Microsoft Word L 2-03-Greenberg-paper doc

Sprachbund
, literally 
a ‘language confederation’. Such an areal classification has the typical hallmarks of 
prototypicality. It is defined by a cluster of traits any one of which may be absent in at 
least one of the languages. A language which contains all or nearly all of them may be 
said to be prototypical, while in other cases we have languages which have relatively 
few of these characteristics and share some with other neighboring areas, as in the 
instance of Serbo-Croatian. We thus have the phenomenon of 
fuzzy sets
, which 
frequently accompanies prototypicality. One should note that all of the languages of the 
Balkan speech area are Indo-European, but the properties that define the speech area 
are all subsequent innovations produced by contact, and not those which result from 
their common Indo-European inheritance. Only in relatively few instances have 
attempts been made to define speech areas; e.g., Sandfeldt 1968 (Balkans), Masica 
1976 (South Asia) and Greenberg 1984 (Africa). 
The above exposition might make it appear that distinguishing those resemblances 
among neighboring languages which are the result of contact from those which are to 
be explained by genetic inheritance and those which are exclusively typological (that is, 
not deriving from either contact or genetic survival) is a straightforward matter. In fact, 
the reasoning, particularly in regard to typological resemblances, is largely 
probabilistic, and it is the existence of a number of these, of which any single one is not 
completely certain, that provides cogent evidence for significant contact phenomena. 
The following example will perhaps help to illustrate this point. 
The Thai group of languages have in almost all instances basic SVO order, are 
prepositional, and have the dependent genitive after the noun. The Khamti language 
belongs to the Thai group, but is geographically isolated from the rest; it is spoken in 
Burma and neighbors languages of the Tibeto-Burman group and Assamese, which is 
an Indic (hence Indo-European) language. Both the Tibeto-Burman and Indic languages 
are SOV, postpositional, GN languages, here and in most other areas. Khamti is SOV, 
has some postpositions and variable genitive order. It is, of course highly plausible that 
these word order properties of Khamti developed through contact with the Burmese 
languages in the vicinity. Yet there are well attested instances of change from the SVO 
to the SOV types through purely internal factors. It simply becomes much more 
probable in this case to attribute the change to linguistic contact factors. We may ask 
why, of the numerous Thai languages, only Khamti has these characteristics at the same 
time that it is the only one which has been in contact with SOV languages. 
It was noted in the preliminary discussion that there are two levels to be 
considered, that of individual traits of resemblance and that of languages as wholes on 
the basis of these individual traits. In the case of genetic classification to which we now 

---

The Methods and Purposes of Linguistic Genetic Classification
121 
turn, the lower level resolves itself into a consideration of what are usually called 
cognate forms. 
In the case of genetic classification, the question of the nature of relevant 
resemblances has been, I believe, a major source of misunderstanding over the 
methodology of classification. Let us consider first the kind of resemblance which was 
in the initial exposition stated to be non-typological, namely those involving sound and 
meaning simultaneously. Thus we may say that English 
tongue
and German 
Zunge
, 
with the same meaning, are similar both in sound and meaning and that this similarity 
derives from a common original that can be reconstructed for Proto-Germanic. Let us 
call this sort of similarity diachronic genetic similarity. The reason for including the 
term genetic is that similarities resulting from borrowing are also diachronic in nature; 
they both involve processes which take place over time. However, in the remaining 
discussion, in order to simplify our terminology, the term