Yngve, Victor (1996) From Grammar to Science: New Foundations for
General Linguistics, Victor. John Benjamins.
Reviewed by Pius ten Hacken, Universitaet Basel, Switzerland
The purpose of this book, as stated in the subtitle, is to found a new
type of linguistics, which, as implied by the main title, is meant to be
more scientific than the existing types. Actually, this is quite an
understatement. The revolution which Yngve tries to set in motion with
this book dwarves anything Chomsky has achieved, sweeping away much more
radically the entire linguistic research of the last two thousand years as
largely irrelevant. To put it briefly, Yngve attempts to convince us that
linguistics should give up the concept of language and study human beings
in their communicative behaviour.
Yngve's book consists of 22 chapters. The structure
is highly linear; no hierarchical grouping of chapters is given or can be
imposed straightforwardly. The ideal way to read the book is one chapter
at a time, as if using it in a course of 22 sessions. In this review I
will not therefore summarize the book by chapter, but attempt to summarize
the main themes while referring to chapters for various details. Since I
am aware of the increased danger of misrepresenting Yngve's theses in this
way, I will rigorously separate my own comments from the summary.
Readers familiar with the early history of Chomskyan linguistics will know
Yngve as the head of a machine translation project at MIT on which Noam
Chomsky and a number of people who played a role in the early rise of
generative linguistics were employed (cf. Newmeyer (1986)). Those
interested will find Yngve's version of this episode, outlining the
project and his conflict with Chomsky, in 4.3 6.1.
The major goals of the book are, first, to demonstrate that linguists have
been wrong to assume that linguistics can be made into a science while
retaining language as its object of study and, second, to sketch the
general outline of an alternative, truly scientific framework for
1.1. Problems with a Language-Based Linguistics
Much of our view of science and linguistics can be traced back to Stoic
philosophy. In Stoic philosophy, reality was divided into physical and
logical domains. Only in the physical domain did empirical evidence play
the role of a truth criterion. Language, however, was attributed to the
logical domain by the Stoics. Linguistic theory is still based on
techniques such as explication, adequate for the logical domain but not
for empirical science (ch. 2). "Language" and "utterance" are not
empirical entities but logical concepts created by assumptions (ch. 1).
Hypotheses depending on these assumptions cannot be scientific because
they are rooted in the logical domain. Throughout the history of
linguistics we observe tendencies to renew the field and make it a real
science, but when they mix the physical and logical domains, they can
never succeed. This is valid for the Neogrammarians, Saussure, Bloomfield,
Chomsky, and HPSG (ch. 3, 4, 6).
If we want to make linguistics a science we have to clear it radically of
all concepts based on the logical domain, most of all "language". What we
can observe is not words but sound waves (ch. 7). We have to determine a
new scope for linguistics, for which Yngve proposes human communicative
behaviour (ch. 9). This is a given entity in the real world, as required
by science (ch. 10). Making this assumption automatically solves a number
of problems which have long haunted linguistics. They include internal
ones, e.g. pragmatics (ch. 6), variation, acquisition, and historical
change (ch. 7), as well as external ones, e.g. public acceptance (ch. 6),
and delimiting the borderlines of the subject matter (ch. 13). As to the
latter, it is no longer important to determine whether linguistics or a
neighbouring discipline treats a borderline phenomenon, because the
methods of two scientific disciplines are the same and will therefore
yield compatible results (ch. 6, 13).
1.2. A Framework for Linguistics Without Language
Communication takes place between people in a group of some kind, although
in boundary cases the group may consist of one person. A group of people
with those elements of their surroundings which are relevant in their
communication is called an "assemblage" (ch. 7). A scientific description
of an assemblage is expressed in terms of properties (features). The
attribution of properties explains similarities and differences observed
between assemblages. The combinatorial explosion accounts for the
differentiation of a number of assemblages larger than is ever necessary,
with a relatively small number of binary features (ch. 11). Properties are
grouped hierarchically. All properties describing an assemblage together
constitute a "linkage". A linkage consists of four constituents:
participants, channels, props, and settings. A participant is a set of
properties, namely the subset of the properties of an individual relevant
to the description of the assemblage. The other three constituents offer
different possibilities of describing elements of the surroundings.
Delineating them is a matter of convenience, as well as delimiting the
linkage from the outside world (ch. 10). Thus we may choose to describe a
single conversation as a linkage or a lasting friendship (ch. 14).
Linkages consist of role parts fulfilled by participants (ch. 15). A
special case occurs when another linkage plays the role part, as when
someone interacts with an institution. Here the linkage describing the
assemblage of the institution plays a link part in the superordinate
linkage. One participant of the linkage describing the institution is the
link participant, representing the linkage in actual communication (ch.
18). Apart from this hierarchical organization of linkages, a community is
structured in that people may take part in several assemblages. The
linkages describing these assemblages have participants that are subsets
of properties from the same individual. Typically these subsets do not
coincide (ch. 16-17). This part of the mechanism serves to describe states
of assemblages at different hierarchical levels.
Another part of the mechanism describes changes. Assuming that properties
do not change without reason, we can specify setting procedures. A setting
procedure for the binary feature F consists of two clauses, one specifying
the conditions under which F changes from 0 to 1, the other the conditions
for the reverse change. A network of setting procedures is a plex (ch.
12-13). The domain of control specifies which region of the plex is
active, accounting for the relevance of expectation in communication (ch.
21). Control procedures describe the reaction to an impulse in a given
context. They consist of a single statement, briefly changing the value of
a feature to its non-default value. Control procedures are combined into
selection procedures. They are active in, e.g., the pronunciation of a
particular sequence of sounds (ch. 19-20).
By means of this system of description, supplemented by a scientific
notation (ch. 13, 19, 20), linguistics can take its place as a science
among the other sciences.
Before turning to some points of content, I would like to say something
about the general appearance and style of the book. I think the author and
the publisher have done a great job in producing this book. Layout and
binding are excellent, there is an index, typographical errors are
extremely rare, and the style is accessible and makes for agreeable
reading. As the publisher assured me, the author delivered the text in a
format ready for photographic reproduction, so he deserves most of the
praise. Given the uncommon philosophical and linguistic positions defended
by the author, it is remarkable that his style can virtually always be
read as expressing sincere enthusiasm rather than aggression. Only where
he discusses his relation to Chomsky and his opinion on science does the
author's style verge on pedantry.
There are two points in Yngve's argument which I want to take issue with
here before evaluating his system in general. They are Yngve's concept of
science and the way he discusses other linguistic theories.
Yngve devotes an entire chapter to the description of what science is (ch.
8) and most of his summary (p. 308-311) lists similarities between his
system of linguistics and his concept of science. The properties he
attributes to sciences may be surprising to readers with a reasonable
background in philosophy of science, such as the present reviewer. Thus in
Yngve's view scientific knowledge is permanent and represents truth.
Science is a coherent, unified enterprise. The purpose of a science is to
describe certain aspects of the real world in a way that is testable. The
real world is given. Yngve seems to be aware that his concept of science
is not the same as what is assumed in philosophy of science, but he tells
his readers to pay no attention to philosophers. He prefers the views of
practising scientists as one would prefer to take violin lessons from a
good violinist rather than a music critic. Perhaps the only time in the
book where Yngve loses his temper is in fn. 53, where he replies to a
reader's criticism that his system is similar to Carnap's logical
To be honest, I think the reader's criticism is incorrect. Yngve's concept
of science is so naive that no serious philosopher would get away with it.
Even logical positivists such as Ayer (1946) acknowledged that strong
verifiability, leading to permanent certainty of a statement, is only
possible for tautologies. Scientific statements are, in Ayer's terms,
weakly verifiable, resulting in a degree of probability below 100%. It is
precisely the observed non-permanence of scientific knowledge which is at
the basis of modern philosophy of science. Kuhn (1957) offers an important
case study of how this works. He shows how the system of astronomy changed
from antiquity to Newton with the gradual acceptance of new ideas which
replaced older ones. Both the old and the new model are scientific. The
model determines the types of possible observations so that exchanging it
means entering a new world. In a trivial way Yngve avoids this particular
point by putting the starting point of modern science in the 17th century.
It is questionable whether we should be happy about excluding Aristotle
and Copernicus, but the implicit claim that Einstein's theory of
relativity is a straight continuation of the permanent foundation laid by
Newton's science is simply not true.
It is true, as Yngve states, that a lot of disagreement exists among
philosophers of science. There is a consensus, however, determining the
boundaries of variation which is seriously taken into account. This
consensus is quite far removed from what Yngve tries to make us believe
science is. I trust that modern philosophers of science would agree that
scientific knowledge is not permanent but consists of hypotheses. A system
of hypotheses is a theory which has the function of explaining certain
aspects of the world. Objective truth does not exist. The world is not
given. Observations of the world are determined in part by our theories
and metatheories. Two valid approaches to the same object may yield
results which are not compatible with each other.
Nevertheless, many practising scientists may agree with Yngve's concept of
science, if not intellectually at least emotionally. This is accounted for
in philosophy of science by setting up "paradigms" (Kuhn (1970)),
"research traditions" (Laudan (1977)), or a similar concept. Practising
scientists need not all concentrate on general epistemological questions
as long as there are some who do. When Yngve explicitly addresses
epistemological questions, however, he cannot hide behind the protective
wall of other people's epistemological work and get away with a naive
view. His analogy with violin lessons is entirely misplaced unless he
wants to commit himself to the view that musical taste is permanent and
universal, holding the same position as truth in his concept of science.
2.2. Linguistic Theory
Throughout the book remarks abound to the effect that "linguistics of
language" is not able to account for various phenomena. Chapter 21 is a
good example. Here the problems of reference, context, deixis, anaphors,
connotations, and metaphors are explained and it is claimed that Yngve's
theory can handle them but "linguistics of language" cannot. There are two
reasons why I find this argument less than convincing. First, there is no
demonstration of how Yngve's theory would account for them. All we get is
vague indications. Second, there is no demonstration of how alternative
theories fail to account for them. In fact, the only alternative account
referred to is by Apollonius Dyskolus.
Because of my feeling that modern linguistic theory had been neglected, I
decided to do a brief analysis of Yngve's list of references. I found that
20 of the 141 entries in the list refer to Yngve's own works, with at
least an additional 6 referring to works by students or collaborators. Of
the remainder, 43 are works of pre-Chomskyan 20th century linguistics,
mainly American structuralism but also Hjelmslev and Jespersen. Only 17
refer to more modern works of linguistics, 7 by Chomsky, 10 from competing
theories. The rest are editions of texts from antiquity, 19th century
school books and other works which I can hardly consider relevant to the
current state of linguistics. The picture becomes even clearer when
looking at the way modern works are referred to. Of Chomsky's works, only
philosophical points are taken up, and as far as theoretical analysis is
mentioned at all it is taken from Post-Bloomfieldians. One gets the
impression that to Yngve, current linguistics means Post-Bloomfieldian
linguistics. If Yngve has more than a superficial knowledge of linguistic
theory of the past forty years, he does not share it with his readers.
While it is clear that any claims as to the non-existence of accounts of
certain phenomena cannot be maintained without a better coverage of the
literature, even at the philosophical level, discussion hardly goes beyond
the exposition of Yngve's ideas and the a priori rejection of competing
ones. Yngve's claims might have been more acceptable if he had addressed
some of the obvious arguments for mentalism as formulated by, for
instance, Jackendoff (1993), or some of the points of discussion relating
to Chomskyan meta-theory as outlined by Botha (1989).
2.3. Conclusion: Yngve's Theory as Science
In the above discussion I have shown that Yngve bases much of the argument
for his new theory of linguistics on a naive view of science. Such a naive
view is legitimate for practising scientists in a science where others
have established the foundations in a more sophisticated way, but not for
scientists who intend to rework these foundations.
Assuming, however, that several different frameworks, approaching
linguistics from different perspectives, may each be legitimate, the
question remains as to how valuable Yngve's theory and framework would be
in competition with existing ones. While introducing a great deal of new
terminology and carefully justifying his notation, Yngve never addresses
the question of what his theory should explain. It seems that he is
satisfied with a descriptive mechanism for human communication.
It is typical of scientific theories such as Newton's theory of astronomy
and Chomsky's theory of linguistics that they propose an abstract model
which is not directly observable in the world as a basis for the
explanation of certain phenomena. Such models are not the only possible
ones. Einstein proposed a different model for astronomy and Bresnan &
Kaplan (1982) propose a different one for linguistics. I would be curious
to know the abstract model underlying Yngve's descriptive mechanism, the
phenomena it can explain, and the type of explanation. I have not found
these in this book, however. Without such a model, Yngve's theory is
incomplete as a scientific approach.
Ayer, Alfred Jules (1946), "Language, Truth and Logic", Dover, New York
(First edition 1935, second edition with new introduction 1946, Dover
Botha, Rudolf P. (1989), "Challenging Chomsky: The Generative Garden
Game", Blackwell, Oxford.
Bresnan, Joan & Kaplan, Ronald M. (1982), 'Introduction: Grammars as
Mental Representations of Language', in Bresnan, Joan (ed.), "The Mental
Representation of Grammatical Relations", MIT Press, Cambridge (Mass.),
Kuhn, Thomas S. (1957), "The Copernican Revolution: Planetary Astronomy in
the Development of Western Thought", Harvard University Press, Cambridge
(Mass.), repr. 1985.
Kuhn, Thomas S. (1970), "The Structure of Scientific Revolutions, Second
Edition, Enlarged", Chicago University Press, Chicago (orig. 1962).
Jackendoff, Ray (1993), "Patterns in the Mind: Language and Human Nature",
Harvester/Wheatsheaf, New York.
Laudan, Larry (1977), "Progress and Its Problems: Towards a Theory of
Scientific Growth", University of California Press, Berkeley.
Newmeyer, Frederick J. (1986), "Linguistic Theory in America", second
edition, Academic Press, New York.
Pius ten Hacken is a researcher and lecturer at the University of Basel,
partly in the computer science department in the faculty of humanities,
partly in general linguistics. His current research concerns the
epistemological foundations of computational linguistics in its relation
to theoretical linguistics. Selected publications are listed on the
following WWW-page: http://www.unibas.ch/LIlab/staff/tenhacken
Dr. Pius ten Hacken
Institut fuer Informatik/Allgemeine Sprachwissenschaft
Petersgraben 51 || Tel. +41-61-267'33'38
CH-4051 Basel || Fax +41-61-267'32'51
Switzerland || email: email@example.com
web page: http://www.unibas.ch/LIlab/staff/tenhacken