which are of some common origin, so they are about 54% related. Of the basic 13 items 10 are related, so that is 77%. Of the 21 which are more resistant to linguistic change 16 are related so that is 76%. So when I look back I see that for the 13 items I tended to select words which I thought were most resistant to word borrowing, and the results from the 21 items seem to suggest varification there (as 77% corresponds to 76%).
What I am leading to is the question of what can we say about the relation between two languages if we only know the first letters of the words. I need to know this to interpret certain results I have obtained in hieroglyphics which has its symbols holding phonetic values based on the first letter of the name of the symbol.
So, for Bole and Gera I have noticed that of the 21 more basic words 11 had a correspondance in the first letter, that is 52%; and of the 35 items 31%, and of the 14 more abstract basic words ( the last 14 besides the first 21) the relation was 0, none of the first letters corresponded. These two languages would have about 28 letters in their alphabet. If two languages are unrelated then the probability that for a particular meaning they both have a corresponding word which begins with the same letter is about 1divided by the number of letters in their alphabet. If Bole and Gera have two unrelated words meaning the same thing the probability that they both begin with the same phonetic value is about 1/28 which is about 3.5% (.035). This kind of probability of first letter correspondance between any two languages will be between about 3 and 5 percent for unrelated languages.
It is certain that for a set of related words, not all of the words will have the same phonetic value. Example, "blood" ndoma (Gera) domi (Bole). In the 35 items, only one noncorrespondance had a match in the first letter by chance it seems, "tree" shafinga (Gera) shoiwi (Bole). On the other hand the following was considered related: "head" ka (Gera) ko?i (similar to 'ko-ah-i') on the basis of the initial letter 'k'. In this case 9 of the 19 correspondances do not have initial corresponding letters. That leads to only a 28% correspondance (10/35). If the two languages were unrelated we would only find about 3 or 4 percent of initial letters corresponding, or in this case one word out of 35.
So, if two languages are say 40% related we could not expect nearly that many corresponding words to have initial phonetic values which correspond.
