Some Truisms.
Let me start with some general points and questions.
- Linguistics comprises many and varied activities.
This should be a truism. But within the field there are conflicting views and
emotions about what "real linguistics" is. (Hudson, 1972, gave an insightful
picture of the hierarchies of prestige in academic and scientific
disciplines.) Part of what I want to do here is call for mutual respect and
cooperation among all who deal with language from various points of view, with
various interests and various agenda.
- The world's languages are surprisingly similar and surprisingly
different.
Whenever I undertake the study of a "new" language I am constantly
struck by two things:
- How often the language conforms to my expectations;
- How often I am surprised by some totally unexpected phenomena.
The similarities and the divergences come in all levels and aspects of
languages. Some samples will be given below in Section IV.
- Linguistic theory must account for the similarities and the
diversities among languages.
- Linguistic theory needs lots of different languages.
This is an obvious point: If only one language were spoken in the world, we
would not know what features of the language were particular and which
universal. There is a kind of argument that goes like this:
Poverty of stimulus argument:Such and such a feature must be universal
because it would be impossible for children to learn it on the basis of the
evidence that they get. (To be found, for example, in many writings of
Chomsky.)
But without some independent knowledge of what can and can't be learned, this
argument has no force (I believe I heard this counter-argument first from
Barbara H. Partee, p.c.).
The inescapable fact about language acquisition is that children and other
people do learn languages, with their expected and unexpected, universal and
particular characteristics.
- Where are all the languages going?
Extreme language loss is a fact of our current life. There are two ways of
responding to this fact, both important for linguists to think about:
Two responses to language loss:
- record, document, analyze languages before they go, call this
"documentary linguistics".
- try to stem the tide, call this "ecological linguistics";
I assume that a third response -- ignoring it -- is not an option.
With limited resources, there can be conflicts between the two aims. I am
heartened by the fact that many younger linguists, equally adept in theory and
in description, devote a large proportion of their time to community-oriented
projects. The profession of linguistics must recognize and support such work.
- Do languages need linguistic theory?
Obviously, they do not, in the simple sense. Can linguistic theories help
attempts to revitalize and preserve endangered languages? Probably not. But
if the aim is documentation, then there are properties of languages that
generally will not be noted unless there is active invocation of the linguistic
theories that have uncovered these properties.
- Do languages need linguists?
Here I think we can give a more positive answer. There are many things that
linguists can do which can contribute to the health of languages as well as
contributing to the scientific aims of documentation. The main thing (in my
opinion) to keep in mind here is the asymmetry between practical concerns and
requirements and scientific practice. Linguists naturally tend to write for
other linguists. To render their results, documentation and descriptions, useful
for community aims, they should be prepared to present them in or translate
them into a form that is accessible to nonlinguists. For example, with
computers as a help it is easy to present dictionary and textual materials in
an adequate practical orthography, or to have parallel representations using
both practical orthographies and more arcane representations.
A little history: Will the real linguist stand up?
One of my first public presentations on linguistics was at the Ninth
International Congress of Linguistics). I mention that not to brag on
how old I am, but to give some sense of the temporal frame for my remarks.
Unfortunately, my vision is narrower than it should be on this occasion, since
I will be largely confining my purview to linguistics as it has gone on in
North America. That Congress took place in Cambridge, Massachusetts, in 1962
(if my memory is correct). It was in the heady early years of the Chomskyan
Revolution or Abberation. Noam Chomsky made a presentation in a plenary
session. The second wave of young turks were mostly graduate students. So
what has happened in the ensuing four decades?
When I first became interested in linguistics, a few years before that,
linguistics was all about going out and describing languages, especially
languages that had not been documented. The "real linguist" was the
fieldworker. The attitude toward "theory" was often apologetic. In the
American linguistic scene, theory was mostly about procedures. Zellig Harris's
1951 book was originally called "Methods in Structural Linguistics." Its main
motivation was theoretical. Harris wanted to be able to compare the
structures of different languages. The point of the book was to try to ensure
that different linguists could produce comparable or intertranslatable
descriptions. Harris tried to define procedures of segmentation and
classification by distribution that would lead to uniform results. Various
choices -- for example, introducing "long components" -- would lead to certain
kinds of statements, but in the ideal case, making different choices would
lead to statements that could be converted readily into equivalent
descriptions given other choices. Chomsky's early characterizations of what he
called "taxonomic linguistics" was not far off the mark for his teacher
Harris. I think it was pretty far off the mark for many other linguists,
including probably most linguists in other parts of the world.
What was most original in Chomsky's early work was the very idea of a
generative or formal grammar, considered as a theory that would specify all
and only the (infinitely many) expressions in a language, and assign
structural descriptions to them. In a kind of Kantian turn, linguistic theory
became then the study of the general structure of the grammars that would be
just adequate to capture natural languages as formal systems. The idea that a
language could be described as a formal system might well be called "Chomsky's
thesis," even though, as is often the case, the idea was "in the air" at the
time (Bar-Hillel, Harwood, Greenberg, Hockett, among others). A decade and a
half later, linguistics saw the formulation of "Montague's Thesis," the idea
that a natural language could be characterized as an interpreted formal
system. ("Chomsky's Thesis" and "Montague's Thesis" were socalled in Bach,
1989.) Again, it was not just Montague who was coming up with this idea
(Keenan, David Lewis, Parsons, among others).
There is no doubt that these new conceptions of theory became dominant in
American linguistics and beyond over the next decades. The cultural clash
between proponents of the two images of what a "real linguist" is and does, was
-- and still is -- vigorous, often acrimonious. The split went along with
differences in fundamental views of what scientific activity should be (see
Bach, 1965, for a then contemporary view of this philosophical disagreement).
In 1967, MIT made a brilliant appointment, when Ken Hale, arguably the best
field linguist of his era and a gifted theoretician as well, joined the
department. I believe a good deal of mending between the two impulses or
sides of linguistic research in the following decades can be attributed to his
influence.
Three kinds of language study.
Think of a diagram, something like a schematic of the solar system. Ask: what
is the Sun and what are the Planets? In one view, some language is in the
center, arranged around it are various theories and subdisciplines. In the
other, linguistic theory is the sun and the planets are languages and
subdisciplines and related areas. These diagrams are supposed to represent
crudely two kinds of linguistics: descriptive and theoretical.
Now another truism:
There is no such thing as a theory-free description.
Whenever you undertake to describe a language you carry with you a set of
expectations about what languages are like. These are really hypotheses
about Universal Grammar. They may be completely formal: you expect phonemes,
recurrent structures of form and meaning. They may include much more
specific, but still formal, hypotheses about the form of a grammar for
arbitrary languages. And they may include substantive hypotheses about the
content of a language, at all levels: phonological, phonetic, syntactic,
semantic.
What are the results of these two kinds of activities?
- Primary Linguistic Description
Primary descriptions of languages are the basis for everything else:
There is no such thing as a language-free theory of language.
Descriptive linguistics traditionally results in descriptive grammars,
dictionaries, texts, recordings -- nowadays, audio and video, one hopes. In
the United States and Canada, the great descriptive grammars, dictionaries,
and text collections of the late nineteenth and twentieth century are the
heritage of Boas and his tradition. True, there are many questions left
unanswered in the products of this stream, but there is not nothing: the
results are a rich mine for successive workers. Think, for example, of the
careers that have been built on the grammars and text collections of this
heritage.
- Linguistic Theory
The results of linguistic theorizing are theories or bits of theories
(hypotheses). In line with the schematic I drew in your minds, languages are
drawn upon to give evidence for or against some point of theory or, in grander
attempts, a whole theory. So the typical result of a graduate work in
linguistic theory might be a dissertation with a title like: "The ABC
Principle in Language X," where X might be some language that had not been
dealt with in depth by theoreticians. Such a study might be followed by a
whole string of papers or dissertations taking the same material to argue
against the ABC Principle in favor of the A'B'C' Principle, or for a whole new
approach or theory. But usually there is no new data, often not even checking
of the original sources, and only in rare cases are these studies based on new
primary work.
It would seem that both of these activities would share a concern about
accuracy, "getting the facts straight." This concern for accuracy is central
to a third kind of actitivity.
- Philology
Besides the activities just mentioned -- descriptive and theoretical
linguistics -- there is another kind of study devoted to languages. Unlike
those two disciplines, philology is devoted not to uncovering the
system of a language or coming to understand the general abilities of humans
to acquire and use such systems. It is devoted rather to the products of
users of the languages, memorable products for the most part. The name is apt:
`lovers of the word.'
The proper cultural matrix for philological efforts should be the community
most directly associated with the texts or oral traditions in question. But
in the colonialistic and post-colonialistic world in which we live, it is
more often than not up to linguists from outside these communities to do the
primary work that is a necessary foundation for such activity. Why is this
so? It is clearly because the healthy functioning of the native traditions of
story-telling or recording have become debilitated for reasons that are too
familiar to need rehearsing. Languages and cultures change and sometimes die
as a matter of human history. But sometimes they are killed.
Ken Hale's life work on many languages was distinguished by one insistent
maxim: the best work on any language could only come from native speakers of
that language. Therefore, it was incumbent on the foreign linguist to try as
hard as possible to teach as well as take, to treat speakers of the language
not as "informants" or "consultants" but as fully equal co-workers, linguists
in their own right.
The activity of fieldwork, primary linguistic description has changed as the
result of political change, the insistence by many First Nations people, that
the work of linguists be responsive to community needs. Thus in the
contemporary world, fieldwork of the old style is usually just not an option.
Homo loquens grammaturgicus
Work within the Chomskyan paradigms has emphasized "linguistic creativity."
What this has generally meant is that adequate theories of Language and of
languages must accomodate the fact that speakers can produce and understand
novel utterances without bound. I believe there is another kind of creativity
in the world of language, at the level of grammar creation. Adequate theories
of Language must make room for this kind of creativity as well. In short,
accounts of Universal Grammar must give room for the quite astonishing variety
that we find in particular grammars. The best way to appreciate this is to
look at a lot of languages in detail. Since none of us can look at all
languages directly, we must make do with descriptions of the languages. As a
matter of strategy, descriptions of languages that are theory-driven must be
balanced by descriptions that are carried out in the spirit of the advice
often given in older and other traditions than those of some of the dominant
theoretical stances:
We need to cherish and study linguistic diversity for reasons that are as
important scientifically as they are politically and ethically. It is not a
bad idea to let a language unfold itself to you on its own terms for a good
long while before you jump to fitting it into your theory or testing your
theories against it. (Bach, [1996])
There are two reasons to believe that the human linguistic organ (UG?) has an
inherent creativity, and hence two reasons to believe that theories of
universal grammar that are too constrained cannot be adequate:
Two reasons for believing in the creativity of the human linguistic organ:
- Persistence
Some very special aspects of languages and language families can persist over
long stretches of time and space.
Examples:
consonantal roots in Semitic, special systems of pronominal marking in Algonquian.
- Diffusion
Examples: areal features in the Pacific Northwest of North America:
phonology, word-grammar, phrase grammar; word order characteristics from
substrata or surrounding languages: Amharic SOV syntax as opposed to general
Semitic patterns. See Thomason, 2001, on areal and contact phenomena.
Three meanings for "language"
In various publications, Chomsky has distinguished between various meanings of
the word `language.' Two notions have been constant: (1) a language
considered as a set of utterances, sentences or whatnot, (2) a more abstract
and idealized object connected to a grammar and more a virtual than an actual
matter. These two meanings went with his early distinction between
`observational' and `descriptive' adequacy, as attributes of linguistic
theories, specifications of the form and content of grammars. The latest
incarnations of these notions appear in these two terms (Chomsky, 1995):
- E-language
..."E is to suggest `external' and `extensional'" (Chomsky, 1995: 16)
- I-languages
..."I is to suggest `internal,' `individual,' and `intensional'" (Chomsky,
1995, p. 15).
In the surrounding text, Chomsky makes an explicit connection to the earlier
concepts of observational and descriptive adequacy.
It is fair, in view of the discussion here to also link the `I' to another
word, `idealised.' Chomsky hypothetical speaker Jones does not `have' such a
language in a pure state. "Rather, Jones will have some jumble of systems,
based on the peculiar pattern of his experience" (p. 19). And:
Furthermore, even if a homogeneous speech community existed, we would not
expect its linguist system to be a "pure case." Rather. all sorts of
accidents of history would have contaminated the system, as in the properties
of (roughly) Romance versus Germanic origin in the lexicon of English. The
proper topic of inquiry, then, should be a theory of the initial state that
abstracts from such accidents, no trivial matter. (ibid.
Some years ago I suggested a third term, to complement Chomsky's two (Bach,
[1996]):
- R-languages.
R is intended to suggest "real." (Bach, 1996)
An R-language is supposed to be precisely
what Jones or an actual speech community might "have." Of course,
idealization is still in order, but dealing with an R-language means dealing
precisely with the results of the accidents of history, and whatever jumble of
systems might result. But theories of acquisition must make room for how
people learn such sets of systems, and to understand language change,
linguistics must deal with just how such jumbles and historical accidents can
become systematized and approach and mold I-languages. Moreover, I believe
that it is only through such study that we can approach an understanding of
the surprising diversity of languages.
Tensions
There is a natural affinity between the three notions of language and the
three kinds of language study distinguished above. The theoretical linguist
is, like Chomsky, focused on I-language(s), the descriptive linguist will be
primarily concerned with R-languages, or at least will not be able to avoid
them, while the philologist will be deeply involved with R-languages, and the
products of users of the language, that is to say, with E-language: sets of
texts, corpora, memorable utterances by particular Joneses. There will
probably always be tensions among the practicioners of these various
activities, often within the heart and mind of individual workers. A few
hours spent on reading articles and reviews in prominent journals of the
various activities will bring this home.
But there is a fourth group that we must add to our consideration: the
native speakers, the members of the communities whose languages we
study. This group also has a notion of language, and it is different from all
the above. I say "notion," but probably we should think of pluralities. In
any case, I am thinking of a language as embedded in a culture, and alive in
the consciousnesses of the members of the culture. The concepts of language
that linguists use often abstract away from the cultural matrix of the
language. I believe that many of the disagreements and tensions that arise
when native speakers and linguists confront each other arise simply from
differences in what the several groups mean by "language."
We can appreciate this point by considering the problem of translation. One of
the cardinal results of the linguistic study of the last two centuries has
been the realization that languages are all completely adequate toolboxes for
expressing anything that their users want to express. This thesis is often
embodied in the slogan:
All languages are created equal!
This means that given time and patience and the possibility of paraphrase it
is possible to translate anything from any language into any other. But what
do we mean by `translate' here? There is an opposite view of expressibility
that is equally often expressed, perhaps more by non-linguists than linguists:
"Translation was never possible" writes Margaret Atwood in a poem `Marsh
Languages' (Atwood, 1995). And this sentiment is equally true and valid but we
must understand here "translation" in the sense of reproduction of a piece of
language in its flesh and bones and skin, not just the discursive content.
A Case Study: Syntactic Categories.
In this section I will take the question of the cross-linguistic
identification of syntactic and morphological categories as a case in point,
where different linguists have made different assumptions about what is to be
expected, i.e. what is a "null hypothesis."
Ideas about syntactic and morphological categories provide a good mirror of
changing stances toward universal and particular grammar and grammars. Put
briefly, the development has been from the assumption that all languages
potentially share the same categories, namely, those familiar from Latin and
Greek, to the opposite assumption that such categories were purely language
specific, then back to the universalist view, but presumably grounded in a
theoretically more defensible foundation, and then back to an emphasis on the
language-particularity of such categories. The history of morphological /
inflectional concepts is similar. These stages or stances can be
characterized by appropriate slogans:
- Universal I: all languages are the same, namely just like Latin.
But we must qualify immediately, as some languages are defective, in lacking
various features of the quasi-Universal model. This is the era, where in the
"Western" world, you find statements like these:
- Language X lacks a clear distinction between Verbs and Adjectives.
- English noun paradigm:
nominative: John
accusative: John
genitive: John's, of John
dative: to John
ablative: from John
vocative: O John
BUT (see below) It is easy to laugh at such examples, but we should not
be too sniffy. The idea may be good and the execution bad, that is, Latin or
Greek do not provide the right set of categories to start with but the idea
that terms like Noun, Verb, Adjective, or names for cases have some
theoretical and universal content is not foolish. The much maligned and
admired Grammar of Port Royal (Anonymous, 1676) contains a beautiful
example of an explanation of very language-particular facts of French based on
assumptions about the universal nature of such categories as Verb, Adjective,
Participle.
- Bloomfieldian Particularism: it is not expected that languages share
categories.
In that era and tradition, linguists were reluctant to use terms from
traditional grammar. Instead of Noun, Verb, and the like, descriptions used
terms like Class I, Class II, and so on.
There is a way of understanding syntactic categories under which the
Bloomfieldian mode makes perfect sense. If a category is simply the name for a
set of expressions which in the strictest construal can stand in all and only
the same environments, then it is completely obvious that Noun in Japanese
grammar and Noun in English grammar, for example, cannot name the same sets.
So either the idea of universal categories has to have some different
interpretation, or we must think of the categories as "really" meaning
`Noun-in-Japanese,' `Verb-in-English,' and so on.
- Universal II: all languages are the same, namely just like English.
The earliest transformational grammars took over with no substantial
justification the categories of traditional grammar, but refined so as to
reflect the existence of phrasal categories or `projections' (speaking
anachronistically) of those traditional categories. The initial empirical
base was English and as this base was broadened to include more and more
different languages these categories were naturally taken over for the `new'
languages.
- Early discussions of categories in transformational grammar
There are two streams in the early years of the transformational-generative
tradition that are directly relevant here.
The first may be traced to a paper by John Lyons (1966) who asked why there
should be phrase structure rules such as "NP ==> Det + N" and not "NP ==> Det
+ V." In other words: Is there any substantive connection between the use of
"N" on both sides of the arrow, or is it just a kind of pun? There is a
connection here to Harris's (Harris, 1952) use of symbols like N,
N2, on the one hand, and to Categorial Grammar, on the
other, and Lyons suggests using a categorial base instead of phrase-structure
for what was then called the "kernel" (the notion of "kernel" was itself
borrowed from Harris). This stream led eventually to socalled X-Bar theory
(Chomsky, 1970; Jackendoff, 1977).
The other was a line leading up to and incorporated into the Universal Base
Hypothesis of yore (see for example Lakoff, 1970[1966]; Bach, 1968). As in
recent discussions (see below), the strongest hypothesis was taken to be that
all languages shared the same set of base rules, and hence categories. This
was enforced by the idea that the base rules provided a direct representation
of meaning. But it was not assumed that the form of the Universal Base was
just what had been posited for English. In Bach (1968), for example, it was
argued that English categories were on closer consideration rather like
those one might posit for a language like Nuuchahnulth (Nootka), which
traditionally had been assumed to have a single category of Contentives or
Predicates in place of the traditional division into Noun, Verb, Adjective.
(The best review of this question that deals directly with the Southern
Wakashan facts as instantiated in Makah is still Jacobsen, 1979.)
James D. McCawley (1982) gave a critical review of these early discussions of
syntactic categories. The whole discussion lost a good deal of its urgency
over the 70's as linguistic theory absorbed model-theoretic semantics as a
more adequate theory of (some aspects of) meaning. Another blow to the
Universal Base Hypothesis was delivered by the demonstration by Peters and
Ritchie that the hypothesis had no empirical force, given the excess power of
then current transformational theories. (As far as I know recent revivals of
the Universal Base Hypothesis have not given the requisite formal attention to
showing that the Peters and Ritchie result no longer applies.)
- Recent discussions
The last few years have seen a revival of interest in the controversy about
basic syntactic categories across languages.
Eloise Jelinek argued on an entirely new basis for the lack of Noun-Verb
distinction in Straits Salish (Jelinek, 1995; Jelinek and Demers, 1994).
Demirdache and Matthewson (1995) argued against Jelinek, but on the basis of a
different Salish language. In my opinion, the verdict on this controversy is
still open, but the question still needs to be clarified:
Is the claim for and against universality relevant to lexical categories or categories of the
syntax proper?
How does the discussion relate to questions about the mappings
from syntax and lexicon to the semantics proper?
For the discussion here let me just focus on one point:
It may be true that the strongest or null hypothesis is that all languages
have the exact same set of syntactic categories, as Demirdache and Matthewson
argue. The immediate next question is: just what is this set? I would no
longer argue for the position of Bach, 1968: English categories are those one
might arrive at from the point of view of a Wakashan language. But on the face
of it that is a priori just as strong and no stronger than the null hypothesis
of Demirdache and Matthewson.
- Semantics of Syntactic Categories
As a final point on the matter of this brief review of discussions about the
universality of syntactic categories let me refer to a different line of
inquiry that relates to semantics. This line asks questions about the
universality of semantics and the models that we use to interpret natural
language. One approach is to say that the basic model structures made
available for interpreting natural language are universal, but that the
mappings that are made from the syntax to these structures may and do vary
from language to language (Bach, [2001]). A fruitful series of inquiries has
dealt with questions about the nominal domains of plurality, mass, count and
so on as related to various languages, see Krifka, 1995; Chierchia, 1998; Cheng and
Sybesma, 1999.
Outlook: the discussion continues.