Microsoft Word L 2-03-Greenberg-paper doc

diachronic similarity
will be 
employed to mean diachronic genetic similarity, unless otherwise indicated. What we 
are interested in is the kind of similarity between a linguistic form involving sound and 
meaning, in its earlier and later forms, whether it occurs within the history of a single 
language or independently from an earlier common form ancestral to a number of 
languages. 
There are certain logical characteristics of diachronic similarity which are different 
from that of similarity as it is understood in its application as a classificational criterion 
in practically all other instances. One of these is that similarity is generally conceived to 
be symmetrical. If 
A
is similar to 
B
, then 
B
is similar to 
A
. In phonetic change we would 
naturally say that a sound will in general change to a similar one. For example an 
unvoiced consonant often changes to a corresponding voiced one. Therefore, the earlier 
and later forms share a set of common features, all except voicing, and it is in these 
shared features that their similarity consists. Moreover, it seems natural to assert that 
this is a symmetrical relationship. If a
 t 
is similar to a 
d 
then surely the similarity must 
hold in the other direction and to the same degree. However, there are instances in 
which a change can occur in one direction but not in the other. Thus there are many 
attested instances of 
s
> 
h
but, as far as I am aware, none of 
h
> 
s
. However, diachronic 
similarity is non-symmetrical, rather than asymmetrical, since the majority of changes 
are symmetrical. Thus both 
e
>
 i 
and
 i
> 
e
are possible changes. 
Further, in synchronic similarity we are free to define degrees of similarity in 
terms of the number of shared features according to some overall phonetic analysis of 
sounds into combinations of features. However, while as empirical fact diachronic 
similarity often coincides with synchronic similarity, this is not always the case. For 
example, as we have seen, sibilants often change into 
h-
sounds, but in every synchronic 
scheme (of which I am aware) they differ by a whole set of features. 

---

Joseph H. Greenberg
122 
These considerations also hold in regard to semantic change, but with an added 
twist, which increases the complications. When one sound replaces another, the first 
normally disappears from the language, with a usual transitional period of free 
variation.
2
In semantic change however the old meaning in the general case persists so 
that, as we can see in looking at the dictionary entry for any common word, there are a 
series of meanings, most of whose interrelationships are apparent in terms of semantic 
similarity based for the most part on metaphorical transfers and metonymic shifts, 
which are the most frequent types of semantic change. However, often some of the 
connecting links no longer exist in that the word in some particular meaning has been 
replaced by another lexical item. In addition, the cumulative effect of a set of changes, 
particularly metonymic, which are often surprising, combined with the replacement of 
certain meanings just mentioned, often leads to a situation in which historically 
connected meanings of the same original form become, viewed synchronically, 
homonyms. 
As a result, a historical arrangement of the varied separate senses of a single term 
resembles a genealogy, in which some members have died. It is then no wonder that the 
search for necessary and sufficient conditions for the definitions of words in natural (as 
opposed to logically devised) languages is often futile. When Wittgenstein made his 
celebrated remark about the various senses of the same word showing a “family 
resemblance”, he created a very apt metaphor, but in his ignorance
 
of historical 
considerations regarding semantic change he did not realize how this had come about. 
To summarize, in regard to individual resemblances, which correspond to the 
notion of trait in the initial discussion, we have in effect asserted that forms are likely to 
have a common origin if they could have descended by known types of change from a 
single original. It may have been noted, particularly by linguists, that in saying this we 
have alluded neither to regular sound correspondences nor to regular sound changes.
3
This is because regular sound change, whether conditioned by neighboring sounds or 
unconditioned, is just one of many processes which are known to occur in sound 
changes. Moreover many sound changes are known to be irregular. 
2
It does happen however that a sound change is incompletely carried out so that, depending on 
the dialect and the word, a particular change is or is not carried out. Sometimes both sounds 
survive and the doublets acquire different meanings. These facts were well known to earlier 
dialect geographers who coined the slogan that each word has its own history. The residues of 
such a process are found in the so-called incomplete satemization of certain branches of Indo-
European in which certain words have fronted the original velars and others have not in a 
manner which differs from branch to branch. The work of Wang (1969, 1977) and his 
associates on “lexical diffusion” belongs here.
3
For a fuller discussion of the relation between evolutionary theory in biology and linguistics 
including historical references, see Greenberg (1959).

---

The Methods and Purposes of Linguistic Genetic Classification
123 
Further, conditioned sound changes may produce regular alternations of sounds in 
grammatically related forms. Such morphophonemic alternations are generally subject 
to the unifying force of analogy in which one of the alternants replaces the other. When 
this occurs the direction of change usually differs in individual cases and in an 
independent manner in related languages which have inherited the alternation. This 
process is called reverse analogy and results in completely sporadic correspondences. 
The Neogrammarians, to whom we are indebted for the general concept of regular 
sound change, were well aware of analogy as the second major factor in sound change. 
Take for example the various subsequent changes in Germanic after the 
alternations in Proto-Germanic due to conditioned changes in consonants, summarized 
in Verner’s law. One of the conditioned changes was an alternation of *