LINGUIST List 5.1409

Wed 07 Dec 1994

Disc: Comparative method in syntax, continued

Editor for this issue: <>


Directory

  1. "Prof.Dr. R.Block, FB13, Tel.:305320", syntax and comparative method
  2. Helge Dyvik, Re: 5.1393 Using the comparative method in syntax

Message 1: syntax and comparative method

Date: Mon, 5 Dec 1994 23:15:50 Msyntax and comparative method
From: "Prof.Dr. R.Block, FB13, Tel.:305320" <BLOCKfhmrz4.rz.fh-muenchen.de>
Subject: syntax and comparative method


Fritz Neumeyer asks for opinions on whether or not syntax can be
reconstructed by the comparative method. He is frankly skeptical that this
is possible and so am I. Let me state my position in rather abbreviated
form.

The problem is that syntax lacks the all important *tertium
comparationis* ('basis for comparison'), which makes the comparative method
possible. For example, we are licensed to compare Greek /pater-/, /pod-/
with Eng. (father), <foot> because _these pairs have the same meaning_.
Since "the sign is arbitrary" (e.g., there are over 40 possibilites for the
initial sound in an English word), it appears unlikely that the /p - f/
correspondence in initial position in these and many other similar pairs is
due to coincidence. Once we have ruled out the possibility of borrowing
(e.g., Germ. (kaufen) 'buy' is not related to, but borrowed from Lat.
caupo) 'tradesman'), we can, with some confidence, reconstruct I-E /p/
corresponding to Lat., Gk. /p/, Germ. /f/.

Syntactic constructions, however, do not mean anything and are not
arbitrary in the same sense that words are. Consider the well-known example
of "basic" word order. There are six mathematical possibilities for the
combination of the three elements subject, object and verb. Apparently,
three of them are as rare as the proverbial hen's teeth so that virtually
all of the world's three thousand odd languages chose between three common
patterns: SOV, SVO, VSO. Thus, the fact that English and Chinese are both
SVO says nothing about any special relationship between them.

But, is sharing some relatively rare syntactic trait proof of genetic
relationship? Unfortunately, here too we must answer no. Consider
Romanian and Bulgarian, which are neighbors and have postposed articles,
but are nevertheless members of different families (Romance and Slavic) and
do not possess this feature by virtue of inheritance. If we knew less about
the external history of European languages than we do, "syntactic
reconstruction" might tempt us to posit a connection to the Skandinavian
languages as well since they too have postposed articles.

It seems to me that the "problem" with the comparative method is not with
any inherent limitations it may have, but rather that some investigators
attempt to apply it where there is not sufficient supporting evidence. This
practical limitation on the comparative method is admittedly frustrating
since we would like to know more than it can tell us, but the alternative is
far worse: Relaxing the standards of scientific proof simply opens the
doors to any number of competing and equally speculative hypotheses about
genetic relationships between languages.

This is not to say that historical syntax is uninteresting or unrevealing,
just that we cannot draw the same kinds of conclusions from historical
reconstruction of syntax as we can from phonology.

Russell Block
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: Re: 5.1393 Using the comparative method in syntax

Date: Tue, 6 Dec 1994 10:51:16 +Re: 5.1393 Using the comparative method in syntax
From: Helge Dyvik <dyvikalf.uib.no>
Subject: Re: 5.1393 Using the comparative method in syntax

Unlike Martin Haspelmath I find much of Lightfoot's criticism of the
concept of diachronic syntax fairly convincing. It is not just a question
of it being "more difficult" to apply the comparative method to syntax than
to phonology. Rather, the question is whether the distinction between
historical and typological comparison can be maintained at all on the level
of syntax. At least we should beware of drawing a close analogy between
diachronic phonology and diachronic syntax.

Let me start with a few elementary insights that we all share.

Phonological systems can be compared typologically or genetically. Any two
language systems can be compared typologically, while, of course, not every
language pair can be compared genetically. Genetically corresponding
phonemes need not share any typological features. What unites them is
occurrence in corresponding positions within etymologically corresponding
morphs. Thus genetic correspondences between phonological systems
presuppose the existence of regular correspondences between a significant
number of word forms from the basic vocabularies of the two languages.
Word form or morph pairs in this kind of relationship are said to be
etymologically 'identical'. Morphs are also minimal units of language
acquisition - they are not constructed productively, but acquired and
remembered en bloc. Therefore they can meaningfully be regarded as
preserving some kind of transtemporal identity through generations and
centuries of language evolution. Thus genetic correspondence between two
elements (like two vowels) presupposes a concept of etymological identity
between entities on a higher level of analysis, i.e., morphs, and the
concept of etymological identity presupposes a kind of en bloc transmission
of the elements in question to new generations of speakers.

Shifting the view to diachronic syntax, it seems evident that a concept of
diachrony within syntax must be quite different from the corresponding
concept within phonology and morphology. Units of syntax - phrases and
clauses - are by definition constructed productively; they are not learnt
and remembered as unanalysed wholes. Hence they are not transmitted from
one generation to the next as unanalysed wholes either. As a consequence
the concept of 'etymological identity' becomes dubious within syntax. But
such a concept is a prerequisite for establishing genetic correspondences,
at least of the traditional type, between language elements. On what basis,
then, can syntactic constructions be placed in a diachronic relationship
with each other? We can, of course, frequently find syntactic constructions
in old and modern languages having identical or similar semantic and formal
properties, but what will motivate a claim to the effect that the modern
construction is 'developed from' the older one? There will be no larger
units of analysis containing the compared constructions and standing in a
relation of etymological identity to each other, as is the case with
phonemes and morphs. Nor is there an unproblematic relation of etymological
identity between the constructions themselves. We miss a unit with a
plausible kind of 'transtemporal identity' on this level, such as the
morph.

As a result it is fairly obvious that syntactic phenomena cannot be used to
establish genetic relations between languages - such relations must be
established on the basis of phonological and morphological data, as before.
But it may even seem dubious to assume that diachronic relations can be
established between individual syntactic constructions, even in cases where
the genetic relationship between the languages has been independently
established. The question is whether there can be any foundation for a
concept of genetic correspondence within syntax, which should not rather
just be seen as a special case of typological correspondence. For given a
case of typological syntactic correspondence, what has to come in addition
in order for us to say that the correspondence is genetic as well? And is
it possible to imagine syntactic constructions corresponding to each other
only genetically, without sharing typological properties at all, as in the
case of phonemes? The latter definitely seems to presuppose a persisting
identity of syntactic constructions evolving through time, similar to the
etymological identity of morphs which persists through quite radical
metamorphoses.

Thus the picture of evolving syntactic constructions is problematic. In the
case of syntax it seems even clearer than in the case of phonology that
what, if anything, can be meaningfully interrelated diachronically is
complete systems, and not individual linguistic items. The construction is
not a basic unit of evolution. Nor, for that matter, is it a basic unit in
modern syntactic theories, and for analogous reasons. A syntactic
construction is a function of interacting rule or principle modules, whose
interactions will have many other consequences as well, outside the
particular construction studied. It seems that it is on the level of the
interacting modules that evolution can be fruitfully studied , if at all.

Of course there are phenomena which historical linguistics traditionally
treats together in the realization that the development can only be
understood if different changes are seen in relation to each other. A
well-known example is the development typical of many Western European
languages, whereby case systems are reduced, constituent order becomes more
fixed, and articles develop. We have at first a stage where the argument
structure of sentences is expressed directly or indirectly by case endings.
Constituent order is free to be used for other purposes, such as the
encoding of information structure, with the topic in initial position, etc.
Then case systems dwindle, which is frequently offered as an explanation
why constituent order must now be used to express argument structure, with
fixed positions for verbal arguments. (I am not discussing the plausibility
of this assumption here.) Now information structure, with its distinctions
between given and new referents, is deprived of expression, which may
partly account for the rise of definite articles in the same languages.
What remains comparatively constant through time here, then, is certain
categories of semantic information which at any point must be expressed
somehow: The identification of the arguments of verbs, and the separation
of given and new referents. When the encoding of some of this information
for some external reason becomes inadequate, other encoding takes over,
leading to the same effect for other kinds of information which is left
with inadequate expression. This kind of development obviously cannot be
grasped if we only look at constructions individually. The comparison is
still basically typological, but we approach the question of possible
constraints on the ways in which one type may replace another historically.

The study of the development of individual constructions may, after all, be
meaningful in cases where the occurrence of etymologically identical form
words create some historical continuity between them. One example is the
emergence of the periphrastic perfect with an auxiliary verb of the type
'have'. It is the etymological connection between the auxiliary 'have' and
the full verb 'have' which allows us to talk meaningfully about the
historical connection between the constructions involved. Thus the
historical development of individual syntactic constructions may be a
meaningful concept in cases where we kan hinge the development on the
etymological development of grammatical form words, like the verb 'have'.
The morph thus confirms its position as the diachronic core unit.

Helge Dyvik
Department of Linguistics and Phonetics
University of Bergen Phone: +47 55 212261
Sydnesplass 9 Fax: +47 55 231897
N-5007 Bergen, Norway E-mail: dyvikfoli.uib.no
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue