Microsoft Word L 2-03-Greenberg-paper doc

criterion problem
. Secondly, assuming that we are clear concerning 
the relevant criteria, there remain concrete problems regarding just how we are to 
proceed. Let us call this the 
methodological problem
. Finally we may ask why of all 
the ways in which we can classify languages, the genetic type should be considered 
classification 
par excellence
. Let us call this the 
justification problem
. 
The answer to this last question, it should be pointed out, does not entail the view 
that other modes of classification might not be useful for other entirely legitimate ends. 
What we do want to know is why genetic classification has had a unique status in 
linguistics. In this regard, the term 
genetic
(which would seem to be metaphorical) 
does, as we shall see, have a justification in that its parallel in biology (evolutionary 
taxonomy) is likewise the preëminent and basic manner of classifying species. We shall 
start, not by a direct attack, but by an enveloping movement, by considering other ways 
of classifying languages in order to highlight by contrast and thus disengage the basic 
properties of genetic classification. 
Let us consider what is, at first blush, a peculiar and indeed even foolish way of 
classifying languages. Yet to specify why it is foolish will, I think, not turn out to be a 
foolish exercise. Let us consider a standardized form of language names as spelled, e.g., 
in the Voegelins’ volumes on the languages of the world. We could then classify all 

---

Joseph H. Greenberg
114 
languages by means of the initial letters of their names. Such a classification, in which 
Amharic
would belong to the same group as 
Atakapa
(an Amerindian language of 
Texas), while 
Zyryan
(a Finno-Ugric language) would go with 
Zulu
, would obviously 
be categorical, since it would be complete and without class overlap. The reason that it 
is of no scientific interest is that the set of languages with the same initial letter in their 
names would have nothing in common except that fact itself. Another consideration is 
that in a sense it is not linguistic because the property of having a certain initial letter 
would not be in itself a fact about the English language, but about the spelling of the 
word 
English
. Such metalinguistic facts are but a variety of a larger set of facts about 
any language which we may call external, as opposed to internal. For example, 
statements such as “spoken by more than one million people” or “used in higher 
education” are examples of external properties that are not metalinguistic, as opposed to 
saying that a language possesses “labial stops”, which is an internal property. 
Clearly it is possible to have useful classifications, such as sociolinguistic ones, 
into standard and non-standard languages, which utilize external criteria. In the case of 
pidgins and Creoles we have an interesting situation. It seems clear that the basic 
definitions are here based on external criteria. A pidgin is a language which is no one’s 
first language, while a Creole language is one which developed out of a pidgin by 
acquiring first-language speakers. However, a central problem of the study of these 
languages is whether there are likewise internal linguistic properties which these 
languages possess and which may in fact be unique so that one would recognize a 
language as a pidgin or a Creole without knowledge of the linguistically external facts 
that have just been mentioned. Among oft-cited characteristics are the absence of 
inflectional morphology and a limited lexicon. 
The aforementioned properties are what would usually be considered typological. 
We shall therefore consider next this important form of classification. We may proceed, 
so to speak, heuristically by enumerating the sorts of criteria which would ordinarily be 
considered typological and then seeking to isolate, if possible, what, if anything, they 
have in common. 
We may start by pointing out that all languages contain numerous items which 
involve the association of a particular sequence of sounds with a particular meaning, 
which, following de Saussure, is often called arbitrary. What is meant here is, I believe, 
not the exclusion of the obvious facts about sound symbolism and the numerous other 
iconic facts about language. We may restate the principle of the arbitrariness of this 
association in the following way. Suppose someone were to describe on the basis of 
first-hand observation a hitherto unstudied language in New Guinea and assert that the 
word for mother was 
papa
. We would not be able to assert that he was wrong because it 
reversed the usual facts regarding sound symbolism for terms designating the female 

---

The Methods and Purposes of Linguistic Genetic Classification
115 
parent. In other words, potentially any sound may designate any meaning, although the 
probabilities of a particular combination may in some instances be very low. However, 
they are never zero. 
When it was stated earlier that in the widest sense a language contained numerous 
pairs in which sound was associated with meaning, the reason for stating it in this 
manner was that we wish to include here not only lexical items in the usual sense, e.g., 
the word