ad libitum
.” Later, Collinder apparently
reconsidered, since in his 1965 work he returned to etymologies unaccompanied by
reconstructions.
Reconstructions change over time, or are not even carried out (Afroasiatic), or
carried out only partially (Uralic), yet the classification remains secure along with a
number of fundamental and obvious etymologies which survive all vicissitudes.
Regarding the details of method, it might appear that since the world is the only
natural unit, multilateral comparison of all the world’s languages should, in principle,
be carried out simultaneously, using the most stable elements of the vocabulary,
including pronouns.
4
Such an approach is clearly impracticable, and in fact unnecessary. In actual
practice we face a situation which varies for different areas of the world. Thus, since
families like Indo-European and Uralic are well established and etymological
dictionaries are readily available, we may use reconstructed forms, or approximations to
them where they are not given. Even where starred reconstructions are supplied, one
will wish to examine the actually attested forms which often provide important clues.
The opposite situation obtains in areas like South America, in which scores of
independent families are stated to exist and comparative works are almost non-existent.
Even here one will not have to consider every last language. For obvious and extensive
groupings like Arawakan a reasonable sample of languages will be adequate. In
carrying out this type of investigation, one should bear in mind that at deeper levels of
classification the same basic principles enunciated earlier still hold. Thus there is no
reason to assume that Indo-European is necessarily a member of a stock with only two
members any more than it proved to be the case for Germanic.
4
That in principle this holds is shown by the interesting example of Arda in Colombia, listed as
an independent stock in early classifications. Rivet (1925), guided by the resemblance of the
name
Arda
to that of an important slave trading port in Dahomey, found that it was virtually
identical to the Niger-Congo languages spoken in that area in West Africa.
The Methods and Purposes of Linguistic Genetic Classification
129
It might seem that there is still a third method of classifying languages genetically,
namely glottochronology. When it was introduced, of course, it was intended for
another purpose, i.e., to measure the period of separation of related languages based on
the assumption of a constant rate of change in fundamental vocabulary. In any specific
instance, the date is derived from a count of shared cognates between two languages on
the assumption of independent loss in both languages. In spite of its well-known
weaknesses, it has been up to now the only reasonably objective method we have to
accomplish this in the absence of written documentation.
However, it later began to be employed as a method of classifying languages
genetically on the assumption that there was a lower limit of chance resemblance and
that a significantly higher percentage indicated genetic relationship. Of course, viewed
in terms of its original procedures, its use for this purpose, since cognate counts were
involved, is circular. By definition there are cognates only when languages are already
related.
This method bears a superficial resemblance to multilateral comparison, since it
compares lexical forms in different languages and the data are often set forth in
comparative tables similar to those used in the latter method, at least in its preliminary
stages. The most important difference is that it employs pairwise percentages, thereby
not taking into account the possible, multiple recurrence of resemblant forms across
many languages by which the genetic groupings become evident. A great part of the
evidence which connects related languages is in only one of the two languages
compared and, it will be argued, in some instances occurs in neither. If, for instance, we
were to compare English and Hindi directly, the percentage of cognates would be very
low. However some of these would be recurrent over most or all of the other Indo-
European languages and hence highly diagnostic. In other instances, English would
show a cognate with, say, Slavic which was not in Hindi, while in other cases it would
be Hindi that agrees with Slavic to the exclusion of English. These independent
agreements of English and Hindi with Slavic are part of the evidence for Indo-European
as a whole, as is, naturally, independent agreements of English and Hindi with still
other branches of Indo-European. It could even be said that agreements between Slavic
and Italic are relevant since they help to establish the overall family to which both
English and Hindi belong.
Put syllogistically, English is a Germanic language; Germanic languages are Indo-
European languages; therefore, English is an Indo-European language. Hindi is an
Indo-Iranian language; Indo-Iranian languages are Indo-European languages; therefore,
Hindi is an Indo-European language. Hence, English and Hindi are related.
To the weakness just discussed we may add that, as languages become more
genetically distant over time, semantic changes occur so that items fall off the
Joseph H. Greenberg
130
comparison list, although they are still present as cognates. Thus English
Dostları ilə paylaş: |