Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Volume 96
John E. Joseph
JOHN E. JOSEPH
University of Edinburgh
J. E. J.
Edinburgh, 2000
CONTENTS
Foreword vii
Introduction
Natural and Unnatural Language 1
Afterword
Linguistics after Naturalism 201
References 205
Index 217
INTRODUCTION
NATURAL AND UNNATURAL LANGUAGE
1. On a noteworthy exception, the French linguist Victor Henry (1850-1907), see Joseph (1996a).
2 LIMITING THE ARBITRARY
answer that won't make it too obvious what a fool you are. With a wry smile in
his voice he'll say that, well, yes, it's an antiquarian thing, decent cocktail party
chat perhaps, but with no serious interest for the professional linguist. The prob
lem was solved back before World War I, by that Frenchman, Chaussure. Lan
guage is arbitrary and conventional. All the introductory textbooks say so.
Now ask what aspects of language he himself works on. He may mention
something about universals, either a Universal Grammar that is physically present
in the brain (à la Chomsky), or just linguistic features that tend statistically to
characterize most of the world's languages (à la Greenberg).2 He may have an
interest in optimality theory and markedness, the notion that some elements of
language (the simplest, stablest, and most widespread) are more basic (unmarked),
while marked elements are derived from these basic ones. Possibly he will have
something to say, if you're still on the line, about cerebral structure and the neural
aspects of cognition, or else about speech processing and computer parsing and
simulation of speech. Alternatively, he may work on how linguistic structure is
determined by discourse functions, or social stratification, or universal features of
child language acquisition, or laws and principles of historical change, or ...
Finally, ask the good professor what kind of language English is, i.e. what
'family' it belongs to. Germanic, he will say without missing a beat. Ask him about
the fact that every estimate of the English vocabulary places it at over 65% non-
Germanic-derived, with most English words coming from the Latin-Romance
family. Ah, — note how his obvious condescension reveals his growing impatience
— words do not matter. It's the grammar, the syntax, the structure that make a
language what it is. Words can be borrowed without limit, but they do not change
the structure, hence they are irrelevant in determining a language's genetic affilia
tion. The essence of a language lies in its grammar, not its lexicon, the word-stock
being more superficial, less resistant to outside influence, in sum less substantial.
Good of you to call. Click.
By my reckoning the eminent (and ubiquitous) Professor Strawperson has
twice contradicted his initial estimation of the nature-convention debate. First,
while it is true that linguists claim to be in consensus about the arbitrariness of
language, if you look at the rest of what they actually teach and write, you find
them treating language as not arbitrary at all, but determined by, or grounded in,
something outside itself. That something may be cerebral structure, neural net
works, the organs of perception, the physical universe. It may be some kind of
metaphysical notion about the 'natural' structure of codes or systems. It may even
be 'society', conceived as another kind of metaphysical force, moving independ-
2. To be precise, nearly all the linguistic universals formulated by Joseph H. Greenberg and his
followers take the form of implications: if a language has x, it will also have y, etc. For example,
an early universal of Greenberg's stated that if a language has the dual number, it also has the
plural, though the reverse is not true. See Greenberg ed. (1966), and for a fine introduction to this
line of research, Comrie (1989).
INTRODUCTION 3
ently of the will of its members. In other words, modern linguistics talks the
conventionalist talk, but walks the naturalist walk.
Secondly, the claim that the nature-convention debate is irrelevant to lin
guist's present-day concerns becomes spurious when we consider such standard
modern linguistic dichotomies as those between the lexical and functional, the
regular and irregular, the unmarked and marked. These dichotomies have, to a large
extent, effectively absorbed the 2500-year-old debate about nature and arbitrariness
that most linguists regard as an antique curiosity rendered obsolete by progress in
linguistic theory. The idea that one part of language is somehow more fundamental,
more real than another recapitulates what the nature-convention debate was about,
even if the old terms are no longer used. Nor is this the only case of a modern
doctrine covertly propagating a dichotomy generally believed to be long dead.
The basic dichotomy has survived for a simple reason: it has proven to be a
satisfying way of bringing order to what would otherwise seem like the chaos of
language. It offers a principle by which linguists can impose systematicity, limiting
what cannot be explained — limiting the arbitrary. Everything that has to do with
languages as systems demands to be approached, we are convinced, from this point
of view, which is hardly of any interest to linguists. The reason is that they almost
never imagine themselves, any more than Saussure did, as being the ones who limit
the arbitrary by deploying their various conceptions of naturalness.
Unquestionably, these conceptions have been successful; but this does not
necessarily mean that they are logically tenable. The history of science is filled
with long abandoned, now ridiculed conceptions, paradigms and research pro
grammes that in their time generated prolific quantities of widely respected re
search.
The purpose of this book is not to refute any of the approaches to language
which I have grouped together here as versions of linguistic naturalism. Nor is to
promote any of them, or to promote or refute any alternative to naturalism. Rather,
it is to explore how versions of naturalness and its opposite took shape in their
locus classicus, Plato's Cratylus, and some of the principal lines of development
they have followed in Western theories of language down to the present day. Each
of them presents its own conceptual difficulties, which in some cases can be
resolved, and paradoxes, which by definition cannot. The hope is that exploring
them may help us eventually to transcend the whole series of dichotomies along
which naturalness and its opposites are constructed in theories of language.
plays (SOV), while another 10% have Plays Mary drums (VSO). In many lan
guages, including Latin and Russian, any order is possible.
Nevertheless, Kayne argues from theoretical principles that all sentences in
all languages start out with the order SVO, after which some languages allow O
or V (or both) to move leftward to produce other orders. While not all linguists
accept Kayne's hypothesis, it is approaching the status of received truth in many
quarters. Perhaps it is correct — but it remains hypothetical, since the linguistic
inner workings of the brain cannot yet be directly observed.
Somewhat surprisingly, linguists have not voiced much caution about a pro
posal which could be read as suggesting that SVO languages are closer to the
'natural' order that proceeds directly out of the hard-wiring of the brain than are
languages with other, less natural orders. The potential is there for this to be
interpreted into a hierarchy of languages and their speakers according to their
cerebral complexity, an outcome few linguists would accept. Probably this does
not occur to them because, in the model Kayne is using, elements such as subjects
have to move about even in SVO languages, in order to check abstract features.
No language has no movement in this model. Also, linguists generally assume that
an extravagance in one part of a language's structure will be made up by an econ
omy elsewhere.
Still, some languages have more movement to undertake than others, and it
is a matter of time before someone suggests that the non-SVO languages are
wasteful of cerebral energy, or more specifically, of the electrical activity that
many have come to view as our first direct window into how language works in
the brain. A 1996 article in Language (Jaeger et al.) showed that electrical activity
is registered in different areas of the brain in the processing of verbs depending on
whether their past tense forms are regular, like waited, or irregular, like ate. This
appears to confirm earlier experimental evidence reported by Stephen Pinker in
a 1991 article in Science, and subsequently developed into the core of his 1999
book, showing that regular plural nouns like houses and irregular ones like mice
are produced by completely different mental operations. As will be discussed in
Chap. 4, these interesting findings raise some difficult questions about whether the
contrasting structures of a language like Chinese, in which verbs have no past
tense forms, and Arabic, in which all verbs (and indeed most nouns) would have
to be classed as 'irregular', do not imply the distinctive existence or operation of
a 'Chinese mind' and an 'Arabic mind'.
The fundamental paradox is this:
1. If languages are analysed by linguists in such a way that some of their
elements are more central than others, and
2. if this centrality is conceived as really real in some way that extends be
yond language to connect with the physical, mental, perceptual, social,
metaphysical, in short, the natural in whatever guise, then
INTRODUCTION 5
3. we are left with the implication that the peripheral elements are some
how less natural, but
4. can we really assert that anything human beings do is more natural than
anything else they do, on a basis that is not ultimately normative, even
if not intended as such?
With regard to (4), there are of course linguists who actually use the term 'natu
ral', for example in 'natural' phonology or morphology, or referring to data drawn
from 'natural' conversations. The concerns expressed here apply to them, but also
to linguists who may not talk about the natural yet base their work upon some
version of it nonetheless. In fact they may be even more a cause for concern, since
it is bound to be harder to convince people of the power of a hidden metaphor than
of one whose terms appear on the surface.
This does however raise the central methodological concern I have had to
bear in mind over the years of researching and writing this book. I am asserting
that there is an intellectual continuity that runs through centuries of Western
writings on language, and that some but not all of those who are part of this
continuity explicitly presented themselves as taking part in the tradition in ques
tion. Such an assertion poses the immediate danger that I may interpret texts as
falling into this tradition when there is insufficient objective grounds for reading
them that way. I have therefore striven to deal with texts which take up questions
of naturalness and arbitrariness as explicitly as possible, and to give evidence,
clearly but unobtrusively, of why I believe it is justified to assert that they are part
of the continuity I am arguing for. This is after all what good history should strive
to do: not just recycle past documents, but look for patterns that are not obvious
yet are convincingly grounded in documentary evidence; and that are meaningful
enough to illuminate our understanding of both past and present. That at least has
been my aim.
attach to their meanings, when it clearly takes the position that both these views
are impossibly oversimplified. There is an even longer history of the Cratylus
being read as an authoritative treatise on etymology, when its attempt at an ency
clopaedic etymological account of Greek unmistakably ends in disappointed
resignation. Plato has laid out the key issues for any consideration of linguistic
arbitrariness and its limitations with devastating precision, and we are still man
aging quite spectacularly to ignore them.
Part Two again consists of three chapters, which trace ideas about linguistic
naturalness and unnaturalness as they have developed within three different
contexts, each defined by what it looks toward in order to 'limit the arbitrary' —
wilfulness, as opposed to compulsion (Chap. 4); spontaneity, as opposed to social
control (Chap. 5); and invisible hierarchies, as opposed to apparent equality or
disorder (Chap. 6).
Chapter 4, "Natural Grammar and Conventional Words, from Aristotle to
Pinker", makes a historical sweep through versions of linguistic naturalism from
ancient times to the present, to examine how conceptions of the natural and its
opposites have been intertwined with such central dichotomies of Western lin
guistic analysis as word and grammar, lexical and functional, and even noun and
verb. A number of the problems signalled here will be explored in more depth in
later chapters, but the real aim of this chapter is to give credibility to the thesis that
there has been a significant continuity in Western linguistic thought concerning
the core question of what is and is not 'natural' in language, even in periods when
the term itself has not necessarily been in use. Indeed, some of the most overt
debates on the subject, such as that between Whitney and Steinthal over nomos
and physis in the 1870s, or that among various members of the Copenhagen and
Geneva schools over the arbitrariness of linguistic signs in the 1930s and '40s,
were merely a sideshow to much more fundamental shifts in linguistic naturalism
that were taking place at the time.
Chapter 5, "Natural Dialect and Artificial Language, from Varro to Chom
sky", considers the construction of the 'standard' in opposition to the natural, and
how this process too, like the preceding one, has been much more closely con
nected with the goals and methods of linguistic theory than is generally recog
nized. Particular attention is given to Dante's De vulgari eloquentia, the founding
text of the Western tradition of language standardization, Saussure's remarks on
'literary language', Orwell on Standard English and Chomsky on E- and I-
language. The chapter concludes with an explanation of why all these constructs
are fictions, and why this is not the same as rejecting them as 'myths'.
Chapter 6, "Invisible Hierarchies, from Jakobson to Optimality Theory", fo
cuses on the turning point in structuralist linguistics in 1930, when Trubetzkoy
and Jakobson effectively discarded Saussurean arbitrariness in favour of a system
not based on pure difference, but one in which the actual content of elements
made some elements 'marked' relative to others. The chapter traces the later
INTRODUCTION 7
that it is as if Plato in the Cratylus were taking time off from his usual interests".3
Yet, as we shall see, the Cratylus shares with several other Platonic works the aim
of defending dialectical method against other, more 'linguistic' ways to knowl
edge. The alternative path most directly addressed in the Cratylus is etymology;
in the Ion, poetics; in the Phaedrus and the Gorgias, rhetoric; and in the Sophist,
grammar.
Somewhat separate traditions of reading the Cratylus have developed among
its various academic 'constituencies'. Classicists, rhetoricians and historians of
linguistics have been quicker than philosophers or linguists to appreciate the great
importance of the dialogue. Yet even historians of linguistics have tended to
characterize it as aporistic, culminating in paradox rather than resolution, and have
denied its significance for the subsequent development of linguistic thought, since
it appears to be limited to the 'nomenclaturist' view of languages as consisting of
names for pre-existing things; and to insist upon the existence of one or more
'law-givers' responsible for assigning them, which is seen as a naively non-
evolutionary view of language origins. Furthermore, its position on the natural vs.
conventional connection between words and their meanings, has always seemed
ambiguous, so that it is not uncommonly cited as a defense of physis, making it
inadmissible to 20th-century linguists who follow the structuralist doctrine of the
arbitrariness of linguistic signs. For these reasons the Cratylus has received rather
less attention from linguists than its historical importance would seem to warrant.
The idea that the Cratylus supports physis over nomos (present for example
in Demos 1964: 595-6; Robinson 1956: 324; see also Weingartner 1970: 17n.)
may be the product of overzealous attempts to polarize Platonic and Aristotelian
philosophical positions, since Aristotle has been read as coming down firmly on
the side of convention (but see Chap. 5 below). Another consideration is that
Socrates' attack on nomos occurs in the beginning of the dialogue, and appears to
be maintained through the long middle section on etymology, hence to predomi
nate. On the other hand, some have been led to view Plato as a conventionalist
because his attack on physis occupies the dialogue's conclusion. A third view
offers an indecisive Plato unable to choose between the two views, or simply too
concerned with more important philosophical issues of knowledge and reality to
bother formulating a coherent linguistic doctrine; this is the impression left by
Robins (1997): "Both views are given due consideration in the mouths of the
participants, without a definite conclusion being reached" (p. 23); "Later scholars
took up more definite positions than we find in Plato" (p. 24). A similar portrait
is found in Crombie (1963: 485-486). Robins (1997: 23-24) states further that
"The naturalist argument leaned, as it must, on the weight of onomatopoeia in a
vocabulary and on a more general sound symbolism in the phonological structure
3. Thus for example Fowler (1926: 4), in the Introduction to his translation: "The Cratylus cannot
be said to be of great importance in the development of the Platonic system, as it treats of a special
subject somewhat apart from general philosophic theory [...]".
INTRODUCTION 9
of some words [...]". And yet, Socrates never once refers to onomatopeia except
to dismiss it as insignificant, while his discussion of sound symbolism (synaesthe-
sia) is surrounded by disclaimers in which he insists that he does not accept it.
A fourth view, this time of an agnositic or nihilistic Plato, dates back at least
to Giussani (1899: 103ff.), who, according to Tagliavini (1970: 24-25), believed
that Plato's intention is not to construct a theory of language, but to refute the
Sophistic belief in language as a means for teaching the nature of things. Kretz-
mann (1971: 138) and Arens (1975: 25-26) on the other hand read the Cratylus as
reconciling the two theories, showing that both are right in their own particular
spheres. A fifth possibility, that Plato is advancing a definite position which
happens to correspond neither to physis nor nomos has been proposed for instance
by Pagliaro (1930: 15) and White (1976). Finally, Lorenz & Mittlestrass (1967:
8) have suggested that Plato was indeed developing an independent theory but left
it unfinished (so also Weingartner 1970: 6); nevertheless they interpret the Craty-
lus "as a programme for a rational philosophy of language" that "could be used
even today as a frame for further research" (Lorenz & Mittelstrass 1967: 4).
Very recently the interpretative tide has begun to turn radically back, to re
join pre-modern views of the dialogue that find its focus in the long central section
on etymologies, rather than treating it as an unfortunate digression from the
philosophical considerations that precede and follow it. Perhaps the most impor
tant article to date in this vein, Sedley (1998), suggests quite plausibly that
It was only towards the beginning of the nineteenth century, when scholars began to
realise what a linguistic science should really look like, that it dawned on them that
Plato must have been joking all along. [... But] If Plato was joking, the joke flopped.
Neither Socrates' pupil Hermogenes within the dialogue nor Plato's pupil Aristotle
outside it shows the least awareness that it is all a gigantic leg-pull. This ancient con
sensus should be enough in itself to shift the burden of proof firmly onto any modern
reader who wishes to downplay Plato's seriousness in the matter. (Sedley 1998: 145,
143)
Barney (1998) begins from the same assumption, though her interpretation of the
etymologies as an 'agonistic display' is far different from Sedley's. Different
again, though starting from the same basic position, is the stance taken in the
present book, and anticipated in earlier work such as Joseph (1990a, 1995).
For readers wishing to undertake further research into the Cratylus, an
excellent place to start is Geneviève Clerico's (1992) rich survey of work on the
dialogue since 1960. Among books specifically on Plato's theory of language,
Timothy Baxter's The Cratylus: Plato's critique of naming (1992) is a close
reading of the dialogue within the context of classical philosophical problems of
names and truth. Jetske Rijlaarsdam's Platon über die Sprache (1978) provides
a close analysis of the relationship between the ideas of the Cratylus and those of
19th and 20th-century language theory, with particular attention to Ferdinand de
Saussure. La grammaire philosophique chez Platon by Antonia Soulez (1991)
10 LIMITING THE ARBITRARY
does a fine job of tying together the threads of Platonic language theory from the
various dialogues. Michael Palmer's Names, Reference, and Correctness in
Plato's Cratylus (1989), while not particularly distinguished, is competently
compiled and written.
Among articles and chapters in collective volumes, besides those mentioned
earlier in this section, Sluiter (1997) succeeds superbly in locating the Cratylus
within the overall Greek tradition of reflection on words and meaning, while
Matthews (1994) attempts the more difficult task of placing it within the tradition
of grammatical analysis. Joly (1986) takes a close look at the relationship between
words/names and letters/sounds in the Cratylus, while Kahn (1986), in the same
volume, makes useful comparisons between the doctrine of ideal forms as it
appears in the Cratylus and in other Platonic works, examines the relationship
between. Burkert (1970) is a useful source on the Derveni papyrus and its rele
vance to the Cratylus (see also Kahn 1997).
For a broader perspective on language in ancient thought and education,
good overall introductions may be found in Jan Pinborg's "Classical Antiquity:
Greece" (1975) and in Chap. 2 ("Greece") of R. H. Robins' solid A Short History
of Linguistics (4th ed. 1997). For more detailed information, L'analyse linguis-
tique dans l'antiquité classique by Marc Baratin and Françoise Desbordes (1981)
is a fine source, as are the five articles by the same authors, and the one by Daniele
Gambarara, in Vol. 1 of the Histoire des idées linguistiques edited by Sylvain
Auroux (1990). Histoire de l'éducation dans l'Antiquité by Henri Marrou (1948)
remains a work of great insight and vividness.
There are only some relevant highlights from the very vast literature on the
subject. It must be said by way of ending this introduction that none of this litera
ture, the present book included, can pretend to offer more than a shadow of the
intellectual excitement and stimulation of reading Plato himself.
Part One
Cratylus
CHAPTER 1
1. Joseph (1995) examines the ideological slant behind 'pre-Socratic linguistic thought' as it has
traditionally been presented since thefirsthistories of linguistics written in 19th-century Germany,
and suggest an alternative reading that is quite as plausible as the traditional one, but that com
pletely upsets the picture of progress from religious 'connectedness' to secular 'detachment' that
the historians were determined to paint.
14 LIMITING THE ARBITRARY
2. The information that Cratylus and Hermogenes were Socrates' teachers comes from Diogenes
Laertius (3.8). That what Cratylus taught was the correctness of words is inferred from Hermo
genes' comment in Cratylus 428b4, cited on p. 64 below.
16 LIMITING THE ARBITRARY
Hermogenes, who upholds a version of the nomos position, and Cratylus, who
believes that language exists and operates by virtue of physis. They invite him to
take part in their debate. Hermogenes lays out their differing opinions regarding
the nature of language:
HERM.: Cratylus here has been saying, Socrates, that for everything there is a natu
rally correct word — that a word isn't simply whatever people agree to call some
thing, applying to it a portion of the sounds of their voice. Rather, he believes in a
kind of natural correctness of words that's the same for all people, Greeks and for
eigners alike. And so I ask him about the word Cratylus: is it truly his name? Yes,
he says, it is. "What is Socrates' name?" I ask. "Socrates", he answers. I say: "And
that goes for everyone else as well? Whatever word we use to call each person, that's
his name?" "No", he says, "Hermogenes isn't your name, even if that's what every
body calls you". (383a4-b7) 3
3. Translations from the Cratylus are my own. Citations from other Platonic dialogues are given
in the Jowett translation.
4. Baxter (1992: 10—11) has suggested that the actual wording of the two passages quoted above
provides a subtle reinforcement of the ideas they contain, or a foreshadowing of positions to be
taken later on. Hermogenes does not begin by referring simply to 'Cratylus', but to Kratylos [...]
hode "Cratylus here" (just as in the opening line of the dialogue he refers to Sōkratei tōde "Socra
tes there"), and Baxter sees in this a subversion by Hermogenes of Cratylus' view that words or
names "can only refer to things by describing them, or rather: if they refer at all, they refer to what
thcy describc". That is, for Cratylus, the word 'Hermogenes' fails to name Hermogenes because
it does not describe him accurately. But when Hermogenes attaches hode and töde to the names
he pronounces, "the demonstratives function as the linguistic equivalents of pointing at a person,
and used with proper names underline the referring function of names". The 'referring function'
is in effect what Hermogenes is arguing for when he says that a word is "whatever people agree
to call something", regardless of how well or poorly that word 'describes' the characteristics of the
thing named. Similarly, when Socrates begins his reply to the question of what Hermogenes' name
actually is, he does so by addressing him precisely as O pai Hipponikou Hermogenes "O Hermo
genes, son of Hipponikos", using the full patronymic form, his legal identification. In this instance,
"there is no doubt to whom the name Hermogenes refers" (Baxter 1992: 10), any more than there
could have been doubt with the addition of demonstratives. Some further significance of the
NATURE AND CONVENTION 17
patronymic will be considered in a later section of this chapter. We should note first however that
when Cratylus asserts the correctness of his own name and that of Socrates, he is also presumably
making a judgement about personal character: Sōkrates has the transparent literal meaning "sound
ruler", and while the literal meaning of Kratylos is not so obvious, it too is based upon the root
krat, suggesting strength or mastery.
5. The interweaving of linguistic and economic themes is noteworthy: the characteristic jab at the
Sophists for teaching for money, the matter of Hermogenes' financial condition and its failure to
correspond with the literal implications of his name. Money was an obsession of Socrates' on
account of its potential for corrupting people. (Ironically, corrupting the youth would be the crime
for which Socrates himself was put to death.) Part of that potential lay in the fact that money
functioned as a rival measure of value to the 'true' inner goodness of men. The wealth a man
possessed was, like his name, commonly taken to be a mark of his inner nature, but in fact was
merely superficial. If it is absurd to think that Hermogenes cannot be this man's name because he
is not lucky, it is no less absurd to take Hermogenes' lack of wealth as an index of his inner
character. Thus it is doubly absurd to think that Hermogenes' name does not accord with his inner
character. So too the absurdity of thinking that a fifty-drachma lecture on linguistic correctness
must be worth fifty times more than a one-drachma lecture. If either lecture contains what purports
to be the truth about this matter, so should the other, or else the teacher is a fraud. The jab at
Prodicus thus embodies a parallel example to that of Hermogenes' name, centred on the correct
ness of the terms 'fifty-drachma lecture' and 'one-drachma lecture'.
6. The dialogue Theaetetus contains an analysis of how it is possible unintentionally to confuse
the identities of two individuals one knows. The possibility that one would do so intentionally is
not even raised in that less playful context.
18 LIMITING THE ARBITRARY
things whose existence and nature are permanent". By this he means 'common'
nouns, which denote classes of things rather than individuals, and in addition, the
names of the gods. "For it is especially there that the establishing of words seems
to have been taken the most seriously — indeed in some cases perhaps with a
power greater than that of human beings". In other words, supposing that a princi
ple of linguistic correctness turns out to exist, the name of a particular person,
having been given by other mere mortals, is not likely to embody that principle.
In Plato's view of language the names of gods and mortals stand in the same
relation to one another as do common and proper nouns, because just as the gods
are immortal and unchanging, so are the classes which common nouns denote,
whereas individual things and human beings perish.
The word translated here as 'word', onoma, is given in earlier English
translations of the Cratylus as 'name'. Either translation is misleading, because
Greek did not make the modern distinction between names and other kinds of
words. As we have just seen, however, Plato does make essentially the modern
distinction between common and proper nouns, except for aligning the names of
gods with the former rather than the latter. Also, at 431b3 Socrates says explicitly
that the conclusions drawn about nouns should also extend to verbs and whole
sentences.
status of nomoi was the same in the legal and the linguistic dimension. In its time,
for its audience, Hermogenes' theory of language must have been politically
charged, representing as it did the most extreme form of linguistic democracy
imaginable. 2500 years on, the slave reference makes it read much less democrati
cally, though no less politically.
The question Socrates takes up is unrelated to these considerations. The
names of slaves do not interest him, being the whims of ordinary mortals just like
the name of Hermogenes. He turns instead to the fact that the two kinds of law or
custom represented by nomos and xynthēkē are not really identical, as Hermogenes
has implied. If linguistic correctness is a matter of contract and agreement, then
it is a kind of social arrangement individuals engage in on a regular basis. But
most people never in their entire lifetimes establish a societal custom, whether it
is formalized as a law or more vaguely internalized into 'culture'. It is precisely
the individual versus the social character of linguistic 'contracts' that Socrates
calls into question:
SOC: Maybe you have a point, Hermogenes. But let's consider it further. Whatever
someone chooses to call something will be the word for that thing?
HERM.: SO it seems to me.
SOC: Whether it is an individual or the whole community which calls it that?
HERM.: Yes. (385al-5)
One cannot introduce a political dimension more directly than by using the word
polis 'community, city, state', from which 'politics' is derived. The question is,
if we are to uphold nomos over physis, then whose nomos do we uphold, that of
the idiōtēs 'individual' or the polis? At issue here are the rights of the individual
versus those of the state, insofar as language is concerned. In the broader political
context these rights figure among Plato's principal concerns, and his position is
familiar from the Republic and the Laws. He stands for the power of the state and
the need for the masses to obey the laws set down, not democratically, but by a
wise ruler, ideally a philosopher-king.7 Socrates presses Hermogenes on the
implications of his individualistic response, asking whether a particular word can
have one meaning for the people (dēmosia) and another for the individual (idia):
SOC: Really? Suppose I make up a word for some entity or other, for instance what
we now call a man — if I were instead to refer to him as horse, and what we now call
a horse, I called man — then will the word for one and the same thing be man for the
population at large, but horse for the individual? And in the other case, man for the
individual, but horse for the population at large? Is that what you're saying?
HERM.: Yes, that's how I see it. (385a6-bl)
7. The Laws (Nomoi), believed to be Plato's final work, lays out the constitution for the 'second-
best' city, in the belief that it may be more practicable than the truly Utopian vision of the Republic
which calls for the rule of the philosopher-king. Laws V 744 c4-dl calls for the creation of a four-
tiered class system (not including slaves and metics) based, surprisingly enough, on wealth. All
classes would be represented in government, so as to achieve "a mean between monarchy and
democracy" (Laws VI 756e9-10). Yet overall the state described in Laws is even more repressive
to individual liberty than the Republic.
20 LIMITING THE ARBITRARY
8. In discussions of the human and the divine, equating humans with lower animals has always
been a powerful technique for reductiones ad absurdum by anti-materialistic philosophers like
Plato. Perhaps his most stinging use of it is in a discussion of Protagoras' "man is the measure of
all things" in Theaetetus 161c2-6, where Socrates says: 'T am charmed with his doctrine, that what
appears is to each one, but I wonder that he did not begin his book on Truth with a declaration that
a pig or a dog-faced baboon, or some other yet stranger monster which has sensation is the measure
of all things [...]". See also the less pointed, though no less amusing, passage in Republic II 375e9-
376b10 on the dog as the "true philosopher".
9. The discussion in Theaetetus 190c1-3 includes the following exchange: "SOC: And do you
suppose that any other man, either in his senses or out of them, ever seriously tried to persuade
NATURE AND CONVENTION 21
For in Socrates' view there is a moral quagmire here as well. If the individ
ual can determine his own nomoi, then he can create not only his own words, but
his own laws. Under those circumstances both 'word' and 'law' cease to have any
meaning, since they cease to fulfil the function for which they are intended. It is
not for hoi polloi, the mob, to question nomoi in either the legal or the linguistic
domain.10
himself that an ox is a horse, or that two are one? THEAT.: Certainly not".
10. Cf. Crito 50a6-b5: "SOC: Imagine that I am about to play truant [...] and the laws and the
government come and interrogate me: 'Tell us, Socrates,' they say; 'what are you about? are you
not going by an act of yours to overturn us — the laws, and the whole state, as far as in you lies?
Do you imagine that a state can subsist and not be overthrown, in which the decisions of law have
no power, but are set aside and trampled upon by individuals?"' [emphasis added]. In the state
described in Laws, where legislative representation is extended (however unequally) to all classes
of citizens, it is still forbidden to all but the eldest citizens even to wonder whether a law should
be amended: "[B]ut with one mouth and one voice they must all agree that the laws are all good,
for they came from God; and any one who says the contrary is not to be listened to. But an old man
who remarks any defect in your laws may communicate his observation to a ruler or an equal in
years when no young men are present" (Laws I 634el-6).
11. Schofield (1972) postulates that this abruptness is due to a copying error, and that the passage
in question (385b2-dl) actually belongs later in the text (at 387c5).
22 LIMITING THE ARBITRARY
we might assume that of the various words that are assigned to a thing, Hermo-
genes considers one to be true and the rest false. There would then no longer be
a contradiction between Cratylus' insistence that things have a correct designation,
and Hermogenes' view that things are called by whatever words (false or true) are
assigned to them. By admitting that false words exist, Hermogenes seems in effect
to have gone over to Cratylus' side. But his answer to Socrates' last question
proceeds as though the whole foregoing discussion of truth and falsehood had not
taken place:
HERM.: Yes, because in my view, Socrates, the correctness of words is nothing other
than this: I have the right to refer to anything by whatever word I choose, and you
by whatever one you choose. Likewise for communities: some, I find, have their own
unique words for things, whether one is comparing Greeks with other Greeks or with
foreigners. (385d7-e3)
For Hermogenes, whether the words used are true or false is finally immaterial to
the issue of correctness, which resides not in the moral, but in the political dimen
sion. Idia and polis are not direct opposites: the state is a projection of the individ
ual, and behaves like an individual relative to other states.12 There is an implicit
comparison here between the relation of the individual to the state, and of the
individual state to the community of states. Hermogenes believes that where
words are concerned the same principles should apply in both cases. In this light
he raises what has always been the most powerful argument for conventionalism.
If norms of language, law, or any other area of human activity are rooted in nature,
then they should be universal across human communities, since nature is the same
everywhere. But in fact norms vary widely from community to community. How
then can anything in language be natural, given the fact that different languages
exist? How can questions of 'true' and 'false' words even apply?
12. Cf. Republic IV 435el-3, where Socrates asks: "Must we not acknowledge [...] that in each of
us there are the same principles and habits which there are in the State; and that from the individual
they pass into the State? — how else can they come there?"
NATURE AND CONVENTION 23
says that "man is the measure of all things" — that however things appear to me,
they really are that way with regard to me, and likewise how they appear to you is
how they actually are for you — or does it seem to you that the reality of things has
a certain stability to it?
HERM.: There have been times, Socrates, when, in search of a way out of this confu
sion, I've found myself swept up in Protagoras' view — though I'm not totally sure
it's correct. (385e4-386a7)
With this invocation of Protagoras and his doctrine of metron anthröpos, 'man is
the measure', Socrates moves the discussion overtly out of the linguistic sphere.
From a philosophical point of view, Socrates and especially Plato found Protago
ras' doctrine repugnant because it eliminated consideration of the divine, and
elevated perception above reality (i.e. mere human perception over true celestial
reality). Its repugnance was greatly intensified by what they saw as the doctrine's
political and moral implications: it provided a justification for relativism and, by
some interpretations at least, anarchy.13
Socrates' reasons for raising 'man is the measure' in the present context are
however much more specific. The key segment of the question is the one which
asks whether "the reality [ousia, also translatable as 'substance' or 'essence'] I4 of
things" has "a certain stability to it" that is not dependent upon human perception.
By Socrates' interpretation, Protagoras' view would deny that this is the case,
holding instead (and incorrectly) that the real existence of things is in people's
opinions. It is as though he were asking Hermogenes: I know you believe words
exist by nomos — but what about the things they name? Do things exist in nature,
or merely in human thought and convention? Given his previously stated views,
one would not be surprised if Hermogenes sided with Protagoras, and in fact he
does admit to having earlier been 'swept up' in the latter's doctrine. While he now
renounces Protagoras, his renunciation is halting and half-hearted. Hermogenes
does not recognize what seems apparent to the reader: that his view on language
may itself represent a case of being carried away into Protagoreanism, since it
would free individuals to identify things according to their perception of them,
independent of any consideration of reality. Nevertheless, Hermogenes denies the
doctrine that appears to underlie his own view. Socrates now addresses the ethical
problem directly for the first time:
SOC: Really? Were you ever so swept up [in Protagoras' doctrine] that it no longer
seemed to you that any human being was evil?
13. Protagoras is treated by Socrates with considerable deference and respect in the dialogue that
is named for him, yet is clearly defeated in their debate. The defeat is not a given: in the Par-
menides, for example, Zeno and Parmenides overwhelm the more Socratic Aristoteles. Regardless
of what Socrates and his disciples may have thought of Protagoras personally, on the plane of ideas
he represented the very antithesis of everything Plato, at least, believed in. As the Athenian
Stranger says in Laws IV 716c4-6, "[...] God ought to be to us the measure of all things, and not
man [...]" (see also n. 8 above).
14. Anagnostopoulos (1971/72: 723, n.34) points out that the meaning of ousia in the Cratylus is
vague, a view shared by Baxter (1992: 4).
24 LIMITING THE ARBITRARY
HERM.: God, no! Just the opposite: I've had plenty of opportunities to see that cer
tain people are very evil, in fact quite a large number of them. (386a8-b4)
Next he draws a parallel between virtue and wisdom that seems obvious, but
actually requires a considerable leap of faith:
SOC: [A]re those who are very good also very wise, whereas those who are very evil
are also very stupid?
HERM.: That's just what I think. (386b9-cl)
Once Hermogenes has acknowledged the existence of wisdom, it is a short step
to proving the existence of truth:
SOC: But then, if Protagoras is right, and the truth is different for each person de
pending on how things appear to that person, can it be that some of us are wise and
others stupid?
HERM.: NO, it can't.
SOC: Then I suppose you'll agree fully that if wisdom and stupidity exist, it is utterly
impossible for Protagoras to be right. For in truth no individual would be more intel
ligent than anyone else if the way things appear to each person is true for that person.
HERM.: That's right. (386c2-d2)
Again, Socrates is leading Hermogenes into agreeing to propositions which, on
their face, appeal to common sense and basic morality — the alignment of wisdom
and goodness, stupidity and evil. Yet the propositions have political underpinnings
of a fundamentally antidemocratic sort. The existence of stupid people has always
served as the most powerful argument against giving everyone an equal political
voice. If both wisdom and stupidity exist, what could be more stupid than to let
anyone but the wisest rule? On the other hand, to deny the existence of wisdom
and stupidity would put one in the most absurd position of all: the denial itself is
so stupid that nothing else the denier says is worthy of attention.
The propositions Hermogenes here accepts again seem to contradict his own
views on the correctness of words. Because of this, many commentators have
described this section as the finishing off of Hermogenes, the reduction of his
views to absurdity. As noted earlier, some have criticized Plato for not letting
Socrates deliver the knock-out blow by pointing out the 'devastating' contradic
tions. But the very fact that he does not do so should give one pause to consider
whether Plato, rather than suffering a failure of will or of art, is endorsing Hermo
genes' view to a limited extent. Hermogenes has after all confounded every one
of Socrates' attempts to lead him into ethical contradiction. He has given the
'right' answers to all of Socrates' questions. The effect of this section is finally not
to deconstruct Hermogenes' logic, but to show that the correctness of words as he
conceives it is fundamentally unconnected to matters of truth, morality, and law.
When Socrates revisits truth and falsehood much later in the dialogue, it is
to demolish Cratylus' position, not Hermogenes'. For it is Cratylus, not Hermo
genes, who will deny the possibility of falsehood. This is a surprising turn of
events, given that the dialogue starts with Cratylus questioning the 'truth' of
NATURE AND CONVENTION 25
Hermogenes' name and Hermogenes denying that there is any truth to be ques
tioned. What this section makes clear from Hermogenes' perspective, and the later
section from Cratylus', is that, paradoxically, the correctness of words is inversely
connected to truth. For someone who believes in truth and falsehood, as Hermo
genes does, the correctness of words is irrelevant, not worth worrying about. Only
someone who has lost his faith in the reality of truth and falsehood will be as
concerned with the correctness of words as Cratylus is.
stupidly. But not only does Socrates assume that morality and intelligence are
linked, he also cites their combined existence as proof of the fixed reality of the
objects of intelligence.
With this notion of fixed reality Plato introduces the doctrine of Ideas or
Ideal Forms that is central to his thought on every subject, political as well as
epistemological.15 An example of a Platonic Idea, one of a handful that concerned
Plato deeply, is Justice (dike). We normally speak and act as if something called
'Justice' exists. We cry out to it or for it, ask where it is, feel satisfied upon per
ceiving that it has been done, evaluate particular interactions according to how
well or poorly they embody it. The belief that Justice really exists, that it has an
existence distinct from individual just actions, that it is not merely a linguistic
abstraction but a force, personal or impersonal, on earth, in heaven, or permeating
the universe, is a Platonic position. The broader philosophical and political conse
quences of a belief in such Ideas will become apparent as the discussion proceeds.
15. Although Plato's later writings downplay the theory of Ideal Forms, they remain sufficiently
close to it in spirit to forestall any supposition of an abandonment of the basic framework of
language and meaning established in the Cratylus.
NATURE AND CONVENTION 27
evident, which is why Socrates draws the analogy between speaking on the one
hand and cutting and burning on the other. These two types of actions seem as
dissimilar to the modern reader as they appear to have seemed to Hermogenes.
Indeed, those modern commentators who take Hermogenes to task for not being
consistent in his replies might ask themselves who in his right mind would insist
on applying the same criteria and judgements to speaking as to cutting and burn
ing. But Socrates will shortly offer as uncontroversial the statement that one of the
two primary purposes of words is "distinguishing among realities" (388b 13), in
other words to cut up reality.
This passage introduces the conception of 'speaking', when the enquiry up
to this point has been about 'words'. The difference between the two, which
reveals subtle information about the Greek conception of language, is made
manifest in the next interchange.
SOC: And isn't word-making a part of speaking? After all, in word-making one
makes statements.
HERM.: Of course.
SOC: And isn't word-making a kind of action, given that speaking is a kind of action
that concerns things?
HERM.: Yes. (387c6-ll)
Evidently, it was not obvious that word-making or naming is an action. This had
to be deduced from the more obvious fact that speaking is an action.16 Socrates
knows that Hermogenes will not automatically take an assertion made about
things and apply it to actions, though neither will he rule out doing so. Neverthe
less, things and actions have different status not only for the grammarian but for
the philosopher. Hermogenes recognizes the difference, and this is what leads him
into the seeming contradiction. Human actions are normally conceived of as
purposeful; things, however, may be so (as in the case of tools) or may not be. To
say that words are 'tools' is to contend, by means of metaphor, that they are
purposeful. The same point can be made about speaking — an action — more
directly and without recourse to metaphor.
The two preceding passages are more easily understood in the light of
Plato's other discussion of cutting and burning, in the Gorgias (476ff.), where it
is people who are being burned or cut. The question is whether it is for their good.
The answer is yes, if the burning or cutting is done with the tekhn 'art' or 'sci
ence' of a physician, performing surgery or cauterizing a wound in order to heal
the body, or by someone using the art of exacting a punishment in order to heal
the soul. But for just anyone to go around cutting or burning others without tek
hnē, simply exercising his own will, would be monstrous. Again, Hermogenes
16. In Greek, onomazein "to name" is a denominal verb, derived by adding a verbal ending onto
a noun, whereas the same is not true of legein "to speak". Thus naming, unlike speaking, has a
direct and automatic association with a thing, namely, a name. 'To name' is derived from 'name',
while conversely 'speech' is derived from 'to speak'.
28 LIMITING THE ARBITRARY
cannot disagree with what Socrates says about cutting and burning without impli
cating himself as amoral. Thus he is seduced into making parallel statements about
language even though it is far from obvious that the same moral consequences
pertain.
If speaking is indeed a purposeful action, the purpose toward which it is di
rected is a constraint upon the sort of linguistic free will Hermogenes advocated
in the example of the words for 'man' and 'horse'. Socrates sees to it that the
importance of sublimating one's will to the nature and purpose of the instrument
is made explicit:
Soc.: And so words must be made with what was naturally given, and in the way
naturally given for naming things and for things to be named, rather than the way we
would wish, if we're to be consistent with what we've said up to now? And in this
way we may accomplish something and make words, otherwise we will not; am I
correct?
HERM.: Clearly, as far as I'm concerned. (387d4-9)
Socrates returns to the 'tool' metaphor, and asks what tool is used for the activity
of naming or word-making. Hermogenes answers, "A word".
SOC: Well said! Then a word too is a kind of instrument.
HERM.: Of course. (388a8-9)
17. This despite Plato's insistence that abstractions are more 'real' than concrete objects. In
Phaedrus 263a5-b9 Socrates says that the rhetorician "ought to make a regular division" between
the classes of things like iron and silver, on whose definition everyone agrees, and things like
justice and goodness, on which they differ. One of the most amusing passages in Plato occurs in
Gorgias 490c8-491a3, when Callicles refuses to countenance any more of Socrates' 'commonplace
parallels'.
NATURE AND CONVENTION 29
pick him out at least from other members of his family, it teaches false informa
tion about him. And Hermogenes could go on holding that the switching of 'man'
and 'horse' would make no difference, so long as the two are distinguished from
one another, since neither word teaches anything about what it names.
First of all, however, words are not created equal with regard to the func
tions of discriminating and teaching. The words 'man' and 'Hermogenes' both
apply to the first speaker in the present dialogue. If we were to ask whether that
individual is properly called 'man', the first thing we should want to know is
whether he is an adult human of the male sex. If so, the common-sense answer to
the question would be that he is properly so called, since the word 'man' correctly
distinguishes him from women, boys, girls, eunuchs, and other animals, vegeta
bles, and minerals. We could also say that calling him 'man' teaches something
true about him, provided the hearer understands the meaning of 'man'. On the
other hand, if we were to ask whether he is properly called 'Hermogenes', the
common-sense reaction would be to treat this as a historical question, and to
enquire whether or not that is the name he was given at birth. In the same way, his
patronymic pai Hipponikou 'son of Hipponikos' distinguishes him from (most)
other men named Hermogenes and teaches another historical fact about him.
Again, the question of the correctness of the patronymic would normally be
understood as a question about Hermogenes' paternity, which of course might
have important legal consequences where inheritance is concerned (Hermogenes'
inheritance will be raised directly by Socrates at 391c2). The patronymic Hip
ponikou is in a sense more like the common noun 'man' than like the given name
Hermogenes, since like 'man' it picks out a class with a physical basis. In the full
name Hermogenes pai Hipponikou it is the patronymic that teaches independent
historical information about the individual, while the given name teaches only the
historical fact that it is the name that was given to this individual at birth.18
Somewhere between the two lies the case of 'inherited' given names, such as
Plato's original name Aristocles, after his grandfather.
The kind of 'discriminating' that diakritikon represents is not ethically neu-
18. The case of modern first and last names is essentially the same. Asking whether my elder son
is really a Joseph is different from asking whether he is really a Julian. The first question concerns
his paternity and legitimacy and could have serious legal ramifications; the second question is
merely about the aesthetics of, and motives behind, a fundamentally arbitrary choice. In telling
Hermogenes that his name is incorrect because he does not display the characteristics of someone
'born of Hermes', Cratylus is first of all treating a given name as if it were a patronymic, and then
assuming that a patronymic functions as a criterion of character. A patronymic can encode some
thing about a person's origin, but one's origin is not equivalent to what one is. Therefore, legal
considerations aside, the patronym finally teaches us no more about the person than does the given
name. Plato recognizes that the same is true with words other than names: some, though not all,
appear to teach us something about the thing they name, but the information they teach is unreli
able. We need corroborating evidence to prove that it is true, and if we have such evidence, then
we do not really need the information which the word purports to teach. The French word for a
dandelion, pissenlit, literally 'piss-in-bed', appears to teach us that this plant is a diuretic; but in
fact what it teaches us is that whoever began calling it that believed it was a diuretic, which may
or may not actually be the case (see further Chap. 2).
30 LIMITING THE ARBITRARY
tral. Rather the word suggests a critical discrimination. The mention of teaching
drives this point home, since a didactic purpose implies both ethics (knowing right
from wrong, true from false) and politics (discriminating between the taught and
the untaught person). These are the same considerations that guarantee the ab
surdity of switching 'man' and 'horse'. The functions of words and word-making,
then, are discriminatory, to pick out the ousia of what is named — that is, its
essence, what belongs to it and it alone — and didactic, to transmit that knowl
edge from the few who are capable of perceiving it to the many who are not. Note
that 'communication' as such does not enter into the discussion, nor will it do so
until much later (434el; see p. 73 below).
Finally, the idea that words serve to pick out the essence of things is not the
same as the naturalistic view that words reproduce the things named. Hence it is
quite compatible with any but the most radical version of conventionalism, which
holds that things and truth are as conventional as the words that name them, and
which Hermogenes clearly does not maintain. He is a linguistic conventionalist
only. His intellectual 'crime' is not radical scepticism, but inconsistency.
19. The metaphor of weaving is developed in far greater detail, and again as part of a discussion
of lawgiving (though not in a linguistic context) in Statesman 279a7-283b5.
NATURE AND CONVENTION 31
SOC: Then whose work will the weaver be using well when he uses the shuttle?
HERM.: The carpenter's work.
SOC: Is just anybody a carpenter, or only those who know the art?
HERM.: Those who know the art.
[...]
SOC: All right. Then whose work will the teacher be using when he uses the word?
HERM.: I don't know the answer to that one.
SOC: Can you answer this one: who provides us with the words we use?
HERM.: I certainly can't!
SOC: Doesn't custom seem to you to be their source?
HERM.: Apparently so.
SOC: Then it will be the work of the establisher of customs and laws that the teacher
is using when he uses a word?
HERM.: That's how I see it. (388c9-e3)
A crucial distinction is made here between nomos 'custom' and the nomothetēs,
the 'establisher of customs and laws'. When Hermogenes cannot say whose work
words are, Socrates asks "who provides us with the words we use". The two are
not the same. The answer to the second question is nomos', Hermogenes has to be
prompted to give it even though he himself asserted back at 384d5 that a word
attaches to a thing only by the nomos and ethos "of those who started calling it
that and those who picked up the habit". In the present passage, Socrates does not
agree to that strong assertion about linguistic correctness, but concedes this much
of it: it is through nomos that words come to us. We acquire language from — it
is given to us by — the customs of our community.
That is not a sufficient answer for anyone who cares about correctness. But
it is sufficient to account for the linguistic behaviour and usage of the mob, who
couldn't care less. Nor does Plato care about how hoi polloi speak; they should
simply follow the conventions of language such as they are, with no thought
toward changing them —just as they should simply obey the existing laws, and
leave any consideration of their correctness to those with knowledge of such
things, namely, the legislators. It is the lawgiver's business to establish correct
language.
The movement from nomos to nomothetēs, from law to lawgiver, is a
movement from present (synchrony) to past (diachrony), a historicization of the
question of linguistic correctness. The focus henceforth will not be on how lan
guage is used or acquired, but how it was created. The origin of language is
equally pertinent (or equally irrelevant) to the physis or the nomos position. Nei
ther Socrates nor Hermogenes believes in nomos as a god or demigod, and the fact
that language is acquired from nomos does not exclude the possibility that it
operates through physis. But it can only operate through physis if it was created
following a principle of physis.
Socrates therefore leads Hermogenes one step further back, to ask who cre
ated those customs from which we acquire language. The answer is obvious, if
tautologous: the custom-creator. However it does not seem tautologous to Hermo-
32 LIMITING THE ARBITRARY
20. While Laws I 624a3-5 notes that the nomothetai of Crete (Zeus) and Lacedaemon (Apollo, via
Minos and Lycourgos) were gods, that of Athens was a man (Solon) whose status was nevertheless
legendary. According to another myth, recounted in Laws III 676a8ff., the first lawgivers emerged
after Deucalion's Flood, when the few surviving communities spread down from the mountaintops
to repopulate the valleys. As they encountered one another, they needed arbitration among their
customs, which had grown apart during the years of isolation.
NATURE AND CONVENTION 33
it: the tekhnë of the nomothetēs. The same point is made in a political context by
the Eleatic Stranger in the Statesman (296e4-297a5):
As the pilot, by watching continually over the interests of the ship and of the crew,
— not by laying down rules, but by making his art [tekhnë] a law [nomos], — pre
serves the lives of his fellow-sailors, even so, and in the self-same way, may there
not be a true form of polity created by those who are able to govern in a similar
spirit, and who show a strength of art which is superior to the law?
The pilot and the nomothetës could, if they wished, simply 'lay down rules' of an
arbitrary sort to maintain order. But unless those rules are informed by the appro-
priate tekhnë, chances are very great that disaster will ensue.
What then is the tekhnë of word-making? It has to do, as Socrates is about
to tell us, with what the lawgiver has in his mind when creating words, the princi
ple that he follows: either physis, something else, or nothing at all (i.e., 'pure'
nomos). The analogy between the lawgiver and the carpenter is drawn out further:
Soc.: All right, then, think: when the lawgiver establishes words, what is he looking
toward? Think about it in light of what we've already said. What does the carpenter
look toward when he makes the shuttle? Isn't it something which has as its nature to
weave?
HERM.: Of course.
SOC: Oh? And if the shuttle should break while he's making it, will he look toward
the broken shuttle when making a new one, or toward the same ideal form he had in
mind when making the one he broke?
HERM.: Toward the form, I'd say.
SOC: Then wouldn't we be right in calling that the real shuttle?
HERM.: I'd say so. (389a5-b7)
This takes us back to 386d8 and the doctrine of Ideas: the 'ideal form' that is the
'real shuttle' is precisely what we would call the Platonic Idea of the shuttle. Any
individual shuttle will be judged as being better or worse, and more or less 'real',
according to how well it approximates to this Ideal Form. This is the essential
Platonic moment: the real shuttle is not the physical shuttle, but the idea of the
shuttle as determined by its nature and purpose. Obviously, a shuttle which breaks
while being made is far from ideal, and if the carpenter were to try to remake the
same shuttle he would simply be repeating his error.22 Socrates reiterates that the
Ideal Form of an instrument is determined — naturally — according to its pur-
pose.
21. In Laws IV 719e7ff. and IX 857b9ff, Plato goes a step further and says that it is not enough
for the legislator or the physician simply to issue orders based upon his tekhnë; the orders must be
combined with a discourse of persuasion and education, so that the citizens or patients will under
stand why it is that the orders should be obeyed.
22. It is not immediately apparent what the analogy would be with language. One could imagine
a word 'breaking' under a theory of nomos if it was no longer able to fulfil its function of separat
ing reality — for example, if the community were unable to agree upon a precise definition of
'justice' or 'murder' — or under a theory of physis if its sounds changed so much that it ceased to
be recognizable. Socrates does not pursue this aspect of the analogy, though the matter of sound
change is taken up later in the dialogue (see Chap. 3 below, especially pp. 69-83).
34 LIMITING THE ARBITRARY
SOC: Having discovered the instrument naturally suited to a particular purpose, one
must render it out of the material one is working with, not in whatever way one
wishes, but in the natural way. It is as though one must know how to put into iron the
drill naturally suited to each particular job.
HERM.: Of course.
SOC: And into the wood the shuttle naturally suited to each job.
HERM.: That's right.
SOC: Because by nature there seems to be a particular shuttle for each particular
type of cloth, and likewise for the other things.
HERM.: Yes. (389c4-d3)
Here Socrates insists again that an artisan must not make instruments in accor
dance with his own will. The individual will must be restrained in favor of fulfill
ing the purpose for which things are made. For example, it is a fact of nature that
a drill must be made of harder material than the thing it is to drill. Thus an iron
drill for use on wood is a good tool, whereas a wooden drill for use on iron would
fail in its purpose, and the 'toolmaker' who made it would really be no toolmaker
at all. Socrates now extends the analogy to language. The passage which follows
is in many ways the crux of this first section of the Cratylus:
SOC: Well then, my good friend, isn't there also a word suited by nature for each
thing, which the lawgiver must know how to put into sounds and syllables? And
mustn't he keep his mind on the very essence of what a word is, with all the words
he creates and applies, if he is to be a master establisher of words? And even if not
every lawgiver puts it into the same syllables, they must not ignore it: for not every
toolmaker puts the drill into the same iron, even when making the same instrument
for the same purpose. However, as long as they reproduce the same idea, even in dif
ferent iron, still it is truly the instrument, whether it is made here or abroad. Would
you agree?
HERM.: Of course.
SOC: Then will you evaluate the lawgiver, domestic or foreign, in the same way? So
long as he reproduces the ideal form of the word proper to each thing, in whatever
syllables, you won't judge him to be any worse a lawgiver, whether he is from here
or anywhere else?
HERM.: Of course not. (389d4-390al0)
Socrates has opted for the third choice, and identified it as Ideal Form. If indeed
the wordmaker looks at something outside language, rather than creating 'pure'
conventions, then the nomos position fails to account for linguistic correctness.
The physis position comes closer than nomos to accounting for linguistic correct
ness because it recognizes the existence of external criteria, but it is mistaken or
at least ambiguous about the identity of those criteria. What the Sophists generally
understood physis to mean was physical, material nature, with no consideration
of the Ideal Form that for Plato was more real than such mere appearances.
Yet Plato himself complicates matters by using the word physis inconsis
tently. Since the 7th century BC physis had served to indicate physical reality, and
to distinguish physical (and secular) philosopher-scientists from religious thinkers.
By the 5th century, when religious thought had been banished from the philo
sophical 'mainstream', physis came to be defined as the antithesis not of the
divine, but of those human creations called nomoi. As such, its meaning was
redirected toward all that was not the product of human creation — the whole
'natural' universe in both its physical and ideal dimensions. It is in this latter sense
that Socrates declares ideal words to be fitted by physis to the things they desig
nate, even though he has just insisted that neither the words nor the things they
designate exist in the physical dimension. In later works Plato will be much more
clear about the difference and will restrict physis to apparent, material nature (cf.
Laws X 889a4-892c7).
Although ideal words are the instruments 'naturally' fitted to distinguish
and teach about particular Ideal Forms, actual words exist by virtue of nomos. The
degree to which actual words correspond to ideal words, hence to Ideal Forms, is
a matter of politics and ethics. This becomes clearer if one transfers nomos to its
everyday meaning of 'law'. Plato believed very firmly in 'natural' (later, divine)
laws that exist on the ideal plane together with concepts like 'justice'. Still, their
existence would not guarantee that any actual existing law is 'natural'. On the
contrary, actual laws exist entirely on the plane of human convention. One hopes
— and again, this is a matter of politics and ethics — that actual conventional
laws will in fact correspond to natural law. But Plato would be the first to declare
that this is the case nowhere but in his imaginary, Utopian Republic.23
23. In this sense I agree with the thrust of Baxter's (1992: 4) view that "Socrates' theory is pre
scriptive, not descriptive [...]". However I fear that this pair of terms may create more misunder
standing than insight, especially since they have a set meaning in linguistics that does not fit the
case at hand. Most linguists, if told that Socrates' position is prescriptive, would imagine him
setting forth rules for socially inferior speakers to conform with socially sanctioned norms. Baxter
makes clear that this is not what is meant, for instance when he writes that for Socrates, "The
language we should all be speaking is only possible to those with at least some divine blood in
them" (p. 12) — an overstatement, but one which nevertheless draws into question the appropriate
ness of 'prescriptive' here. Still more confusing is Baxter's attempt to extend the dichotomy, when
he says that while "Cratylus' theory certainly looks like a prescriptive theory", in fact "Cratylus
does not recognize the description/prescription distinction: what Greek ought to be is what it
(broadly speaking) is [...]. Cratylus' view seems to be that Greek is fundamentally sound, that the
ideal language need not be distinguished from the vernacular, merely that some so-called 'names'
36 LIMITING THE ARBITRARY
like Hermogenes are not, thus used, names at all" (p. 11). The textual evidence does not support
the characterization of Cratylus as believing that Greek is fundamentally sound, that the ideal
language need not be distinguished from the vernacular, or that it is 'merely' 'some' so-called
names that are really not names at all. Baxter is here extrapolating from things Cratylus fails to say.
But to conclude from Cratylus' rejection of 'incorrect' words as not really being words at all that
his position is therefore descriptive, not prescriptive, seems misguided and potentially misleading.
24. "Cratylus is right" is implicitly part of the 'then' result of an 'if'-clause: if all the preceding
contentions are correct, then Cratylus is right; and Socrates will ultimately show that they are not.
The statement marks the conclusion of his argument against Hermogenes and the turning point of
the dialogue. After creating the illusion that he is defending Cratylus' position, Socrates now sets
out the groundwork for its undoing. Note too that in fact all Cratylus ever contended, according
to Hermogenes, was that "there is such a thing as right and wrong where words are concerned, and
that it is connected to the nature of the things they name [...] that is in the words themselves, and
that is the same for all people, Greeks as well as foreigners" (383a4).
NATURE AND CONVENTION 37
25. In addition, false representations of the gods are banned from books and even from the stories
mothers and nurses tell children {Republic II 377b11-c5), and the practitioners of any art who
cannot conform to the 'simple' style are banished from the state {Republic III 399e8ff.). In Laws
the works of poets {including hymns to the gods, Laws VIII 829dl-e4) are subject to the most
extreme censorship. In the case of the tragic poets, this is defended on the grounds that the lawgiv
ers are themselves tragic poets, inventing a state that is "an imitation of the best and noblest life,
which we affirm to be indeed the very truth of tragedy", and that therefore the lawgivers would be
mad to give those 'other' tragic poets license to voice their rival vision without first determining
whether their poetry was fit for publication {Laws VII 817a2-d3). It is not poetry as such that Plato
objects to, but the corrupting influence of the untruths it contains. Amid the calls for censorship
and banishment, Socrates is profusely apologetic to the poets: "And we must beg Homer and the
other poets not to be angry if we strike out these and similar passages, not because they are
unpoetical, or unattractive to the popular ear, but because the greater the poetical charm of them,
the less they are meet for the ears of boys and men who are meant to be free, and who should fear
slavery more than death" {Republic III 387b1-6). At his trial, Socrates says: "What would not a
man give if he might converse with Orpheus and Musaeus and Hesiod and Homer? Nay, if this be
true, let me die again and again" {Apology 41a6-8). And in prison awaiting death Socrates himself
wrote verses, in obedience to a recurring dream telling him to "compose music" {Phaedo 60e4-7).
26. Ironically, given the present context, one of Plato's strongest indictments against the poets is
put in Protagoras' own mouth: "Now the art of the Sophist is, as I believe, of great antiquity; but
in ancient times those who practised it, fearing this odium, veiled and disguised themselves under
various names, some under that of poets, as Homer, Hesiod, and Simonides [...]. But that is not my
way, for I do not believe that they effected their purpose, which was to deceive the government
[...]" {Protagoras 316d3-317a4).
38 LIMITING THE ARBITRARY
This completes the first part of Socrates' response to the debate between
Hermogenes and Cratylus as to whether linguistic correctness is a function of
nature or convention. His answer is: both, and neither. If it is true that actual
words exist and operate by a sort of convention or agreement, it is not like the
typical agreement of everyday experience, which requires simply that two or more
people share a common will, which can be changed at will. Rather, they are like
instruments, created for a specific purpose, and modelled after ideal words. If it
is true that these ideal words exist in 'nature', it is not the physical, material nature
of everyday experience, as the Sophistic backers of physis take it to be, but a
celestial form perceptible only to the wise. Actual words are correct or incorrect
insofar as they approximate to the ideal words; but that is a matter for considera
tion only by the wordmaker, the person skilled in dialectic, and the teacher. Thus,
Socrates historicizes and politicizes (more specifically: undemocratizes) the
notion of convention, and desecularizes that of nature.
The ability to judge how well words embody Ideal Form is exceedingly rare,
and that rarity is the seed from which the political dimension grows. We can say,
paraphrasing Heraclitus, that language is one place where physis really likes to
hide itself. As a result, sublimating one's will to physis (understood broadly)
where language is concerned actually means bowing to the will of those empow
ered as the interpreters of physis, whoever they may be at a given time: priests,
poets, philosophers, scientists, rhetoricians or politicians. Furthermore, if an actual
word fails to embody the corresponding ideal word, it is because one or more
persons have allowed their own will to take precedence over their cultural duty to
maintain the instrument of discriminating and teaching. This is the ethical or
moral dimension of linguistic correctness.
Socrates has indirectly responded to the difficult question suggested by
Hermogenes back at 385d9: how can language be natural when different lan
guages exist? The implication of Socrates' view is that language is 'natural' on a
completely different plane of reality from the one on which actual languages
operate. Given that word-making is a function with a purpose, there must be one
ideal way of carrying out that function, together with an infinite number of ways
of actually carrying it out. None of the latter can be exactly identical with the ideal
way, since an ideal is by definition an idea and not a material thing or action. But
the actual ways of carrying out the function are good or bad, right or wrong, true
or false, according to how well they embody the ideal way. In linguistic terms, this
means that there exists an ideal language, to which actual languages approximate
insofar as their words, which are human creations, embody its forms. As to what
kind of existence this ideal language and the ideal words which constitute it might
have, the answer cannot be broached until the next chapter. The various strands
of enquiry which Socrates has pursued up to this point will now give way to a very
different kind of investigation, though some of them will resurface toward the end
of the dialogue.
CHAPTER 2
WORDS AND TRUTH
CRATYLUS 391D2-422E1
men call the Skamandros (391e4). The name Xánthos is transparently derived
from xanthós 'yellow', thus 'Yellow River', whereas Skamandros is a proper
name with no obvious source. Hence, the gods use a name that is motivated by
the colour, and therefore the nature, of the river, while men use an arbitrary,
conventional name. The importance of this passage can hardly be overesti
mated, yet has been completely overlooked. Commentators treat the passage as
a simple statement by Socrates that whatever word the gods use must be cor
rect, without noting what distinguishes the divine from the mortal name of the
river: only the divine name indicates something about its nature. Skipping for
the moment past the more complex second example, we may note that the third
one, the hill on the plain of Troy which "men call Batieia and the Immortals
the tomb of the bounding Myrina [an Amazon]" (Iliad ii.813-14), resembles
the first in that here again the gods use a transparently motivated designation
while men use a word that is opaque, and thus presumably arbitrary.
The second example, from Iliad xiv.291, is the bird which "The gods call
khalkis, but men call kymindis" (392a3). Here again, as in the first case, the
gods use a word transparently motivated by the bird's being the colour of
khalkos "copper". However, the human word also suggests a colour, that of
kyminon "cumin". If it is really "much more correct [...] for this bird to be
called khalkis than kymindis", that is presumably because the bird's colour
really is more like that of copper than like that of cumin. Again, the link be
tween this passage and the much later exchange about imitation (422el-425b5,
430a8-434b9) has been generally ignored. But what Socrates implies here is
consistent with what he states openly there: words are correct insofar as they
accurately reproduce some part of the essential nature of what they designate.
The human words Skamandros and Batieia do not even attempt to reproduce
anything; kymindis does, but apparently succeeds less well than the word used
by the gods. Socrates asks, referring to Homer:
Soc.: Don't you find that he gives great and wonderful information there about the
correctness of words? For surely it's evident that the gods call these things by
words which are naturally correct. Or do you disagree?
HERM.: I know perfectly well that if they call things they do it correctly. (391d6-
e3)
Hermogenes has finessed his way nicely out of another Socratic corner, giving
an answer that is sceptical yet pious. His answer implies that, while the gods
cannot err, perhaps the information given by the poets is not entirely accurate.
In view of Socrates' own reservations about the poets (see above p. 37), he is
not about to argue otherwise. However lofty the status of Homer and Hesiod, it
would be hubris to claim that these poets really knew the true language of the
gods — unless, conceivably, their poems were like oracles uttered in a trance,
as Socrates will claim of his own etymological speculations later in this sec
tion.
WORDS AND TRUTH 41
1. The choice among several conceivably 'correct' names would likely have been dismissed by
Plato as merely an accidental matter, judging by the following passage from Laws IV 704,
where the Athenian Stranger asks about the city they are planning: "And now, what will this
city be? I do not mean to ask what is or will hereafter be the name of the place; that may be
determined by the accident of locality or of the original settlement, — a river or fountain, or
some local deity may give the sanction of a name to the newly-founded city [...]".
42 LIMITING THE ARBITRARY
regularly used for any group of mixed sex. The supposed deduction that the
women must have called the boy Skamandrios is therefore patently absurd in
itself;2 but it also happens to contradict the passage of the Iliad (vi.402-403) in
which the boy makes his first appearance: he is described as the one "whom
Hector called Skamandrios, but the others Astyanax". We can infer that the
latter name was bestowed upon the boy as a tribute to Hector by his fellow
Trojans, so that it would be unbecoming for Hector himself to use the name. It
is possible that Socrates is alluding to a different textual tradition of Homer
than the one that has come down to us; but the other Homeric allusions in the
Cratylus are consistent with the text we have. Together with the clearly hu
morous intent of the reference to the Trojan men and women, this suggests in
stead that Plato was making a joke which few in his contemporary audience
would have missed.
The relative 'correctness' of the names Astyanax and Skamandrios thus
appears in a new light. Since no one would dispute Hector's standing as the
'wisest' of all the Trojans, his name for his son, Skamandrios, would seem to
be the more correct. Furthermore, Astyanax would never be 'Lord of the City'
(any more than Hermogenes is really 'Lucky'), since like his father he was
killed during the invasion of Troy by the Achaeans. Yet how can Skamandrios
be a 'correct' name when it is derived from the 'incorrect' name Skamandros,
which men use for the river that is correctly called Xanthos? Plato has devised,
doubtless in parody, the perfect literary-philological 'proof' of something his
audience would have found patently false.
Another paradox underlying this discussion is now brought to light by
Socrates:
SOC: Well then, my noble friend — didn't Homer also give Hector his name?
HERM.: Why do you ask?
SOC: Because that name seems to me similar to Astyanax, and both names appear
to be Greek: anax 'lord' and hektōr 'holder' mean almost the same, both being
words for a king. (393al-b7)
These names "appear to be Greek", rather than Trojan, which suggests that
they are Homer's inventions or perhaps translations. This point serves to re
mind us that it is at best only Homer's opinion about the correctness of names
that is under investigation here. Hermogenes agrees with Socrates that he has
"got onto the track of Homer's doctrine concerning the correctness of words"
(393b3). And so he has: for this section has revealed how hopelessly muddled,
self-contradictory, and absurd is Homer's opinion about linguistic correctness.
That conclusion is quite consistent with Plato's view of the poets as expressed
2. On the other hand, Socrates and Hermogenes were far from unusual in believing that "the
women [...] of a city, regarded as a class in general" are less wise than the men. This opinion is
reflected in the political absence of women from the democracy of Athens. Note, however, the
special qualification "regarded as a class in general" that Socrates goes out of his way to add;
the implication being that individual women may be wiser than individual men.
WORDS AND TRUTH 43
elsewhere in his work. He has also opened the way toward a later suggestion
that the making of language is always dependent upon the opinion of the lan
guage-maker. Moreover, he has shown that literary-philological evidence and
arguments can be used to construct an argument that, while patently contrary to
the truth, can nevertheless force another party into conceding absurd conclu
sions, especially if that party does not know his classics as well as he ought to,
and is loathe to admit it.3
These exchanges over Homer on the words used by gods and men is a
microcosm of the etymological section as a whole. Socrates begins apparently
in earnest, establishing explicitly that the words used by gods must be correct,
and implicitly that they alone accurately embody the nature of what they des
ignate. However he then proceeds to cases that show that we cannot really
learn anything about the correctness of words from the poets — contrary, we
can surmise, to the claims of certain Sophists about the power of their philol
ogical arguments.4
3. Cf. Protagoras 347b9-e7, where Socrates says: "[...] I would rather have done with poems
and odes [...]. The talk about the poets seems to me like a commonplace entertainment to
which a vulgar company have recourse [...]. And a company like this of ours, and men such as
we profess to be, do not require the help of another's voice, or of the poet whom you cannot
interrogate about the meaning of what they are saying; people who cite them declaring, some
that the poet has one meaning, and others that he has another, and the point which is in dispute
can never be decided [...]".
4. See also the dialogue Ion, a critique of a rhapsode who specializes in explicating Homer.
44 LIMITING THE ARBITRARY
why do noble fathers pay Sophists to teach virtue to their sons? Plato believes
that the sons of kings, even virtuous kings, are likely to be degenerate on ac
count of "the evil life which is generally led by the sons of very rich and royal
persons" (Laws III 695e5-696al).
In the midst of this discussion Socrates opens an entirely new line of lin
guistic argumentation. Following his admonition to Hermogenes that he not let
himself be tricked, he continues:
SOC: And whether this is signified in these syllables or others doesn't matter. Nor
does it matter at all if some sound is added or taken away. That means nothing, so
long as the reality of the thing remains firmly evident in the word.
HERM.: What do you mean?
SOC: Nothing complicated. You know for example that we have words for the
letters of the alphabet; we don't just say the sound of the letter itself, except in four
cases: e, u, o, and ō. For the other vowels and consonants, as you know, we create
names by adding other letters. But so long as we include the actual sound in ques
tion and make it evident, whatever word indicates that sound to us is correct. For
instance the letter bēta: you see that adding ē, t, and a, as the lawgiver wished,
does not keep the nature of this letter from being evident in the whole word. That's
how well he knew how to establish words for the letters. (393dl-e9)5
So long as the reality of the thing is made evident in the word, it does not mat
ter whether it is expressed in one set of syllables or another. This represents a
significant amendment to — even a jettisoning of — the view originally put
forward at 389d4 (see p. 34 above), that an "ideal word" is "suited by nature
for each thing". Now Socrates is saying that any number of correct words are
conceivable for an object, so long as they capture its essence and make it plain.
This is Socrates' answer to the objection Hermogenes raised to linguistic cor
rectness, when he asked how there can be correctness when different languages
exist. The fact that words can be correct or incorrect does not entail that only
one correct word exists. One might have assumed otherwise when in an earlier
passage Socrates spoke of the 'ideal word' for each thing. But henceforth he
will not refer to ideal words or an ideal language, only to the non-linguistic
Ideal Form of what is to be named. It is this that the language-maker must keep
in view and embody through imitation in sound.
It is probably no accident that to illustrate how it matters not at all "if
some sound is added or taken away" Socrates chooses a word for a letter of the
alphabet. The reality or essence of the word bēta is b, and the other three
sounds are contingent or accidental. Presumably any word for this letter (in
cluding English bee and French be) is correct so long as it contains that es
sence; oosh would be an incorrect word for b. The remainder of the word bēta
is mere ornamentation, pure nomos, that "means nothing". It is an excellent
example; but it falls into the category of metalanguage, language about lan-
5. These letters too later acquired 'names' of their own: epsilon, upsilon, omicron ("short o")
and omega ("long o").
46 LIMITING THE ARBITRARY
(The start of this passage has already been commented upon above.) The im
plication is that the essence of the word basileus 'king' should be contained in
the word for the king's son, as a reflection of the shared royal essence of the
two men. But the contingent letters may be entirely different, just as bēta and
bee both express the essence of b yet have no other sound in common between
them. Socrates will give some examples momentarily, but first he offers a
pharmaceutical metaphor.6 The physician knows what is the active ingredient
of a particular drug and what its effect should be, whereas the layman is apt to
be misled by dyes and perfumes — superficial, contingent elements — into
paying much more than the drug is worth, or indeed into buying a worthless
placebo instead of real medicine. The role of the physician includes protecting
the layman from the unscrupulous pharmacist. And just so, the role of the
teacher or dialectician is to protect the public from those who would delude
them into believing 'empty' rhetoric, words with no real value but full of col
ours and perfumes. Indeed, this is just what Socrates is in the process of doing
now, by showing Hermogenes how easy it is to conjure up such worthless dis
course. He returns to the earlier example:
SOC: As we just now said, Astyanax and Hector have no sound in common except
t, and yet they signify the exact same thing. And what sound does Arkhepolis
'Lord of the City' share with the two of them? Yet it too means the same [...]. And
we might find still others which, while sounding different in their syllables and
letters, express it with the same effect. (394b7-c9)
Socrates has just taken his argument in a perfect circle. The original question
was whether such a thing as linguistic correctness exists; to this Socrates first
answered that for everything there is an ideal word which expresses its es-
sence, then amended his answer to say that any word which expresses the es
sence is correct. One would nevertheless expect that two things with the same
essence, even if they do not share an ideal word, are at least called in ways
which partly resemble one another, as in the case of Greek beta, French be, and
English bee. But here is an instance of how metalanguage is different. Words
for letters of the alphabet are exceptional: elsewhere in language, synonyms
and partial synonyms are far from rare, and it is seldom the case that they share
part of their form in common, in parallel with their shared meaning.
At the end of this argumentative circle, however, a trap lies waiting. To
assert that words are correct insofar as they embody the essence or Ideal Form
of the thing named, and then to add that this reproduction may be only on the
level of meaning, not sound, is to leap headlong into conventionalism, as
shown by the following argumentative chain:
1. the mare and the foal share the same equine essence;
2. to be correct, the terms which designate them should both reflect this
equine essence;
3. words reflect essence independently of the identity of any particular
syllable or letter;
4. therefore, the words for mare and foal may reflect the equine essence
only on the level of meaning;
5. therefore (what Socrates does not say) any words for mare and foal,
simply by virtue of having these meanings, reflect the equine essence
and are therefore correct. This is in effect identical to the position
initially taken by Hermogenes.
The problem lies in that, at least insofar as ordinary (non-philosophical) lan
guage is concerned, the 'meaning' and the 'essence' of a thing are not clearly
distinguishable. To say that the connection between a word and its meaning is
correct if the word expresses the essence of the thing named makes sense only
so long as 'word', 'meaning' and 'essence of thing named' occupy three dis
tinct existential planes. But that is not what Plato has suggested. Rather, he has
insisted that the essence of the thing named is the true meaning, in effect col
lapsing any distinction between the two. That leaves us with two components
of words: first, form; and second, meaning = essence. Now that he goes a step
further, and says that finally the form does not matter, words are reduced to
meaning = essence alone. So the earlier hypothesis that "the connection be
tween a word and its meaning is correct if the word expresses the essence of
the thing named" now becomes: "the connection between the meaning of a
word and its meaning = essence is correct if the meaning expresses the essence
of the meaning", which passes beyond tautology and into nonsense.
But the logical incongruity goes unremarked, and this section serves as a
prologue to the long series of investigations into word origins which follows.
48 LIMITING THE ARBITRARY
There Socrates will add, subtract, and transpose sounds as necessary to arrive
at cogent etymological explanations, presumably reflecting the practice of
etymologists of his time. One more point about this section requires mention,
though. What Socrates has proposed here is a historicization of physis, the
physical reality that constitutes meaning, in parallel with the historicization of
nomos (as tekhnē) discussed at p. 31 above. There he said that in order to un
derstand conventions we need to look to their historical creators and the proc
ess of creation. Here he asks whether in order to understand nature we must
likewise look to the process of creation and engenderment and to the historical
parent. This time, the answer is no: while convention must be approached his
torically, nature must be approached synchronically, without consideration of
history. History could only mislead us into thinking that physical entities are
like the entities which engender them, which is far from always the case.
Further etymologies
Socrates now embarks on a brief series of investigations into names of
heroes and gods that serves as a sort of thematic and methodological prologue
to the more systematic investigation that will begin at 397c4. The sections
which follow offer a unique glimpse into Greek etymological practice of the
5th and 4th centuries BC. To the eyes of a modern reader, the etymologies pro
ceed from the relatively credible (Agamemnōn from agastos kata tēn epimonēn
"admirable for remaining", 395a5-b2) to the dubious (Tantalos coming from
both talantaton "most wretched" and talanteia "balancing", because in Hades
he balanced a stone above his head, 395d3-e5) to the preposterous (Ouranos
from horōsa ta anō "looking at the things above", 396b7-c3). But Hermogenes,
rather than growing doubtful, instead becomes more and more enraptured, fi
nally declaring that Socrates seems to be "uttering oracles, exactly like those
who are possessed by gods" (396d2).7
The very first name for which Socrates deduces an etymology is typical
of most that follow. It is Orestēs, which has the transparent meaning "man of
the mountain". Socrates asserts that Orestes "is correctly named [...] whether
some accident gave him the name, or some poet expressing in the name the
thēriōdes "savagery", agrion "brutality", and oreinon "mountain wildness", of
his nature" (394e8). If we follow Socrates in ignoring the distinction between
unaspirated t (t) and aspirated t (th), we see that all the sounds of Orestēs are
distributed across the three suggested source words: thëriödes, agrion, oreinon;
actually the first and third words suffice to cover them all. But they are distrib
uted anagramatically, and in fact bear a close resemblance to the anagrams
which many centuries later the linguist Ferdinand de Saussure would claim to
find hidden in Latin poetry (see Starobinski 1971).
7. Hermogenes made a similar remark about Cratylus in the beginning, when he asked Socrates
to interpret "Cratylus' oracular speech" for him (384a4).
WORDS AND TRUTH 49
8. Barney (1998: 71-74) likewise reads the 'inspiration episode' as overtly one of divine origin.
She interprets it as an attempt to distance the etymologies from Socrates: as "they are fated to
be surpassed and undermined", they are "not to be attributed to Socrates himself' (p. 73).
Rather, she believes they are intended as a "rational reconstruction" of a kind of "strong ety
mology" that is "a method of discovery, teaching, and persuasion not just about language but
about the world. And it is a method which tends naturally to support normative judgements
about names [...]" (p. 69). She also makes the following guess, which is more intriguing than it
is well-founded: "I think that we are probably to understand Cratylus as a practitioner of strong
etymology: in fact, since he is not strongly characterized as an individual, Cratylus is probably
meant as the generic representative of people who do this sort of thing" (ibid.).
WORDS AND TRUTH 51
5. Virtues (Part 1)
6. Interlude: Art and contrivance
7. Virtues (Part 2) and their opposites
8. Sensations and emotions
9. Belief and will
10. The 'highest things' — and the topics of this dialogue
Sedley (1998) suggests that there are two superordinate categories here: one
for 'cosmology' (corresponding to 1-4 in my list above) and another for
'knowledge, value and truth' (my 5-10), and that this division is crucial for the
question of the 'philosophical correctness' of the etymologies, as will be ex
plained shortly.
For the name of Hephaistos in the opening category, Socrates introduces
his first 'contrivance' (see below), a letter added 'by attraction'. Hermes is of
particular interest since it is over the name of Hermogenes ("born of Hermes")
that the whole debate got started. When Socrates explains 'Hermes' as 'he who
contrived speech', Hermogenes remarks: "By god, it seems to me that Cratylus
was right to say that I am not 'Hermogenes'! For I am not at all a good con
triver of language" (408b4). This looks like his ultimate concession to Craty
lus' view, but it does not get any serious reaction from Socrates, who simply
carries on with the next group of words.
In category 4, the most significant of the etymologies are those for pyr
'fire' and hydōr 'water'. Socrates admits that these words are 'too much' for
him, and alerts Hermogenes to "note the contrivance [mēkhanē] I introduce in
all cases like this which are too much for me" — namely, setting them aside as
'foreign' borrowings. On this account, "we must not propose forced explana
tions of these words" (409d3), an interesting declaration in the context of all
the forced explanations about to be given (410al-el). Socrates will have re
course to this contrivance several more times. The obvious problem is that it
functions as an escape hatch for the etymologist, allowing him to 'explain'
anything that defies whatever explanatory principles he claims to be adhering
to. Etymologies produced under such circumstances cannot be reckoned as
having scientific value. Again, Sedley's (1998) thesis that Socrates believes in
the exegetical correctness of virtually all the etymologies requires us to over
look the significance of his overt comments on etymological practice.
Categories 5 and 7 are the crux of the entire etymological section, for
they deal with the words that matter more to Socrates than any others. When
Hermogenes asks him to delve into this group, Socrates temporarily stops the
action:
SOC: Well now, by the dog, my impression is that I divined not too badly a little
while ago [397c8], when I had the idea that those very ancient men who estab
lished the words for everything were just like most of the wise men of today: from
their constant spinning and revolving in their search for what reality is like, it then
52 LIMITING THE ARBITRARY
appears that things are being whirled around and altogether set in motion. And
they attribute the cause of this belief not to their own internal condition, but to the
nature of things themselves, none of which do they take to be stable or permanent,
but always flowing and set in motion and always filled with every sort of move
ment and generation. (411b3-c6)
And so the section on the virtues becomes a tour de force, as Socrates attempts
to construct a whole series of etymologies on the basis of a unified and rather
surprising theory — that all the virtues (especially in the examples from epis-
tërnë on) were originally connected with the idea of flux. In the Theaetetus
(157b3-8) Socrates suggests that earlier philosophers had argued that the the
ory of flux ought to be embodied in language:
SOC: But great philosophers tell us that we are not to allow either the word
'something,' or 'belonging to something,' or 'to me,' or 'this' or 'that,' or any
other detaining name to be used; in the language of nature all things are being cre
ated and destroyed, coming into being and passing into new forms; nor can any
name fix or detain them; he who attempts to fix them is easily refuted [...].
Socrates does not share the view of the philosophers he refers to, but he does
appear to be serious in hypothesizing that the earliest language-makers may
have been under spell of the theory of flux. This whole notion will return with
great flourish at the end of the dialogue.9
By the time he reaches dikaion 'justice', Socrates can carry on no longer
without making clear what he really thinks about this sort of etymological en
quiry. After explicating the etymology given above, he says:
SOC: Up to this point, as I was saying, there is agreement among many that this is
what is to dikaion, the just. But I, Hermogenes, being so persistent about this, have
found out all the following things by interrogating those with secret knowledge:
that this dikaion is also the aition 'cause' — for that di' ho 'through which' some
thing is generated is its cause — and someone told me that for this reason Zeus is
correctly named Dia. And when after being told this I nevertheless gently question
them again — "Then, o excellent one, what in the world is just, if this is so?" —
they think I'm pushing my questions too far and am leaping over the barriers. For
they say that I have already been told enough [...] (412e3-413bl)
For the etymologist (specifically the 'weak' etymologist, by contrast with
Barney's 'strong' one — see n. 8 above), it is enough to be told the origin of
the word for justice. To ask 'What then is justice?' is to pass beyond the
boundaries they have set for themselves. But for Socrates, etymological en
quiry is useless as an end in itself. It has value only if it leads to a better under
standing, not just of words as collections of sounds or letters, but of the con-
cepts they designate. What, after all, is Socratic dialectic except an enterprise
of etymology — the study of truth — in its etymological sense? And what kind
of linguistic investigation can rightfully call itself the 'study of truth' if it cares
9. Theaetetus 160d6-8 and 179e4-5 suggest that Homer subscribed to the theory of flux.
WORDS AND TRUTH 53
nothing about the true meanings of the words it studies? The question Socrates
asks in Theaetetus 147b2-3 is the harshest indictment the linguist has ever had
to face: "How can a man understand the name of anything, when he does not
know the nature of it?"
Socrates continues his description of how the etymologists react when he
asks about the meaning of justice:
SOC.: [T]hen when they try to satisfy me by telling me more, each one says some
thing different, and they're no longer in agreement. One says, "Here's what is just:
the sun — for it alone governs all beings by diaïonta kai kaonta "passing through
and burning" them. Then, all pleased at having learned something good, I tell this
to someone else, and he jeers at me, asking whether I think nothing is just among
men after sundown! So when I insist that he tell me what he thinks it is, he says
it's fire. Now, this isn't easy to see. Another says it's not the fire itself, but only
the heat which is in the fire. And yet another says that all these notions are ridicu
lous, that the just is what Anaxagoras says it is, namely nous "mind". He says that
mind is autonomous, never mixing with any other beings, but keeping things in or
der by passing through them all. By this time, my friend, I'm much more confused
than I was before I set out to learn about what the just might be. But as regards
what we were investigating, the word dikaion appears to have been established for
these reasons. (413bl-d2)
Socrates' story recapitulates in brief the history of Greek philosophy, from the
(mythical) prehistoric phase of sun worship, to Heraclitus' equating the logos
with fire, to Anaxagoras and nous.10 The point is that etymological enquiry is
of no help at all in evaluating these ideas, since it contributes nothing toward
understanding the nature of justice. Indeed, Socrates feels more perplexed than
before he began the enquiry.
If this were an early Platonic dialogue, we might expect that Socrates is
leading up to the conclusion that we simply do not know anything about jus
tice. But the very fact that he emphasizes so strongly the possibility and neces
sity of knowing suggests that this is not such a dialogue. The point is not that
we cannot know about the nature of justice — that issue is left moot — but that
etymological enquiry of a purely linguistic sort cannot provide such knowl
edge. For as Socrates says in the Republic VII (533c7-e2):
SOC.: [D]ialectic, and dialectic alone, goes directly to the first principle and is the
only science which does away with hypotheses in order to make her ground secure
[...] and she uses as handmaids and helpers [...] the sciences which we have been
discussing [arithmetic, astronomy, geometry; but also true a fortiori of rhetoric,
poetics, grammar, etymology]. Custom terms them sciences, but they ought to
have some other name, implying greater clearness than opinion and less clearness
than science: and this, in our previous sketch, was called understanding. But why
should we dispute about names when we have realities of such importance to con-
sider? (Emphasis added; see also Philebos 57a9ff.)
10. Here, unlike in Aristotle's Metaphysics, there is no sense of a historical progression from
one theory to the other. Socrates treats them as in principle equally valid accounts of the uni
verse.
54 LIMITING THE ARBITRARY
The passage from Cratylus 412e3-413d2 ends with Socrates saying "But I
think the word [...] was given for the reasons I have mentioned". This sounds
too much in earnest to cast aside as irony. Its tone suggests that an alternative
position is forming in Socrates' mind: why words were given — Sedley's
'exegetical correctness' — is a superficially interesting question, but one an
swer is finally as good as another so long as strictly formal criteria are consid
ered. No formal answer can add anything to our understanding of things; only
substantive, i.e., dialectically derived, 'philosophically correct' answers can do
so. In the absence of real significance, formal etymology reduces to a word
game, and at that point there is no reason to prefer etymologies that do without
ad hoc contrivances to those that use them. Nor is it any longer apparent that
etymology is a tekhnē at all. For Socrates it is essential to distinguish this kind
of formal etymological game from the substantive etymological science that is
dialectic.
Plato now inserts yet another disclaimer regarding the etymologies:
HERM.: YOU appear to me, Socrates, to have heard these things from someone, and
not to be making them up as you go along.
SOC: And the other things as well?
H E R M . : No, not at all.
SOC: Keep listening, then, and perhaps I may fool you into thinking that the rest
of what I say are likewise things I haven't heard before. (413d3-8)
Curiously, the preceding passage is the only one Hermogenes suspects Socra
tes of having heard elsewhere, yet it is the most typically Socratic segment of
the whole etymological section. Socrates' comments here are transparently
ironic and made at Hermogenes' expense. Hermogenes, for his part, cannot
even remember that Socrates has already identified the source of his 'inspira
tions'. It may be that the discussion of dikaion recapitulates some source well
known to Plato's audience, which this is a subtle way of acknowledging. But
the interchange also serves the function of turning the discussion inward, to the
nature of the investigation at hand. This will become the next topic after an-
dreia 'manliness' and the digression to which it gives rise. Actually, the di
gression itself provides the immediate motive for discussing the investigation;
for after moving quickly in the etymological chain courage → man → woman
→ teat → flourish → run + jump, Socrates remarks, with an obvious play on
the last words he has discussed:
SOC: But you don't notice how I race right off the running track once I get onto
smooth ground ! There still remain for us many things generally deemed worthy of
consideration.
HERM.: You're right.
SOC: And one of these is to see what tekhnë 'art' might mean. (414b2-8)
the other being the investigation of onoma (see p. 57). It is ironic, and no doubt
intentionally so, that Socrates chooses tekhnë as the object of his most far
fetched etymology, the first one to elicit a criticism from Hermogenes:
SOC: Doesn't tekhnë signify hexis nou 'possession of mind', when we remove the
t and insert an o between the kh and the n and between the n and the ë (giving
ekho-noē 'I-have-mind-ness')?
HERM.: Well, extremely roughly, Socrates. (414bl0-c3)
Socrates recognizes that language change may or may not be wilful, since it
may occur merely 'in the course of time'. But who are these people who "care
nothing about the truth, only about how they shape their mouths"? It is obvi
ous: those who value linguistic form over content, namely, the poets and the
rhetoricians. They are the enemies of truth and understanding. And the formal
etymologist, who attempts to unravel what the poets and rhetoricians have
done to words, merely compounds their errors and brings still more confusion.
For he too, unlike the dialectician, values linguistic form over content, and
cares nothing for the truth.
Socrates points out what these facts mean for the tekhnë of etymology:
SOC.: If one is allowed to insert and remove whatever one might wish from words,
it will be very easy to fit any word to any thing.
HERM.: You're right.
SOC: Right indeed. But I think, as a wise overseer, you must keep to what is mod
erate and reasonable.
HERM.: That's what I would wish.
SOC: And I share your wish, Hermogenes. But not too much argumentative rigour,
o wondrous one, lest you 'sap my strength' [Homer, Iliad vi.265]. (414d7-415a2)
This passage leaves no doubt that Socrates does not consider etymology to be
an absolute tekhnë. Insofar as its rule is moderation and reasonableness, rather
than precision, it is closer in spirit to poetry than to dialectic. Sedley (1998:
144), who reads this passage as a denial by Socrates that he is succumbing to
56 LIMITING THE ARBITRARY
The rest of the etymological section continues essentially as before the digres
sion and interlude. Socrates pursues the hypothesis that words for virtues and
their opposites, and in many cases for emotions, embody the theory of flux.
One of the etymologies in this list is particularly noteworthy. Socrates
derives to kalon — the good, knowledge of which is the ultimate goal of all
philosophy — not from a word suggesting motion, but from to kaloun 'to call'.
The latter word has already appeared dozens of times in the dialogue in con
nection with how things are called and who does the calling. Socrates finds the
link between calling and the good in the nous, the mind or intellect, in terms
reminiscent of Anaxagoras, who identified nous as the universal arkhē, the
originating principle. Identification of nous as the good and as the 'cause' of all
language is not typically Socratic or Platonic, so the appearance of this idea
here likely represents a subtle kind of etymological mëkhanë: the ad hoc ap-
11. In support of this Sedley (ibid.) cites Socrates' "quips" about inspiration from Euthyphro,
and makes the comment cited on p. 50 above about 'inspiration' supposedly referring to any
intuitive rather than rule-bound method.
WORDS AND TRUTH 57
someone were to ask unceasingly about the phrases from which words might take
their meaning, and then in turn about where those phrases might take their mean
ing from, and went on like this without stopping, wouldn't the person giving the
answers finally have to bring it to a close? [...] When, then, would the quitter be
right to cease and desist? Won't it be when he gets to those words which are like
the elements of other words and sentences? For these can no longer rightly be seen
as composed of other words, if they are elements. (421c3-422a5)
Socrates proposes a theory of language similar in spirit to the atomism then
current in analysis of physical reality. Words are composed of elements, sub
stances that cannot be further broken down. In the case of language he identi
fies these as the prōta onomata 'original, primal words'. He will later declare
how central they are to any etymological enquiry —
SOC: For whatever reason a person might be ignorant of the correctness of the
original words, it is quite impossible for him to know about later ones, which can
only be explained by the original words he is ignorant of. In any event it's obvious
that anyone who alleges to be competent in these matters must be able to make
proofs concerning the original words that are absolutely precise and clear. If not,
it's evident that he will be talking nonsense about the later ones. (426a4-b3)
— thereby tossing into the philosophical rubbish bin all existing etymology,
including his own. Now the task at hand is to investigate the correctness of the
primal words, and Socrates sets out to do this, but runs straight into an episte-
mological barrier.
SOC: Now then, I think you'll agree that for every word, from the earliest to the
most recent, there is one and only one kind of correctness, and none of them dif
fers from any other in being a word.
HERM.: Of course.
SOC: But now, in all the words we've just examined, correctness professed to be
something which revealed what sort of thing each being is.
HERM.: There's no denying it.
SOC: Now, this must hold equally in the original words and in later ones, if they
are indeed words.
HERM.: Of course.
SOC: But the later words, it seems, managed to complete this task by means of the
earlier ones.
HERM.: Apparently.
SOC: Well! Then how can the original words, which are not as yet based upon
others, make things in any way more evident to us, which they must do if they are
to be words at all? (422c7-el)
The etymologists are shown no mercy. Not satisfied with exposing the soft un
derbelly of its methodology, Socrates now suggests that the whole enterprise of
formal etymology is philosophically empty. A tekhnē with no hope of attaining
ultimate causes is unworthy of the name, and its practitioners are no better than
physicians concerned only with relieving their patients' immediate symptoms.
Socrates shows that an etymology with any pretension to logic must extend
beyond words. There we enter the realm, not just of prehistory, but of myth.
CHAPTER 3
IMITATION AND ESSENCE
CRATYLUS 422E1-440E7
1. And vice-versa: it is just because non-arbitrary vocal imitation of non-vocal concepts is im
possible that language must be conventional. Moreover, once convention exists, all language
can be arbitrary. Note however that arbitrariness and conventionality pertain to two different
(though relentlessly interrelated) realms: arbitrariness is a semiotic condition, conventionality a
political condition, of the bond between sound and meaning.
60 LIMITING THE ARBITRARY
2. On the absurd example of people who imitate animal sounds, cf. Republic III 397al-b2.
IMITATION AND ESSENCE 61
than a cloven-hoofed camel), or 'a blue horse', or 'a blue hump-backed horse
that sounds like a cow', so long as it was essentially a horse, that is, possessed
the essence of horsiness.
On the other hand, it is clear from the discussion in Chap. 2 (pp. 43-44)
that even an animal born from the union of two horses, if it is hump-backed,
not cloven-hoofed, and chews the cud, would be called not a horse but a camel,
because its essence overrides its history. Imitating the sound the animal makes,
then, does not count as language-making because it deals with a contingent, not
an essential feature. Socrates further points out (423e2) that sound, shape, and
colour are each things in themselves, with their own essential nature, which
other things happen (contingently) to possess.
Socrates returns to the main thread of the discussion, the correctness of
the original words, armed now with two conclusions: (1) the original words
must function through imitation, but (2) this imitation must be of the essence of
the things named and not of any contingent features such as sound, shape, or
colour. He recalls three primal words that emerged during the etymological
enquiry, rhoē "flow", ienai "go", and skhesis "retention" (see p. 57 above) and
suggests that there are probably others as well. He introduces a crucial new
idea:
SOC: But what would be the method of analysis from which the imitator begins
imitating? Since the imitation of essence takes place with sounds and syllables,
wouldn't it be most correct to distinguish the elements first — just as those who
set to work at rhythms first distinguish the force of the elementary sounds, then of
the syllables, and only then come to the examination of rhythms?
HERM.: Yes.
SOC: Well then, in the same way we too must first distinguish the vowels, and
then among the other sounds according to their kind, consonants and mutes — for
some such thing is what they're called by the experts in these matters — as well as
those which are neither vowels nor mutes, mustn't we? And among the vowels
themselves, what different kinds exist relative to one another? And once we've
properly distinguished all the subtypes, then we must attach words to them — if
indeed there are any words which can cover all the members of each subtype (as
the word 'sound' covers the class as a whole) and from which one can see what
they are and why they fall into the various groupings that they do (as is the case
with the higher-level terms 'consonant', 'vowel', and 'mute'). (424b7-d5)
The first step in inquiring into whether the original words imitate the essential
nature of the things namcd is to separate those words into their component
sounds. This is surprising at first blush, given Socrates' very recent dismissal
of any consideration of the imitation of sound as dealing with contingency
rather than essence. But sound is a contingent feature of everything except
sounds. For Plato, grammata, the elements of language, are sounds, therefore
sound is their essence. Hence the importance of the correct classification of
these elements: it means determining their essence.
This passage is remarkable for what it tells us about the state of linguistic
62 LIMITING THE ARBITRARY
SOC: [J]ust as the picture was fashioned by painting, we'll fashion the language by
word-making or rhetoric or whatever the art may be. Or rather, not 'we'; in saying
that, I got carried away. For it was the ancients who put it together in just the way
it is composed. As for us, if we know how to investigate all these matters artfully,
taking them apart, this is how we must look at them: to see whether both the origi
nal words and the later ones are fitly given or not. For them to have been strung
together in any other way would be careless and not methodical, Hermogenes.
(425a3-b4)
Socrates suggests that the outcome of the investigation will be a verdict on the
ancient creators of language, and so it will be. But it will also be a verdict on
two groups of his contemporaries. One is the formal etymologists, for if words
are 'strung together' at random, the attempt to determine patterns is a fool's
game. The other is the rhetoricians (whose tekhnē he tosses in here rather gra
tuitously), because again, if it is found that words do not reflect the essential
nature of the things named, then their manipulation is the game, not of fools,
but of charlatans.
This is an amusing passage, the joke being that r is "an instrument of all
movement", when the word for movement itself, kinēsis, does not contain the
sound r! And so Socrates is right back where he started, resorting to the 'gods
on machines' which suggest that language really is haphazard. Yet his hy
potheses about the nature of sounds are not outlandish, but typical of the asso
ciations traditionally made with the physical characteristics of certain sounds.
Greek r (rhō) was a trilled /r/ in which, as Socrates points out, "the tongue is
minimally stationary, and maximally agitated" (426e4).
These hypotheses cover fifteen letters (though g only as part of the clus
ter gl) out of the twenty-four in the Greek alphabet. Again, neither in spirit nor
64 LIMITING THE ARBITRARY
in letter (no pun intended) are these remarks unique to Plato, for up to the pres
ent day people have searched for direct semantic content in the individual
sounds of a language, and have often turned up patterns in which at least indi
rect associations appear indisputable. The set of English words beginning with
sl- (slip, slide, slick, slime, sludge, etc.), with their general semantic sense of an
unpleasant oiliness, is a classic case in point. Socrates concludes:
SOC: And in this way the lawgiver appears to bring the others too into accordance
with sounds and with syllables, making for each entity a sign and a word; and from
these he thenceforth puts together the remaining words for other things by imitat
ing. This appears to me, Hermogenes, to be the meaning of the correctness of
words, unless Cratylus there has something different to say.
HERM.: Truly, Socrates, as I said in the beginning, Cratylus often troubles me a
good deal; he declares that there is such a thing as correctness of words, but does
not say clearly what it is [...]. (427c6-d7)
Now comes a hiatus in the argument. The hypotheses regarding the meanings
of sounds is abandoned until 434bl0, as Socrates pulls together the previous
IMITATION AND ESSENCE 65
Cratylus' attitude is found among those for whom there is no such thing as
'good' or 'bad' science, because science as intrinsically good, and 'bad' sci
ence is just not science at all. As will be discussed below, this view proceeds
directly from his acceptance of the Eleatic doctrine of The One.
Part of Cratylus' argument is perfectly congenial to Plato, namely the
importance of distinguishing between a person's (or thing's) true nature and
what he (or it) appears to be. This distinction underlay Socrates' insistence that
a horse born of a cow must be called a foal and not a calf. Where they fail to
agree, however, is on how to deal with the case of someone who does other
wise. In contesting Hermogenes' view an individual could make 'horse' the
word for man and vice-versa, Socrates argued that this would negate the possi
bility of truth and falsehood. He now shows that Cratylus' view, though on its
face directly opposite to that of Hermogenes, would nevertheless lead to the
very same absurd conclusion.
SOC: And isn't it then a lie if someone says that he is Hermogenes? But on the
other hand, might it not be a lie at all to say that he is Hermogenes, if he isn't?
CRAT.: What do you mean?
SOC: Is it that speaking falsely does not exist at all, is this the force of your argu
ment? [...]
CRAT.: But how, Socrates, could someone saying what he says, not say what ex
ists? Or isn't that what it is to speak falsely, to say things that do not exist? (429c7-
d6)
Cratylus is here expounding the doctrine of Eleatic unity. Since all the universe
is one, thought and being are one. Hence 'untrue' thought or speech is a logical
impossibility. In the rather enigmatic words of Parmenides, "What is there to
be said and thought must needs be: for it is there for being, but nothing is not"
(Fr. 6, Simplicius in Phys. 86, 27-28; 117, 4-13; transl, in Kirk, Raven & Scho-
field 1983: 247). Anything 'untrue' must simply be disqualified from counting
as thought or speech.3 Socrates asks whether someone who greeted Cratylus as
'Hermogenes, son of Smikrion' would be saying these words to Cratylus, to
Hermogenes, or to nobody. Many commentators fail to recall at this point that
Hermogenes is the son not of Smikrion, but of Hipponikos (see 384a8, 406b8).
Presumably (though there is no external evidence for or against this), it is Cra
tylus who is 'son of Smikrion'. In that case Socrates' question is not simply
about calling a person by the wrong name, a point that will be taken up in the
next passage (see the following section). Rather it is about the more subtle
matter of calling a person by a name that is partly true (son of Smikrion) and
3. Cf. Sophist 260b1 Off.; also Euthydemus 283e7ff., where the Sophist brothers Euthydemus
and Dionysodorus also develop the argument that it is impossible to speak falsely, prompting
Socrates to call the argument "quite wonderful, and suicidal as well as destructive" (286c4-6).
The brothers fare much better in their debate with Socrates than Cratylus does in his, but Soc
rates still gets in his swipe at the Sophist who professes the impossibility of falsehood: "[...] I
will ask my stupid question: If there is no such thing as error in deed, word, or thought, then
what, in the name of goodness, do you come hither to teach?".
IMITATION AND ESSENCE 67
partly false (Hermogenes), a point to be developed in the passage after that (see
'The image of Cratylus, or two Cratyli?' below). The same point would be
made if the hypothetical greeter were to call Cratylus 'Cratylus, son of Hip-
ponikos'. Cratylus answers Socrates thusly:
CRAT.: It seems to me, Socrates, that this fool would be making sounds with no
purpose. [...] I'd say that such a fool was just making noise, setting himself vi
brating without reason, just as if he set some bronze pot vibrating by striking it.
(429e8-430a7)
Cratylus' view seems absurdly extreme: if someone calls him by a name that is
(wholly or partly) other than his own, that person is not performing any lin
guistic act, any more than a pot that makes a noise when struck. But this ab
surdity is the ultimate consequence of a naturalistic view of language, and
Plato was the first, so far as we know, to recognize it. Any speech which fails
to embody true reason must be dismissed by the linguistic naturalist as not
speech at all. And if speech is true by definition, the whole issue of truth and
falsehood in speech dissolves into nothingness.
Cratylus reasserts the distinction he has drawn between painting and language-
making:
CRAT.: But maybe, Socrates, it's possible to distribute incorrectly in the case of
paintings, but not in the case of words, where instead it is necessarily always cor
rect. (430d8-e2)
However, Socrates shows Cratylus that the shared imitative purpose of paint
ings and words, which Cratylus does not dispute, disallows making such a dis
tinction:
SOC: What do you mean? What distinguishes the one case from the other? Isn't it
possible to go up to a man and say to him, "This is your picture", and to show him
perhaps an image of him, or perhaps one of a woman? By 'show' I mean to set be
fore the sense of sight.
CRAT.: Of course it's possible.
SOC: Well? Then couldn't one go up to this same man again and say "This is your
name"? [...] and then set before his sense of hearing, perhaps a copy of him, saying
that he is a man, perhaps that of the female of the human species, saying that he is
a woman? Doesn't this seem to you to be possible and even to happen sometimes?
CRAT.: I'm willing to concede you this, Socrates; let it be so.
SOC: It's well for you to do this, my friend, if it actually is so; for now we don't
have to fight about this. (430e3-431a9)
That is the end of Cratylus' Eleatic denial of falsehood — and, for Plato's pur
poses, of the linguistic doctrine of physis as maintained by the Sophists. Soc
rates reduces it to absurdity in much more resolute terms than he applied to the
nomos view as maintained by Hermogenes, perhaps only because the failure of
physis is so much more disappointing for Plato. But the possibility remains that
language is constructed partly by physis, and that naturalness, measured in de
grees, still constitutes the best criterion for what is correct and incorrect in lan
guage. To explore this possibility, Socrates continues to elaborate the compari
son with painting.
The reason for Cratylus' equivocal last response becomes clear shortly. He has
given up his initial insistence that there are no 'bad' words, only real words and
noise, and has agreed that there can be 'right' and 'wrong' words for things.
But he still refuses to acknowledge that words might be right or wrong by de
grees, depending on whether one 'omits or adds a little'. Apparently his earlier
'change of opinion' was not much of a change at all, since his absolutism re
mains untempered.
CRAT.: But you see, Socrates, when we assign these sounds, a, b, and so on, to
words according to the art of sounds and letters [grammatikē tekhnē], if we take
out or stick in or misplace something, it isn't the case that we have written the
word but haven't done so correctly. Rather, it hasn't been written at all, but is im
mediately a different word if it has undergone something of this sort. (431e9-
432a4)
Given this statement, it is difficult to see why Cratylus found Socrates' at
tempts at etymology so much to his liking, since almost without exception they
depended upon letters being taken away or added or transposed. Rather than
point this out, Socrates returns one last time to the comparison with painting as
another form of mimēsis, and brings in another comparison, a negative one this
time, with arithmetic.
SOC: Perhaps what you say might hold for things whose existence or nonexistence
depends entirely on a number, for example ten or whatever other number you like:
if you take away or add something, it immediately becomes another number. But
for anything qualitative and for images in general, surely this is not correctness.
Quite the opposite: the image absolutely must not render everything in what one is
portraying, if it is meant to be an image. (432a8-b4)
The question here is: what is the essential nature of mimēsis itself? That of
numbers is such that adding or subtracting changes them completely. But the
same is not true of things, so long as only contingent features are added or
subtracted. A horse with three or five legs is still a horse. Socrates begins by
considering painting:
SOC: Now, consider whether I'm speaking sense. Would there be two things,
Cratylus and Cratylus' image, if some god not only copied your complexion and
shape, as painters do, but also made all the insides just like yours, and rendered the
same softness and heat, and put into them motion and soul and intelligence just
like your own — in a word, if he placed beside you all your same qualities in an
other being? In such a case, would there be Cratylus and Cratylus' image, or two
Cratyli?
CRAT.: TWO Cratyli, it would seem to me, Socrates. (432b4-c6)
In his discussion with Hermogenes about whether someone who imitated the
sound of an animal would be naming that animal (423c4), Socrates concluded
that the imitation of sound is the domain of music, the imitation of shape and
colour the domain of graphic art, and the imitation of essential nature the do
main of naming. There he was concerned simply to match each aspect of na-
70 LIMITING THE ARBITRARY
ture with the appropriate art. Here, however, his aim is slightly different: to
show that each of these arts is limited to just that aspect of nature and no other.
A painting of Cratylus will be good or bad according to how well it imitates his
shape and colour. A painting that expresses more than his shape and colour
might still be an image of Cratylus, but such expression taken to its logical ex
treme leads us out of the realm of mimēsis entirely, since a 'perfect' imitation
of Cratylus would be no imitation at all, but another Cratylus.4 Socrates now
makes the connection to language:
SOC.: Then, Cratylus, words would have ridiculous effects on the things they're
names of if they resembled them completely in every regard. For all things would
be double, and no one could say which of them was the thing itself, and which the
word.
CRAT.: That's true.
SOC: Then believing with confidence, my friend, that those things are words both
which name well and which do not, don't demand that they have all the necessary
sounds, so that they would be exactly like the things they name, but let them also
bring in an inappropriate sound. And if a sound then also a word in the sentence,
and if a word then also a sentence in the discourse, can be brought in although in
appropriate to the things referred to, and nonetheless can name the thing and speak
of it, so long as they contain the type of the thing about which the discourse is con
cerned. This was the case with the words for the letters of the alphabet, if you'll
recall what Hermogenes and I were talking about a while ago. (432d5-433a2)
Cratylus needs confidence not only to admit his error, but to deal with the fact
that acknowledging the imperfection of words places language within the
sphere of human responsibility. This means that Cratylus can no longer give
black-and-white answers to linguistic questions, saying that words are either
naturally correct or not words at all. Instead he is condemned to dabble in
shades of grey, bouncing relentlessly from one to the other of Socrates' four
poles: Ideal Form, material nature, tekhnē, and nomos.
SOC: For so long as they contain this [type of the thing named], even if not all the
appropriate sounds are there, the thing is spoken of — well if all of them are there,
badly if only a few. So let's admit that they're spoken of, my friend... Otherwise,
you should seek out some other correctness of words, and you shouldn't agree that
a word is a manifestation of a thing in syllables and sounds. For if you hold both
these views you won't be in accord with yourself.
CRAT.: Well, what you say seems fair to me, Socrates, and I hold likewise. (433a4-
b7)
The two contradictory positions Socrates refers to are: (1) words are correct if
they imitate things in sounds, and (2) words cease absolutely to be correct if
any sound is changed. (An example of the sort of contradiction to which these
4. Cf. the long discussion of imitation and truth in the Republic X (595al-604a9), where a
three-way distinction is drawn between (1) the painter of a picture of a bed, (2) the maker of an
actual bed, and (3) the maker of the idea of a bed, who is God. Related discussions are found in
Sophist 233d8ff. and 265a1 Off., and in Laws II 667c9-669b4.
IMITATION AND ESSENCE 71
views give rise will occur at 434d7.) Note that (1) is a semantic and functional
claim about language, while (2) is a purely formal one. Socrates points out in
effect that the functional aspect of language inherently contradicts any absolute
claim of a purely formal nature — an insight as timely today as it was 2500
years ago. He also insists that nomos must be allowed to play a role in lan
guage at the formal level.
CRAT.: Of course.
SOC: Well, are both r and s like the same thing, and does the final r indicate to
them the same thing as the s does to us? Or is it not indicated to one or the other of
us?
CRAT.: It is indicated to both.
SOC: Is this insofar as r and s resemble one another, or or insofar as they do not?
CRAT.: Insofar as they resemble one another.
SOC: Do they resemble one another in every regard?
CRAT.: AS far as indicating rapid movement in the same way is concerned. (434c7-
d6)
In the earlier discussion, Socrates had equated sigma not with motion as such,
but with 'blowing'. Yet the connection made by Cratylus here is not outland
ish, and he heads off the different-languages argument successfully. Moreover,
just as Socrates has borrowed an argument from Hermogenes, Cratylus will
now borrow one from Socrates himself:
SOC.: And the / in the word (sklērotēs)? Doesn't it indicate the opposite of hard
ness?
CRAT.: Well, maybe it is in there incorrectly, Socrates; just as when you were
speaking with Hermogenes a while back, you removed and stuck in sounds as nec
essary, and correctly in my view. And here perhaps instead of the / we ought to say
r. (434d7-12)
Cratylus' earlier insistence that "if we leave out or stick in or misplace some
thing [...] the word [...] hasn't been written at all" (432al) appeared to contra
dict his approval of Socrates' etymologies. This new statement contradicts the
contradiction. Cratylus here faces exactly the dilemma Socrates warned him of
at 433b4: to be consistent with his earlier position, he would have to deny that
sklërotës functions as a word for hardness, the only possible word for which is
skrērotēs.
count of language.5 That in fact seems to be Cratylus' view in his reply, which
is meant to put understanding in its place, to trivialize it by saying it is merely a
matter of ethos. Yet in so doing, he puts his argument into checkmate.
SOC: But when you say habit, do you suppose you're saying anything different
from convention [xynthēkēs]? By habit, do you mean anything other than that I,
when I speak a word, have something in mind, and that you discern that I have this
in mind? Isn't that what you mean?
CRAT.: Yes. (434e5-435al)
5. Cf. Timaeus 46e6-47e2, which suggests that the gods gave speech and hearing (as well as
sight) with the following principal end in view: "that we might behold the courses of intelli
gence in the heaven and apply them to the courses of our own intelligence which are akin to
them [...] that we, learning them and partaking of the natural truth of reason, might imitate the
absolutely unerring courses of God and regulate our own vagaries". Intercourse with the di
vine, not communication with one another, is then the primary purpose of human language.
Later essentialist theories would likewise deny that communication is the ultimate reason for
language; such is currently the case, for instance, with Chomskyan generative grammar. Gor-
gias 481 suggests that even when we do communicate, it is as a function of feelings that are not
truly individual, but universal (hence ideal) in nature: "[...] if there were not some community
of feelings among mankind, however varying in different persons — I mean to say, if every
man's feelings were peculiar to himself and were not shared by the rest of his species — I do
not see how we could ever communicate our impressions to one another".
6. In much the same way, Saussurean structuralists would use the wastebasket of parole, and
Chomskyan generativists the wastebasket of 'performance' and later 'periphery' (on which see
Chap. 6).
IMITATION AND ESSENCE 75
7. Cf. the Eleatic Stranger's castigation of nomos in the Statesman (294b2-c4): "The differ
ences of men and actions, and the endless irregular movements of human things, do not admit
of any universal and simple rule. And no art whatsoever can lay down a rule which will last for
all time [...]. But the law is always striving to make one; — like an obstinate and ignorant ty
rant, who will not allow anything to be done contrary to his appointment, or any question to be
asked [...]". By following nomos mindlessly, the many allow themselves to be tyrannized — a
base thing indeed. On the other hand, the nomos that will not allow any question to be asked
sounds very much like the nomoi of Plato's own later book by that name (the Laws), in which
the Athenian Stranger twice glorifies nomos by pointing out its similarity to the word nous
"mind".
76 LIMITING THE ARBITRARY
Socrates never suggested that the language-maker was infallible; on the con
trary, he made clear that the dialectician was needed as his overseer.8 So, he
asks, what if the fallible workman made his product without the proper super
vision?
8. Cf. Laws VI 769d3-7, where the Athenian Stranger asks: "Do you imagine that there ever
was a legislator so foolish as not to know that many things are necessarily omitted, which
someone coming after him must correct [...] ?".
78 LIMITING THE ARBITRARY
and their opposites. Cratylus seems not to have understood that it did not fare
well in the course and aftermath of the etymological enquiry, and this will
prove fatal for his argument. Socrates attacks Cratylus' statement about the
supposed consistency of words, both on principle and with references to the
cases already examined.
SOC: But my dear Cratylus, this is no defense. For if at first the establisher erred
and from then on forced the other words in that direction and demanded that they
agree with him, it wouldn't be surprising — just as with geometrical diagrams,
sometimes from a first, small, unseen error, all the rest follow in agreement. [...]
I'd be astonished if words were consistent with one another. Let's go back and ex
amine what we went through earlier. Given that all is going and moving and
flowing, we said that words signify reality to us. Does it seem so to you?
CRAT.: Yes, yes, absolutely; and they signify correctly. (436c7-437al)
According to Aristotle's Metaphysics C5, the most extreme proponent of uni
versal flux was "Cratylus [...], who finally thought one ought not to speak at
all, but who simply pointed his finger and censured Heraclitus for saying that it
is impossible to step twice into the same river — for he himself believed that
one could not do this even once" (Aristotle 1960: 79). Metaphysics A6 informs
us that "in his youth Plato became familiar through Cratylus with the Heraclit-
ean doctrines" (Aristotle 1960: 19). The choice of Cratylus as a participant in
the dialogue and the prominence of the theory of flux in the etymological en
quiry cannot have been coincidental. The Socratics recognized the flux of
earthly things, but their concern was to establish the eternal ideas that underlie
them (a concern more closely aligned to Heraclitus' conception of the logos).
Socrates now proceeds to take apart his earlier proposals on the role of flux in
etymology:
SOC: Then let's re-examine some of these words, starting with epistēmē 'knowl
edge'. It's ambiguous, and appears to signify that it histēsin 'stops' our soul epi
'at' things, rather than carrying it around with them [...]. (437a2-5)
The etymon Socrates originally gave for epistëmë was hepomenē "accompa
nying", because "the soul worthy of consideration is accompanying things set
in motion" (412al). He now points out that, on the level of form, the word for
the opposite of motion, histësin, makes just as convincing an etymon. He pro
ceeds to give etymologies for a new set of words, all of which appear to con
tradict the consistent basis of the earlier ones. First, four words which like the
reconsidered epistëmë designate good things and qualities through the idea of
motionlessness; then four words for evil things which appear to be based on the
idea of motion.
Robin (1950: 1347-8) has noted that, for Plato, to be bebaion "firm" is a
necessary quality of epistëmë "knowledge", and that to be piston "trustworthy"
is a necessary quality of a historia "enquiry". This suggests that these final
etymologies embody another ironical commentary on the dialogue itself, since
IMITATION AND ESSENCE 79
Socrates' point here is precisely that the knowledge he has been producing is
not firm (and therefore not really knowledge) and his enquiry not trustworthy
(and therefore not really an enquiry). Indeed, they are characterized by nothing
so much as their amathia "ignorance" and akolasia "lack of restraint".
The proof of the language-maker's infallibility which Cratylus based upon
the consistency of etymological patterns is thus laid to rest. But as usual he
does not give up easily:
CRAT.: However, Socrates, you see that most words do signify movement.
SOC: And so what, Cratylus? Should we count words the way we do ballots, and
correctness will reside in this? Whatever the majority of words appear to signify,
that will be the truth?
CRAT.: That's certainly not reasonable.
Soc.: Not in any way whatsoever, my friend. (437dl-8)
Socrates casts his disagreement with Cratylus in overtly anti-democratic terms.
He has already made clear that communication, where 'base, vulgar' conven
tion is the rule, is the 'democratic' sphere of language. Meaning, however
'sticky' it may be, nevertheless adheres to the eternal and unchanging truths
that are beyond the will of the masses.
'All that we said before' was that words for virtues show no consistent pattern
of being composed from words for motion, and indeed in some cases appear to
be composed from their opposites. Cratylus tries to fall back yet again on the
argument that badly composed words are not really words, only to have Soc
rates turn it against him.
CRAT.: But one of the groups of words in contradiction aren't really words at all!
SOC: Which group, my friend? Those which lead back to standing still, or to
movement? For we said a little while back that it won't be decided by numbers.
CRAT.: That would indeed be unjust, Socrates.
SOC: So, with the words split into two factions, with each side claiming to be
those that resemble the truth, how do we make a decision, and what do we base it
on? Surely not on other words than these; for there are none! [...] (438c6-d5)
Socrates here conjures up the image of a court of law, where parties represent
ing two sides of a dispute each claim to be telling the truth, and a judge or jury
must decide between them. But as he points out, the analogy does not hold.
When two words each claim to represent the truth, we cannot call on other
words to decide, for they too are always implicated in the same question of
truth. There can be no 'impartial jury'. Thus linguistic enquiry is hopelessly
circular — unless recourse can be had to something outside of language.
SOC: Rather, it's obvious that something beyond words must be sought out, which
will show us wordlessly which of them are the true ones, the ones that reveal the
truth about existing things.
CRAT. : So it seems to me.
SOC: Then it would seem to be possible, Cratylus, to learn about existing things
without the help of words, if this is so.
CRAT.: Apparently. (438d5-e4)
If one has to go outside of language to learn about things, then, Socrates asks,
why bother with language at all? If etymology is a circular play of words un
less and until considerations of meaning are brought in — that is, if etymology
is methodologically dependent upon meaning — how can we expect etymol
ogy to teach us about meaning? Why not simply study meaning directly?
SOC: Then if it's possible to learn about things either through words or through
the things themselves, which would be the better and surer kind of learning? To
learn from the image both whether it is itself a good representation, and the truth
about what it represents? Or to learn from the truth itself both what it is and
whether its image is suitably made?
CRAT.: I think it has to be from the truth.
SOC: Now, what we must do to learn about or discover the reality of things is per
haps beyond you or me. We'll have to be content with agreeing on this, that it's
much better to learn about and seek after things from the things themselves rather
than from the words that name them. (439a6-b8)
This then is the very Socratic conclusion of this segment: we may not in fact be
IMITATION AND ESSENCE 81
able to learn anything about ultimate realities (except, as the early dialogues
are fond of pointing out, how little we know). But at least dialectic provides
some hope of learning. Etymology, rhetoric, grammar and the rest of linguistic
study do not.
Socrates' dream
One task remains, to determine just what those ultimate realities are. Soc
rates says we should do this "to keep us from being fooled by the fact that so
many of these words tend in the same direction" (439bl0). This statement is
crucial to a full understanding of Plato's view of language. He recognizes that
patterns exist on the level of both form and meaning; they are what makes
etymology and the rest of linguistic study possible. But such patterns are de
ceptive. Contrary to appearances, the perceptible patterns of language do not
reproduce the logos, the hidden patterns of the universe, even if such repro
duction was the intention of the language-maker. The reason is that no human
language-maker has direct access to the logos. He has only his opinion.
SOC.: [W]hat if, on the one hand, those who established [words] really believed
that all things are always in motion and in flux — for it would appear to me that
they believed this — and what if, on the other hand, it turns out that this isn't so,
but instead it's as though they themselves have fallen into some kind of whirlpool
and are spinning about and pulling us down into it with them. (439c 1-6).
This is one of Plato's finest literary images. What the patterns of language ac
tually reflect, according to Socrates, is not the reality of things, just one par
ticular set of perceptions — those of the language-makers, who were under the
influence of the theory of flux. That would explain both why language is sub
ject to change and why it is so difficult for us to perceive what is not moving in
the universe. Language, far from revealing the essence of things, emerges as
the great obstacle to our understanding. The next passage is the most poetic of
the entire dialogue:
SOC: For consider, o wonderful Cratylus, what I often dream about: whether we
should say that there are such things as beauty and goodness and so on for every
one of the existing things, or not?
CRAT.: I think we should, Socrates.
SOC: Then let's investigate beauty itself, not whether some particular face is
beautiful or anything else of that sort, for all such things seem to be in flux. But
beauty itself, shouldn't we say that it is always exactly what it is?
CRAT.: We must. (439c6-d7)
Similarly to what was noted on pp. 26 and 52-53 concerning 'justice', even
persons who do not consider themselves Platonists routinely talk about
'beauty' in a Platonic way, as though it were a thing existing separately from
any of the entities which possess it. The fact that a particular beautiful face is
in flux and will eventually cease to be beautiful takes nothing away from
beauty itself. Neither does the fact that the word for beauty might change with
82 LIMITING THE ARBITRARY
the passage of time. Beauty in this sense is eternal; it is the most real beauty,
the one most worth knowing, the only one which can be known.9 Socrates' first
concern is to establish conclusively that this ideal of beauty cannot change:
SOC: Well, is it possible to declare correctly what it is, if it's always slipping away
from us, first being this, then that? Or must it at the very moment we are speaking
become something completely different, and sneak away, and no longer be as
we're saying it is?
CRAT.: It must.
SOC: Then how can there be anything that is never such as it is? Anytime it stays
such as it is, for that time obviously it isn't changing; and if it is always such as it
is and is the same thing, how could it change or move without abandoning its own
ideal form?
CRAT.: Impossible!
SOC: But neither could it be known by anyone. (439d8-e7)
If beauty were not an unchanging standard but itself subject to flux, one could
never call anything beautiful, for beauty itself would be liable to change in the
course of the statement. This is perhaps not the best example for our time, ac
customed as we are to aesthetic 'contingencies of value' (to borrow the title of
Smith 1988) and to the notion of an absolute 'justice' that is served well or ill
by a particular legal case. We could hardly tolerate a legal system in which
justice was held to be constantly in flux, so that cases were decided on a com
pletely ad-hoc basis with no recourse to precedent and stare decisis, and where
it was impossible to think afterwards that 'justice' was or was not done. The
last point Socrates made in the preceding excerpt was that in such a situation it
would be impossible for us to know what justice is.
SOC: For in the very moment when someone would approach it in order to know
it, it would become something else, something different, so that one could no
longer know what it's like or how it is. Presumably, no knowledge knows what is
known to have no stable existence at all.
CRAT.: You're right.
S o c : But it isn't even possible to say that there is any knowledge, Cratylus, if all
things are changing and nothing remains. For if this thing itself, knowledge, does
not change and cease to be knowledge, then it would always be knowledge and
knowledge would exist. But if the very essence of knowledge changes, then at the
moment when it changes into another essence of knowledge there would again be
no knowledge; and if it is always changing, there would always exist no knowl
edge, and following this logic there would be neither anyone to know nor anything
to be known. (439e7-440b4)
9. Cf. Republic V 476c2-7: "SOC: And he who, having a sense of beautiful things has no sense
of absolute beauty, or who, if another lead him to a knowledge of that beauty is unable to fol
low -of such a one I ask, Is he awake or in a dream only? Reflect: is not the dreamer, sleeping
or waking, one who likens dissimilar things, who puts the copy in the place of the real object?"
In Republic V 479el-5, Socrates adds that "those who see the many beautiful, and who yet
neither see absolute beauty, nor can follow any guide who points the way thither; who see the
many just, and not absolute justice, and the like, — such persons may be said to have opinions
but not knowledge".
IMITATION AND ESSENCE 83
Here Socrates turns the argument inward onto itself. If everything were in flux,
not only could we not know anything, but knowledge itself would be in flux.
Yet since knowledge demands by its very nature a kind of standing still in the
mind, this means that knowledge itself could not exist. What is more, there
must be someone to possess knowledge, but that person would be constantly
changing too, leaving the whole concept of gnōsis empty and absurd. The con
sequence, for Socrates, is that knowledge must inhabit a realm where flux does
not pertain, the heaven of the ideal.10 Language, however, is bound to the re
lentlessly fluid realm of appearance and opinion.
SOC.: But if there always exists the knower on the one hand, and on the other, the
thing known, and the beautiful, and the good, and every one of the existing things,
then these things we have just been discussing do not appear to me to be like either
flux or movement. Now whether these things are like this, or are as the followers
of Heraclitus and many others say, is not at all an easy matter to investigate. But
certainly no rational person should entrust the care of himself and his soul to
words, believing in them and those who established them to the point of affirming
that he knows something [...]. (440b4-c6, emphasis added: JEJ)
Summary
Plato's response to the physis-nomos dialectic of the Sophists is to reject
it as overly simplistic. As argued by the Sophists, the question is whether
words (assumed to be conventions) are or are not connected to things (such as
10. Cf. Phaedrus 247c3-dl on "the heaven which is above the heavens [...]. There abides the
very being with which true knowledge is concerned; the colourless, formless, intangible es
sence, visible only to mind, the pilot of the soul".
84 LIMITING THE ARBITRARY
they appear to us). Plato's political philosophy demands that nomos be brokei
down into what is appropriate for the small ruling class and what is appropriate
for the mindless mob, including slaves and women. His epistemological phi
losophy demands that physis be broken down into what nature appears to be
and — since appearances are proverbially deceptive — what it really is. In fact
politics and epistemology are inseparable for Plato, who believes that the phi
losopher, he who knows things as they really are, should rule.11
In its Sophistic form, the physis-nomos debate effectively limits to twc
the possible connections between words and the things they name. Either the
connection is material (residing in the shared physis of word and thing) or non
existent (opening it up to free, arbitrary will). Given these two choices, some
one with an essentially religious outlook would be inclined to opt for physis,
since it at least says that language and other human activity connect to some
thing outside themselves. But to someone as deeply (though unconventionally)
religious as Plato, the possibilities this dichotomy defines are absurdly limited.
A true dialectic must deal with ultimate causes, which the Sophists claim phy
sis and nomos to be. Plato shows instead that physis and nomos as the Sophists
conceive them are not ultimate causes at all, but merely the products, the shad
ows, of things more distant and impalpable, yet more real. Physical things are
shadows, or copies, of Ideal Forms. Nomoi, including words, are tools fash
ioned by a legislator applying the tekhnē of language-making. It is clear that
the search for linguistic correctness must move beyond physis as appearance
and nomos as convention. It must also shift from the many who use words to
the one who created them, which means a move from the present to the past —
a historicization. The question 'What is linguistic correctness?' has become:
What did the lawgiver have in mind when making language, and how did he
embody it in the words he created?
The exploration begins using the standard investigative process for the
history of words: formal etymology. In the present context, etymological pro
cedure would suggest starting with words currently in use {nomos as conven
tion) and determining, through inductive analysis, how the lawgiver embodied
their meaning in sound. However, a number of obstacles arise:
1. Nomos as convention does not immediately reveal tekhnē, because of
changes that have come about since the words were created, either through
11. Neither is the moral dimension separable from the other two. In the Gorgias (482c4-
484c3), Callicles, after accusing Socrates of manipulating natural and conventional definitions
of terms in order to confound his interlocutors, asserts that conventions — nomoi — were in
troduced by the many who are weak to protect themselves against the few who are strong, and
who therefore deserve by nature to have political and economic control. Thus conventional law
is unnatural, and by implication, evil. Socrates undoes the argument by introducing the moral
dimension: 'strong' and 'weak' must rest principally on a moral rather than a physical judge
ment. If nomoi ensure that the morally strong rule the morally weak, then they operate in har
mony, not in conflict, with nature.
IMITATION AND ESSENCE 85
the passage of time or the intervention of people who "care nothing about
the truth, only about how they shape their mouths".
2. No matter how wise the lawgiver may have been, he did not have direct
access to Ideal Form. It is always someone's perception that is embodied in
language.
3. Etymology appears to work only so long as it is not dealing with ultimate
elements, but merely breaking words down into other words. Upon reach
ing the elements, the etymologist has to resort to contrivances to explain
them away.
Much as the opening section demonstrates that linguistic correctness cannot be
understood in terms of a simple physis-nomos dichotomy, the middle section
shows that the relationships among the terms are far too complex to permit
simple yes-no answers to questions about the nature of language, or to sustain
the superficial conclusions of formal etymology. Having rejected the etymo
logical practice of his time, Plato goes on to sketch out what he believes a
proper etymological enterprise should consist of. But it is important to recog
nize that he does not toss formal etymology aside so glibly as he does other
linguistically-based approaches to knowledge, such as rhetoric and poetics. He
abandons it only after a fully-blown attempt to construct a synthesis of it. He
does not declare but demonstrates its failure as a tekhnē, and that suggests that
he may at one time have taken it very seriously as a possible approach to
knowledge.12
Why would Plato have taken etymology so seriously, only then to turn
around and repudiate it? The preceding pages have hinted at an answer. The
key point is that words share a common purpose with Ideal Forms: both the
word shuttle and the Ideal Form of the shuttle serve to group all individual
shuttles together, no matter how different they may appear, by identifying their
common essence. They accomplish this in different ways, to be sure: the word
indicates the essence, whereas the Form is the essence. Yet even indicating the
essence makes words a potentially important route to knowledge, for they are
immediately available as objects of investigation, while Ideal Forms are not. In
the Parmenides (130e5-131a3), Socrates is asked by Parmenides whether he
believes "that there are certain ideas of which all other things partake, and from
which they derive their names". Socrates answers, "Yes, certainly". And in the
Republic X (596a6-7), Socrates begins a discussion of imitation and Ideal
Form by saying that "Whenever a number of individuals have a common name,
we assume them to have also a corresponding idea or form [...]".
12. In fact, despite the conclusions reached in the Cratylus, etymological speculations do occa
sionally appear in other dialogues, including late ones (Laws I 626d3, II 654a4, IV 714al, VII
799el0, XII 957c6; Phaedrus 237a7, 244b6, 251c7, 255cl; Sophist 221a7; Timaeus 43c5,
45b4, 62a2), though always as asides, never as integral to an argument.
86 LIMITING THE ARBITRARY
the masses to do otherwise, since they do not have the tekhnē of the lan
guage-maker or of his supervisor the dialectician. Hence any meddling the
mob does in language can only result in decreasing correctness by moving
words away from their original forms given by tekhnë. Furthermore, since
the masses have communion only with one another, nomos is a sufficient
principle to provide and account for their communication. It is only when
one has communion with something higher — in effect, with heaven —
that physis becomes a relevant principle. But communion with heaven re
quires that we first know heaven, and heaven is the end of all knowledge.
So the study of language cannot lead to knowledge, and therefore — contra
the etymologists, grammarians, philologists and rhetoricians — it is finally
trivial. Dialectic is the only 'science' properly so called.
We can, then, identify the operating principles of language and its correctness
as physis in the ideal and nomos in the actual, though, technically speaking, to
do so is to focus upon secondary effects rather than primary causes. The ulti
mate operating principles of language, for what they are worth, are Ideal Form
and the tekhnë of embodying it in sound.
Part Two
After Cratylus
CHAPTER 4
many words have undergone phonetic change, which means that even if they
were originally natural imitations of true meanings, they are no longer so.
Hence 'mere convention' is more powerful than nature when it comes to lan
guage use.
SOC: But when you say habit, do you suppose you're saying anything different
from convention? By habit, do you mean anything other than that I, when I speak
a word, have something in mind, and that you discern that I have this thing in
mind? [...] Now, if habit is completely different from convention, one could no
longer say that resemblance is the indication, only habit: for habit, it seems, indi
cates both by likeness and by unlikeness. [...] Now, I myself also like for words
to resemble things insofar as possible; but beware, for in truth, as Hermogenes
says, this force of attraction by resemblance is a meagre, sticky thing; and one
has to make use of this vulgar business of convention in regard to the correctness
of words. (434e5-435c)
as Plato's nomothetēs 'giver of rules'. A few centuries later the Epicurean Dio
genes of Oenoanda would object strenuously to the logical impossibility of
agreement about words preceding the existence of the very words needed for
agreement to be reached:
As for the words, I mean the nouns and verbs, of which the men who sprang from
the earth made the first utterance, let us not adopt Hermes as our teacher, as some
say he was: that is manifest nonsense. Nor should we believe those philosophers
who say that it was by coining and teaching that names were assigned to things,
in order that men might have signs to facilitate their communication with each
other. For it is absurd, indeed absurder than any absurdity, not to mention impos
sible, that someone should all on his own have assembled all those multitudes —
there were no rulers at that time, and certainly no letters, seeing that there were
no words, since it was about these <that they were meeting, so that it was not> by
edict that their assembly was brought about — and having assembled them, in
structed them like a schoolteacher, holding a rod, and touching each thing have
said "Let this be called 'stone', this 'stick', this 'man', or 'dog' [...]". (Diogenes
of Oenoanda 10, 2.11-5.15, trans, in Long & Sedley 1987: 98; see also Chilton
1962)
Aristotle's formulation was also unsatisfying in its inability to answer two
other questions that struck later generations as obvious. First, where, in Aris
totle's account, does 'mind' (as psykhē or noēsis) come from, and why does it
differ between humans and other animals? Secondly, why do different lan
guages exist? If the passions of the mind/body are universal, why should men
arrive at different conventions for signifying them? Aristotle leaves us to sur
mise that it just depended on who was or was not talking to whom. Words
would be purely historical and arbitrary signs for purely natural meanings. Nor
does Aristotle make any distinction between concrete and abstract meanings; in
this he can once again be taken as consonant with Plato, for whom 'concrete'
meanings are illusions, and dangerous ones, arising out of the mistaking of
perceptible things, which are mutable, as being on the same order as meaning
and truth, which are unchanging and permanent. But apart from this non-
distinction of the concrete and abstract, there is still the very great difference
that Aristotle's meanings are in perceptible things and the minds and bodies of
all people, not in the heaven of the ideal.
both dimensions, his explanation makes crucial use of the concept of ethnos,
translated below as 'nationality', elsewhere as 'race': 1
And so names too were not at first deliberately given to things, but men's natures
[physeis] according to their different nationalities [ethnē] had their own peculiar
feelings and received their peculiar impressions], and so each in their own way
emitted air formed into shape by each of these feelings and impressions, according
to the differences made in the different nations by the places of their abode as well.
And then later on by common consent in each nationality special names were de
liberately given in order to make their meanings less ambiguous to one another and
more briefly demonstrated. And sometimes those who were acquainted with them
brought in things hitherto unknown and introduced sounds for them, on some oc
casions being naturally constrained to utter them, and on others choosing them by
reasoning in accordance with the prevailing mode of formation, and thus making
their meaning clear. (Epicurus, Letter to Herodotus 75-76, translation by Bailey
1926)
Certain of the positions taken in this letter were to have a very long life indeed,
often but not always in conjunction with an overt 'Epicureanism', so that it is
difficult to know how much Epicurus deserves credit for them, or indeed
blame. According to Epicurus, this exhalation takes different form, on the one
hand because of ethnic differences per se, on the other because pathē 'feelings'
(or 'passions') and phantasmata 'impressions' (or 'images' or 'imaginings')
vary by nationality. These differences are what today would be called 'racial'.
Some of them — and by implication, not all of them — have been caused by
the environment in which the people live. When the environment changes, by
the importation of 'things hitherto unknown', the language changes as well.
The second and perhaps most significant thing to note is that Epicurus
divides the creation of language into two distinct stages. The nature of the peo
ple in a particular ethnos, having been shaped by the combination of race and
environment, determines a first 'rough' language which is flawed by impreci
sion and extravagance, its words being ambiguous and not concisely expressed.
This rough language, spoken not by reasoning but by natural constraint, pro
vides the basis upon which social agreement and rational thought are con
structed. It makes it possible for people to become aware of the flaws in the
language and wilfully introduce changes to remedy them. In this second stage,
ambiguity and lack of conciseness are improved by coinages made by 'com
mon consent' within the ethnos.
There has long been dispute over the phrase Bailey translates as 'hitherto
unknown' in the last sentence of the citation from Epicurus' letter to Herodotus
above. The Greek is ou synorōmena pragmata, and synorōmena literally means
'seen together'. In a note to the passage Bailey glosses the phrase as '"things
previously unseen', i.e. either introduced from foreign tribes or invented for the
1. E.g., in the same translator's note on this passage (Bailey 1926: 248). In their translation of
Epicurus' letter, Long & Sedley (1987: 97) render ethnos alternately as 'tribe' and 'race'.
NATURAL GRAMMAR AND CONVENTIONAL WORDS 97
first time. Surely not 'tried to introduce the notion of things not visible' (Hicks
[1925: 605])" (Bailey 1926: 248). But the sense is much less clear than this
'surely not' would suggest. Long & Sedley (p. 97) opt for the ambiguous 'un
seen':
Also, the men who shared knowledge introduced certain unseen entities, and
brought words for them into usage. <Hence some> men gave utterance under
compulsion, and others chose words rationally, and it is thus, as far as the princi
pal cause is concerned, that they achieved self-expression.
What Hicks' reading implies, and Long & Sedley's suggests, is that abstrac
tions did not exist in the first stage of rough language, but were introduced in
the second stage by 'men who shared knowledge'. Whether or not it is a justi
fiable reading of Epicurus, the notion would reappear across the centuries that
abstractions, words for 'unseen entities' are less original and natural than
words with 'concrete' meanings. It is still a commonly taught principle in his
torical linguistics that abstract terms derive metaphorically from concrete ones
— this despite the fact that the principle is difficult to defend, given that we
cannot determine factually that it is so or that indeed the reverse was not the
case. Moreover, the abstract-concrete division is already problematic, since, as
Jean-Jacques Rousseau (1712-1778) would point out in his Discours sur
Vorigine et les fondemens de l' inégalité parmi les hommes (1817 [1753]: 272-
273; see futher p. 115 below), apart from proper names, the meanings of words
cannot be concrete things, but only classes of things, and classes are abstract
by definition.
It is not obvious that the Epicurean theory would have been taken by the
Sophists as falling on the physis side of their physis-nomos polarization. What
Epicurus calls tas physeis tön anthröpön "men's natures" is actually not far
removed from some of the early conceptions of nomos, before semantic
change and conflation with other antitheses took it from indicating the guiding
force behind human actions to the often arbitrary products of those actions (see
Joseph 1990a). In Sophistic times, a physis position was taken to be one that
held language to be determined by something outside of the human sphere. In
Epicurus, the entire physis-thesis distinction (see p. 15) has been shifted to
within the human sphere, and the dividing line drawn according to the pres
ence or absence of volition, will, choices made 'under compulsion' (i.e., from
within) versus those made 'rationally'. In the next section we shall see a later
version of this distinction deployed by Varro.
On only one point, to be discussed in the next paragraph, does Epicurus
directly contradict Aristotle's statements on language. Otherwise, he fills in
blanks, or answers questions which Aristotle implied but failed to answer. We
noted in the preceding section how Aristotle's view that words signify kata
synthēkēn, by agreed upon convention, was unsatisfying for three reasons, be
cause it raised the logical problem of how conventions could be made prior to
98 LIMITING THE ARBITRARY
the existence of a language in which to make them, and failed to address the
questions of why different languages should exist, and how mind came into
existence. Epicurus solves these problems by taking Aristotle's distinction
between human phōnē 'unarticulated voice' and gramma/dialektos/logos 'ar
ticulated speech' and historicizing it. Again, it is perhaps nothing more than a
common sense reading of Aristotle to divide language as Epicurus does into
two periods, first one in which human voice, like animal voice, is propelled
through the windpipe by emotions, then later one in which it is refined into
speech. Aristotle may not have presented these as separate historical periods,
but neither did he deny that they were such.
Where Epicurus contradicts Aristotle is in saying that "men's natures ac
cording to their different nationalities had their own peculiar feelings and re
ceived their peculiar impressions". For Aristotle, the mental/bodily passions,
the pathēmata, are "the same for all" (On Interpretation 16a3-9). Epicurus im
plies that from the bodies of the members of a race there arise racially distinct
feelings and impressions, out of which arises in turn the natural, pre-historical,
unconscious period of the 'rough' language. This is followed by the conven
tional, rational, historical period in which sound-meaning correspondences are
adjusted and (perhaps) abstractions are formed. In the pre-historic period,
meaning and vocal symbol are inseparable; indeed, body-mind-will-meaning-
sound are all one. In the historical period they are separated into distinct ele
ments. Thus does it become possible to make abstractions, concepts whose
meaning is mental rather than physical, and to choose wilfully the vocal signs
by which to represent them.
Despite the difficulty of reconciling Epicurus' approach with the 'classic'
view of physis maintained by the Sophists, in later antiquity Epicurean lan
guage theory was taken to be physis-based. For proponents such as Diogenes
of Oenoanda (cited on p. 95), it avoidance of what they saw as the historical
absurdity of a conventionalist account of the origin of language overrode any
other consideration. Another Epicurean, the Roman poet Lucretius (94?-55?
BC), makes some of the same points as Diogenes while modifying Epicurus'
views slightly and expanding on them in a way that implies a connection be
tween bodily theories of language and our animal nature — a connection that
would be a sub-text for the suppression of the body over the centuries to come.
Lucretius begins by reworking the Epicurean distinction between a natural first
stage of language and a second stage of rational refinement:
It was nature that compelled the utterance of the various noises of the tongue, and
usefulness that forged them into the names of things. (De rerum natura 5.1028-
1029)2
The first part of this sentence appears to agree with Epicurus, apart from the
2. "At varios linguae sonitus natura subegit / mittere et utilitas expressit nomina rerum [...]".
NATURAL GRAMMAR AND CONVENTIONAL WORDS 99
shift of the physical focus from the breath to the tongue. The second part,
however, does not say quite the same thing. Epicurus said that particular
coinings were made by consensus, in order to eliminate ambiguity and increase
conciseness. These aims certainly have to do with the 'usefulness' of words.
But where Epicurus stresses the element of rational choice, Lucretius turns
usefulness into an impersonal force. Out of context this might be taken for a
poetic use of metaphor, but in the rest of the passage Lucretius explicitly
denies the role of human rationality in the creation of language:
It was rather in the way that children's inarticulateness itself seems to impel them
to use gestures, when it causes them to point out with a finger what things are
present. For everyone can feel the extent to which he can use his powers. The calf
angrily butts and charges with his incipient horns before they have even
protruded from his forehead [...]. So to think that someone in those days
assigned names to things, and that that is how men learnt their first words, is
crazy. Why should he have been able to indicate all things with sounds, and to
utter the various noises of the tongue, yet others be supposed not to have had that
ability at the same time? Besides, if others had not already used sounds to each
other, how did he get the preconception of their usefulness implanted in him?
How did he get the initial capacity to know and see with his mind what he
wanted to do? Again, one person could not subdue many and compel them to
want to learn the names of things. Nor is it easy to find a way of teaching and
persuading a deaf audience of what needs to be done: they would utterly refuse to
tolerate any further his drumming into their ears the unfamiliar sounds of his
voice. (De rerum natura 5.1030-35, 1041-55)3
What is unclear is whether Lucretius still has in mind the two Epicurean stages
of the formation of language and is staking out a new position regarding the
second stage, or has forgotten about the second stage and is simply making the
same argument as Diogenes of Oenoanda about the impossibility of rational
language-making in the first stage. Lucretius invokes examples of the bodies of
animals to suggest that language is both as natural and as useful as a calf's
horns, a lion's teeth and claws, or a bird's wings. The idea that someone
invented language and taught it to others is as absurd to the Epicureans as it
would be to suggest that some calf invented horns and taught other calves to
grow them. Lest the objection be made that language qua vocal sounds is not
commensurate with a part of the body, Lucretius goes on to draw analogies
3. "[•••] non alia longe ratione atque ipsa videtur / protrahere ad gestum pueros infantia lin
guae, / cum facit ut digito quae sint praesentia monstrent. / sentit enim vis quisque suas quoad
possit abuti. / cornua nata prius vitulo quam frontibus extent, / illis iratus petit atque infestus
inurget. / [...] / proinde putare aliquem turn nomina distribuisse / rebus et inde homines didi-
cisse vocabula prima, / desiperest. nam cur hie posset cuneta notare / voeibus et varios sonitus
emittere linguae, / tempore eodem alii facere id non quisse putentur? / praeterea si non alii
quoque vocibus usi / inter se fuerant, unde insita notities est / utilitatis et unde data est huic
prima potestas, / quid vellet facere ut sciret animoque videret? / cogere item pluris unus vic-
tosque domare / non poterat, rerum ut perdiscere nomina vellent; / nec ratione docere ulla
suadereque surdis, / quid sit opus facto, facilest; neque enim paterentur / nec ratione ulla sibi
ferrent amplius auris / vocis inauditos sonitus obtundere frustra".
100 LIMITING THE ARBITRARY
4. "postremo quid in hac mirabile tantoperest re, / si genus humanum, cui vox et lingua vigeret,
/ pro vario sensu varia res voce notaret? / cum pecudes mutae, cum denique saecla ferarum /
dissimilis soleant voces variasque ciere, / cum metus aut dolor est et cum iam gaudia gliscunt. /
[...] / ergo si varii sensus animalia cogunt, / muta tarnen cum sint, varias emittere voces, /
quanto mortalis magis aecumst tum potuisse / dissimilis alia atque alia res voce notare?"
NATURAL GRAMMAR AND CONVENTIONAL WORDS 101
5. On Stoic and Alexandrian treatments of language see Baratin (1982) and Malmberg (1991:
72-88).
6. "[...] voluntarium est, quo ut cuiusque tulit voluntas declinavit. Sic tres cum emerunt Ephesi
singulos servos, nonnunquam alius declinat nomen ab eo qui vendit Artemidorus, atque
102 LIMITING THE ARBITRARY
So the name as a lexical entry is subject to the buyer's free choice. But when it
comes time to add morphological inflections to the name, free choice no longer
applies; inflection is a matter of 'natural' derivation. The preceding passage
continues:
In contrast I call natural derivation that which is determined not by an individ
ual's will but by common consent. Just so, once the names are given, all three de
rive their case-forms in like fashion, and in the same way say in the genitive case
Artemidori, lonis, Ephesi [...]. (De lingua latina 8.22)7
Varro further declares that the two types are not mutually exclusive:
It sometimes happens that both types are found together, so that in voluntary
derivation the effects of nature may be observed, and in natural derivation the ef
fects of will [...]. (Ibid., 8.23)8
Whereas Book 8 says that analogy must exist everywhere in language if it is to
be considered a principle, Book 9 says that it is sufficient merely that it exists
at all (9.35); that "not every utterance is naturally fitted" for it (9.38); and that
— up to the point where it offends against general usage (9.35) — it ought to
be followed and increased (9.6).
Thus, holds Varro, analogists and anomalists both have it right, but for
different aspects of language. In Varro's compromise, anomaly accounts for
just those parts of language which are wilfully chosen by the individual
speaker, analogy for those parts where no individual choice enters because the
'choice' has already been made by common consent. The example Varro
chooses for the non-voluntary, 'natural' type is inflectional morphology, the
same one Pinker would adopt for very similar purposes over 2000 years later
(see Introduction). Varro's example of the voluntary type is from an area of
language, naming, which modern linguistics has marginalized starting at least
from Saussure, who accused those ancient language theorists who took naming
into account of falling victim to a fallacy he called 'nomenclaturism', the false
assumption that languages consist of an inventory of names for things. It is
clear from this passage alone that Varro did not think of language in that way.9
As I read Varro, voluntary derivation holds for the link between word and
meaning generally — it is the equivalent of what Chomsky now calls 'Saus-
surean arbitrariness' — and more generally still for all those lexical choices
Artemam appellat, alius a regione quod ibi emit, ab Ionia Iona, alius quod Ephesi Ephesium
[...]"
7. "Contra naturalem declinationem dico, quae non a singulorum oritur voluntate, sed a comuni
consensu. Itaque omnes impositis nominibus eorum item declinant casus atque eodem modo
dicunt huius Artemidori et huius lonis et huius Ephesi [...]".
8. "Cum utrumque nonnunquam accidat, et ut in voluntaria declinatione animadvertatur natura
et in naturali voluntas [...]".
9. Indeed, I am not sure how many of the theorists of language accused of nomenclaturism by
Saussure or by Roy Harris, who has revived interest in the concept (see Harris 1980), were
actually guilty of it, as I have discussed at length in Joseph (1997).
NATURAL GRAMMAR AND CONVENTIONAL WORDS 103
which go into deciding what to talk about, which Saussure would consign to
the linguistique de la parole. Perhaps this is reading too much into Varro's
words; but the alternative is that he is talking strictly about the naming of
slaves, which is too restrictive as a topic to make sense in this context. The
passage seems to be best interpreted with voluntary derivation referring more
or less to the modern concept of the lexical module, and natural derivation to
the functional module, of which inflectional morphology provides the best ex
ample.
This is to my knowledge the origin of the notion of a bipartite linguistic
mind, part of which is open to free choice, and part of which is closed. We find
already in Plato the notion of a bipartite mind split essentially along the terms
of the modern conscious-unconscious divide, which Plato likens to a birdcage
in which some of the birds we can reach and others we cannot. But he does not
imagine language itself as being divided for us mentally in this way. Similarly,
discussions about the 'arbitrary' or 'voluntary' nature of language prior to
Varro are about whether or not all of language is so, not whether some part of
it might be so.
In Varro's compromise, the link between word and meaning is voluntary,
the grammatical processes that words undergo are compulsory. In modern
terms, this means that the lexical meanings of words exist on a higher plane of
consciousness than do the rules of morphology and syntax. As a result the for
mer are more subject to the historical and political processes we call conven
tion, while the latter, being unconscious, are more natural in their origin and
operation.
10. "Figura in nominibus aut simplex est aut conposita. simplex est a natura, id est quae de una
re constat, ut doctus; conposita, quae de duabus constat rebus, quae accipit ornatum, ut indoc
tus. simplex est, quae unam rem habet, duplex, quae de duabus rebus constat, id est conposita".
11. See Baratin & Desbordes (1981: 62-64). The idea that 'natural' composition-derivation
runs through all language may seem to be implicit in Priscian's analogy between the composi
tion of 'simple' words by letters and syllables, and the composition of 'compound' words by
simple ones and of sentences by words of both types, since this analogy suggests that both
simple and compound words are grounded in nature. However, other comments by Priscian
concerning the 'chance' generation of certain accidental forms make this less clear, and leave
open the possibility of ultimately arbitrary derivation of those simple forms that cannot be co
gently derived in a natural way from some other base form (see Amsler 1989: 76).
NATURAL GRAMMAR AND CONVENTIONAL WORDS 105
the period. For Servius, the noun is the principal part of speech, and all others
are 'procreated' from it (Keil ed., 4.406). For Virgilius Maro Grammaticus in
the late 7th century,
However, the relationships of nouns and verbs seem to me to be similar to human
genealogy, in which we speak of the father, the son, the grandson, and the great-
grandson. (Virgilius Maro Grammaticus, Epitomae 5.14)12
Thus, writes Amsler (1989: 202), "the father-noun glorificatio begets the son-
verb glorifico (the verb nascitur a patre) ['is born from the father'], who be
gets the grandson-primitive noun gloria, who begets the greatgrandson-
adjective gloriosus-a-um". One might have expected the 'primitive noun' glo-
ria to be the 'father', but for Virgilius it is the noun forms ending in -o that are
the originalia nomina, because from them the verbs, also ending in -o, are born
(Epitomae 5.14). Amsler comments further (ibid.), "We read here an echo of
the Patristic trinitarian debate regarding the derivation of the Son from the Fa
ther [...]", a debate which would go on to generate sectarian schisms for centu
ries to come.
The relative naturalness of nouns and other parts of speech under such a
view depends on whether derivation is itself conceived of as natural or not:
thus for Servius, who associates derivation with 'art', only nouns could con
ceivably be fully natural. For those thinkers who follow Priscian in seeing
derivation as the (super)natural core of language, however, there is a potential
tension between the status of the noun as the starting point of derivation, which
seems to fix its place in nature, and the relatively conventional character of
whatever precedes derivation. This may provide an additional explanation for
Virgilius Maro Grammaticus' identification above of the father-noun as the
derived form glorificatio: the fact that it includes derivation means that the
starting point of language is grounded firmly in nature, hence less exposed to
arbitrariness.
12. "Vidantur autem mihi nominum et verborum adfinitates humanae genialogiae similitu-
dinem habere, qua dicimus patrem et filium et nepotem pronepotemque".
106 LIMITING THE ARBITRARY
This may also help explain why such a high proportion of Renaissance
borrowings from the classical into the Romance languages were of nouns and
adjectives. Borrowing of verbs was far rarer, of syntactic devices rarer still,
though it does occur and has important linguistic consequences (see further
Joseph 1987a). Borrowing of prepositions was virtually non-existent.
These developments might well be connected in part to the continuing as
sociation of nouns with real substance discussed in the preceding sections.
They may also have been driven by the fact that, by the Renaissance, the Ro
mance languages (except Rumanian) had lost all traces of case marking on
nouns and adjectives, with the result that these parts of speech were no longer
formally implicated in grammar in the same way as verbs, which continued to
show extensive inflection for person, tense, and so on. Thus, a borrowed noun
did not threaten to do violence to the grammatical system, hence to the 'nature'
of the language, in the same way that a borrowed verb did.
The effects of the preferential treatment of nouns in Renaissance linguis
tic thought and language use continue to be felt in the present day. Modern
rhetoric has long been at war against a 'nominal' style, characteristic of aca
demic and legal writing and considered to be too 'heavy' in nouns, with a
minimum of verbs other than the copula.13 The nouns in question tend over
whelmingly to be precisely those which were borrowed from the classical lan
guages in the Renaissance. Legal stylistics is caught in a conflict between the
democratic desire for more 'natural' style and the extreme 'precision' of
meaning which the law requires, and which it recognizes above all in nouns,
followed by adjectives, followed distantly by verbs — the same hierarchy that
prompted the rather radical legal stylistics of Jeremy Bentham (1748-1832),
discussed later in this chapter. Again, these rhetorical-stylistic reactions to
parts of speech are inseparable from the trends established in cross-linguistic
borrowing in the 16th and 17th centuries.
Nature v. convention reborn: from ideas and experience to usage and genius
By the late middle ages the nature-convention debate had not died out,
but had ceased to appear immediately relevant to the concern of grammarians.
Yet the realism-nominalism debate which dominated philosophy during the
period, and which lay at the convergence point of logic, language theory, and
theology, can be seen as a version of nature-convention restricted to universal
categories, asking not how it is that their names attach to them, but whether in
fact there is anything in nature for the names to attach to. If not, then presuma
bly those names are purely conventional, both in terms of the sounds which
compose them and of their meanings: this was the nominalist position, whose
most vigorous defender was Peter Abelard (1079-1142). It was associated with
13. For one of the rare treatments of this or any other issue in rhetoric by a modern American
linguist, see Wells (1960).
NATURAL GRAMMAR AND CONVENTIONAL WORDS 107
agreed that it is 'not by any natural connexion, that there is between particular ar
ticulate Sounds and certain Ideas, for then there would be but one Language
amongst all Men; but by a voluntary Imposition, whereby such a Word is made
arbitrarily the Mark of such an Idea.'
THEOPHILUS. I know that the Scholastics and everyone else are given to saying
that the significations of words are arbitrary {ex instituto), and it is true they are
not settled by natural necessity; but they are settled by reasons — sometimes
natural ones in which chance plays some part, sometimes moral which involve
choice. (Leibniz 1981 [1765]: 278)14
For Locke in his Essay Concerning Humane Understanding (1975 [1690]),
III.i.1, language has a natural aspect in the simple fact that the vocal apparatus
was provided by Nature as a mechanism for expression. But as soon as words
"came to be made use of by Men, as the Signs of their Ideas" they entered the
realm of the voluntary and the arbitrary (ibid.; see further Taylor 1992, Chap.
2). This is not so clearly the case for a believer in innate ideas like Descartes or
Leibniz (speaking here through his mouthpiece Theophilus, the 'lover of
God'); for if ideas themselves are not arbitrary, neither can the language which
expresses them be wholly so. If the mind is the mirror of nature, and language
the mirror of mind, then language too should mirror nature.
Much as Aristotle refigured the eternal Ideas of Plato's heaven as con
cepts existing in the mind, Locke refigured the innate ideas of Descartes as
constructs of the individual mind — but acquired and knowable only through
the body. 15 By giving the individual will this role in the construction of ideas,
Locke is able to explain how it is that the conception of justice, for example,
differs from individual to individual. He notes, however, that this construction
14. Words in single quotes are from Locke's original text. I have restored Locke's italics and
capitals to Remnant and Bennett's translation, following the standard edition by Nidditch
(1975) of Locke's Essay. In the original French text of Leibniz which follows, the words in
single quotes were taken directly by Leibniz from Coste's (1700) translation of Locke:
PHILALÈTHE. Maintenant, les mots étant employés par les hommes pour être signes de
leurs idées, on peut demander d'abord comment ces mots y ont été déterminés; et l'on
convient que c'est 'non par aucune connexion naturelle qu'il y ait entre certains sons
articulés et certaines idées (car en ce cas il n'y aurait qu'une langue parmi les hommes)
mais par une institution arbitraire en vertu de laquelle un tel mot a été volontairement
le signe d'une telle idée.'
THÉOPHILE. Je sais qu'on a coutume de dire dans les écoles et partout ailleurs que les
significations des mots sont arbitraires (ex instituto) et il est vrai qu'elles ne sont point
déterminées par une nécessité naturelle, mais elles ne laissent pas de l'être par des rai
sons tantôt naturelles où le hasard a quelque part, tantôt morales, où il y entre du
choix" (III.ii.1).
In the last sentence the translator Coste has made a transposition in Locke's text; the original
reads, " [...] by a voluntary imposition, whereby such a word is made arbitrarily [...]".
15. Aarsleff (1982: 173-175) points out the danger of imposing too great a polarization be
tween Cartesian 'innateness' and the views of Locke, who did after all accept that the capaci
ties for sense and reason were innate. In the paragraphs which follow I shall be pointing out
still other points of partial agreement between Descartes and Locke. However, the tradition of
strongly opposing them is a venerable one dating back at least to Leibniz, and has an authori
tative weight of its own that cannot simply be thrown off by recourse to the ipsissima verba.
NATURAL GRAMMAR AND CONVENTIONAL WORDS 109
Where language is concerned, Locke agrees with Descartes that language has a
'natural' aspect in the simple fact that the vocal apparatus was provided by
Nature as a mechanism for expression (Essay, III.i.1), but as soon as soon as
sounds "come to be made use of by Men, as the Signs of their Ideas" (Essay,
III.ii.1), they too enter, like the complex ideas themselves, into the realm of the
voluntary and the arbitrary.
Words in their primary or immediate Signification, stand for nothing, but the
Ideas in the Mind of him that uses them [...]. (Essay, III.ii.2)
Again like Descartes, Locke believes strongly in free will. Their views part
ways when it comes to individuality and ideas. The individualistic nature of
language is such that, for Locke, a speaker cannot 'impose' his meaning on any
other speaker (III.ii.8):
And every Man has so inviolable a Liberty, to make Words stand for what Ideas
he pleases, that no one hath the Power to make others have the same Ideas in
their Minds, that he has, when they use the same Words, that he does.
As Taylor (1992: 41) points out, this individualistic aspect separates Locke not
only from earlier linguistic thought, but also from later 'conventionalist' posi
tions, which treat language as a set system of conventions the individual can
merely make use of, whereas for Locke it is really the individual who creates
the convention which each act of signification. Nevertheless, he says,
[T]he necessity of Communication by Language, brings Men to an agreement in
the signification of common Words, within some tolerable latitude, that may
serve for ordinary Conversation: and so a man cannot be supposed wholly igno
rant of the Ideas, which are annexed to words by common Use, in a Language
familiar to him. But common Use, being but a very uncertain Rule, which re
duces it self at last to the Ideas of particular Men, proves often but a very variable
Standard. (Essay, III.xi.25)
to
110 LIMITING THE ARBITRARY
I, when I speak a word, have something in mind, and you believe you discern that I have
this thing in mind
— whether or not you actually do. Unlike Plato, Locke goes on to give a series
of prescriptive 'Rules' for making language correspond to its ideal condition
(Essay, III.xi.8; see Taylor 1992: 43), which include making sure that one's
words (and therefore ideas) are distinct, correspond to real things, and are in
variable in meaning.
The Names of simple Ideas, Substances, and mixed Modes, have also this differ
ence; That those of mixed Modes stand for Ideas perfectly arbitrary: Those of
Substances, are not perfectly so; but referr to a pattern, though with some lati-
tude: and those of simple Ideas are perfectly taken from the existence of things,
and are not arbitrary at all. (Essay, III.iv.17)
'Real' things for Locke, and for empiricism generally, are those that are know-
able by the body together with the mind. Abstractions as such are knowable
only to the mind, and therein lies the breakdown in 'understanding' which
Locke would remedy by breaking abstractions down into their sensible compo
nents, what is surely and universally knowable because it is known through the
body.
Grammarians took positions corresponding to those of the philosophers.
The post-Cartesian view of the relationship between language and mind gave
rise to projects for the writing of a grammaire générale in 17th-century France,
the most famous being that undertaking at the Jansenist school of Port Royal
by Claude Lancelot (1615-1695) and Antoine Arnauld (1612-1694) and pub
lished in 1660 as the Grammaire generale et raisonnée. It was a 'general'
grammar in that its aim was to analyze the grammar of French not for its own
sake, but as a source of insight into the correlations between, on the one hand,
the structure of the human language faculty generally, and on the other, logic,
the structure of thought. Lancelot and Arnauld did not think that French or
any other language represented a 'perfect' embodiment of logic; rather, they
believed that one had to have an understanding of how logic was embodied in
any particular language for one's use of that language to be rational, and for
one's use of logic not to be obfuscated through contradiction with the inherent
logical structure of the language one was using. In the Port-Royal perspective,
which would have at least as great an impact on rhetoric as on language theory
in France and elsewhere on the continent, those elements of any language
which failed to correspond with logic were deemed arbitrary, in the sense of
being capricious, random accidents that contravened the logical essence of the
language and that were therefore best avoided.
The concerns of Port Royal are reminiscent of the analogy-anomaly de-
16. Thus it was linked in spirit with speculative grammar, though its much greater attention to
actual linguistic structure makes it finally much less subservient to logic than its medieval
predecessor.
NATURAL GRAMMAR AND CONVENTIONAL WORDS 111
bate as recounted by Varro, but draw opposite conclusions from similar prem
ises: Lancelot and Arnauld agree with the Stoics that language is a direct mani
festation of logic, but do not therefore take the anomalist position that language
should be accepted as it is. Instead they arrive at the analogist position of the
Alexandrian grammarians, that language should be made to accord with our
ideas about what logic should be. What has changed is above all the loss of
faith in logic as an absolute, monolithic entity, a universal logos that is either
present or absent. For Port Royal, logic is no longer a given. Indeed, the
teaching of logic is their essential problem, and they see language as the key to
solving it. Insofar as they construct that solution as a division of language into
arbitrary and non-arbitrary elements, and exercise a preference for the latter
over the former, they would fall on the naturalist side of the nature-convention
dichotomy; again, such an observation is of limited historical validity, yet can
be helpful in trying to construct an overall perspective on the evolution of lin
guistic thought.
Within a few years, the language theory of Port Royal would be appro
priated for the same project Joachim Du Bellay (1522-1560) had been con
cerned with in the previous century, the 'defence' of French (discussed in
Chap. 5 below), even though in principle this contradicted the Port-Royal view
of particular languages as equally amenable to logical improvement. In his
Avantages de la langue françoise sur la langue latine (1669), Louis Le Labou
reur (1615-1679) declares that there is a logical order in which things appear
to the mind; that languages differ in their word order, and so differ in how
closely they represent this natural order; and that French is better than Latin in
this regard, because its syntax follows the 'logical' order sub
ject-verb-complement, whereas Latin does not. Thus the French sentence le
discours couloit plus doux que miel de sa langue 'the speech flowed more
sweetly than honey from his tongue' (the word order is precisely the same in
the English translation) is intrinsically superior to its Latin equivalent Ex eius
lingua melle dulcior fluebat oratio (Cicero, De senectute X.31, speaking of the
eloquence of Nestor according to Homer), word for word: 'From his tongue
than-honey sweeter flowed speech'. In the Varronian tradition, Le Laboureur
states that words are human inventions, and the only criterion of evaluation is
that a language should have all the words it needs. Syntax, however, is
grounded:
It is always true that if one changes the natural course of thought by inverting the
words which are its signs, this causes a disorder which produces an effect wholly
contrary to the institution of speech, since it sometimes suspends and disturbs un
derstanding of the discourse. (Le Laboureur 1669)17
17. "Or il est constant que si l'on change le cours naturel de la pensée par l'inversion des mots
qui en sont les signes, cela cause un desordre qui produit un effet tout contraire à l'institution
112 LIMITING THE ARBITRARY
Louis Petit (1615-1693), on the other hand, holds that word order is as arbi
trary as words themselves:
All words, or at least most of them, are the children of chance and caprice. One
can say the same about their endings, their genders, and the manner of arranging
them [...]. Each nation has arranged its own [language] in accordance with its
genius [...]. But to come back to words, mostly children of chance and caprice, as
I was saying a moment ago, do they not show how little rationality there is in the
composition of languages? They are whatever was pleasing to men, whose minds
can be called a perpetual motion. (Petit 1688, Dialogue XII)18
Just as the followers of Port-Royal found in that school's Stoic-like views a
justification for not anomaly, but analogy, Petit and those who shared his Al
exandrian-like views found in it a justification for anomaly, enunciated by
Claude Favre, seigneur de Vaugelas (1585-1650) as the doctrine of bon usage,
that correctness in language is to be determined not by abstract notions of
grammar or logic but following the usage of the noblest element of the popula
tion.19 Locke's own rules of usage are of the sort that in modern times would
normally be classed as rhetorical rather than grammatical; but in fact it may
well be that Locke's dictum to follow common usage where possible would
cover almost the entire domain of grammar and syntax, in which case he is
very much on the side of le bon usage in all but the lexical domain.
But reading back through the quotation from Petit, we find one very un-
Lockean element. How capricious and arbitrary is usage if each nation has ar
ranged its language 'in accordance with its genius'? In this seemingly strong
profession of conventionalist faith, we glimpse a new manifestation of natural
ness that will not come fully into its own until the Romantic period further glo
rifies that notion of 'genius', the spiritual quintessence of the gens, the people,
the race, occasionally incarnated in an individual artist or thinker but really
belonging to the race as a whole and with an existence apart.
de la parole, entant qu'il suspend et qu'il trouble méme quelque fois l'intelligence du dis
cours".
18. (The interlocutor Cléante is speaking:) "Tous les mots, au moins la plus part, sont enfans
du hazard et du caprice. On peut dire le mesme de leurs terminaisons, de leurs genres, et de la
maniere de les arranger [...]. Chaque nation a disposé la sienne [langue] selon son genie [...].
Mais pour revenir aux mots, la plus part enfans du hazard et du caprice, comme je le disois tout
à l'heure, ne font-ils pas voir qu'il y a peu de raison en la composition des langues? Elles sont
ce qu'il a plû aux hommes, dont l'esprit se peut appeller le mouvement perpetuel".
19. For fuller versions and insightful discussions of these texts see Rickard (1991).
NATURAL GRAMMAR AND CONVENTIONAL WORDS 113
forward invoking the authority of Locke did not adhere to his views on the ar
bitrary and the natural in language. One of those who took up and developed
Lockean themes concerning language and thought was Etienne Bonnot, the
abbot of Condillac (1714-1780), especially in his 1746 Essai sur l'origine des
connoissances humaines. Early in that work Condillac proposes a biblically-
based compromise between the Cartesian view of mind as possessing innate
ideas and controlling the senses, and the Lockean view of it as a tabula rasa
entirely dependent on the senses for the knowledge it acquires. (He never cites
Descartes' Passions de l'ame, with its emphasis on habit formation, especially
where language is concerned.) Condillac believes that the Cartesian view is
true of the pre-lapsarian mind, the state of Adam and Eve before original sin,
and to which we will return when we die; while the Lockean view is true of the
mind of all living human beings since the Fall.
Thus, when I say that we have no ideas which do not come from the senses, it
must be remembered that I am only talking about the state we have been in since
the Fall. Applied to the mind in the state of innocence, or after its separation from
the, this proposition would be completely false. (Essai sur l'origine 1.1.1 [1746]
= Condillac 1947: 7-8. My translation; italics in original.) 20
20. "Ainsi, quand je dirai que nous n'avons point d'idées qui ne nous viennent pas des sens, il
faut bien se souvenir que je ne parle que de l'état où nous sommes depuis le péché. Cette
proposition appliquée à l'ame dans l'état de l'innocence, ou après sa séparation du corps, seroit
tout-à-fait fausse".
21."Je me borne donc [...] à l'état présent. Ainsi il ne s'agit pas de considérer l'ame comme
indépendante du corps [...] ni comme unie à un corps dans un systême différent de celui où
nous sommes. Notre unique object doit être de consulter l'expérience [...]".
114 LIMITING THE ARBITRARY
ondly because the individual can freely choose to use the particular symbol or
leave it unused. It is not however individual in the Lockean sense that each in
dividual speaker 'chooses' to create it with each usage.
Condillac makes a primary distinction between the 'language of action'
(gestural language) and conventional language on the grounds that the latter,
but not the former, is successive, that is, it signifies in bits spread across time.
The sounds make up the word, the words make up the sentence, the sentences
make up the discourse. Both in the species and in the individual, it is with the
acquisition of conventional language that thought, too, changes from an atem-
poral, synthetic mode, to a temporal, analytical one.
In later works, Condillac stressed that artificial signs must be based on
the analogy of natural and accidental ones in order to be functional; they can
never be arbitrary.22
But I have by now said enough to demonstrate to you that languages are the work
of Nature; that they have formed themselves, so to speak, without us; and that in
working with them, we have only obeyed slavishly our manner of seeing and
feeling. (Cours d'Etudes pour l'instruction du Prince de Parme, 2: Grammaire
[1775] = Condillac 1947: 432)23
Languages are not a bunch of expressions taken at random, or which are used
only because we have agreed to use them. If the use of each word presupposes a
convention, the convention presupposes a reason for the adoption of each word.
And analogy, the source of law, and without which it would be impossible to un
derstand one other, does not permit an absolutely arbitrary choice. But, because
different analogies lead to different expressions, we believe we are choosing, and
we are wrong: for the more we consider ourselves to be in control of the choice,
the more we choose arbitrarily, and the worse are the choices we make. (La
langue des calculs [1798] = Condillac 1948: 419.)24
22. Aarsleff (1982) ignores these later texts in order to paint an unambiguous picture of Con
dillac as the key source of the modern tradition of the arbitrariness of linguistic signs, based on
the Essay. Even if it is true that the Essay was Condillac's most often cited text — even if we
were to go further and agree with Aarsleff that it is the most influential text on language of the
whole 18th century — Aarsleff s procedure is methodologically inadequate by the very criteria
he invokes in his stern criticism of others who take a particular work out of the context of its
author's whole output. Moreover, as we have seen above, the Essay even taken in isolation is
much more ambivalent on the question of arbitrariness than Aarsleff portrays it. The problem
is that Aarsleff has a blatant agenda of proving that all the leading ideas of 18th, 19th and 20th-
century linguistics come from Condillac, and more precisely that none of them originates with
any of the German thinkers to whom they are usually credited. One of Aarsleff's achievements
is to have helped restore Condillac to a place in the pantheon of linguistic thought, and perhaps
to do so required that he exaggerate Condillac's role and suppress that of others in this way. If
so, it is certainly time to restore the balance.
23. "Mais j'en ai assez dit, Monseigneur, pour vous faire voir que les langues sont l'ouvrage de
la nature; qu'elles se sont formées, pour ainsi dire, sans nous; et qu'en travaillant, nous n'avons
fait qu'obéir servilement à notre maniêre de voir et de sentir".
24. "Les langues ne sont pas un ramas d'expressions prises au hasard, ou dont on ne se sert que
parce qu'on est convenu de s'en servir. Si l'usage de chaque mot suppose une convention, la
convention suppose une raison qui fait adopter chaque mot, et l'analogie, qui donne la loi, et
sans laquelle il seroit impossible de s'entendre, ne permet pas un choix absolument arbitraire.
NATURAL GRAMMAR AND CONVENTIONAL WORDS 115
Mais, parce que différentes analogies conduisent à des expressions différentes, nous croyons
choisir, et c'est une erreur: car plus nous nous jugeons maîtres du choix, plus nous choisissons
arbitrairement, et nous en choisissons plus mal".
25. "[...] une machine ingénieuse, à qui la nature a donné des sens pour se remonter elle-même
[...]".
26. "J'aperçois précisément les mêmes choses dans la machine humaine, avec cette différence,
que [...] la nature seule fait tout dans les opérations de la bête, au lieu que l'homme concourt
aux siennes en qualité d'agent libre. L'une choisit ou rejette par instinct, et l'autre par un acte
de liberté [...]".
27. "[I]l a supposé ce que je mets en question, savoir, une sorte de société déjà établie entre les
inventeurs du langage [...]".
116 LIMITING THE ARBITRARY
this state cannot have continued beyond the establishment of any social bond,
not even that of the family. He imagines primitive humans as entirely solitary
wanderers, with no need to communicate. They did however feel the urge to
express themselves, and because they had no social intercourse, there were as
many 'languages' as there were individuals. Only when humans finally settled
down into living together as families, bringing an end to the primeval dawn,
did "speech become established and improved" (p. 286).28
The Discourse does not claim to answer the problem of the origin of lan-
guage, only to establish its Lucretian paradoxicality. Rousseau later read-
dresses it in his Essay on the Origins of Language, written in 1761 but not
published until 1781, as result of which it did not have anything like the impact
of the Discourse. Rousseau takes exception with Condillac's idea that the ori-
gin of language is in the cries and signs that arise from physical causes and the
need to communicate them. The Essay holds that although "needs dictated the
first gestures, [...] the passions drew out the first words" (Rousseau 1817
[1761]: 505).29 In the opening paragraph of the Essay, Rousseau makes a
statement that is very reminiscent of Lucretius, before posing the essential Epi-
curean question and giving his own view, derived from that enunciated by by
Charles-Louis de Secondat, Baron de Montesquieu (1689-1755) in De l'Esprit
des lois (1748):
Usage and need cause each person to learn the language of his country, but what
is it that makes this language that of his country and not of another? To answer
this, one must find some reason that connects to the locale, and that is anterior to
the customs themselves (Rousseau 1817 [1761]: 501).30
The conditions in a harsh northern country provoke very different human re-
sponses from those in a lush southern one, whence the difference in character
between European and other language: "[S]peech, being the first social institu-
tion, owes its form entirely to natural causes" (ibid.).31 Thus, where language is
concerned, convention is nature: climate, landscape, race, national character,
and language are intimately and inseparably bound, and the choice of conven-
tions is never free to individuals. Even in their origin they were not arbitrary,
but determined by that conjunction of causes which defines the genius of a
people.
The idea of a direct effect of 'locale' on language is anticipated by Epicu-
rus' remark about "the differences made in the different nations by the places
of their abode". In Rousseau and his German contemporaries like Johann
Georg Hamann (1730-1788) and Johann Gottfried Herder (1744-1803, dis
cussed below), there emerges a sense of climate, landscape, race, national
character and language all being intimately and inseparably bound together, so
that, as with Condillac, whatever conventional character languages exhibit is
ultimately superficial. The choice of conventions is never in any sense free to
individuals, and even in its origin it was not arbitrary, being determined by that
conjunction of causes which together define what will eventually be called the
Volksgeist, the National Spirit, the 'genius' of a people that is reflected in their
language and other 'folk' creations.
The fact that Rousseau wrote the Essai as part of an ongoing polemic
with the composer and music theoretician Jean-Philippe Rameau (1683-1764)
helps explains why most of the second half is about music, and why the section
on language is especially concerned with sounds. In Rousseau's view, the pri
mary difference between the northern and southern languages (the latter repre
sented particularly by Chinese and Arabic) has to do with the predominance in
the former of consonants, which he sees as an intrusion of the analytical into
the more natural kind of language represented by vowels. Vowels, accents,
tones, and rhythm predominated in the original human language and have con
tinued to do so in southern languages unless and until they have fallen under
the influence of the northern cultures, which Rousseau believes happened pri
marily through the extension of phonetic writing.
The simple sounds issue naturally from the throat, the mouth is naturally more or
less open; but the modifications of the tongue and the palate, which create
articulation, require attention and exercise. No one performs them without
meaning to; all children have to learn them and some do not achieve this easily.
32. "Les simples sons sortent naturellement du gosier, la bouche est naturellement plus ou
moins ouverte; mais les modifications de la langue et du palais, qui font articuler, exigent de
l'attention, de l'exercice; on ne les fait point sans vouloir les faire; tous les enfans ont besoin
de les apprendre et plusieurs n'y parviennent pas aisément. [...]
"Je ne doute point qu'independamment du vocabulaire et de la syntaxe, la première
langue, si elle existait encore, n'eût gardé des caractères originaux qui la distingeraient de to
utes les autres. [...]
118 LIMITING THE ARBITRARY
The third paragraph of this citation harks back directly to Aristotle's view of
the 'naturalness' of voice prior to articulation. In History of Animals iv.9,
Aristotle states that voice becomes dialektos "speech" when it is articulated by
the tongue and other organs to distinguish consonants from vowels. From this
Rousseau deduces that the original human language, being close to the state of
nature, had few consonants. He treats consonants as though they represent the
language of the mind imposing itself on that of the body, culture imposing
itself on nature.
Rousseau depicts the original language as southern rather than northern
in type, consisting entirely of tropes, images, and other forms of figurative
speech, with the force of onomatopoeia felt continually. This for him is the
'natural' state of language:
This [original] language would have many synonyms for expressing the same
being through its different relationships; it would have few adverbs and abstract
words for expressing these same relationships. It would have many augmenta-
tives, diminutives, compound words, expletive particles for giving a cadence to
periods and roundness to sentences. It would have many irregularities and anom
alies, would neglect grammatical analogy in favor of euphony, number, harmony,
and the beauty of sounds. Instead of arguments it would have proverbs, it would
persuade without convincing and depict without reasoning. It would resemble the
Chinese language in certain regards, Greek in others, Arabic in others. Extend
these ideas to all their ramifications, and you will find that Plato's Cratylus is not
so ridiculous as he appears. (Rousseau 1817 [1761]: 507)33
Rousseau has revived the traditional lexicon-grammar distinction, which had
become largely irrelevant in Enlightenment language theory, and restored its
traditional association with a nature-convention dyad. But he has reversed the
traditional affiliations, so that words predominate in the natural state — hence
the remark on synonyms — and the reference to "augmentatives, diminutives,
and compound words", suggest that it is especially nouns and adjectives he has
in mind. Grammar as it exists in the European languages is the unnecessary,
analytical, arbitrary supplement, and the closing invocation of Cratylus encap-
"Comme les voix naturelles sont inarticulées, les mots auraient peu d'articulations;
quelques consonnes interposées, effaçant l'hiatus des voyelles, suffiraient pour les rendre cou
lantes et faciles à prononcer [...L]es voix, les sons, l'accent, le nombre, qui sont de la nature,
laissant peu de chose à faire aux articulations, qui sont de convention, l'on chanterait au lieu de
parler [...]".
33. "Cette [première] langue aurait beaucoup de synonymes pour exprimer le même être par
ses différens rapports; elle aurait peu d'adverbes et de mots abstraits pour exprimer ces mêmes
rapports. Elle aurait beaucoup d'augmentatifs, de diminutifs, de mots composés, de particules
explétives pour donner de la cadence aux périodes et de la rondeur aux phrases; elle aurait
beaucoup d'irrégularités et d'anomalies; elle négligerait l'analogie grammaticale pour
s'attacher à l'euphonie, au nombre, à l'harmonie, et à la beauté des sons. Au lieu d'argumens
elle aurait des sentences; elle persuaderait sans convaincre, et peindrait sans raisonner; elle
ressemblerait à la langue chinoise à certains égards; à la grecque, à d'autres; à l'arabe, à
d'autres. Étendez ces idées dans toutes leurs branches, et vous trouverez que le Cratyle de
Platon n'est pas si ridicule qu'il parait l'être".
NATURAL GRAMMAR AND CONVENTIONAL WORDS 119
34. In this context see the now classic examination of Rousseau on language in Derrida (1967).
120 LIMITING THE ARBITRARY
that speakers and writers all across Europe did without thinking every day, and
not only lawyers, though perhaps they most especially. Certainly no one ever
overtly glorified the nominal style to the extent that Bentham did, and it may
have been with his critics that the struggle against that style began. It is still
being waged and may well go on forever, as battles against ancient, deep-
rooted assumptions tend to do. C. K. Ogden (1889-1957), author of several
books on Bentham (Ogden 1932a, 1932b, 1950), took inspiration from these
views in creating Basic English (on which see further p. 158 below and Joseph
1999a). As Mack points out, Bentham's use of 'give' and 'take' parallel pre
cisely the function of the 'operators' in Basic English.
'Reflexion ' and language structure: from Herder and Schlegel to Renan
Herder's Abhandlung über den Ursprung der Sprache won the Berlin
Academy Prize in 1770 and thus was very widely read, unlike Rousseau's Es-
sai. Early on Herder positions himself in opposition to both Condillac and
Rousseau by asserting that "the former turned animals into men and the latter
men into animals" (Herder 1772: 31).35 He rejects any attempt to draw an anal
ogy from the 'naturalness' of animal language to human language (ibid., p. 37).
Human language was invented, when man was "placed in the state of reflection
which is peculiar to him", and when the mind in this state of reflection was
"for the first time given full freedom of action" (p. 52).36 Neither this reflection
nor this freedom is shared by animals, whereas for man they are "essential to
his species; and so is language" (ibid.).37 Man does not at all 'speak by nature'
in the way that animals do because his essential nature consists in this reflec
tion and freedom. In that human sense of nature, language was "invented as
naturally and to man as necessarily as man was man" (p. 56).38 No aspect of its
invention was arbitrary — Herder considers the very idea of the arbitrariness of
language to be nonsensical (pp. 92-93).
The long tradition of philosophical reflections on the origin of language
and its relation to mind and body (of which this chapter has examined a few
high points) was about to be ruptured. What occasioned the rupture was the
popularization of the idea of a genetic relationship of Sanskrit, the sacred lan
guage of Hinduism, with Latin, Greek and most of the other languages of
Europe. The event traditionally cited as establishing this idea is the paper by
Sir William Jones (1746-1794) in 1786. Over the next thirty years Europeans
found various ways of coming to terms with the evidence that the 'mother'
35. "[...] da jener die Thier zu Menschen, und dieser die Menschen zu Thieren machte".
36. "[...] in den Zustand von Besonnenheit gesetzt, der ihm eigen ist"; "[...] zum erstenmal frei
würfend".
37. "Diese Besonnenheit ist ihm Charakteristisch eigen, und seiner Gattung wesentlich: so
auch Sprache und eigne Erfindung der Sprache".
38. "[...] eben so natürlich und dem Menschen notwendig erfunden, als der Mensch ein
Mensch war".
NATURAL GRAMMAR AND CONVENTIONAL WORDS 121
In this period too the scientific world was coming to appreciate the power
of the 'unconscious' mind more fully than the Enlightenment had done. In
deed, among the principal historical causes we could cite for the focusing of
scientific linguistics on grammar-syntax is the Varronian tradition whereby
sounds and grammatical forms, unlike words, inhabit the domain of the invol
untary, which is more or less equivalent to what in modern times would be
called the unconscious. Institutional prestige was already shifting toward the
natural sciences (see Koerner 1995), creating a motivation for linguistics to be
constituted as the study of unequivocally 'natural' phenomena. This could be
the case if it studied the products of the unconscious mind, as opposed to those
of the will. The human will, being maximally individual and minimally pre
dictable, had to be eliminated as thoroughly as possible from any object of sci
entific study.
The role of the Varronian tradition in the establishment of scientific lin
guistics would crystallize with Wilhelm von Humboldt (1767-1835), whose
work made a deep impact on the 19th-century study of language, and who is
chiefly remembered today for his posthumously published introduction to his
study of Kawi, the sacred language of Java (Humboldt 1836). Kawi showed
heavy lexical borrowing from Sanskrit, but had a grammatical-syntactic system
that was purely Malayan. To the basic question — is Kawi a Malayan or a
Sanskritic (Indo-European) language, or a combination of the two? — Hum
boldt answered that the grammatical system of a language is what determines
its historical affiliation and its 'genius'. The word stock means very little in
this regard. Hence Kawi is a Malayan language with Sanskritic borrowings, not
the other way round, nor is it a mixed Sanskritic-Malayan language.
This looks like staying true to the Varronian line in more than one way:
for like Varro, Humboldt ties the lexical-functional divide directly to the op
erations of the mind:
We have everywhere set out at first from the structure of languages alone, and in
forming a judgement about it have also confined ourselves solely to this. Now that
this structure is better in one than another, is more excellent in Sanscrit than in
Chinese, and in Greek than in Arabic, could hardly be disputed by any impartial
scholar. However we might try to weigh off their respective virtues, we should al
ways have to admit that one of these languages is animated by a more fruitful
principle of mental development than the other. (Humboldt 1988: 217 [1836: 318-
319])39
The great difference between Humboldt and Varro however is that Humboldt
39. "Überall sind wir zuerst rein von dem Baue der Sprachen ausgegangen, und zur Bildung
eines Urtheils über ihn auch nur bei ihm selbst stehen geblieben. Dass nun dieser Bau, dem
Grade nach, vorzüglicher in der einen, als in der andren, sei, im Sanskrit mehr, als im Chine
sischen, im Griechischen mehr, als im Arabischen, dürste von unparteiischen Forschern
schwerlich geläugnet werden. Wie man es auch versuchen möchte, Vorzüge gegen Vorzüge
abzuwägen, so würde man doch immer gestehen müssen, dass ein fruchtbareres Princip der
Geistesentwickelung die einen, als die anderen dieser Sprachen, beseelt".
NATURAL GRAMMAR AND CONVENTIONAL WORDS 123
is no longer thinking strictly in terms of Greek and Latin. He knows that root-
and-morphology is not the way in which all, perhaps not even most languages
instantiate the lexical-functional divide. But he considers it the natural way to
do so, for root-and-morphology recapitulates in language the actual functioning
of the human mind. Languages which depart from this do so at the cost of re
ducing their intellectual and expressive capacity. His views are complex, how
ever, for just a few pages after the preceding quote, he says the following:
On this showing, we ought at first sight to consider Chinese as departing the fur
thest from the natural demands of language, as being the most imperfect of all.
But this attitude evaporates on closer inspection. Chinese, on the contrary, has a
high degree of excellence, and exerts a powerful, albeit one-sided, influence on
the mental faculties [...]. In the first place, there is no contesting the great consis-
tency of its structure [...]. And there the very nature of the means employed in it,
for the understanding of everything formal, without the support of significant
sounds, provided an impulse towards stricter attention to the various formal rela
tionships, and a systematic method of ordering them.
[...] However paradoxical it may sound, therefore, I consider it established
nonetheless, that the seeming absence of all grammar in Chinese is precisely what
enhances, in the national mind, the acuteness of the ability to recognize the for
mal linkage of speech; whereas, on the contrary, those languages which attempt,
without success, to designate grammatical relationships, are liable, rather, to put
the mind to sleep, and to cloud the grammatical sense by a mingling of the mate
rially and formally meaningful. (Humboldt 1988: 230-231 [1836: 339-341])40
As argued in Joseph (1999b), Humboldt's resolution of this seeming paradox
involves a distinction between the realm of pure ideas, in which the unsur
passed excellence of the purely lexical Chinese resides, and the realm of hu
man thinking, where the inflecting languages excel. Scholars are still sorting
out Humboldt's views (cf. Mueller-Vollmer 1993); many of his writings are
only now being prepared for publication for the first time. Humboldt (1836)
does not explicitly say that grammar-syntax matters more than words, but fur
ther praises the inflecting languages for the perfect balance they achieve be
tween lexical designation and grammatical arrangement. Yet the most immedi-
40. "Hiernach sollte man auf den ersten Anblick die Chinesische Sprache für die von der natur-
gemässen Forderung der Sprache am meisten abweichende, für die unvollkommenste unter
allen halten. Diese Ansicht verschwindet aber vor der genaueren Betrachtung. Sie besitzt im
Gegentheil einen hohen Grad der Trefflichkeit, und übt eine, wenn gleich einseitige, doch
mächtige Einwirkung auf das geistige Vermögen aus [...]. Zuerst kann ihr die grosse Conse-
quenz ihres Baues nicht bestritten werden [...]. Dann trieb gerade die Natur der in ihr zum Ver-
ständniss alles Formalen angewandten Mittel, ohne Unterstützung bedeutsamer Laute, darauf
hin, die verschiedenen formalen Verhältnisse strenger zu beachten, und systematisch zu ord
nen.
"[....] Wie paradox es daher klingt, so halte ich es dennoch für ausgemacht, dass im Chine
sischen gerade die scheinbare Abwesenheit aller Grammatik die Schärfe des Sinnes, den for
malen Zusammenhang der Rede zu erkennen, im Geiste der Nation erhöht, da im Gegentheil
die Sprachen mit versuchter, aber nicht gelingender Bezeichnung der grammatischen Verhält
nisse den Geist vielmehr einschläfern, und den grammatischen Sinn durch Vermi-schung des
materiell und formal Bedeutsamen eher verdunkeln".
124 LIMITING THE ARBITRARY
ate and enduring legacy of his work is his classification of languages (in the
wake of the progress made by Friedrich Schlegel and his brother August
Wilhelm Schlegel [1767-1845]) into inflecting, agglutinating, and isolating
types, in which it is exclusively the grammatical mechanisms, morphology and
syntax, that matter.
The most famous linguist of mid-19th century France would turn out to
be an 'Orientalist' as well, though neither a Sinologist nor a Sanskritist, but a
specialist in Semitic languages who was at the same time a leading archaeolo
gist, historian and moral and political philosopher: Joseph-Ernest Renan
(1823-1892). His linguistic heroes were nevertheless Schlegel and Humboldt,
whom he raises above their 18th-century predecessors who were content to
philosophize without detailed philological evidence. Nevertheless Renan draws
a good deal from Herder, starting with his view that language is 'natural' to
man in a uniquely human sense of naturalness.
It was not for convenience, nor through imitation of the animals, that man chose
speech to formulate and communicate his thought, but rather because speech is
natural in him, both in terms of its organic production and its expressive value.
[...] It is thus dreaming to imagine an initial state in which man did not speak,
followed by another state where he attained the use of speech. Man is naturally
speaking, as he is naturally thinking [...]. (Renan 1858 [1848]: 90-92)41
The linguistic consequence of this 'naturalness' for Renan is that "The joining
of meaning and word is never necessary, never arbitrary; it is always moti-
vated" (ibid., p. 149; italics in original).42 In other words, the meaning of Eng
lish mutton might have been expressed by the word sheep, but this does not
mean that the connection between mutton and its meaning is arbitrary. What
ever word that meaning is expressed by, there will be a reason for the connec
tion, even if it lies in 'secret and often ungraspable analogies' (p. 148). Logi
cally, this is an unprovable assertion, actually a declaration of faith — but a
faith that would long endure in linguistics, as will be shown particularly in
Chap. 6.
Renan agrees with Humboldt that the structure of languages must have
been already fully fixed at the moment of their creation. "All these parts [of
speech] are primitive, all coexisted in the language of the patriarchs of each
race, less distinct, no doubt, but with the principle of their individuality [...].
The grammar of each race [...] was complete from the very start" (1858 [1848]:
41. "Ce n'est ni par une vue de convenance ou de commodité, ni par imitation des animaux,
que l'homme a choisi la parole pour formuler et communiquer sa pensée, mais bien parce que
la parole est chez lui naturelle, et quant à sa production organique, et quant à sa valeur expres
sive. [...] C'est donc un rêve d'imaginer un premier état où l'homme ne parla pas, suivi d'un
autre état où il conquit l'usage de la parole. L'homme est naturellement parlant, comme il est
naturellement pensant [...]".
42. "La liaison du sens et du mot n'est jamais necessaire, jamais arbitraire; toujours elle est
motivée".
NATURAL GRAMMAR AND CONVENTIONAL WORDS 125
105-106).43 Like the child that he was, primitive man created language effort
lessly (p. 98), not employing his will, but by letting language flow spontane
ously from the structure of his faculties, physical and mental. "The error of the
18th century [...] was to attribute to combination, and to a free and self-
possessed will, what was the natural product of the faculties" (pp. 92-93).44
Renan rejects Herder's view that reflection was the key to the origin of lan
guage, returning instead to something like the Epicurean idea of the language
proceeding from the body — and indeed, from the ethnic body. Like Hum
boldt, Renan believed that "The mind of each people is in the closest connec
tion with its language [...]" (p. 190).45 He makes allusion to the 'sensitivity' of
certain races, by which he means their greater reliance on the senses than the
intellect.
Thus onomatopoeia is by no means found in all languages to the same degree.
Almost exclusively dominant among the sensitive races, as among the Semites, it
appears much less in the Indo-European languages, (p. 144)46
Isn't the religious and sensitive race of the Semitic peoples painted stroke by
stroke in these totally physical languages, in which abstraction is unknown and
metaphysics impossible? (p. 190)47
Dominated by the direct sensory imitation of onomatopoeia, wholly physical
and devoid of abstraction, and therefore suited for irrational religion but not for
the rational philosophy that "metaphysics" means for Renan, the Semitic lan
guages (and by implication the Semitic peoples) are made into the direct
'other' of the Indo-European. Hebrew in particular "represents a very ancient
state of language" (p. 124), close to the primitive state, as Renan elaborates
with a quotation from Herder's Von Geist der Ebräischen Poesie {On the Spirit
of Hebrew Poetry):
Everything in it [the Hebrew language] proclaims: I live, and move, and act. The
senses and the passions, not abstract reasoners and philosophers were my creators
[...T]he language was moulded and uttered with a fuller expiration from the
lungs, with organs yet pliable and vigorous, but at the same time under a clear
and luminous heaven, with powers of vision acute, and seizing as it were upon
the very objects themselves, and almost always with some mark of emotion and
43. "Toutes ces parties [du discours] sont primitives, toutes coexistèrent dans la langue des
patriarches de chaque race, moins distinctes, sans doute, mais avec le principe de leur indi
vidualité. [...] La grammaire de chaque race [...] a été faite du premier coup".
44. "L'erreur du XVIIIe siècle [...] fut d'attribuer à la combinaison, à une volonté libre et se
possédant elle-même, ce qui était le produit naturel des facultés".
45. "L'esprit de chaque peuple et sa langue sont dans la plus étroite connexité [...]".
46. "Ainsi l'onomatopée est loin de se trouver dans toutes les langues au même degré. Presque
exclusivement dominante chez les races sensitives, comme chez les Sémites, elle apparaît
beaucoup moins dans les langues indo-européennes".
47. "La race religieuse et sensitive des peuples sémitiques ne se peint-elle pas trait pour trait
dans ces langues toutes physiques, auxquelles l'abstraction est inconnue et la métaphysique
impossible?"
126 LIMITING THE ARBITRARY
48. "Alles in ihr ruft: 'ich lebe, bewege mich, wirke. Mich erschuffen Sinne und Leidenschaf
ten, nicht abstrakte Denker und Philosophen [...D]ie Sprache ist mit voller Brust, mit noch
anausgebrauchten starken Organen, aber unter einem reinen und leichten Himmel, mit schar
fem Blick, immer gleichsam die Sache selbst erfaßend und fast nie ohne Spur der Leidenschaft
gebildet worden".
49. It was pieced together after Saussure's death by two of his colleagues, Charles Bally and
Albert Sechehaye, using scattered manuscript fragments plus the notes taken by Saussure's
students during his three courses of lectures at the University of Geneva in 1907, 1908-1909
and 1910-1911. A partial critical edition of the source materials has been published by Engler
(1967-74).
50. "Le lien unissant le signifiant au signifié est arbitraire, ou encore, puisque nous entendons
par signe le total résultant de l'association d'un signifiant à un signifié, nous pouvons dire plus
simplement: le signe linguistique est arbitraire".
51. "Le mot arbitraire appelle aussi une remarque. Il ne doit pas donner l'idée que le signifiant
dépend du libre choix du sujet parlant (on verra plus bas qu'il n'est pas au pouvoir de
l'individu de rien changer à un signe une fois établi dans un groupe linguistique); nous voulons
dire qu'il est immotivé, c'est-à-dire arbitraire par rapport au signifié, avec lequel il n'a aucune
attache naturelle dans la réalité". Although this passage has been stitched together by the edi
tors of the CLG, all its elements are multiply attested in the source materials.
NATURAL GRAMMAR AND CONVENTIONAL WORDS 127
voluntary and the determined. While Saussure sees the link between the signi
fier and signified, i.e. the sound pattern and the concept, as arbitrary, he makes
clear that because this link is arbitrary from the point of view of the speech
community as a whole, it cannot be arbitrary for the individual speaker. He
makes no distinction between lexical and functional categories in this regard.
Whatever is voluntary for the individual belongs for Saussure in the category
oí parole, the use of language. The system itself, langue, is socially determined
from top to bottom.
Thus the social and the unconscious — connected in ways that are never
made explicit — take over many functions that the natural previously filled, in
particular the imperviousness of language to individual will. 'Natural' may
disappear as a word, perhaps even as a concept insofar as lexical signs are con
cerned, but it is never structurally eliminated from early structuralist thought.
So the nature-other dyad survives the 19th century, this time as a division be
tween the arbitrariness of the sign on the one hand, and its status as a 'social
fact', an autonomous and ahistorical entity, on the other. This can be seen in
the following passage, which aims to show that change in language always oc
curs unconsciously, never as the result of a wilful decision, either by an indi
vidual or by the language community as a whole:52
If, with respect to the idea it represents, the signifier appears to be freely chosen,
then, on the contrary, with respect to the linguistic community which uses it, it is
not free, it is imposed. The social mass is not consulted, and the signifier chosen
by the language could not be replaced by another. This fact, which seems to en
velope a contradiction, could be called familiarly "the forced card." One says to
the language: "Choose!," but adds: "It'll be this sign and no other." Not only
would an individual be incapable, if he wanted to, of modifying in any way what
soever the choice which has been made, but the mass itself cannot exercise its
sovereignty over a single word; it is bound to the language just as it is. (CLG
104)53
Saussure uses the metaphor of 'the forced card', a fin-de-siècle parlour trick, to
create the illusion that the paradox of language being a system of arbitrary
signs that are nevertheless wholly determined is actually not a paradox. But the
metaphor itself is forced, the sleight-of-hand too obvious to be convincing.
Yet this is the 'hard core' of Saussure's course which had the founding
impact on 20th-century structuralism; and indeed for most of this century the
52. The source materials suggest that Saussure's actual words were even more truculent than
the Editors' version which follows.
53. "Si par rapport à l'idée qu'il représente, le signifiant apparaît comme librement choisi, en
revanche, par rapport à la communauté linguistique qui l'emploie, il n'est pas libre, il est im
posé. La masse sociale n'est point consultée, et le signifiant choisi par la langue, ne pourrait
pas être remplacé par un autre. Ce fait, qui semble envelopper une contradiction, pourrait être
appelé familièrement «la carte forcée». On dit à la langue: «Choisissez!» mais on ajoute: «Ce
sera ce signe et non un autre.» Non seulement un individu serait incapable, s'il le voulait, de
modifier en quoi que ce soit le choix qui a été fait, mais la masse elle-même ne peut exercer sa
souveraineté sur un seul mot; elle est liée à la langue telle qu'elle est".
128 LIMITING THE ARBITRARY
notion of separate minds for the lexical and functional fades into the back
ground. This is especially true of course for those theorists of language who
give up on the notion of mind entirely, but even holds for those who do not.
With the CLG, the Varronian compromise is broken in an unprecedented way.
All elements of language are subjugated to the system, which besides being
completely organized — to the point that the actual elements of the system
count for nothing, only the network of organization that holds among them —
is also subject to no force outside itself. In langue, the language system, Saus
sure posits an entity of such power, completely beyond the control of humans
individually or as a group, as to constitute a natural force of its own.
But elsewhere Saussure appears to realize, as Varro did, that what applies
to grammar and syntax may be absurd in the domain of lexicon, and vice versa.
If the last passage were limited to grammar-syntax, it might at least appear in
tuitively correct, but instead it is directed at words (or signs, which are here
inevitably exemplified with words and their meanings, never with grammatical
or syntactic elements), and again with nouns like 'horse' and 'tree' as proto
typical examples. Consequently, the statements are patently false, running di
rectly counter to everyday experience. As an example of how an individual or
the community can in fact change or modify a word, what better place to start
than with Saussure's own innovations, including diachrony and semiology, as
well as the special uses of the triad langue, parole, langage associated with his
name? Indeed artificial languages like Esperanto were much in vogue in his
day — in fact his brother René (1868-1943) was an extremely prolific writer
on the subject. A later passage of the CLG subjects them too to social determi
nation by pointing out that even though they may be created, once they are
spoken they are subject to the effects of unconscious change just like any other
language. Even so, their very ability to exist and function disproves the strong
claims in the last passage quoted above.
It is also interesting to compare that passage with the following extract
from an 1879 essay by the English historian Edward Augustus Freeman (1823—
1892):
Every word that a man speaks is the result of a real, though doubtless uncon
scious, act of his free will. We are apt to speak of gradual changes in languages,
as in institutions or anything else, as if they were the result of a physical law,
acting upon beings who had no choice in the matter. Yet every change of the kind
is simply the aggregate of various acts of the will on the part of all concerned.
Every change in speech, every introduction of a new sound or a new word, was
really the result of an act of the will of some one or other. The choice may have
been unconscious; circumstances may have been such as practically to give him
but one choice; still he did choose; he spoke in one way, when there was no
physical hinderance to his speaking in another way, when there was no physical
compulsion to speak at all [...].
The study of men's skulls then is a study which is strictly physical, a study of
facts over which the will of man has no direct control. The study of men's łan-
NATURAL GRAMMAR AND CONVENTIONAL WORDS 129
guages is strictly an historical study, a study of facts over which the will of man
has a direct control. (Freeman 1909 [1879]: 237-238).
Freeman sees the situation much as does Saussure in the parable of the forced
card — the person choosing may have "practically [...] but one choice" — yet
he draws from it the opposite conclusion: "still he did choose". One thing sepa
rating Saussure from Freeman is the changing view of the unconscious in the
thirty years between Freeman's essay and Saussure's course. In the wake of
Saussure's close contemporary Sigmund Freud (1856-1939), Freeman's 'un
conscious act of free will' seems an oxymoron, for we have become accus
tomed to associating the unconscious with the id and think of it as a barrier to
the free will as manifested in the ego. But this historical development does not
mean that Saussure had no choice in the matter. His bold assertion of arbitrari
ness ran counter to the prevailing thought of the time, so it is not unthinkable
that he might have stayed consistent with it by maintaining something like
Freeman's view on language change. Indeed, Freeman's statement brings to
mind a fact Saussure neglects to mention, but that even the most skilled magi
cian knows all too well: the person offered 'the forced card' sometimes
chooses another, maybe five times in a hundred.
Some 80 pages later, the editors of the CLG introduce a section on "Ab
solute and Relative Arbitrariness", in which Saussure returns to the orthodox
Varronian fold:
Only some signs are absolutely arbitrary; with others, a phenomenon intervenes
which permits the recognition of degrees of arbitrariness without doing away
with it: the sign can be relatively motivated.
Thus vingt 'twenty' is unmotivated, but dix-neuf 'nineteen' is less so, since it
evokes the terms which compose it and others associated with it, for instance dix
'10', neuf '9', vingt-neuf '29', dix-huit '18' [...]. Likewise for poirier 'pear tree',
which recalls the simple word poire 'pear', and whose suffix -ier brings to mind
cerisier 'cherry tree', pommier 'apple tree', etc. [...]. The English plural ships re
calls through its formation the whole series flags, birds, books etc., whereas men,
sheep recall nothing [...].
This is not the place to look for the factors conditioning the motivation in
each case; but the degree of motivation is always proportional to the ease of syn
tactic analysis [...]. (CLG 180-181) 54
The examples Saussure gives here of absolute and relative arbitrariness con-
54. "Une partie seulement des signes est absolument arbitraire; chez d'autres intervient un
phénomène qui permet de reconnaître des degrés dans l'arbitraire sans le supprimer: le signe
peut être relativement motivé.
"Ainsi vingt est immotivé, mais dix-neuf ne, l'est pas au même degré, parce qu'il évoque
les termes dont il se compose et d'autres qui lui sont associés, par exemple dix, neuf, vingt-
neuf, dix-huit [...] Il en est de même pour poirier, qui rappelle le mot simple poire et dont le
suffixe -ier fait penser à cerisier, pommier, etc. [...].
"Ce n'est pas le lieu de rechercher les facteurs qui conditionnent dans chaque cas la moti
vation; mais celle-ci est toujours d'autant plus complète que l'analyse syntagmatique est plus
aisée [...]".
130 LIMITING THE ARBITRARY
trast lexical with functional categories: lexical vingt with the syntactic joining
of dix-neuf, lexical poire with the derivational morpheme -ier, and in line with
Pinker, the 'lexical' irregular plurals with the functional regular ones. The no
tion of the functional morphemes 'recalling' the whole series of lexical signs to
which they are attached is the basis for what Saussure calls les rapports asso-
ciatifs "associative relations" (CLG 173), what would later come to be known
as "paradigmatic relations", and which he defines as les groupes formés par
association mentale "groups formed by mental association". Soon afterwards
he makes a comment recognizing that further division must be made and, it is
precisely the traditional one that he proceeds to re-establish:
In a certain sense — which must not be pushed too far, but which illustrates one
of the forms of this opposition — we could say that languages in which unmoti-
vatedness reaches its maximum are more lexicological, and those in which it
drops to the minimum, more grammatical. Not that 'lexicon' and 'arbitrariness'
on the one hand, 'grammar' and 'relative motivation' on the other, are always
synonymous; but there is a common element in the principle. These are like two
poles between which moves the entire system, two opposite currents which share
the movement of the language: the tendency to employ the lexicological instru
ment, the unmotivated sign, and the preference accorded to the grammatical in
strument, i.e. the rule of construction.
It can be seen for instance that English gives a much greater place to the un
motivated than German; but the best example of the ultra-lexicological is Chi
nese, whereas Indo-European and Sanskrit are specimens of the ultra-
grammatical. (CLG 183)55
While he would seem to agree with Humboldt as to the identity of the ultimate
'grammatical' and 'lexicological' languages, for Saussure it is the unmotivated
sign which represents the higher principle: [...] les signes entièrement arbi-
traires réalisent mieux que les autres l'idéal du procédé sémiologique ("com
pletely arbitrary signs realize better than do any others the ideal of the semi-
ological process", CLG 101). So clearly Chinese is the semiologically ideal
language — except that Saussure never says so, and in fact in the sentence pre
ceding the last quote he makes a Chinese reference as his example of natural
expressivity of signs, the direct opposite of arbitrariness:
Signs of politeness, for example, often endowed with a certain natural expressiv-
55. "En un certain sens — qu'il ne faut pas serrer de trop près, mais qui rend sensible une des
formes de cette opposition — on pourrait dire que les langues où l'immotivité atteint son
maximum sont plus lexicologues, et celles où il s'abaisse au minimum, plus grammaticales.
Non que «lexique» et «arbitraire» d'une part, «grammaire» et «motivation relative» de l'autre,
soient toujours synonymes; mais il y a quelque chose de commun dans le principe. Ce sont
comme deux pôles entre lesquels se meut tout le système, deux courants opposés qui se parta
gent le mouvement de la langue: la tendance à employer l'instrument lexicologique, le signe
immotivé, et la préférence accordée à l'instrument grammatical, c'est-à-dire à la règle de con
struction.
"On verrait par exemple que l'anglais donne une place beaucoup plus considérable à
l'immotivé que l'allemand; mais le type de l'ultra-lexicologique est le chinois, tandis que
l'indo-européen et le sanscrit sont des spécimens de l'ultra-grammatical".
NATURAL GRAMMAR AND CONVENTIONAL WORDS 131
ity (think of the Chinese who greets his emperor by bowing to the ground nine
times), are nonetheless fixed by a rule [...]. (CLG 101).56
This is curious, because he does not often give Chinese examples, and he could
perfectly well have referred to the fact that Europeans bow and curtsy to their
emperors, one time only, with the same 'natural expressivity' and 'fixed by a
rule'.
If we ask why at the start of the last quote Saussure interjects the warning
against pushing his idea too far, the obvious answer would appear to be that the
correlations lexical-arbitrary and grammatical-motivated will collapse. Yet as
we have seen, they had already held for the better part of 2000 years, and
would reassert themselves in the wake of the CLG. What in fact would not
hold if we pushed these correlations too far is the Saussurean system, that Le-
viathan clutching lexicon in one hand and grammar in the other. The CLG will
not let the system go, but does set it in motion, shuttling between the twin
'poles' of lexicon-arbitrariness and grammar-motivation. On what sort of ex-
istential plane these poles exist the CLG does not say: it is waxing mystical
again, drawing rhetorical power this time from the image of an electrical side-
show marvel shuttling visible currents between its poles.
On the other hand, if we did push the correlations to the extreme, it is
hard to see how we might avoid cornering ourselves into saying with Hum
boldt that Sanskrit and Chinese form the basis for two different mental struc
tures, and that the Chinese structure is the ideal one. This suggests a paradox in
Saussure's thought: for it means that languages differ from one another in their
internal systematicity, contrary to Saussure's own assertion that in every lan
guage tout se tient, everything is connected to everything else. If there is in
deed this hierarchy from the ultra-grammatical Sanskrit to German to English
to the ultra-lexicological Chinese, the network of associative relations within
each language progresses from thick to thin along the hierarchy. If tout se tient
in Chinese, it does not do so in the same way as in Sanskrit, and specifically it
does so without 'groups formed by mental association'. Presumably the ele
ments of Chinese hang together more purely by the kind of difference which
establishes value in the Saussurean system — again suggesting that it is the
ideal language from a Saussurean point of view, and Sanskrit the direct oppo
site of that ideal, a position we never find the Professor of Indo-European Lan
guages and Sanskrit at the University of Geneva actively taking.57
56. "Les signes de politesse, par exemple, doués souvent d'une certaine expressivité naturelle
(qu'on pense au Chinois qui salue son empereur en se prosternant neuf fois jusqu'à terre), n'en
sont pas moins fixés par une règle [...]."
57. We do find it being put forward a decade and a half after Saussure's lectures by Edward
Sapir (1884-1939), writing on the ideal form for an international auxiliary language: "[T]he
ideal of effective simplicity is attained by a completely analytic language, one in which the
whole machinery of formal grammar is reduced to carefully defined word order and to the op
tional use of 'empty' independent words (like 'several', 'did', 'of'). Inflection is reduced to
132 LIMITING THE ARBITRARY
Chomsky's fix
From the 1920s, both European and American structuralism drop Saus
sure's concerns with sign theory and words and focus almost entirely on pho
nology-grammar-syntax as system, fixed either socially or, with Noam Chom
sky (b. 1928), in the physical make-up of the brain, hence genetically —
bringing us full circle back to 'nature' in the truest sense.
To be more precise, structuralist linguistics at mid-century was very pho
nology-centred, and was only just getting around to taking morphology seri
ously. One way of looking at Chomsky's early impact is to say that he rejected
the morphological approach to the functional side of the lexical-functional di
vide. In Leonard Bloomfield's (1887-1949) later work it was no longer neces
sary, as in his magnum opus of 1933, to deal only with what was directly ob
servable in language. Phonemes, after all, were only indirectly observable, and
by the time of his "Menomini Morphophonemics" (1939) Bloomfield had
come a long way from his earlier behaviourist rhetoric. By the early 1950s
Charles Hockett (b. 1916), Zellig Harris (1909-1992) and Roman Jakobson
(1896-1982) were all involved in rather abstract analyses of morphology and
syntax; but it was Chomsky who took that a step further to say in effect that the
functional side of language is largely 'invisible' — that its essence will be
found not in morphemes, but in their syntactic arrangement. Since then, the
lexical-functional divide has come step-by-step to coincide with a division
between what is directly observable and only indirectly observable in speakers'
knowledge of language, though overlaid by something like the more traditional
division between words or morphemes with identifiable referents and those
which serve only to establish the relations among those words.
In later years, however, something like the older conception of morphol
ogy kept rearing its head. It began with Charles Fillmore's (b. 1929) case
grammar in the late 1960s, which Chomsky did not immediately take on board.
But piece by piece, his syntactic engine acquired thematic roles, case theory,
even Tense and Agreement as universal functional heads, with in the end the
sentence itself being identified as AgrP, the agreement phrase, a projection of
what corresponds to part of the inflection in inflectional languages — but
which, being part of Universal Grammar, has to be posited as part of the core
linguistic knowledge with which speakers of any human language are born.
During the 1970s, generativists shifted focus from the grammars of indi
vidual languages to the principles and parameters which constitute innate Uni
versal Grammar. In order to deal with the undeniable fact that all languages
contain clearly idiosyncratic (read: arbitrary) features, Chomsky posited a di-
zero. This is the ideal that English has been slowly evolving toward for centuries and that Chi
nese attained many centuries ago after passing through a more synthetic prehistoric phase"
(Sapir 1925: 248).
NATURAL GRAMMAR AND CONVENTIONAL WORDS 133
chotomy between 'core grammar' and 'periphery', equating this with the 'un
marked' and 'marked' categories of Jakobson (see further Chap. 6). In addi
tion, Chomsky proposed a view still prominent in the generative primers of the
1990s, namely that all language-specific features are part of lexicon, while the
'configurational system' — the principles and parameters of syntax — is liter-
ally universal, being physically innate. By this view, writes Chomsky "[...]
there is only one human language [...] and language acquisition is in essence a
matter of determining lexical idiosyncracies" (Chomsky 1991: 419). That is,
everything that appears to differentiate one language from another resides
within the words of those languages, as lexical specifications, including lan
guage-specific features of grammar and syntax. The real grammar and syntax
are hard-wired into the brain, and since brains do not differ, neither can the
grammar that is grounded in them.
Ever since the Cratylus, the strongest, most obvious argument against the
naturalness of language has been the existence of different languages. Like
Saussure, Chomsky does not usually discourse on the 'natural'; but natural and
innate share more than a historical root, and the lexicalist move gains intuitive
support from its covert alignment with a set of assumptions as old as history.
Now lexicon is the repository of everything arbitrary in language, meaning
everything that might historically have been something other than it is, includ
ing rules of syntax. And 'grammar' proper, the Universal Grammar that is the
essence, the core of every actual grammar, is natural in the strongest and most
literal of senses, being innate, physical, and wholly determined.
We can identify as 'Humboldt's fix' the following paradox: the inner
form of language determines a nation's mental power; the cultural achieve
ments of the Indo-European nations suggest that the inflecting languages pro
duce incomparable mental power, because they directly reproduce the mecha
nism of the human mind; but the language most unlike Sanskrit, Chinese,
belongs not to a primitive culture but to one of the first rank. In contrast,
'Saussure's fix' stems from his nexus of assertions that every language consti
tutes a completely internally-organized system; that the purely arbitrary sign is
superior to the motivated one; that the arbitrary equates largely with the lexical
and the motivated with the grammatical; and that the grammatical and moti
vated are what create associative relations. From this it would appear to follow
that languages like Chinese and Sanskrit reflect wholly different mental or
ganizations, and in particular that Chinese lacks full systematicity, which con
tradicts Saussure's description of langue as a universal category. Yet at the
same time Chinese is the language which best realizes Saussure's ideal of the
58. Chomsky has written repeatedly of a 'language organ', making clear that he means it liter-
ally rather than metaphorically — a classic example of a metaphor run amuck. On the role of
metaphors in linguistics, see Joseph (1989a). Botha (1989) is a lively and unsparing critique of
Chomsky's views on language.
134 LIMITING THE ARBITRARY
arbitrary sign. 'Chomsky's fix' is that Universal Grammar, which accounts for
the simplicity of language structure that explains the efficiency of children's
language acquisition, must simultaneously account for the overwhelming di
versity of language structures encountered around the world. Universal Gram
mar has to account for Sanskrit and Chinese simultaneously, and it cannot
project anything into the brain of the Sanskrit speaker which is not also in the
brain of the Chinese speaker and vice-versa. In this respect Chomsky's univer
salist project is the opposite of Humboldt's, despite the affinity to Humboldt
claimed by Chomsky in the 1960s (see Joseph 1990b).
Something else Chomsky has been saying about his project since the
1960s is that its universalist assumptions are actually already implicit in any
linguistics which claims to take all the world's languages potentially within its
scope, including traditional Western grammar. That is, any approach to lan
guages which assumes that they have the same basic structure of sound, order
and meaning, used to perform essentially similar functions, and describable
starting from the same basic group of descriptors like noun, verb and so on, is
already implicitly committed to all that Chomsky is claiming a priori. I think
that he is right, and what is more, I think he is right in the way I believe he in
tends the claim, as a downplaying of his own project's aims.
Part of what is interesting about the Minimalist Program is that it is trying
to find a way out of Chomsky's fix, and has pinpointed morphology as the es
cape route. In the Principles-and-Parameters theory, inflectional morphology is
'part of syntax proper' (unlike derivational morphology, which is 'part of the
lexicon', Chomsky 1995: 133). The sentence John likes Mary has the follow
ing D-structure (formerly 'deep structure'):
How -s gets attached to like is a controversial question. The view which was
standard up to the late 1980s was that a movement rule called 'affix hopping'
then applies, whereby the I(nflection) or Agr(eement) morpheme -s lowers to
V. At least that is the case for languages like English. For other languages, in
cluding French, the opposite kind of movement occurs, verb-raising, so that the
V like moves up into I. This is an example of a 'parametric' choice, like a
NATURAL GRAMMAR AND CONVENTIONAL WORDS 135
switch in our innate Universal Grammar which has to be set one way or the
other by our experience. This account of verb inflection was thrown into ques
tion by Pollock (1989) and remained unresolved as the Minimalist Program (or
MP) refocused attention onto other questions (see Chomsky 1995: 133-145). In
MP, or at least in one relatively simple version of it, the above sentence would
have the following base-derivation:
affixes are attached to items of the lexical categories in the lexicon. The functional
nodes of AGR and T serve only to carry the morphological (inflectional) features
necessary to check off the features on Ns and Vs. (363)
The syntactic engine itself — the autonomous principles of composition and ma
nipulation Chomsky now labels 'the computational system' — has begun to fade
into the background. Syntax reduces to a simple description of how constituents
drawn from the lexicon can be combined and how movement is possible [...]. A
vision of the end of syntax — the end of the sub-field of linguistics that takes the
computational system, between the interfaces, as its primary object of study —
this vision encompasses the completion rather than the disappearance of syntax.
(380-381)
Chomsky has confirmed to me in correspondence that as far as he is concerned
a number of different possible accounts appear potentially valid, and that it is
not yet clear to him what are the empirical issues involved. But the starting
point seems clear enough: "parametric differences must be reduced to mor
phological properties" (Chomsky 1995: 192). Chomsky here takes his inspira
tion from a slightly younger contemporary of Saussure's who was directly in
volved with Sapir (see n. 57) in the international auxiliary language movement,
namely Otto Jespersen (1860-1943). Chomsky (1995) quotes repeatedly from
this passage in Jespersen's Philosophy of Grammar (1924):
We are now in a position to return to the problem of the possibility of a Universal
Grammar. No one ever dreamed of a universal morphology, for it is clear that all
actually found formatives, as well as their functions and importance, vary from
language to language to such an extent that everything about them must be re
served for special grammars, with the possible exception of a few generalities on
the role of sentence-stress and intonation. It is only with regard to syntax that
people have been inclined to think that there must be something in common to all
human speech, something immediately based on the nature of human thought, in
other words on logic [...]. (Jespersen 1924: 52)
And yet if the most recent version of Chomsky's theory is right, this Universal
Grammar restricted to syntax and immediately connected to mind is not only
minimal in size and scope, but its operation is not obviously linguistic (in
Chomsky's very restricted sense of 'linguistic'), consisting as it does of de
vices like feature-checking which could find obvious analogues in general in
telligence much more readily than language-specific features — i.e., morphol
ogy — might. Moreover, it means that nearly all of syntax is learned, which
once upon a time was the behaviourist position so thoroughly demolished by
Chomsky that no one any longer remembers the need to avoid moving toward
it.
In the 1960s Chomsky claimed to be the restorer of the great tradition of
linguistics with which his structuralist predecessors had broken faith.59 In the
59. See especially Chomsky (1966). On the evolution of Chomsky's views regarding his intel-
lectual pedigree, see Joseph (1990b) and (1999c). None of the positions discussed in this nec-
NATURAL GRAMMAR AND CONVENTIONAL WORDS 137
Pinker's fix
Again, Tinker's fix' is that if regular and irregular forms are processed
by different mental modules or physical parts of the brain, those modules or
parts, hence the linguistic mind/brain itself, are differently developed in speak
ers of languages which do not have regular and irregular forms. The result of
an attempt to move from Chomskyan nativism toward a common ground with,
if not behaviourism, at least developmentalism, Pinker's fix actually has more
in common with Humboldt's than with Chomsky's fix.
Pinker's approach has come as part of a new or renewed importance be
ing accorded over the last decade to the so-called lexical-functional split within
syntax, semantics, psycholinguistics and applied linguistics, sub-fields which
do not otherwise see themselves as having a great deal in common. It suggests
that languages are not made up of blocks of sound-meaning correspondence
(signs or morphemes) which have the same status as far as the human mind are
concerned. Rather there are two qualitatively different types of units which we
learn, store and process in wholly different ways, perhaps using different mod
ules of the mind/brain.
From experimental data of various sorts which show different results for
the time taken to process past tenses like sang on the one hand and patted on
the other, Pinker concludes that English irregular verbs are handled by the
lexical module, which includes each irregular verb form as a separate lexical
entry. Regular verbs on the other hand are handled by a syntactic process
which suffixes a functional unit, realized phonetically as -ed, onto the lexical
unit, the verb stem. Thus:
Linguistic research suggests an information flow of lexicon to derivational mor
phology (complex word-formation) to regular inflection, with regular and irregular
processes encapsulated within different subcomponents. If irregular past-tense
forms are stored in memory as entries in the mental lexicon, then like other stored
words they should be the input to rules of complex word formation. If regular
essarily rapid historical survey were the only position held in their time. Some were not even
the dominant position. Various forms of generativism continue to define the mainstream of
linguistics, especially in North America; yet when one looks at contemporary approaches to
linguistics, whether their adherents see themselves as Chomskyans or in opposition to Chom
sky, one does not find much variation in the level of their commitment to some version of lin
guistic naturalism.
138 LIMITING THE ARBITRARY
past-tense forms are computed from words by a rule acting as a default, they
should be formed from the outputs of complex word formation rules. Two phe
nomena illustrate this organization [...]. (Pinker 1991: 533)
As many others have done before him, Pinker has in effect taken a particular
way of analyzing English words — in this case perhaps the most neat and effi
cient analysis for English verbs — and has projected this analysis onto or into
the human mind. He has done so with the support of experimental data which
suggest that speakers of English really do perform an analysis of this sort with
out realizing it when processing these words. But this is the whole extent of the
evidence: nothing else in direct or indirect observation of the brain and its
workings and pathology suggests the presence of a separate 'module' for lan
guage, let alone of separate functional and lexical modules. Even the mental
imaging data referred to earlier remain far from conclusive. That does not
mean that the modules might not be there, or that we might not do best to pro
ceed as if they were there.
Pinker's elegant analysis does however give rise to a nasty problem
which will have to be explained away. If regular and irregular verbs in English
and other languages which are like English in this respect are to be tied to
separate mental modules, what about those languages which do not have regu
lar and irregular verbs? Chinese, for instance, has no verbs which could be dis
tinguished from one another as 'regular' or 'irregular'; there are no 'verb
forms' as such, since person, number, tense, voice, mood, aspect and so on are
conveyed not by adding or changing something in the verbal word itself but by
adding one or more additional words. In a sense, Chinese verbs are all 'irregu
lar' from the point of view of the Indo-European languages. Arabic words, on
the other hand, do undergo morphological changes that are extremely complex,
making them too look irregular to a European linguist, though in a different
sense.
Within the Indo-European family the Germanic languages stand out as
having rather a large contingent of irregular verbs in the 'strong' verbs, which
do not follow general rules in their past and participial forms. Even the most
'irregular' Latin or French verbs, like esse or être "to be", still show quite a lot
of inflectional regularity throughout their paradigms when compared with their
English or German counterparts.60
60. To be precise, a regular English past tense verb like patted consists of a lexical, substantive
unit pat and a functional unit which is not -(t)ed as such but an abstract element we can call
Tense, and which at a late stage of derivation will be realized phonetically as -(t)ed. In English
and other Indo-European languages, functional units which have a phonetic realization may
appear as bound morphemes or separate words. In Chinese, functional units can only be real
ized as separate words from the Chinese point of view, where word units are measured by
written characters, just as for English word units are measured by the spaces between written
letters. Chinese functional words include noun classifiers, sentence-final particles, a comple
tive marker, a possessive marker and the like. The fact that, for speakers of both these lan
guages, the perception of what constitutes a word is dependent upon writing, gets ignored in
NATURAL GRAMMAR AND CONVENTIONAL WORDS 139
holds. Rather, it has been absorbed into some of the most solidly established
tenets of present-day theory, with other effects being discernible in the struc
ture and use of languages themselves.
It is only by acknowledging our engagement in this age-old debate that
we can genuinely begin to pass beyond it. Until then we, like so many of our
predecessors, are apt to make little real progress in understanding the most ba
sic questions about language — only the rhetorical veneer of progress, as we
trip into the old intellectual ruts while our attention is distracted by the gleam
of the latest models and metaphors.
It is not enough simply to declare the debate irrelevant, or to claim that
nature is itself a conventional construct, or to say that language operates partly
through natural and partly through conventional means and leave it at that.
Until we have wrestled with this debate in all its historical manifestations and
metamorphoses, any position we take is liable to recapitulate some earlier ver
sion of the debate. To get beyond it will require a major conceptual leap, and
as any broad-jumper knows, preparing for a leap requires taking a few steps
backwards.
I hope that readers will take from this chapter a number of basic ques
tions to bear in mind not just in reading the rest of the book, but in coming to
grips, either sympathetically or critically, with any theory of language that re
lies on a conception of linguistic naturalness and its stated or implied opposite:
Can any version of naturalness itself be 'natural', or must it be known to
us always and only as a historical product?
For any particular conception of the natural, is it based on criteria that are
directly observable? If not, how can it be verified, or falsified?
If some part of language is to be reckoned as somehow less natural than
another, how can it have come into being except through processes that them
selves ultimately depend upon natural faculties?
Given that the languages and dialects of the world vary to the extent that
for any structure we postulate for a particular dialect, we cannot rule out the
possibility of finding its exact opposite in some other dialect, how can any
'natural' structure be postulated without ultimately leading to the implication
that some dialects are less natural than others?
CHAPTER 5
concerned, he argues for distinct analytical principles, not only for the 'nature of
words' and their 'actual use in speaking' (something like Saussure's langue-
parole distinction), but with a third principle applying to 'poets'.
That Analogía or Regularity which is directed toward the nature of the words is not
to be defined in the same way as that which is directed toward the actual use in
speaking. For the former should be defined thus: Analogía is the like inflection of like
words; and the latter thus: Analogía is the like inflection of like words, not inconsis
tent with common usage. But when to the end of these two there has been added
"within a certain range", then poetic Analogía will be defined. The first of these is
that which the people ought to follow; the second is that which all the individuals in
the people ought to follow; and the third is that which the poets ought to follow. {De
lingua latina 10.74)
The last statement accords intuitively with the readily observable fact that literary
and poetical language tends to preserve irregularities and archaic forms. But
analogy-based Alexandrian grammar originally had mainly philological applica
tions, which is to say that it was aimed at establishing the basis for a 'scientific'
interpretation of literary texts. In that light, Varro's call to let poets be the most
free from the principle of analogy is a striking departure from the Alexandrian
project. What Varro establishes is, in fact, a hierarchization of will in the linguistic
domain. Poets should have the most freedom. Other individuals in ordinary speech
should follow 'common usage', which it is assumed will usually follow the anal
ogy principle, but is not obliged to do so. The people as a whole should follow the
principle of analogy. But who is the people, as distinct from the individuals within
the people? The answer to this question has both a metaphysical and a political
dimension. The metaphysics of 'the people' is bound up with the discussion of
'the nature of the words' as distinct from anyone's use of them. Each is an ideali
sation, useful and seemingly harmless enough on its own, though with dangerous
potential.
One way of understanding this passage is to take Varro's concern as being
with establishing separate domains of application for grammar (the nature of the
words) and rhetoric (actual use in speaking). He says that grammar need not take
account of rhetoric, whereas rhetoric does need to take account of grammar but is
not totally dominated by it, since it is also subject to customary usage. Moreover,
the rhetorical practice of poets is subject to customary usage only to a certain
extent, which means that some corners of the province of rhetoric are genuinely
autonomous. What I have called grammar, and what Varro calls the nature of
words, the domain of pure analogy, should (he says) be followed by the people:
1. "Analogia non item ea definienda quae derigitur ad naturam verborum atque illa quae ad usum
loquendi. Nam prior definienda sic: analogia est verborum similium declinatio similis, posterior
sic: analogia est verborum similium declinatio similis non repugnante consuetudine comuni. At
quom harum duarum ad extremum additum erit hoc 'ex quadam parte', poetica analogia erit
definita. Harum primam sequi debet populus, secundam omnes singuli e populo, tertiam poetae".
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 143
the nation as a whole should speak in accordance with the logos-informed nature
of words. What I have called rhetoric, and what he calls use in speaking, the
domain of analogy modified by common usage, should be followed by individuals
when it follows common usage. In this case poets too should follow it, though
only to a limited extent. The analogy with law is apt: the laws of the nation should
follow natural law; the individual should follow the laws of the nation, and should
not violate them on grounds of appeal to a higher natural law; only jurists (here
the equivalent of poets) have this privilege, and only to a limited extent.
Plato's solution to the problem of linguistic correctness relied in part on re
jecting the will of the demos, the people, collectively or (especially) individually,
as the basis for evaluating language. Rather, the will of the informed leader must
be the basis of nomos, just because he will know best how to make nomos accord
with physis. Varro, on the other hand, establishes anomaly as a place for individ
ual will in the form of 'voluntary' derivation, and does not reject it categorically,
but suggests that its products be evaluated in accordance with the customary usage
of the people, just as this customary usage should itself be evaluated in accordance
with nature. This solution is ultimately no more relativistic than Plato's, but
potentially more democratic. Along with most other aspects of Varronian linguis
tic thought, this one would have a long and fruitful life: the struggle between
'logic' and 'usage' is a constant refrain through the history of Western linguistic
traditions up to the present day.
'Secondary' appears at first glance to have simply the temporal meaning that this
type of speech is acquired second. But Dante then states that the classical standard
is also second in nobility to the vernacular:
Of these two the nobler is the vernacular, because it was the first used by the human
race; because the whole world uses it, even if it is divided into different words and
utterances; and because it is natural to us, whereas the other is artificial. (Ibid.)
Latin is the language of the Church, a sacred language, and it would seem to
border on heresy to suggest that the vernacular is nobler. But Dante makes his
appeal to the 'natural' as opposed to the 'artificial', what is made by art. Artful
ness is usually a positive quality in this period. Yet art is human, after all, while
nature is divine.
The 'artificial' — the product of human will, the historical element of lan
guage — is marginalized in favour of a 'natural' language that is reckoned to be
above the human will and therefore more 'real' than us. I am using 'historical'
here in the sense of having human beings as active subjects. One of Dante's
arguments for the greater nobility of the vernacular is that it "was the first used by
the human race", which acknowledges the value of historical priority in the purely
temporal sense. But it is merely 'used' (usitatd) by people, and is not itself the
product of history, understood as the intervention of wilful individuals, capricious
and arbitrary and politically motivated as they may be. The vernacular is more
naϊve — yet another word that shares the root of 'natural'. Dante's view of lan
guage, or at least his rhetoric, is an inherently naturalistic one, that recognizes the
co-existence of natural and artificial language types but strongly values the former
over the latter.
It is surprising that he speaks of one vernacular used by all though divided
into different words and utterances, rather than of diverse vernaculars. The diver
sity of vernacular dialects was the main practical motive for keeping Latin as the
language of the Italian peninsula and Europe generally. Dante ignores this and
says that these dialects are actually one. Yet the problem remains that vernacular
speakers cannot always understand one another. The reason for this is historical.
Separated over long centuries in distinct communities, the descendants of those
who spoke the 'original' vernacular have grown apart linguistically. What Dante
will propose, in De vulgari eloquentia 1.16, is a programme for undoing this
history so as to solve the problem of the localness of vernaculars.
Dante surveys the various Italian dialects to determine which of them is best
suited to serve as the volgare illustre, the vernacular that is both illuminated and
illuminating and that will be the best possible vehicle for poetry in a pan-Italian
context. His verdict is that none of the actually existing dialects is suited to this
end. Instead, the volgare illustre is an ideal language that will have to be found not
4. "Harum quoque duarum nobilior est vulgaris: turn quia prima fuit humano generi usitata; turn
quia totus orbis ipsa perfruitur, licet in diversas prolationes et vocabula sit divisa; tum qui naturalis
est nobis, cum illa potius artificialis existat".
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 145
5. "Postquam venati saltus et pascua sumus Ytalie nec panteram quam sequimur adin-venimus, ut
ipsam reperire possimus, rationabilius investigemus de illa, ut solerti studio redolentem ubique et
necubi apparentem nostris penitus irretiamus tenticulis".
6. "Resumentes igitur venabula nostra, dicimus quod in omni rerum genere unum oportet esse quo
generis illius omnia comparentur et ponderentur et quod velut aliorum omnium mensuram accipia-
mus; sicut in numero cuneta mensurantur uno, et plura vel pauciora dicuntur secundum quod
distant ab uno vel ei propinquant; et sicut in coloribus omnes albo mensurantur; nam visibiles
magis dicuntur et minus, secundum quod accedunt vel recedunt ab albo".
7. "Et quemadmodum de hiis dicimus que quantitatem ostendunt, de predicamentorum quolibet,
et etiam de substantia posse dici putamus; scilicet, unumquodque mensurabile fit secundum quod
in genere est, illo quod simplicissimum est in ipso genere. Quapropter in actionibus nostris,
146 LIMITING THE ARBITRARY
quantumcunque dividantur in species, hoc signum inveniri oportet quo et ipse mensurentur".
8. "[... I]n quantum ut homines latini agimus, quedam habemus simplicissima signa, et morum et
habituum et locutionis, quibus latine actiones ponderantur et mensurantur".
9. "Que quidem nobilissima sunt earum que Latinorum sunt actiones, hec nullius civitatis Ytalie
propria sunt et in omnibus comunia sunt: inter que nunc potest illud discerni vulgare quod superius
venabamur, quod in qualibet redolet civitate nec cubat in ulla [...]".
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 147
scription: gramatica, Latin, but it is ruled out by definition. It is not noble enough,
because although common to all the cities of Italy it is not common to all the
people. We want something common to all the people but specific to none of the
towns; what all of them do, yet what none of them does. Something that cannot
be heard anywhere in Italy, only felt or sniffed, like some kind of mystical fart.
To the modern reader this all seems a fiction, a pretence of discovery in what
will actually be Dante's invention of an illustrious vernacular — which will in
turn camouflage how much of it is actually based on his native Tuscan. But if
invented it would have none of the features demanded by Dante, being neither
original nor common nor natural, nor enjoying the nobility these features confer.
On what grounds then could it possibly be preferable to Latin?
He proceeds as though toward the discovery of a natural element, which he
will then use for his own art, never acknowledging that the element itself might
in any way be the product of art. Whereas gramatica is artificial because it is the
product of human history, the volgare illustre is the product of anti-history. For
what is common to all the people of Italy yet what none of them do is their past,
what they were when they were one. Of course they were one in the time when
Latin was formed, but that oneness also included what would become the Spanish,
the French, the Occitans and so on. Dante's panther is found by reversing history
just far enough to reach a specifically Italian oneness. History is what has undone
the common Italian language, and the volgare illustre will be found precisely by
taking away what history has added to each local dialect as a superfluous defor
mation. For Dante, the problem of history could only be compounded, not solved,
by the use of a 'gramatica' which was itself a historical product — historical in the
worst sense because artificial, a wilful distortion of nature, a sin of commission.
The historical divergence of dialects is a sin of omission, the passive distortion of
nature by failing to abide by the elemental signs.
Dante's illustrious vernacular is anti-historical in its opposition both to
dialectal diversity and to the classical standard language. It aims instead to estab
lish an alternative history that is, inevitably, deeply mythical, creating a pan-
national unity under the pretence of rediscovering and restoring it. This became
the template upon which other modern European standard languages were mod
elled.
writing in 6th and 5th-century Greece. In the debate over the use of classical and
vernacular languages, language theory and linguistic ideology came head to head
with another force, the economic interests of publishers; and the simultaneous
spread of Protestantism may well have played a role in loosening any subliminal
hold that 'sacred' Latin had on the minds of the intellectual elite in many parts of
Europe.
This was also the beginning of a new era of imperialism, with the discovery
of the New World and the beginning of African and Asian colonization, and just
as the focus of linguistic interest had changed with the founding of Alexander's
empire, and again with that of the Romans, Europeans now found themselves in
the position of thinking about their native vernaculars as tools for the management
of empires. In the Preface, addressed to Queen Isabella, of his Gramática castel-
lana (1492), the first important grammar of a modern European language, Antonio
de Nebrija (1444?-1522) wrote:
[] I have resolved before all else to reduce our Castilian language to artifice, so that
that which is written in it now and in the future can follow a standard, and be ex
tended for all time to come, as we see has been done in the Greek and Latin language,
which, on account of having been subjected to art, remain in uniformity even though
they have passed through many centuries. (Nebrija 1946 [1492]: 9)
Both 'reduce to artifice' (reduzir en artificio) and 'subjected to art' (debaxo de
arte) mean precisely to submit language to what the Greeks called tekhnē, so as
to give it the stability over time which 'natural' idioms do not have, but also to
instill order, both in the structure of the language itself and in the minds of those
who use it.
This same period saw the first large-scale spread of printed Greek philo
sophical texts (in Latin translation), and increasing allusions to classical themes
begin appearing in the works of every side of the classical vs. vernacular language
question. Typical of the period is the Dejfence et illustration de la langue fran-
çoyse (1549), written by the poet and scholar Joachim Du Bellay (1522-1560)
with the intention of proving that French was as capable and worthy of use in both
literary and scientific writing as were Latin and Greek. The work opens by posing
the problem in terms that invoke both nature and the human will:
If nature (about which a certain person of great renown [Pliny the Elder] has rightly
wondered whether she should be called mother or wicked stepmother) had given men
a common will and agreement, then, besides the innumerable benefits that would have
followed, human inconstancy would not have had to create so many ways of speaking.
This diversity and confusion is justly called the Tower of Babel. Thus languages were
not born on their own in the manner of plants, roots, and trees: some weak and sickly
10. "[A]corde ante todas las otras cosas reduzir en artificio este nuestro lenguaje castellano, para
que lo que agora i de aqui adelante en el se escriviere pueda quedar en un tenor, i estenderse en
toda la duracion delos tiempos que estan por venir, como vemos que se a hecho enla lengua griega
i latina, las cuales, por aver estado debaxo de arte, aunque sobre ellas an passado muchos siglos,
toda via quedan en una uniformidad".
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 149
in appearance, others healthy, robust, and better able to bear the weight of human
thoughts; but all their qualities were born in the world of the will and volition of
mortals.This seems to me an important reason for not praising one language and criti
cizing another, since all come from the same source and origin, the caprice of men;
and all have been formed by the same reasoning faculty toward the same end, so that
we might signify to one another the conceptions and understandings of the mind. It
is true that over time, some languages, on account of having been more carefully
regulated, have become richer than others, but this must be attributed not to any in
trinsic quality of the languages, but only to the craft and industry of men. (Du Bellay
1914 [1549]: 41; my translation: JEJ)11
The passage is notorious for the apparent lack of logical connections from one
sentence to the next, and most of its arguments had been anticipated in
15th-century Italy by Sperone Speroni (1500-1588), and by earlier 16th-century
French writers like Geoffroy Tory (1480?-1533) in Champ fleury (1529). But this
did not prevent Du Bellay's Deffence et illustration from having an enormous
impact in its time, and it remains to this day a staple of the French educational
canon. It is as interesting to note what Du Bellay takes for granted as what he
asserts: that languages are the product of human will; that nature ultimately cre
ated the will, and that she is an ambiguously good and evil force; that the multi
plicity of human wills is the cause of the existence of different languages. Note
that he does not feel obliged to discuss the Biblical account of the divine origin
of language, other than inserting a passing reference to the Tower of Babel, a
reference that does so much to undermine the logical flow of the paragraph that
one suspects it may have been inserted as a politically prudent afterthought. To
underline the non-naturalness of languages, Du Bellay contrasts them with an
incontestable product of nature, plants. And yet, just two chapters later, he is
already using plants as a metaphor for language:
Perhaps the day will come — and I hope it will, along with a happy destiny for France
— when this noble and powerful Realm will in its turn take the reins of world domi
nance, and when our language (if it has not been entirely buried along with François
I [d.1547]), which is still just beginning to sprout roots, will burst forth from the
ground and rise up to a height and size to rival the Greeks and Romans themselves
11. "Si la nature (dont quelque personnage de grand'renommée non sans raison a douté, si on la
devait appeller mere ou marastre) eust donné aux hommes un commun vouloir et consentement,
outre les innumerables commoditez qui en fussent procedées, l'inconstance humaine n'eust eu
besoin de se forger tant de manieres de parler. Laquelle diversité et confusion se peut à bon droit
appeller la tour de Babel. Doncques les langues ne sont nées d'elles mesmes en façon d'herbes,
racines et arbres, les unes infirmes et débiles en leurs especes, les autres saines et robustes, et plus
aptes à porter le fais des conceptions humaines: mais toute leur vertu est née au monde du vouloir
et arbitre des mortels. Cela (ce me semble) est une grande raison pourquoy on ne doit ainsi louer
une langue et blasmer l'autre, veu qu'elles viennent toutes d'une mesme source et origine, c'est
la fantasie des hommes, et ont esté formées d'un mesme jugement, à une mesme fin: c'est pour
signifier entre nous les conceptions et intelligences de l'esprit. Il est vray que par succession de
temps, les unes, pour avoir esté plus curieusement reiglées, sont devenues plus riches que les
autres; mais cela ne se doit attribuer à la felicité desdites langues, ains au seul artifice et industrie
des hommes".
150 LIMITING THE ARBITRARY
12. "Le temps viendra (peut estre) et je l'espère moyennant la bonne destinée françoise, que ce
noble et puissant royaume obtiendra à son tour les resnes de la monarchie, et que nostre langue (si
avecques François n'est du tout ensevelie la langue françoise) qui commence encore à jetter ses
racines, sortira de terre, et s'eslevera en telle hauteur et grosseur, qu'elle se pourra egaler aux
mesmes Grecs et Romains [...]".
13. "[N]otre langue françoise n'est si pauvre qu'elle ne puisse rendre fidelement ce qu'elle em
prunte des autres; si infertile qu'elle ne puisse produire de soy quelque fruict de bonne invention,
au moyen de l'industrie et diligence des cultivateurs d'icelle, si quelques uns se trouvent tant amis
de leur pays et d'eux mesmes, qu'ils s'y veuillent employer".
14. "[C]hacune langue a je ne scay quoi propre seulement à elle, dont si vous efforcez exprimer
le naïf dans une autre langue, observant la loi de traduire, qui est n'espacier point hors des limites
de l'auteur, votre diction sera contrainte, froide et de mauvaise grace".
15. "[E]t ne les [fidèles traducteurs] doit retarder, s'ils rencontrent quelquefois des mots qui ne
peuvent estre receus en la famille françoise, veu que les Latins ne se sont point efforcez de traduire
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 151
Thus the language and the culture are like republics, populated by words in the
one case and ideas in the other. Not every foreign element entering the republic
will be granted citizenship, of course, but those offering some substantial benefit
to the republic will be welcomed, and will, like transplanted seeds, thrive on
French soil. Here again, political and natural images and parallels intertwine, in
virtually dialectical fashion. Note, however, that the political metaphor is exempli
fied entirely in the form of nouns (rhetoric, music, etc.) — recalling the medieval
preference for both etymological and grammatical derivation from the noun (see
Chap. 4) — and, what is more, abstract nouns, with the very first example being
nothing other than rhetoric.
The Deffence et illustration de la langue françoyse is clearly a political
rather than a philosophical linguistic treatise, and the classical themes are mostly
used toward rhetorical ends. Even so, the disjunction between Du Bellay's open
ing statement on the conventionality of language and his later use of rhetorical
devices suggesting that it is a natural phenomenon, and that between his treatment
of nouns and adjectives on the one hand and verbs and prepositions on the other,
are enlightening and entirely characteristic of modern writing on language. We
shall see that such political and rhetorical concerns were not lost as philosophical
enquiry into language gained new momentum in the years that followed.
tous les vocables grecs, comme rhetorique, musique, arithmétique, géométrie, philosophie [...] et
generalement la plus grand' part des termes usitez aux sciences naturelles et mathematiques. Ces
mots là doncques seront en nostre langue comme estrangers en une cité [...]. Doncques si la
philosophie, semée par Aristote et Platon au fertile champ attique, estoit replantée en notre plaine
françoise, ce ne seroit la jeter entre les ronces et espines, où elle devinst sterile: mais ce seroit la
faire de lointaine, prochaine, et d'estrangere, citadine de nostre république".
152 LIMITING THE ARBITRARY
way, any more than most linguists see standard languages this way, but this fact
hardly makes a dent in the general cultural conception of either.
The way in which all of us become and remain most aware of a standard
language is through the teaching and enforcement of specific language standards.
When we judge whether or not a particular text is in Standard English, for exam
ple, we do not proceed on the basis that there is only one 'standard' way of ex
pressing or communicating anything in English. On the contrary, we assume a
great multiplicity of ways of expressing or communicating anything within Stan
dard English. However, on a very large number of specific points concerning
pronunciation, spelling, grammar, syntax and lexicon, standards are recognized
that fix the range of what is reckoned good and bad,rightand wrong, standard and
non-standard.
The content of any particular language standard is ultimately arbitrary. This
is evident from the language-specificity of such standards. An often cited example
is the non-standardness of English 'double negatives' like I didn't say nothing (in
the sense of Standard English I didn't say anything), which however correspond
exactly to the Standard Italian form Non ho detto niente (literally Not I-have said
nothing). In French, an important language standard distinguishes between the
single-word negation as in J'ai rien dit (literally I' have nothing said), which is
colloquial and non-standard, and Je n'ai rien dit (literally I not have nothing said)
which is standard. (Here the situation is historically more complex than for Italian
in that French rien derives from the Latin rem 'thing', not originally negative.)
With regard to pronunciation, it cannot be the case that standard forms are intrin
sically clearer to perceive than non-standard ones, as is often assumed, since it is
frequently the standard forms that are the more complex in articulatory and acous
tic terms by the phoneticist's analysis and in systematic terms by the phonolo-
gist's. RP, the 'received pronunciation' of English, is a cardinal example of a
standard spoken form of a much higher order of complexity by any measure than
many of the forms reckoned non-standard in opposition to it.
But we can go still further and say that arbitrariness of content is a require-
ment for language standards to exist and function. Insofar as languages may have
content that is not arbitrary, but motivated by the structure of the mind or the
world, such content must be characterized by universality. This precludes by
definition any possibility of variation, which is a prerequisite for a language
standard to be established. A rule which forbids something that no one ever does
would be an absurdity. If anything, rules forbid those things which people are
'naturally' inclined to do. So if, for example, a language showed considerable
variation between two ways of narrating events, allowing them to be recounted
either in the order in which they occurred (as is said to be the normal, unmarked
order in known languages) or in reverse order; and if this variation became the
basis of a language standard, it would likely be the reverse order that was the
standard form, with the order of occurrence (which other languages suggest
NATURAL DIALECT AND ARTIFICIAL LANGUAGE ¡53
16. "[L]'unité linguistique peut être détruite quand un idiome naturel subit l'influence d'une langue
littéraire. Cela se produit infailliblement toutes les fois qu'un peuple arrive à un certain degré de
civilisation. Par «langue littéraire» nous entendons non seulement la langue de la littérature, mais,
dans un sens plus générale, toute espèce de langue cultivée, officielle ou non, au service de la
communauté tout entière".
17. "Livrée à elle-même, la langue ne connaît que des dialectes dont aucun n'empiète sur les
autres, et par là elle est vouée à un fractionnement indéfini. Mais comme la civilisation, en se
développant, multiplie les communications, on choisit, par une sorte de convention tacite, l'un des
dialectes existants pour en faire le véhicule de tout ce qui intéresse la nation dans son ensemble".
154 LIMITING THE ARBITRARY
that gets established between literary language and dialects. But even so, he is
describing the very same situation as one of unity and infinitesimal subdivision,
which may strike you as contradictory or mystical, or simply Dantesque, akin to
the vernacular which "the whole world makes use of [...] though it has been di
vided into different forms of utterance and words".
But what could be less Dantesque than to imagine that the standard dialect
is ever chosen by a 'tacit convention'? Throughout history, the conventions by
which particular dialects have emerged as standard have never been tacit, but
always accompanied by questioni della lingua, battles vehemently argued. In an
effort to marginalize standard languages for purposes of linguistic study, Saussure
erases the entire history of standardization as a historical process. He further
dehistoricizes it by depicting it as the inevitable consequence of civilization,
brought about not directly but only secondarily by wilful human action. This
seems to justify his decision to eliminate any consideration of 'literary language'
from linguistic study, though he admits that to do so removes his 'social' concep
tion of language to a still higher sphere of abstraction:
The facts discussed in this chapter are so common that they might seem to be normal
in the history of languages. However, we shall here set aside everything that obscures
a clear view of natural geographical diversity, in order to consider the basic phe
nomenon unalloyed by any importation of foreign languages or formation of a literary
language. This schematic simplification may seem to distort reality; but the natural
state of affairs must first be studied in its own right. (CLG 269)
What is the 'basic phenomenon' (phénomène primordial) and 'natural state of
affairs' (fait naturel) to be reached by this abstraction? It seems surprising coming
from the man who declared that "in linguistics [...] one might say that it is the
viewpoint adopted which creates the object" of study (CLG 23).19 This means that
in linguistics there is no 'basic phenomenon'. But when it comes to 'literary
languages' Saussure could pass for Dante discussing gramatica. Like the volgare
illustre, Saussure's langue is removed from history; it is common to all; and here,
though not elsewhere, it is treated as a fait naturel, literally a 'natural fact'.
18. "Les faits dont il a été question dans ce chapitre sont si fréquents qu'ils pourraient passer pour
un facteur normal dans l'histoire des langues. Cependant nous ferons ici abstraction de tout ce qui
trouble la vue de la diversité géographique naturelle, pour considérer le phénomène primordial,
en dehors de toute importation d'une langue étrangère et de toute formation d'une langue littéraire.
Cette simplification schématique semble faire tort à la réalité ; mais le fait naturel doit être d'abord
étudié en lui-même".
19. "[D]ans notre domaine [...] on dirait que c'est le point de vue qui crée l'objet".
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 155
The word standard is ambiguous between an older sense of a high mark of quality
that must be met, and a newer sense of everything being identical. The latter sense
was strengthened with the spread of the verb standardize in the early 20th century.
Clearly what 'standard English' means to Orwell is a way of speaking that is meant
to be identical for all, and that although it is being imposed from the top down,
represents no mark of quality, only homogeneity. Orwell believed that the power
of language to promote clear thinking and combat tyranny is inherent to the
language of the working classes. The tendencies of language and thought he
believes must be resisted are those he associates with the educated middle and
upper classes.
'Educated' English has grown anaemic because for long past it has not been reinvig-
orated from below. The people likeliest to use simple concrete language, and to think
of metaphors that really call up a visual image, are those who are in contact with
physical reality [...] (ibid., p. 27)
This theme was taken up in Orwell's article "Politics and the English Language",
which appeared in 1946 in the prominent London literary review Horizon. De
scribed as "his most influential essay" by his biographer Michael Shelden (1991:
430), it is interesting for how it anticipates the core problem of language he would
address so memorably in Nineteen Eighty-Four. Given his remarks in The English
People about the dangers of Standard English, and the fact that the satirical New-
speak of Nineteen Eighty-Four is an engineered language, it may be surprising that
the 1946 article opens with an earnest call for conscious action to engineer current
English.
Most people who bother with the matter at all would admit that the English language
is in a bad way, but it is generally assumed that we cannot by conscious action do
anything about it [...]. Underneath this lies the half-conscious belief that language is
a natural growth and not an instrument which we shape for our own purposes.
[...] The point is that the process is reversible. Modern English, especially written
English, is full of bad habits which spread by imitation and which can be avoided if
one is willing to take the necessary trouble. If one gets rid of these habits one can
think more clearly, and to think clearly is a necessary first step towards political re
generation [...]. (Orwell 1946: 252-253)
156 LIMITING THE ARBITRARY
The linguistic 'bad habits' Orwell refers to, and the 'clear thinking' he opposes to
it, have to do with what comes first in the mind of the speaker or writer, words or
images. The healthy way is to start from mental pictures, then find words to
describe them. For if one does the opposite, it is tempting to let the words string
themselves together in well-worn patterns, which lets the words determine the
meaning rather than the other way round.
[M]odern writing at its worst does not consist in picking out words for the sake of
their meaning and inventing images in order to make the meaning clearer. It consists
in gumming together long strips of words which have already been set in order by
someone else, and making the results presentable by sheer humbug. The attraction of
this way of writing is that it is easy [...]. (ibid., p. 259)
This invasion of one's mind by ready-made phrases ([...]) can only be prevented if one
is constantly on guard against them, and every such phrase anaesthetizes a portion of
one's brain, (ibid., p. 263)
On one level, Orwell's view is similar to the advice generally offered to students
of musical composition. They are warned against working at the keyboard, where
it is too easy to let their fingers do the composing by falling into familiar and
comfortable patterns. Composing mentally is likelier to produce music that is
original rather than derivative, and cerebral rather than emotional. Beyond this,
however, Orwell is concerned about the fact that "if thought corrupts language,
language can also corrupt thought" (p. 262). If we begin from mental images,
those images will be of concrete things, whereas starting from words is likelier to
produce purely abstract thinking. Orwell, realist that he is, is not against abstract
thinking so long as it is grounded in observable reality.
When you think of a concrete object, you think wordlessly, and then, if you want to
describe the thing you have been visualizing, you probably hunt about till you find the
exact words that seem to fit it. When you think of something abstract you are more
inclined to use words from the start [...] Probably it is better to put off using words as
long as possible and get one's meaning as clear as one can through pictures or sensa
tions, (ibid., p. 264)
This discussion has links with the long-standing Western philosophical debate
about realism and nominalism — whether what words mean connects to things
outside language or not. But where do the 'politics' come in? The answer is that
this detachment of language from observable reality is what makes it possible for
a political party to maintain an orthodoxy among its followers, and to dupe those
whom it wishes to enslave. If the party manages to use language in a way that
prevents concrete mental pictures from being called up, people will not understand
what is happening to them, and they cannot rebel against what they do not under
stand.
In our time it is broadly true that political writing is bad writing. Where it is not true,
it will generally be found that the writer is some kind of rebel, expressing his private
opinions and not a 'party line'. Orthodoxy, of whatever colour, seems to demand a
lifeless, imitative style, (ibid., pp. 260-261)
NATURAL DIALECT AND ARTIFICIAL LANGUAGE \ 57
In our time, political speech and writing are largely the defence of the indefensible
[...]. Such phraseology is needed if one wants to name things without calling up men
tal pictures of them, (ibid., pp. 261-262)
The linguistic intervention Orwell calls for is not a restructuring of the language,
just a change in how its elements are put to use. One should always start with
thought rather than words, and with thought about what is concrete and empiri
cally observable, and therefore verifiable. Then and only then can language hope
to serve the interests of truth, rather than merely those of power.
Toward the end of the article he makes clear that his call for abstractions to
be grounded is by no means to reject them entirely. On the contrary, too great a
distrust of abstractions can have catastrophic political consequences of its own.
Stuart Chase and others have come near to claiming that all abstract words are mean
ingless, and have used this as a pretext for advocating a kind of political quietism.
Since you don't know what Fascism is, how can you struggle against Fascism? (ibid.,
p. 265)
Stuart Chase (1888-1985) was the author of The Tyranny of Words (1938), a
widely read book that helped to popularize General Semantics, a movement
concerned with how metaphysical traps encoded into language lead us into false
modes of thought. The fact that Chase encouraged direct intervention into lan
guage use in order to produce clear thinking would seem to link him to Orwell's
programme. But as the preceding quote suggests, Chase was so sceptical about
abstract words as to delude himself that their 'tyranny' was more real than Hit
ler' s. Early in his book he writes:
Abstract terms are personified to become burning, fighting realities. Yet if the knowl
edge of semantics were general [...] the conflagration could hardly start [...].
[...] Bad language is now the mightiest weapon in the arsenal of despots and
demagogues. Witness Dr. Goebbels. Indeed, it is doubtful if a people learned in se
mantics would tolerate any sort of supreme political dictator [...]. A typical speech by
an aspiring Hitler would be translated into its intrinsic meaning, if any. Abstract words
and phrases without discoverable referents would register a semantic blank, noises
without meaning. For instance:
The Aryan Fatherland, which has nursed the souls of heroes, calls upon
you for the supreme sacrifice which you, in whom flows heoic blood, will
not fail, and which will echo forever down the corridors of history.
This would be translated:
The blab blab, which has nursed the blabs of blabs, calls upon you for
the blab blab which you, in whom flows blab blood, will not fail, and
which will echo blab down the blabs of blab.
The 'blab' is not an attempt to be funny; it is a semantic blank. Nothing comes
through. The hearer, versed in reducing high-order abstractions to either nil or a series
of roughly similar events in the real world of experience, and protected from emotive
associations with such words, simply hears nothing comprehensible. The demagogue
might as well have used Sanskrit. (Chase 1938: 14)
158 LIMITING THE ARBITRARY
The 'blab' paragraph, which is all the more hilarious for Chase's claim that it is
not meant to be funny, prefigures what Orwell will term 'duckspeak' {Nineteen
Eighty-Four, p. 322). If Chase thought that "Bad language is now the mightiest
weapon in the arsenal of despots and demagogues", Orwell had a neck wound to
remind him that enormous military-industrial complexes such as Hitler and Stalin
possessed were not so easily 'blabbed' away. Hitler's rhetoric and Goebbels'
propaganda may have played a key role in the Nazi rise to power, but now that the
power was theirs to lose, the way to combat it was not to proclaim their abstrac
tions empty. On the contrary, the urgent need was to show people how the use of
abstract words by despots filled them with concrete and terrible meanings.
In Orwell's Nineteen Eighty-Four, Newspeak is the re-engineered English
of Oceania, a country comprising the Americas, the Atlantic islands including the
British Isles, Australasia and the southern portion of Africa. Oceania is perpetually
at war with one of the world's other two countries, Eurasia and Eastasia. It is
controlled by the Party, whose head, Big Brother, is a symbol rather than an actual
person. There is an Inner Party of a privileged 2% of the population; a larger Outer
Party which does not enjoy anything like the same privileges; and the proles, the
remaining 85% whose lives have not changed radically since before the Revolu
tion, except that in material terms they are considerably worse off. The mind
control described in the opening quotation from the novel is directed almost
entirely at the members of the Outer Party, and is enforced by the Thought Police,
which the Inner Party controls. The proles are considered not worth bothering
about.
The idea of re-engineering the English language by reducing its vocabulary
had already been prominently put into practice by C. K. Ogden (1889-1957), who
was mentioned in the previous chapter (p. 120) in connection with Bentham.. The
chapter on Definitions in The Meaning of Meaning (1923), co-authored by Ogden
and I. A. Richards (1893-1979), had led Ogden to formulate the idea of a 'Basic
English' that would be capable of expressing anything with a vocabulary of just
850 words (see further Joseph 1999a). Orwell was interested in Basic and wrote
about it on a couple of occasions in the 1940s. The feature of Basic trumpeted
most loudly by Ogden, the fact that it had done away with verbs, has a direct
parallel in Newspeak {Nineteen Eighty-Four, p. 165). Another feature of Basic,
its replacement of certain negative adjectives by their positive counterpart pre
ceded by un-, is exaggerated to the point of absurdity in Newspeak, where for
example the equivalent of the Oldspeak form 'terrible' is doubleplusungood.
Like the project for The Meaning of Meaning from which it sprang, Basic
was an attempt to solve a perceived crisis of meaning in the modern world. In
Ogden's view, the First World War was itself the result of the misuse of complex
abstract words like democracy and freedom for purposes of propaganda, and any
hope of future world peace depended upon the ability of thinking people to control
the meanings of such words so that they could not be abused. The Meaning of
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 159
Meaning opens with a long historical survey of attempts to do this, including the
solution proposed by Locke (see pp. 109-110 above). Locke classified ideas into
the simple and the complex, and among complex ideas he believed that those he
called 'mixed modes', including all moral terms, were the likeliest to create
misunderstanding — unless they were always carefully defined in terms of the
simple ideas, derived from direct sensory experience, that combined to produce
them. For essentially the same reason, Ogden believed that paring down the
language to 850 words, a large portion of them referring to concrete substances,
would make it virtually impossible to use language in such a way as to deceive
people for propagandistic purposes.
But Orwell realised that it might actually have the opposite effect. Propa
ganda can only be combated by rational analysis and argument. This entails
rephrasing propagandistic statements in a different form. If such rephrasing were
made impossible through the loss of alternative words in which the same idea
might be given a different linguistic shape, then it might no longer be possible to
question the truth of any statement. Orwell made this into the precise aim of
Newspeak: "to make all other modes of thought impossible". For instance, ac
cording to the Party, 2 + 2 = 5. The hero of the novel, Winston Smith, realises
from the evidence of his own eyes that this is wrong, but the Party already has
enough control over his thought and language that he cannot put together the
argument he intuitively knows would prove its falsity. The same is true with the
Party's operation for rewriting history, in which Winston himself is engaged, and
indeed with its three slogans:
WAR IS PEACE
FREEDOM IS SLAVERY
IGNORANCE IS STRENGTH
Winston's estranged wife Katharine "had not a thought in her head that was not
a slogan" {Nineteen Eighty-Four, p. 69) - that is, a collocation of words and
thought pre-packaged by the Party. By reducing the number of words and their
possible collocations, the Party strictly limits the occurrence of original thought,
whether based on empirical observation or individual reasoning. For Winston, this
stranglehold on sensory evidence and creativity in combining words represents
what is most evil and oppressive about the Party.
The Party told you to reject the evidence of your eyes and ears. It was theirfinal,most
essential command. His heart sank as he thought of the enormous power arrayed
against him, the ease with which any Party intellectual would overthrow him in de
bate, the subtle arguments which he would not be able to understand, must less an
swer. And yet he was in the right! [...] Stones are hard, water is wet, objects unsup
ported fall towards the earth's centre. With the feeling that he was [...] setting forth
an important axiom, he wrote:
Freedom is the freedom to say that two plus two make four. If that is
granted, all else follows. {Nineteen Eighty-Four, p. 84)
160 LIMITING THE ARBITRARY
It is because of the way his command of language has been controlled by the Party
that he could not hope to understand or answer the arguments. At the end of the
novel, Winston, his mind broken by torture, signals how completely he has sub
mitted to the Party's doctrine when he traces "almost unconsciously" in the dust
on the table: 2 + 2 = 5 (ibid., p. 303).
Newspeak is directly connected to the ideas expressed by Orwell in "Politics
and the English Language". As the ultimate language for the suppression of
thought, Newspeak represents the horrific end of the road Orwell describes Eng
lish as travelling, the point at which it is too late to get rid of the linguistic bad
habits that prevent clear thinking and political regeneration because they have
become structurally ingrained. The appeal of Basic, which Orwell himself had felt,
is perhaps just a further symptom of how far this development has gone. Origi
nally proposed as a way of grounding language in observable reality, Ogden's
Basic aimed to do this by intervening directly into the structure of English, paring
it down to a fraction of its traditional form. But was not this already a form of
linguistic tyranny, limiting rather than expanding people's freedom to speak and
think as they pleased? If so, Orwell the interventionist in linguistic usage could not
support it any more than Orwell the socialist could stomach the excesses of Sta
linism.
The point made at the end of the 1946 essay, about excessive distrust of ab
stractions leading to an inability to recognise or combat Fascism, is echoed in the
description of the word free in Newspeak
The word free still existed in Newspeak, but it could only be used in such statements
as 'This dog is free from lice' or 'This field is free from weeds'. It could not be used
in its old sense of 'politically free' or 'intellectually free', since political and intellec
tual freedom no longer existed even as concepts, and were therefore of necessity
nameless... Newspeak was designed not to extend but to diminish the range of
thought, and this purpose was indirectly assisted by cutting the choice of words down
to a minimum. (George Orwell, "Appendix: The Principles of Newspeak", from
Nineteen Eighty-Four, p. 313)
Free has been limited to just its concrete meaning. "This dog is free from lice"
certainly calls up a clearer mental image than do "politically free" or "intellectu
ally free". But again, while abstraction without a concrete anchor remains ex
tremely dangerous, the failure to abstract away from certain key concrete anchors
is no less threatening.
Orwell's error is to have romanticized the working classes. He himself rec
ognized this, and his attitude toward the proletariat is far more ambivalent in his
last two novels than the two essays on language cited at the start of this section
would lead one to expect. In Oceania, only the proles have "stayed human"
{Nineteen Eighty-Four, p. 172), and we see from the occasional glimpses of their
dialogue that their language is Oldspeak, as in this conversation about the lottery
which Winston overhears in a pub:
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 161
'Can't you bleeding well listen to what I say? I tell you no number ending in
seven ain't won for over fourteen months!'
'Yes it 'as, then!'
'No, it 'as not! Back 'ome I got the 'ole lot of 'em for over two years wrote down
on a piece of paper. I takes 'em down reg'lar as the clock. An' I tell you, no number
ending in seven — '
'Yes, a seven 'as won! [...]' {Nineteen Eighty-Four, p. 88)
Each of the two proles is capable of independent thought, and one of them makes
an argument based on historical evidence that would be beyond the ability of any
Party member. For its members, the Party is rewriting history every day and
making sure they cannot perceive it happening. What is more, the fact that these
proles are arguing about numbers contrasts with the inability of Party members to
argue about the sum of 2 + 2. And the numerous non-standard features of their
'Oldspeak' cause it to ring with freedom to Orwell's ears. Part of the reason they
have stayed human is that they have clung to 'real' language. The ambivalence is
that they are a faint hope, but the only hope, against mind control by the Party,
operated in large measure through Newspeak, the ultimate form of 'standard
English'.
Orwell's belief about Standard English that "Its characteristic is its reliance
on ready-made phrases" that are "thought-saving devices" (quoted on p. 155
above), is certainly wrong. A reliance on ready-made phrases cannot be the defin
ing characteristic of Standard English, because ready-made phrases characterize
all English, standard and non-standard. Whether they are more characteristic of
the one than the other might be made into an empirical question — if we could
come up with independent and objective definitions of where 'standard English'
and 'ready-made phrases' begin and end, which is no easy task. But the evidence
from corpus linguistic studies over the last decade, including the COBUILD project
and other large corpus-based studies, has been unequivocally that language is
'chunkier' than was previously believed, in the sense that if a given English word
has just been uttered, we can predict with a surprising degree of accuracy which
words, or which of a limited number of options, will follow it. Our processes of
speaking and writing, that is, do not proceed word by word, but in larger 'pre
packaged' chunks. This observation has potentially important implications for how
we imagine language being 'stored' in the brain. It has long been imagined in the
form of a grammar and a lexicon being in our heads. This is wildly metaphorical
— grammars and lexica are books, after all, and it is astonishing how thoroughly
the metaphorical nature of their projection into our heads has been forgotten by
some people. But the basic idea is that in one part of our brains is an inventory of
words, understood as sound-meaning correspondences, and in another part are
rules for putting the words together.
Metaphors notwithstanding, this picture makes a certain amount of ex
planatory sense so long as we accept that words (or morphemes) are the basic unit
from which utterances our built. The notion of a mental lexicon and grammar are,
162 LIMITING THE ARBITRARY
ments of speech, not only for their rational content but for other content that we
use to evaluate the truth value of what is being said to us, and indeed to evaluate
the whole person with whom we are speaking. That is above all what standard
language tries to control, by eliminating all but the 'rational' elements.
What Orwell realized is that when human language is no longer grounded
in our deeper humanity, particularly in the senses which we all share, we lose our
ability to know when we are being deceived by others, and ultimately our protec
tion against having our minds taken over. 'Ready-made phrases' sounded to him
like evidence of this happening; what he failed to recognize is that some such
phrases are more equal than others (to use a memorable ready-made phrase from
Animal Farm) and indeed that many features of non-standard English that might
be described by this exact term are squarely among the defences against mind
control. Their interpretation is not 'ready-made', and it is on their basis that we
make decisions concerning the truth value of the 'rational' content of what is said
to us. And therein lies the possibility of that infinite linguistic creativity we still
want to believe in so desperately.
dialects" are as diverse as the several Romance languages. We speak of Dutch and
German as two separate languages, although some dialects of German are very close
to dialects that we call "Dutch" and are not mutually intelligible with others that we
call "German". A standard remark in introductory linguistics courses is that a lan
guage is a dialect with an army and a navy (attributed to Max Weinreich). (Chomsky
1986a: 15)
Up to the second sentence of this paragraph he has been talking about 'language'.
He switches to talking about 'a language', a count rather than a mass noun, in the
course of introducing an old problem which led linguists as far back as the 19th
century to avoid the 'commonsense' language/dialect distinction and use only
'dialect' in the description of linguistic systems. Chomsky's point is that 'lan
guages', the count noun, could be a scientific concept if you could predict them
by looking at language structure or mutual intelligibility, but in fact it comes down
to how a people has come, or been persuaded, to think of itself and its way of
speaking and writing relative to those of its neighbours — a matter of rhetoric.
This problem with language the count noun infects language the 'commonsense'
mass noun.
He goes on to say that these problems have prompted earlier linguists to
develop "various technical notions of language to replace the commonsense
notion" (1986a: 19). Chomsky now conflates these with the common-sense con
cept into 'E-language', described as a "construct [...] understood independently of
the properties of the mind/brain" (p. 20), "language regarded as an externalized
object", hence the 'E' (p. 24). In contrast to this, his linguistics studies 'I-
language', which is internalized, individual and intensional, "some element of the
mind of the person who knows the language, acquired by the learner, and used by
the speaker-hearer" (p. 22), "the system of knowledge of language attained and
internally represented in the mind/brain" (p. 24). He makes clear that only I-
language is really real (p. 26):
The technical concept of E-language is a dubious one in at least two respects. In the
first place, as just observed, languages in this sense are not real-world objects but
are artificial, somewhat arbitrary, and perhaps not very interesting constructs. In
contrast, the steady state of knowledge attained and the initial state S0 are real ele
ments of particular mind/brains, aspects of the physical world, where we understand
states and representations to be physically encoded in some manner. The I-language
is abstracted directly as a component of the state attained.
This is an interesting use of 'artificial', first because it is being put in contrast with
'real-world objects' as if artificial objects did not belong to the real world. We do
often contrast the 'artificial' with the 'real' — artificial versus real flowers, for
instance — but without implying that artificial flowers are not 'real' in the sense
of not being 'real-world objects'. So too calling E-languages 'somewhat arbitrary'
in contraposition to 'real-world' is surprising. The expected opposite of both
'artificial' and 'arbitrary' in this context would be natural', I-language is clearly
being opposed to E-language as a natural object to a human construct. The Greek
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 165
word for nature is physis, and Chomsky goes on to stress the physical reality of I-
language. But note how subtly the pluralization appears: "real elements of par
ticular mind/brains" — not of a particular mind/brain. Again, he has previously
characterised I-language as "some element of the mind of the person who knows
the language" (p. 22), "the system of knowledge of language attained and inter
nally represented in the mind/brain" (p. 24). Chomsky has established the 'physi
cal reality' of I-language on the premise that he is talking about a part of some
person's brain, which is convincing enough. But when he slips in that plural,
making it some persons' brains, but continues to talk about 'I-language' rather
than I-languages, everything changes. The I-language too has become a generali
sation, an abstraction, as he goes on to say: "The I-language is abstracted directly
as a component of the state attained". By a masterful use of pragmatics, attention
is drawn to 'directly' and away from 'abstracted'. Remember that what is sup
posed to distinguish I-language from E-language is the physical reality of the
former versus the 'artificial', 'arbitrary', and obviously 'abstract' nature of the
latter. Now we are told that I-language is abstracted — but directly abstracted
from physical reality, whereas (p. 27)
[...] E-language, however construed, is further removed from mechanisms than I-
language, at a higher order of abstraction.
Further removed from mechanisms? Higher order of abstraction? No explanation
of these mysterious descriptions is offered here, though what is meant can be
inferred from other writings by Chomsky, for example the following:
'[I]dealization' is a very misleading term, because it really means a move towards
reality. When you talk about idealization or abstraction, it is an effort to find the re
ality. When you roll a ball down a frictionless plane, that is called an idealization,
but what you are really doing is finding the real principle by which things attract one
another. It is the phenomena that are a nuisance: they are unreal in a way, because
they are too complicated. Reality hides behind the phenomena, as it were, so you
have to get rid of a lot of the phenomena to find it. (Chomsky, in Dillinger & Palácio
1997: 184)
I-language would be like the principle of attraction that is 'directly abstracted'
from the ball rolling down the frictionless plane, whereas E-language would
include all the nuisances like friction which are "unreal in a way, because they are
too complicated". Unreal in a way — to be precise, a way that is directly contra
dictory to the normal usage of 'real' and 'unreal'. Related to this is a statement in
Chapter One of the 1986 book:
Generative grammar ['s ...] standpoint is that of individual psychology. It is con
cerned with those aspects of form and meaning that are determined by the 'language
faculty,' which is understood to be a particular component of the human mind
(Chomsky 1986a: 3).
The problem is that 'individual psychology' is not concerned with 'the human
mind', but with human minds. As soon as we speak of the human mind we have
166 LIMITING THE ARBITRARY
abandoned the realm of the individual for the collective idealisation, what Chom
sky will call the abstract and unreal when his goal is to marginalize E-language.
If generative grammar's standpoint were really that of individual psychology, it
would take any 'language faculty' to be a particular component of human minds,
or of a human mind, but would have no truck with the human mind. Individual
psychology, moreover, is truly about individuals, and focuses on what makes them
different to other individuals; the goal of generative grammar is quite the opposite,
the discovery of what is universal.
Having started out by associating E-language with the common-sense
understanding of language, which is disqualified from scientific consideration,
then rejecting E-language for its own purposes because it is not real, which is to
say that it occupies a higher level of abstraction than I-language, Chomsky quite
astonishingly says (pp. 27-28) that:
The shift of focus is also, arguably, a shift toward the commonsense notion of lan
guage [...]. Of the various technical notions that have been developed in the study
of language, the concept of I-language seems closer to the commonsense notion than
others.
The shift of perspective from the technical concept E-language to the technical
concept I-language is therefore a shift toward realism in two respects: toward the
study of a real object rather than an artificial construct, and toward the study of what
we really mean by "a language" or "knowledge of language" in informal usage
(again, abstracting away from sociopolitical and normative-teleological factors).
I-language is now described as closer to the common-sense understanding of
language — and then this is offered as another argument f or its greater reality!
Chomsky appears to be reacting to criticisms of his earlier 'idealized native
speaker-hearer in a homogenous speech community' as being precisely an ab
stract, artificial construct corresponding to no real-world object (ibid., p. 17). His
response is, in effect: Look, before I came along most linguistics was concerned
with 'languages', and where do they exist in the world? I turned attention to the
physical — a part of the brain of a particular speaker-hearer. Now that is a very
powerful argument, because most of us today believe that something we can see
and touch is more real than something invisible. We are not Platonists who dis
miss physical things as mere shadows of what is really real, namely, the ideal
forms of things that exist in heaven. But the moment Chomsky has us nodding in
agreement, the I-language of a particular speaker-hearer turns into the I-language
(singular) of particular speaker-hearers, and the Gates of Heaven swing open. The
clear distinction between E-language as artificial abstraction and I-language as
physical reality is blurred.
Chomsky cannot stick to a particular speaker-hearer for much the same
reason he cannot imagine a conception of language involving socio-political
elements to be scientific. He is wedded to a very conservative conception of 'hard'
and 'soft' sciences, where the hard sciences are those in which predictions of
absolute certainty and a high degree of generalisation can be made. That, to him,
NATURAL DIALECT AND ARTIFICIAL LANGUAGE 167
is the ideal for his undertaking. He has formulated the concept of I-language
ostensibly in order to identify precisely what he believes can and must be the
object of study of such an approach. The problem is that, if he were to follow his
logic through, the speaker's knowledge of language would have to be studied one
individual at a time, which is neither practical nor intellectually acceptable within
the research framework Chomsky has chosen. And it is thus that he must get out
of the corner he has painted himself into with the kinds of rhetorical nuances
discussed here, which have the effect of turning I-language into just the sort of
'higher-order abstraction' that E-language is supposed to represent, and that
supposedly cannot be studied scientifically.
INVISIBLE HIERARCHIES
FROM JAKOBSON TO OPTIMALITY THEORY
Linguistes" (1928b), written in October of 1927, gained him his first recognition
among linguists world-wide. The third, a monograph entitled Remarques sur
I' évolution phonologique du russe (1929), written in 1927-28, is an application of
the theoretical programme outlined in the first two papers.
Jakobson's project was nothing less than to restructure historical linguistics
upon the findings of synchronic enquiry, making it into the diachronic pro
gramme Saussure had called for. He insisted that the development of elements of
the linguistic system cannot be explained without reference to their function in
the system at any given point in time, i.e., teleologically. Synchronic function
thus appears as the 'goal' (1962 [1928a]: 1; 1962 [1928b]: 3, 6) of diachronic
change. Jakobson soon replaces the term 'goal' with 'result':
The relationship between a diachronic law and a synchronic law (cf. CLG 131) can
be defined as the relationship between a means and a result obtained. We use the term
result rather than goal not in order to deny that diachronic laws are tendencies,
teleological in nature, but because in many cases the outcome does not coincide with
the original problem; just as in other areas of human activity, especially collective
2
ones, goals are not always attained. (1962 [1929]: 106; my translation: JEJ)
The reasoning here is not altogether consistent: the fact that "goals are not always
attained" would seem to justify retaining the term but, not abandoning it. The real
problem with 'goal' is that it is highly metaphorical, suggesting a view of lan
guage as something capable of establishing targets. Jakobson, contrasting his
teleological approach with the 'mechanistic' neogrammarian concept of language
change as fortuitous, involuntary, blind (1962 [1928b]: 5), may have foreseen the
accusation of an equally fallacious 'anthropomorphism' conceiving language
systems as possessed of will and vision.
Jakobson further specified that the investigation of an historical change
should be "limited to a linguistic system characterized by one and the same func
tion, i.e., to linguistic entities which are functionally equivalent" (1962 [1928a]:
1). At the level of sound, this would restrict enquiry to the functional unit, the
phoneme, excluding mere phonetic changes which do not alter the unit's sole
function of making meaningful distinctions possible.
In studying phonetic changes in a language, the first problem is knowing to what ex
tent and in what way they target and strike the phonological system; in other words,
are meaningful differences affected, or only extra-grammatical elements? (1962
1. Viel (1984: 31-50) has portrayed in vivid detail the atmosphere of change and excitement at
the First International Congress of Linguists held at The Hague in April of 1928. Nothing
stirred quite so much excitement as Jakobson's 'Proposition,' the manifesto of Praguean struc
turalism, which was countersigned by Trubetzkoy and Karcevskij (see Viel 1984: 35-36, 39-41
on the disputed authorship of the Proposition.
2. "La relation d'une loi diachronique à une loi synchronique (cf. Saussure [= CLG] 131) peut être
définie comme la relation du moyen au résultat obtenu. Nous disons au résultat, et non au but,
non pas que nous niions le caractère de tendance, l'esprit téléologique des lois diachroniques, mais
parce que, maintes fois la realisation ne coincide pas avec le probleme posé; de même que dans les
autres domaines de l'activité humaine, collective en particulier, le but n'est pas toujours atteint".
172 LIMITING THE ARBITRARY
[1928b]: 6)3
Within these limits, he said, the neogrammarian dictum that "The sound laws
admit of no exceptions" (then generally under attack) would prove valid (1962
[1928a]: 1; 1962 [1928b]: 6).
He claimed that this same method of focusing on changes in systemic
function and the synchronic states to which they give rise could also account for
the 'social' character of change — "why a speech community accepts and sanc
tions individual slips" — which the neogrammarians ignored. Borrowing Saus
sure's most famous analogy, Jakobson noted that as in a game of chess, "[I]f the
order within a linguistic system is disturbed, there follows a cycle of sound
changes aiming at its renewed stabilization [...]" (1962 [1928a]: 2; see also 1962
[1928b]: 5-6). These changes are dictated by the needs of the system; it is the
system's movement from stability to instability and back again that determines
which of the slips made by individual speakers (including children acquiring the
language) will be accepted by the speech community as a whole. The example
he studied in the greatest detail was the loss of the weak jer in Proto-Slavic,
which he characterized as having 'profoundly revolutionary consequences', by
provoking a radical reworking of the phonological system in all the Slavic dia
lects, as well as essential modifications in how phonemes are combined, plus a
whole series of morphological innovations (Jakobson 1929: 55).
Jakobson credited Saussure and the Geneva School for their pioneering ef
forts in synchronic linguistics, and indicated that his own intention was to draw
the remaining conclusions from their view of language as a system of relative
values. Yet he rebuked them that "as to the field of language history they re
mained in the neo-grammarian rut" (1962 [1928a]: 1-2). Saussure's doctrine, he
3. "Quand on étudie les changements phonétiques d'une langue, le premier problème qui se pose,
c'est de savoir à quel point et dans quel sens ils visent et atteignent le système phonologique; en
d'autres termes: sont-ce les différences significatives ou bien les éléments extragrammaticaux
seuls qui en sont affectés?".
4. Jakobson notes further that besides their system-internal and social function, linguistic ele
ments perform a connotative function whenever they are associated with a particular region,
social class, or area of endeavour. "The overlapping between territorially, socially or function
ally distinct linguistic patterns can be fully comprehended only from a teleological point of
view, since every transition from one system to another necessarily bears a linguistic function"
(Jakobson 1962 [1928a]: 2). Analysis must take account of what is communicated by the sim
ple fact of choosing one pattern over another where two or more are available. Jakobson would
not often touch upon such matters in his later work, but several of the Czech members of the
Prague Linguistic Circle would concentrate on this question in the early 1930s (see further Jo
seph 1987).
5. Jakobson does not cite evidence for this viewpoint. It is true that practically all of Saussure's
non-posthumous publications were historical studies of a neogrammarian bent, and that his
1879 thesis remains one a high point of the neogrammarian movement; but the divorce of syn
chrony and diachrony in Saussure (1916) is neither so absolute nor so simplistic as Jakobson
would imply. One detailed study of the question concludes that Saussure was "[...] tout à la fois
bénéficiaire et prisonnier de sa filiation néo-grammairienne, qui lui a servi à éliminer la con
ception de la langue comme organisme naturel, mais qui l'a poussé, dans le domaine de la
diachronie, à un certain mécanisme [...]" (Buridant 1984: 46). It is doubtful that Jakobson, in
INVISIBLE HIERARCHIES 173
wrote, "contaminates the idea of phonetic changes being accidental and involun
tary and a teleological way of treating synchronic phonology" (1962 [1928b]: 5).6
According to Jakobson there could be no purely diachronic phonetics without
reference to synchronic phonology; this led him to go even further and reject
Saussure's fundamental separation of the synchronic and diachronic domains
(1962 [1928b]: 3; 1962 [1929]: 106).
The structuralism of Prague thus began, or at least established its distinct
identity, with an act of rejection of Saussure that was itself rooted in (according
to Jakobson, the consequence of) Saussurean doctrine. It was a critique from
within; the same might be said of linguistics generally, whose practitioners are
paradoxically bound to conduct and formulate their analyses in the very sub
stance, language, which they are seeking to analyze.
whose 'demonology' the neogrammarians occupied the first rank (Viel 1984: 39n.), would have
conceded the positive influence.
6. "La doctrine de F. de Saussure contamine et la conception en question [selon laquelle les
changements phonetiques sont fortuits et involontaires] et une façon téléologique de traiter la pho
nologie synchronique"
7. The last objects of his criticism, after the neogrammarians and the Saussureans, are the phoneti
cians, who have erred in focusing their attention upon the articulatory mechanisms of speech pro
duction. Presaging by many years what would be an important facet of his distinctive feature
analysis, Jakobson underlines the greater importance of the acoustical than of the articulatory for
the teleological approach, "since not the motor but the acoustical aspect of speech sounds, aimed
at by the speaker, has a social value [...]".
8. "Le signe est en lui-même fortuit et arbitraire"; "[C]e sont des valeurs arbitraires ayant une
existence dans la communauté qui [...] font l'objet [de la phonologie]".
174 LIMITING THE ARBITRARY
the vocal cords (along with minor ancillary differences in some languages, such
as greater tension or aspiration). In many linguistic systems, they noted, the
distinction between /t/ and lál (and other pairs of unvoiced-voiced consonants)
is 'neutralized' in certain positions, for instance at the end of a syllable or
word: German Räder "wheels" has as its singular Rad, pronounced not *[rad],
but [rat], the same as Rat "council".
Again, the possibility of such a deeper connection contradicts the Saus-
surean view that the phonetic substance of /t/ and /d/ is inconsequential, and all
that matters is the fact that they differ in some perceivable way. Jakobson and
Trubetzkoy proposed the term correlation for the type of relationship holding
between /t/ and /d/, Any pair of elements which do not exist in a correlation,
such as /d/ and /f/, form instead a disjunction. As their work progressed, a new
perspective developed. They realized that the correlation /t/-/d/ consists of a
core of features common to the two sounds, plus a principium divisionis, the
factor which distinguishes them, vocal cord vibration (voicing). They created
the term archiphoneme for the core of features common to lil and lál (symbol
ized /T/). This allowed them to specify that the alternation between German
Rad and Räder does not involve simply a change of phonemes; it is a realiza
tion of the same archiphoneme, but with the principium divisionis deleted in
word-final position.
The term 'correlation' was introduced in Jakobson (1962 [1928b]: 3), dis
cussed above with regard to its historical concerns. The 'Proposition' was written
in response to the question "What are the most appropriate methods for a com
plete and practical analysis of the phonology of any language?", one of several
queries circulated to participants prior to the 1928 Congrès International de Lin
guistes in The Hague. After stressing that it is the system of significant
phonological differences which must be considered, Jakobson pointed out that
these differences are not all of the same type, and that the phonological correla
tion deserves separate consideration. In subsequently introducing the term 'dis
junction' (Jakobson [1962 [1929]: 9], he attributed it to Wundt (1906: 126).
Jakobson did not state precisely what is the special importance of corre
lations for the linguistic system, or what role they play in historical change. He
did note that the absence of a particular correlation in a linguistic system is of
ten tied to the absence or presence of another correlation in the same system. A
language lacking the correlation of long-short vowels must also lack the corre
lation of rising-falling vowel intonation; a language lacking this latter must also
lack the correlation of presence-absence of musical accent. If a language has
the correlation presence-absence of dynamic accent, it will not have the corre
lation of soft-hard consonants. Relationships such as these, Jakobson said, are
important determinants of phonetic changes, since the loss or creation of one
correlation can necessitate a radical reconstruction of the phonological system.
For various reasons discussed by Viel (1984: 36ff.), Jakobson (1928b)
INVISIBLE HIERARCHIES 175
Underlying /t/ and /d/ there is an archiphoneme /T/,10 which consists of what is
left when the principium divisionis of voicing is abstracted away or, to use a later
term, neutralized. In Russian, both the voicing correlation and the palatalization
correlation operate in like manner, with the result that one archiphoneme /T/ un
derlies four phonemes (/t t' d dV).
10. I have used the symbol /T/ rather than /T/, /D/, or /D/ to represent the archiphoneme be
cause /t/ is the 'fundamental variant', the one which, in Jakobson's words, "se trouve dans la
plus faible dépendance des conditions extrinsèques", that is, "qui se rencontre dans les circon
stances les plus variées" (1962 [1929]: 15). The other phonemes, "qui s'associe[nt] invariable
ment à une seule et même circonstance phonétique", are termed 'accessory variants'. To return
to the German example, /t/ has the more varied environment, since it can occur word-finally as
well as initially and medially, while /d/'s environment is more restricted, /t/ is therefore the
fundamental variant, lál the accessory one. The principle applies within the phoneme as well:
English vowels have non-distinctive oral and nasal allophones, the latter occurring only before
nasal consonants, the former everywhere else. The nasal allophones, with their single environ
ment, are accessory variants of the vowel phonemes, while the oral allophones are fundamental.
INVISIBLE HIERARCHIES \ 77
11. The fact that in his later work he would limit the mark to one type of oppositions, the 'priva
tive', in no way diminished its significance: that happened to be its most clearly defined, proto
typical niche, and if other types of oppositions functioned differently, so much the better that his
apparatus should succeed in accounting for this. Jakobson & Waugh (1979: 91) state that "Tru
betzkoy's discovery was first applied in the lecture on sound systems which he addressed to the
International Phonological Congress held in Prague in December 1930" (Trubetzkoy 1931). If so,
then Trubetzkoy must have beaten Jakobson to the punch by a matter of hours, since Jakobson's
own paper to the same Congress is built upon the concept (Jakobson 1931). Trubetzkoy's paper,
like his earlier letters, aims in a conservative way to establish a solid base for the new principle.
He points out that both the marked and the unmarked member of an opposition are determined by
criteria that are ganz spezifische und dabei ganz objektive ('wholly specific and therefore wholly
objective', 1931: 98) — a methodological manifesto rather than a statement of fact.
12. It is perhaps no coincidence that their literary interests diverged as well: Trubetzkoy wrote a
monograph-length treatise on the novels of Dostoevsky, while Jakobson is best remembered for
his close analyses of the internal patterns in sonnets and other short poetical genres. On the reac
tions of French linguists to the contrasts between Jakobson and Trubetzkoy, see Chevalier (1997),
in a volume reviewed in Joseph (1998).
13. Trubetzkoy's letters and the now barely legible copies of Jakobson's letters are part of the
Roman O. Jakobson Collection of the Institute Archives and Special Collections, MIT Librar
ies, Cambridge, Mass.
INVISIBLE HIERARCHIES 179
One can only imagine Trubetzkoy's astonishment at this reply. His simple reflec
tion on the structure of consonantal systems had mushroomed into an explanation
virtually of the meaning of life. Furthermore, Trubetzkoy's statements had un
dergone significant permutation in Jakobson's reply:
[...] your thought about correlation as a constant mutual connection [...]:
Not one of the words translated here as 'constant', 'mutual' or 'connection' ap
peared in Trubetzkoy's letter, and while the latter two are perhaps implicit, there
is nothing to suggest constancy; indeed, Trubetzkoy's parenthetical '(or might its
not be "any"?)' implies quite the contrary.
[...] between a marked and unmarked type [...]: It may seem a minor point,
but Trubetzkoy had only spoken of a 'mark.' It is Jakobson who here creates the
passive participle form 'marked' as well as the negative 'unmarked'.
[...] such historical-cultural correlations as life ~ death [...] are always
confined to relations a ~ non-3. [...]: Besides the fact that he has extrapolated from
the narrowest of linguistic spheres, phonology, to the broadest of existential do
mains, Jakobson has swept aside another of Trubetzkoy's careful qualifications:
Trubetzkoy had not said that correlations were "always confined to relations a
non-a", but suggested that they took "the form of a contraposition of the presence
of a certain mark to its absence (or of the maximum of a certain mark to its
minimum)". Jakobson has simply ignored the parenthetical suggestion of a con-
14. Discussing these missives in his later years (see Jakobson & Waugh 1979: 90-92; Jakobson &
Pomorska 1983: 93-98), Jakobson never remarked on the enormous distance in the content of
Trubetzkoy's letter and his own reply.
180 LIMITING THE ARBITRARY
less fixed and consistent than anything in the Saussurean system, but whose sub
stance was context-sensitive.
Already the notions of correlation and archiphoneme had presented a chal
lenge to the Saussurean system: individual pairs of elements, or series of such
pairs, between which a non-abstract, non-arbitrary relationship could be shown to
exist. One could conceive a solution faithful to Saussure, positing the archipho-
nemic level as the fundamental abstract one. But the mark implied in a very
broad and general way that elements of a linguistic system existed in a basic- de
rived relationship relative to one another, the end result of which would be the
complete hierarchization of a language's sound system (most notably in Chom
sky & Halle 1968). Such hierarchization is the very antithesis of Saussurean radi
cal relationism.
Trubetzkoy suggested that the operation of a phonological correlation in the
linguistic 'consciousness' (which, whatever it may be, is certainly unconscious)
does not depend entirely upon the individual properties of the correlation in
question, but upon the structure of correlations generally. The mark is formulated
as a generalization about the operation of all correlations, and the implication is
that their structure is such as to interact in a certain way with the structure of the
mind. We can deduce that, prior to this letter, two correlations such as palataliza
tion and voicing were thought to share nothing of substance other than the fact of
belonging to the category 'correlation': their relationship appeared to be strictly
taxonomic. Trubetzkoy's idea of active modification by a certain mark versus
passive modification by its absence suggested instead a real functional similarity.
The correlation was promoted from an analytical category imposed upon linguis
tic data to a universal category of structure with its own 'intrinsic content'.
Jakobson's search for external validation for his concept of the mark would
also influence understanding of its nature. His influential (1941) study estab
lished as a general (and still controversial) precept that marked elements are the
last to be acquired by child language learners, and the first to be lost in aphasia,
while unmarked elements are acquired earlier and lost later. In other studies he
proved the greater diachronic durability of unmarked over marked elements in
the evolution of languages. He demonstrated the connection of markedness with
ease of perception and articulation on the level of sound, with simplicity and
complexity of concept on the level of meaning. Most significantly, starting in the
mid-1960s he proposed a direct connection between the distribution of marked
and unmarked categories at the three levels of sound, linguistic meaning, and
real-world phenomena (see Jakobson 1966).
During the same period, and largely under his influence, linguistic analysis
the world over has shifted from movements based upon the view of language put
forth in Saussure's Course to views which distinguish unmarked 'core grammar'
from marked 'periphery', to 'Natural' Phonology, Morphology, and Syntax, and
even to methods founded upon the 'iconicity' — non-arbitrariness — of the
182 LIMITING THE ARBITRARY
that would defined the mainstream of the humanities, the social sciences, and
even certain of the physical sciences at some point from the late 1940s through
the late 1970s (see further Joseph forthcoming). There is little indication that
Jakobson had significant influence on the mainstream American linguistics of
the time, dominated by the students of Bloomfield, who with their behaviourist
tenets rejected any notion of 'mentalism' in language as phantasmagoric non
sense, and who moreover felt professionally threatened by the new wave of
European immigrant linguists of whom Jakobson was the most prominent, the
most ambitious, the most vocal, and, in their view, the most obnoxious. How
ever, by the 1950s much of this resentment had died down, and while the neo-
Bloomfieldians would never really open themselves to Jakobson's brand of
structuralism, some of their students did. In particular, Chomsky, who had done
a Ph.D. under Bloomfield's student Zellig Harris (1909-1992), came into con
tact with Jakobson at Harvard in the early 1950s, through the intermediary of
Morris Halle (b.1923), Jakobson's student who remains Chomsky's close asso
ciate to the present day. Chomsky has long denied that the major elements of
his theories of language and mind owe anything to Jakobson's direct influence;
and in that he may well be right — but indirect influence, which the person in
fluenced is not well placed to judge objectively, is another matter. Jakobson's
'naturalism', as described above, was in a general way the key influence on the
next generation of American linguists, including those like Joseph H. Green-
berg (b.1915) who acknowledge it, and those like Chomsky, who do not. The
form this influence took is examined in the following sections.
In sum, Jakobson's impact was to resolve the cognitive dissonance be
tween the arbitrary and the natural as it existed in the programme for structur
alism announced by the CLG, and to resolve it unambiguously on the side of
the natural, by claiming that finally every linguistic structure, no matter how
apparently arbitrary, is somehow shaped or determined by the human language
faculty. Jakobson's late work in particular suggests a desire to return to a ho
lism of language, literature, philosophy, neuroscience and physics that would
have been as much at home in the Greek 5th century as in the European and
American 20th. We are still awaiting its realization at the dawn of the 21st.
given or already attained — serves to fix the parameters of UG, providing a core
grammar, guided perhaps by a structure of preferences and implicational relations
among the parameters of the core theory. If so, then considerations of markedness
enter into the theory of core grammar. (Chomsky 1981: 7)
So independent is core grammar that the existence of markedness — which in
heres within the parameters of UG itself — is now stated as merely possible.
Chomsky goes on to introduce 'periphery' and its rationale (ibid., pp. 7-8):
But it is hardly to be expected that what are called "languages" or "dialects" or
even "idiolects" will conform precisely or perhaps even very closely to the systems
determined by fixing the parameters of UG. This could only happen under ideal
ized conditions that are never realized in fact in the real world of heterogeneous
speech communities. Furthermore, each actual "language" will incorporate a pe
riphery of borrowings, historical residues, inventions, and so on, which we can
hardly expect to — and indeed would not want to — incorporate within a princi
pled theory of UG. For such reasons as these, it is reasonable to suppose that UG
determines a set of core grammars and that what is actually represented in the mind
of an individual even under the idealization to a homogeneous speech community
would be a core grammar with a periphery of marked elements and constructions.
Of particular interest here are the 'commonsense' examples used to illustrate
periphery: borrowings, historical residues, inventions. Each has to do, not with
the intrinsic nature of an element or its place or function within the system, but
with its origin, a fact we shall consider again further on. Chomsky then reverts
to equating periphery with irregularity in order to refute this very equation:
[O]utside the domain of core grammar we do not expect to find chaos. Marked
structures have to be learned on the basis of slender evidence too, so there should
be further structure to the system outside of core grammar. We might expect that
the structure of these further systems relates to the theory of core grammar by such
devices as relaxing certain conditions of core grammar, processes of analogy in
some sense to be made precise, and so on, though there will presumably be inde
pendent structure as well: hierarchies of accessibility, etc. (Chomsky 1981: 8)
Note in particular the reference to analogy as organizing what is not internally
systematic, with its resonance of the neogrammarian doctrine of exceptionless
sound laws. This is a turnaround from Chomsky & Lasnik (1977), which set
aside periphery as a kind of tribal homeland for unrestructured standard-theory
transformationalists. But indeed, even a grammar which allows unprincipled
rules only in its periphery does not help much toward explanatory simplicity.
On the other hand, giving periphery a structure derivative of core raises the
problem of adequately distinguishing between the two. What would be irregu
lar about irregular morphology if it is related to regular morphology in system
atic ways? It just means that the rule system is complex. But we have posited
that it is simple. And after all, the whole point of periphery was to be a limbo
for the apparently unsystematic.
Chomsky expressed dubiety about the model as early as 1979-1980, in
the interviews published as On the Generative Enterprise:
INVISIBLE HIERARCHIES 187
I am sure that the periphems and the theory of markedness have structure, but I just
do not have any good ideas about what it should be. I suggested something in the
Pisa paper which I do not think is correct, viz. relaxing some of the conditions of
core grammar. Maybe that is somewhat the right idea, but I do not really feel that
there is any evidence. I do not even think it is clear whether we should make a
sharp distinction between core and periphery. Maybe these are more closely re
lated notions of some sort. But whatever the answer to that is, it is obviously going
to be the case that learning of exceptions is a highly structured matter, you cannot
learn just any class of exceptions. (Chomsky 1982: 108)
'Exceptions' indicates that Chomsky was still thinking of periphery as the do
main of what cannot be brought under the constraint of highly structured, uni
versal rules. In the next passage he will equate it with rules that are not pro-
ductive, i.e., currently in operation:
Or take the Vowel Shift; these are obviously not productive rules, but they are
rules that might very well organize the memory. They impose strong patterns and
give principled ways of organizing materials that are otherwise quite chaotic. I
would imagine that if they turn out to be real, as I suppose they are, it is because
they provide a kind of organization of non-core areas, where core-areas might be
the productive rules. And these would have to do with ways of imposing a tight
and interesting organization on systems. (Ibid., p. 109)
Yet despite these doubts, by Knowledge of Language (1986a) Chomsky was
still proclaiming the distinction as central to his theory. His examples of pe
riphery this time included irregular morphology — precisely what he and Las-
nik had used as a metaphor for the concept itself back in 1977.
The systems called "languages" in common sense usage tolerate exceptions: ir
regular morphology, idioms, and so forth. These exceptions do not fall naturally
under the principles-and-parameters of UG. Suppose we distinguish core language
from periphery, where a core language is a system determined by fixing values for
the parameters of UG, and the periphery is whatever is added on in the system ac
tually represented in the mind/brain of a speaker-hearer [...T]here are, no doubt,
significant regularities even in departures from the core principles (for example, in
irregular verb morphology in English), and it may be that peripheral constructions
are related to the core in systematic ways, say, by relaxing certain conditions of
core grammar [...]. (Chomsky 1986a: 147)
The last clause, suggesting that peripheral constructions are related to the core
in systematic ways, is the very thing he did not think was correct in the earlier
interview.
Chomsky's diffidence concerning the core-periphery distinction would be
brought to a head by Jean-Yves Pollock's 1989 article "Verb Movement, Uni
versal Grammar, and the Structure of IP". This article was a watershed for sev
eral reasons, not the least of them being that it signalled the end of the first
phase of GB theory, characterized by a fairly stable degree of economy even
through the somewhat complexifying reforms introduced by Chomsky in his
book Barriers (1986b). Pollock demonstrated from data on adverb and nega-
188 LIMITING THE ARBITRARY
tive position in English and French that IP, the Inflection Phrase (the equivalent
of older S or Sentence) really consisted of at least two subcomponents, TP and
AgrP, where T and Agr denote elements that convey tense and agreement. The
fact that the most basic unit of generative grammar had to be complexified us
ing data not from any exotic language, but from the two languages already most
thoroughly studied for over two decades, signalled that the minimalist aesthetic
of GB might not long endure.
As it transpired, two of the most prominent (and conflicting) proposals to
appear in the wake of Pollock's article agreed on the need for a major reforma
tion of 1980s ideas on core and periphery. Baker (1991) argued that the concept
of core grammar embodies a form of the explanation-description paradox dis
cussed above. In the face of irreconcilable differences in language structures,
one either complexifies core grammar to account for both structures, thus com
promising its entire raison d'être, or one abandons those structures to the pe
riphery, thus reducing the scope and importance of core grammar. Baker argues
for the latter course:
[T]here is a serious risk [...] in pressing too hard to view every particular linguistic
fact as epiphenomenal, that is, to insist on believing that everything follows from
general principles. The danger is that we will develop a theory of core grammar
containing many principles that do not really belong there, a theory that will re
quire qualification and augmentation in virtually every encounter with new lin
guistic data. Such extraneous principles could, by their very presence within our
theory of core grammar, obstruct our view, so to speak, and thus inhibit the dis
covery of those general principles that truly belong to the core.
In this connection, one possibility that strikes me as deserving of further study
is that the parametric choices of core grammar are both simpler in character and
fewer in number than many current accounts would suggest. (Baker 1991: 427)
This is reminiscent of Chomsky & Lasnik's 1977 description of "core grammar
with highly restricted options, limited expressive power, and a few parame
ters". In effect Baker was advocating a return to the earlier mode of generative
grammar which viewed the determination of language-specific rules as a
worthwhile effort; he would change the core and periphery model by reducing
the importance attached to the core both in how much work it is expected to do
and how much work by linguists it should generate.
Chomsky (1991), however, endorsed Pollock's position — not altogether
unsurprisingly, given his earlier ambivalence about core and periphery — and
tried to show that it does further the project of theoretical economy despite ap
pearances to the contrary. At the same time, he was sufficiently dissatisfied
with late 1980s GB theory to sketch out a further evolved version; and he fi
nally makes clear that the core and periphery model cannot stand:
The core-periphery distinction, in my view, should be regarded as an expository
device, reflecting a level of understanding that should be superseded as clarifica
tion of the nature of linguistic enquiry advances. (Chomsky 1991: 449, n.3)
INVISIBLE HIERARCHIES 189
Indeed, in the revised model which Chomsky here sketches out in a preliminary
way, the basic purpose for which the core-periphery distinction was designed is
transferred to another dichotomy, already needed in his model for independent
reasons: that between lexicon and 'grammar' (or as Chomsky now begins to
call it, the 'computational system'), the long pedigree of which has been traced
in Chap. 4.
It has been suggested that parameters of UG relate, not to the computational sys
tem [i.e. syntax in a broad sense (including phonology)], but only to the lexicon.
We might take this to mean that each parameter refers to properties of specific
elements of the lexicon or to categories of lexical items — canonical government,
for example. If this proposal can be maintained in a natural form, there is only one
human language, apart from the lexicon, and language acquisition is in essence a
matter of determining lexical idiosyncracies. Properties of the lexicon too are
sharply constrained, by UG or other systems of the mind/brain. If substantive ele
ments (verbs, nouns, and so on) are drawn from an invariant universal vocabulary,
then only functional elements will be parameterized. The narrower assumption ap
pears plausible; what follows is consistent with it. (Ibid., p. 419)
In terms of systemic economy this was certainly an improvement over the older
model, which demanded at least two levels of abstract pre-grammar (un-
parameterized UG and parameterized core grammar) both of which required
further experiential development to become functional, as well as the acquisi
tion of a lexicon. The new model had just one level of pre-grammar, fully uni
versal, only one part of one component of which — the functional elements of
the lexicon — had parameters needing to be set by experience.
The 'lexicalist' hypothesis had long held that lexicon is much more than
an inventory of items to be plugged into a pre-set syntactic frame, but that these
items themselves set the frame through the syntactic specifications they contain
along with their phonological and semantic specifications. Resistance to this
notion had weakened over the years, with an acceleration in the GB period, as
lexical specification became a convenient means of shunting aside difficult
data that could not convincingly be relegated to periphery, and that would un
duly complicate core grammar. No one had ever denied that lexical items have
to be learned by children on a piecemeal basis; and it is clear that at least some
grammatical specifications (e.g. transitivity of verbs) are part of what children
have to learn about lexical items. This cannot be built out of the system. Why
not, then, transfer all the work of parameter setting — all that which distin
guishes one language from another — to the part of the lexicon from which
such work cannot be eliminated?
The solution to the paradox of how language must be so systematic yet is
so unsystematic is therefore no longer that it consists of a systematic core and a
less systematic periphery, that overlays another distinction of lexicon and com
putational system; but simply that the explanation for the paradox lies within
that second distinction itself. The computational system and much of the lexi-
190 LIMITING THE ARBITRARY
con is universal, hence the systematicity; some of the lexicon consists of uni
versal parameters to be set by experience, hence the differences among lan
guages. One could continue to speak of core and periphery within the latter
domain, but it was clearly a secondary, epiphenomenal effect, not a primary
foundational division.
To summarize what the problem was that Chomsky came to perceive
with the core-periphery model, let us review the various descriptions of periph
ery from 1977 to 1986:
(a) marked; non-optimal in terms of evaluation metric (1977-1981)
(b) what is added on after fixing of core grammar (1986)
(c) syntactic analogue of irregular verbs (1977)
(d) not determined by parameters of UG (1981, 1986)
(e) examples: borrowings, historical residues, inventions (1981)
(f) relaxing certain conditions of core grammar (1981, ?1982?, 1986)
(g) processes of analogy (1981)
(h) exceptions, non-productive rules (1982)
(i) examples: hierarchies of accessibility (1982)
(j) examples: irregular morphology, idioms (1986)
Some of these were attempts at technical descriptions, others metaphors, others
common-sense examples used to describe periphery in a non-technical way.
Their very different bases are striking: (a) pertains to the status of a structure
within the language system; (b) to the stage in the language acquisition process
at which the structure is added; (c) to the relationship of periphery as a whole
to the core component; (d) and (g) to the means by which a structure is ac
quired — natural-universal or psychological-historical; (e) to the historical
source of a structure; (f) and (i) to how periphery is organized; (h) and (j) to the
systematicity of a structure. We can organize these into the four categories
shown below, with the working labels universality, typology, historicity, and
systematicity.
universality (b d f)
typology (c h j) systematicity (a i)
historicity (e g)
That these sundry considerations could all be lumped together was never ques
tioned, simply assumed — largely on the basis of invoking markedness. Jakob-
son's contention that markedness, the status of a structure within the language
system, is related to stage of acquisition, and to degree of universality, was
widely accepted; and Chomsky of course maintained that stage of acquisition
and degree of universality are directly related through UG. But when we look
closely at the version of markedness that underlies GB, we find that it is fun
damentally different from the classical Praguean version. This becomes clear
INVISIBLE HIERARCHIES 191
11. Although Greenberg's work was directly in line with Chomsky's both in the overall pro
gramme of a search for the universal and in the more specific result of collapsing traditional
distinctions into mega-categories like government, Chomsky and most other generative lin
guists have from the beginning refused to admit a meaningful connection between the two pro
grammes, arguing that Greenberg dealt with mere surface structure phenomena and drew
meaningless conclusions from statistical tendencies. For the Bloomfieldians, whose main con
cern had been to avoid creating pseudo-universals by imposing the categories and structures of
their native language on the very different languages they were investigating, and for European
structuralists generally, the empirical nature of Greenberg's work and that of others in the same
vein, including Bernard Comrie and John Hawkins, made it much more appealing than Chom
sky's, which appeared to admit of no disproof. See the comments below on recent efforts to
ward reconciling the two programmes.
194 LIMITING THE ARBITRARY
more openly, again in parallel with Chomsky and Halle's development of the
theory of markedness in their phonological analysis of English.
Several prominent members of the next generation of researchers inspired
by Greenberg, together with some of their own students, contributed episte-
mological reflections on the enterprise to an important book edited by John
Hawkins (1988). Rather than a unified theory, the book presented an extraordi
narily wide range of ideas on the connectedness of language to things outside
itself, the general condition which all the contributors agreed (implicitly if not
explicitly) to be the fundamental explanation for why universals of the type in
vestigated by Greenberg should exist at all. In several of the contributions, uni
versal-typological research appears to have merged with iconicity studies (see
below) to the point that no clear line of demarcation between them is apparent.
In addition, some of the papers pursue the idea that the search for universals of
language is fundamentally tied to the question of universals of perception in
the domain of physics, which has been actively debated at least since the time
of Kant (see Hatfield 1990), and which also has significant roots in Greek
thought, as we have seen. For Heraclitus, and for Plato, perception was the op-
posite of reality; it was the point at which reality as captured and transfigured
by the individual will. For Locke and Condillac, however, it is the contact be
tween the organs of sense and perception and the external world that provide
the natural starting point of language.
The following chart attempts to organize the main categories of explana
tions that have been offered for language universals.
of the most intense contemporary theoretical disputes. Not all the labels denote
theoretically incompatible positions; many combined explanations exist.
The linguistic 'mainstream' has been dominated for the last 40 years by the
human-psychological-individual-specifically linguistic line, particularly through
the generativist conception of innate Universal Grammar. The idea that language
might be traced to human mental structures was shared by Jean Piaget (1896-
1980) and other developmentalists, who however denied that these structures
were innate or specific to language. Rather, Piaget believed that structures
adapted to general intelligence could account fully for language development.
Meanwhile, those working in the tradition of empirical studies of language
universals founded by Greenberg in the late 1950s began to form common cause
with some of these developmentalist approaches. By the late 1980s attention was
being directed toward the possibility, suggested by contemporary work in
experimental psychology and anthropology, that linguistic universals — common
features, parameters, and especially implicational patterns — exist because all
human beings have the same organs of perception and the same innate mental
structures for the organizing perceptual data. The range of existing language
structures would be limited by the fact that all speakers use the same sensory
organs to perceive an essentially identical reality, and the same general cerebral
structures of cognition to organize incoming perceptual data (see, e.g., Lee 1988,
Dunlea 1989). Since the universality of sensory organs and basic cerebral
structure is not contested, and the existence of some innate mental organization
for perception is assumed by those carrying out the perception research, the
appeal for linguistics is clear. These widely acknowledged, solidly physical
structures could be taken up at no additional explanatory cost, as against the
massive theoretical cost of another innate faculty specifically for language.
These explanations combine the best elements of several categories from
the chart above: physical universe or world structure as mediated by human
physiology yields psychological universals of a general cognitive sort. But from
the mid-70s perceptual cognition was undergoing a revolution of its own, due
largely to the work of psychologist Eleanor Rosch (see Rosch & Lloyd 1978) and
her research into the process of category formation, later extended to the analysis
of visual perception by Marr (1982). As much as anyone since Wittgenstein,
Rosch upset the rule-based tradition of analyzing categories as sets characteri-
zable by necessary and sufficient conditions. She offered data from several cul
tures suggesting that in natural languages categories are not formed deductively
from rules, but grow outward from around a prototype. Rosch found it to be uni
versally true that the members of a category include better and worse exemplars,
16. Chomsky attempted to contradict Piaget's views head-on (see Piatelli-Palmarini 1980); but
by that time other projects similar in spirit to Piaget's were underway in various quarters (see
further Joseph 1999e: 56-58).
196 LIMITING THE ARBITRARY
so that for instance a robin is a more prototypical member of the category bird
than is an ostrich. Yet the precise identity of the prototype, and to an even greater
degree how the category forms around it, differ from culture to culture — and
this constellation of facts held the promise of containing, if not explaining, cul
tural universals and differences in categorization at both the perceptual and the
linguistic level. Rosch's and Marr's work opened a new path into a fundamental
epistemological schism in the analysis of perception. Philosophy and science
have for centuries been divided on the question of whether perception itself is
'natural' or 'normative', in the preferred terms of Hatfield (1990), a history of
this dispute focusing on the mid-19th century work in physiological optics of
Hermann von Helmholtz (1821-1894), which succeeded in reestablishing the
normative view of perception in mainstream science after its eclipse in the late
18th century by Kant's naturalism (even though, in Hatfield's view, Helmholtz
failed to understand Kant's central distinction between transcendental and em
pirical arguments). Yet with the subsequent academic institutionalization that
would follow, the schism would be institutionalized as a division of labour: psy
chology became the domain of empirical and natural study, while matters of
normativity were left to the philosophers, a division that still largely holds.17
The attempts by Greenbergian universalists to explain the trends and fea
tures disclosed by their empirical research has connected them more intimately
than ever to the entire history described in this book, a history whose lessons for
their work few of them, however, appear as yet to appreciate. It has also allowed
them to move away from the metaphysical mode in which Jakobson and Green-
berg sometimes treated such phenomena, certainly a crucial move in the quest for
contemporary scientific legitimacy. However, this same quest has driven a good
number of them to concentrate on seeking resolutions with Chomsky's genera-
tivism. One thing seems clear: any such resolution will hinge upon one or both of
these approaches making a significant shift in its conception of the natural basis
of language (something Greenbergians appear readier to do than Chomskyans).
The potential for this may well exist, especially in light of their common Jakob-
sonian heritage.
Iconicity studies
A third strain of thought inspired by Jakobson's ideas on the mark, and fu-
17. But by reinvigorating the normativistic dimension within empirical research, Rosch, Marr,
and company have lent a strong impetus to the development of connectionist models of percep
tion. The kind of processes Rosch and Marr find to be at work in the mind/brain do not at all fit
with the elegant rule construction of the symbolists, but have much more of the soft, fuzzy look
of weight distributed, self-teaching, recursive connectionist networks. And this has both neces
sitated and spawned a major ongoing change in the scientific aesthetic: the hard and dry of the
computer is gradually giving way to the soft and wet of the brain as the ideal of scientific
knowledge. The most economical explanation is no longer assumed to be the correct one; the
brain, after all, operates with infinite redundancy and complexity. Occam has lain aside his
razor and is growing a beard.
INVISIBLE HIERARCHIES 197
elled by the impetus of both Chomsky and Greenberg, and in some cases devel
opments in semiotics, has given to the group of approaches that I am grouping
together as 'iconicity studies'. From the mid-1970s through the mid-80s, research
programmes with the names Natural Phonology (see Donegan & Stampe 1979),
Natural Generative Phonology ([Bybee] Hooper 1979), Natural Morphology
(Mayerthaler 1981, Wurzel 1984), and Natural Syntax (Haiman 1985) were pro
claimed, with the specific aim of analyzing how human physiological and mental
structures determine and explain the existence of universal features in languages.
(For a further selection of work in this vein, including in historical linguistics, see
Haiman ed. 1985, Landsberg ed. 1995, Simone ed. 1995; and for a historical
overview, Swiggers 1993). Something like the Chomskyan core-periphery di
chotomy is generally assumed: 'crazy' rules in phonology (e.g. Celtic lenition)
are recognized as existing, but are predicted to be unstable over time, acquired
late by learners and lost early in aphasia, etc. The same is true of 'capricious'
structures in morphology and syntax, such as the fact that in Standard English the
only form of the present indicative non-modal verb to receive an inflectional
mark is the third person singular — I/you/we/they go, you go vs. he goes — yet
the third person singular is conceptually the unmarked person par excellence.
Adapting the terminology created by Charles Sanders Peirce, Jakobson
proposed that language structure (conceived on a purely formal level) is iconic of
meaning. As examples he cited the fact that in no language does the regular sin
gular form of nouns contain more sounds than the regular plural form, whereas
the opposite is often the case: hence the greater length of the noun is iconic of the
difference in number expressed by the singular plural distinction, quite apart
from whatever 'conventional' sounds happen to be involved. He noted as well
that all languages have an iconicity between the order of actions related in the
discourse and the order of real-world events which the discourse is expressing:
the classic example Veni, vidi, vici reproduces linguistically the order in which
Caesar actually came, saw, and conquered, and this iconicity appears to hold
across languages.
But in all such cases exceptions are bound to appear. For example, the pro
posed singular-plural iconicity fails to hold for those Old French masculine
nominatives which take a final -s as a mark of the singular, and no mark for the
plural (thus murs "wall", mur "walls"); or those Russian noun conjugations
which have as their shortest form (the bare stem) the genitive plural. One is
18. The fact that the proponents of these theories chose the term natural is not a bow to the
Sophists, but a reflection of the fact that in the 1970s the word reacquired an extremely positive
connotation, and shed most of the sense of being in opposition to human, which it possessed in
the minds of some at least since the 17th century. By the mid-'80s, as 'natural' ceased to be a
buzzword in advertizing, it simultaneously lost its éclat in linguistics — as good a case as one
might hope to find of how science, even the science of language, is influenced by linguistic
fads, in this case a fad prompted mainly by individuals (pace Saussure) working on Madison
Avenue.
198 LIMITING THE ARBITRARY
obliged to 'peripheralize' such cases somehow; for the most part this has been
done by classifying them as historically unstable. Thus it becomes a piece of evi
dence in support of this theory that the Old French nominatives did not survive
beyond Old French, but gave way to the accusative forms, which show the 'uni
versal' distribution {mur 'wall', murs 'walls'). As in the French 17th century,
what is arbitrary in language is associated with what is capricious, and it is un
derstood that, although a linguistic community could in principle adopt any rule,
no matter how capricious and 'crazy', over the long haul nature will separate the
wheat from the chaff, and eliminate any element of the language that is not
grounded in the principles of iconicity and markedness. In the most extreme for
mulation I know of, that of Shapiro (1983), the entire structure of language is
configured as a network of markedness patterns, and exceptions are treated as
'markedness reversals', apparently as needed, without recourse to any principle
that might limit them.
Iconicity studies have had considerably less impact than generativism and
quite a few other approaches. But they are significant in the context of this book
for what they represent concerning 20th-century treatments of language in terms
of nature and its opposites. At least in the realm of grammar/syntax, they are at
the far end of the pendulum swing away from what the CLG at least claimed to
maintain about the arbitrariness of linguistic signs. If structuralists were Cratyli in
Hermogenes' clothing, then by the time of iconicity studies, linguists had got na
ked. This is not necessarily recognized to be so, even within iconicity studies:
most of its practitioners would probably continue to maintain that lexicon, at
least, shows a fundamentally arbitrary relationship between sound and meaning,
and that their enterprise is directed at tracing out the limits of the arbitrary in the
realm of grammar, syntax, and discourse structure. Were that all there was to it,
iconicity studies would be no different from any other area of endeavour classifi
able by the name of linguistics (note for instance that the CLG describes the
whole activity of linguistics as limiting the arbitrary). But in maintaining the view
that grammar, which has usually occupied the 'natural' pole since the compro
mise of Varro, not only can but does reproduce the structure of the real world,
and that anytime it fails to do so, this is the result of an arbitrary human choice
that will in the course of time 'naturally' give way, iconicity studies can truly be
said to resurrect something very close to the Sophistic doctrine of physis; and all
the objections of Plato can, mutatis mutandis, effectively be raised against it.
Optimality Theory
I shall conclude with a brief consideration of an approach to linguistic
analysis that has garnered a great deal of attention in the last few years, and is
being taken by a growing number of linguists as a route out of Chomskyan GB/
Minimalism as it loses the clear sense of having a theoretical centre that for many
was its prime attraction from the 1960s through the 1980s. Optimality Theory, or
INVISIBLE HIERARCHIES 199
OT, has its origins in a paper on phonology given by Prince & Smolensky in
1991. Since then attempts have been made to apply it to virtually every aspect of
the study of language, a number of which are covered in the collective volumes
edited by Archangeli & Langendoen (1997) and Barbosa et al. (1998). For syn
tax, OT is generally seen as connecting to issues of economy raised by Chomsky
after 1989, but approaching them in a radically different way.
The thrust of OT is that innate Universal Grammar consists of a set of con
straints, each of which can be violated, and which are ranked differently in differ
ent languages. Even within a language, the constraints can be ranked differently
for different speakers — this is the basis of the OT account of language variation.
The constraints themselves are 'markedness statements' (Archangeli 1997: 17).
Markedness is here defined in terms of the continuum between language-
universal and language-specific properties, "with completely unmarked proper
ties being those found in virtually all languages and extremely marked properties
being found quite rarely" (ibid., p. 2).
A very simple example showing how OT differs from the GB approach is
the following, drawn from Speas (1997: 180ff.). GB has a principle of 'Full
Interpretation', according to which there can be no superfluous symbols in an
output representation — i.e., everything in every utterance must be fully inter
pretable phonetically and semantically. Among the exceptions to this are 'ex
pletives' such as the it of English It will rain tomorrow or the there of There
are three cats on the porch. In GB Theory, where principles are treated as in
violable, an ad hoc 'hedge' must be posited to account for cases such as this:
expletives are deleted before semantic interpretation takes place, but are nev
ertheless realized phonetically. "In other words", writes Speas (1997: 181),
"there actually are superfluous symbols in a representation, when those sym
bols are necessary to fulfil some other grammatical principle", in this case the
need for the verbs to assign nominative case.
In OT, on the other hand, FULL INTERPRETATION can be treated as a con
straint that is, like all OT constraints, assumed to be violable. It would co-exist
with another constraint called SUBJECT, constraining all clauses to have a
grammatical subject. In languages with expletives such as English and French
(where the equivalent of It will rain tomorrow is ll va pleuvoir demain with
expletive il), SUBJECT outranks FULL INTERPRETATION. On the other hand, in
languages without expletive subjects — such as Italian, where the equivalent of
It will rain tomorrow is
Pioverà domani
rain-FUT tomorrow
or the Uto-Aztecan language Yaqui, where it is
200 LIMITING THE ARBITRARY
yooko yuk-ne
tomorrow rain-FUT
— the FULL INTERPRETATION constraint outranks the SUBJECT constraint. For
Speas, there is an advantage here in that OT is up front about the fact that its
'constraints' are never inviolable, whereas GB keeps up the pretence that its prin
ciples, being innate and physical, are absolute, while (hypocritically, some might
charge) building in hedges (including the whole realm of 'periphery') to deal
with the inevitable exceptions.
The objection from the other side is that OT, like some earlier attempts at
markedness-based analysis (see the preceding section, especially the reference to
Shapiro 1983), has bought its more satisfying descriptive capacity at the expense
of explanatory power. It is not clear that any linguistic phenomenon will pose a
genuine problem for OT to explain; it is simply a matter of determining plausible
constraints, hopefully ones that are not ad hoc or language specific, and ordering
them so as to come out with the desired answer. A case that is unusual will turn
out to be marked, one with analogues in numerous other languages will be un
marked.
The challenge for OT is to escape this inherent circularity, and at least some
of those who comprehend the problem are trying to do this through restricting the
positing of constraints and grounding recognized constraints in psycholinguistic,
if not neurolinguistic, 'reality'. The problem now, as it was in Plato's time, is
knowing what that reality is independently of language. And like Saussure, the
adherents of OT want to be able to deal with the observable fact that human lan
guages (let alone human speech) show such an extraordinary range of structural
configurations as to suggest that the make-up of a language is fundamentally ar
bitrary, while at the same time to analyze it as a system so rigorous and autono
mous as to demand to be approached from the point of view of limiting the arbi
trary.
This is a profound paradox, which needs to be recognized as such before
there will be any hope of resolving it. One thing certain is that resolution will not
come through any new reworking of linguistic naturalism; this will only help the
paradox to endure. I believe that any hope for progress depends upon a thorough
reconsideration of what we understand language to be, particularly our under
standing of it as a 'system' with its 'real' (and somehow homogeneous) existence
in the unconscious mind of 'the individual'. To be sure, reformulating these con
ceptions will not in itself provide answers. It may in the end lead us only to a new
aporia, or a new version of our age-old aporia, or even to new invisible hierar
chies, ranks of angels we shall embrace as our own because we suppose we have
discovered them, even if all we have done is to give them new names.
AFTERWORD
The alert reader, and especially the alert disgruntled reader, will have no
ticed that the second half of this book has in common with Plato's Cratylus the
fact that it asks more questions than it answers. The criticism will be levelled
that it throws into question various versions of (what I take to be) linguistic
naturalism without offering a clear alternative to put in its place — unless that
alternative is to do the history of linguistics rather than linguistics as such. That
is not in fact the alternative I am suggesting, my belief being that the 'history'
and 'doing' of linguistics are, or ought to be, indistinguishable from one an
other. It would not therefore be possible to do one in place of the other.
To the charge that I have not here offered a clear alternative, I am liable,
and can only impose upon readers the unfair burden of consulting other work
of mine on the notion of 'standard' languages, language ideologies and lan
guage and identity, which lay out various avenues for the scientific investiga
tion of human language that attempt to pass beyond naturalistic conceptions.
Yet I am not convinced that they have so far succeeded in passing beyond,
rather than merely by-passing. The historical and conceptual enquiry carried
out in this book is, I believe, a necessary step before such investigation can
genuinely be achieved.
Neither the physis-nomos debate nor any version of the naturalism-other
dichotomy has ever involved language theory alone. There have always been
significant political and cultural connections of many sorts. This continues to
be true in our own time, and in discussing this project with colleagues from
other disciplines, I have found many of them to be aware of facing the same
problem of coming to grips with, and getting past, some notion of the natural
and its opposite, be it the unnatural, the cultural, the conventional, the arbi
trary. Indeed, some believe they have succeeded in getting past this binary di
vision, by virtue of their recognition that our conceptions of nature and the
natural are always no less 'constructed' than are its opposites. Some of them
have asked me whether, in continuing to work on a project aimed at under
standing the nature-convention dichotomy in language theory by focusing on
the Cratylus, I have not in fact contributed to keeping the dichotomy alive.
Why not, they ask, simply pass beyond it? Why, they seem to imply, continue
to dwell in the dark, when you can see ahead to the light (in which all of them
are standing)?
202 LIMITING THE ARBITRARY
My response is that they may be the ones who are continuing to propa
gate the nature vs. x divide, by claiming to have passed beyond it, when in fact
the entire Western framework of perceiving and organizing information, of
thinking and theorizing — in short of knowledge — is founded upon that di
vide. Can one pass beyond it at this point in history? Or can one simply stop
thinking about it, and thereby allow it to kick back in as the default mode of
knowing? If the latter is the case, then by refusing to grapple with the problem,
to roll around in it for a few years (or decades or centuries), we are allowing
superficial changes to fool us from recognizing us a much more fundamental
continuity beneath.
Other, more conservative colleagues claim that it is naive to think that
one could ever pass beyond a division of the world into nature vs. x, and that it
is, if not dangerous, at least a waste of time, to suggest otherwise. My answer
to them is that such is the case if and only if 'nature' is itself a natural and not a
conventional concept: in other words, their denial of the problem begs its es
sential question. But it is inconceivable to me that our conceptions of the natu
ral are not at least to some degree the products of convention, the principal evi
dence being the same as that cited in the Cratylus: the fact that those
conceptions differ among different peoples and different individuals, and that
they change over time. And that is as true with regard to the physical universe
as to human behaviour: again, there is no scientific truth that is not falsifiable,
so that any 'natural law', no matter how seemingly grounded in the reality of
the universe, will eventually be shown up as at least partly conventional, so
long as science continues to pursue progress in the form of revised, and appar
ently deeper, understandings.
It would be foolhardy to imagine that we can genuinely pass beyond the
nature vs. x divide before having understood all the forms it has taken and
continues to take in our knowledge and thought. That will be the preliminary
work to the still deeper understanding of, first, why we have come to think and
know in this way, and second, what thinking and knowing would actually be
like in the absence of this dichotomy or perhaps of all such binary dichotomies.
And then the ultimate goal will be, not to do away with dichotomized thinking,
not to live without nature and convention (or whatever); but something much
harder still, to live both with and without such dichotomies, simultaneously. To
understand how our actions and being can at one and the same time be part of
nature, the opposite of nature, and the conceiving force of nature, which, inso
far as it is not a part of us, is unknowable to us.
And here we have come full circle, for that was precisely the central
problematic of Plato, who in the Cratylus tries to determine whether the study
of language can be the path to such 'unknowable' knowledge. His verdict is
that it cannot. Again, the Cratylus provides the strongest possible argument
against those who claim that we can pass simply and easily beyond a division
AFTERWORD 203
of the world into nature and convention. Just look at the outcome of Plato's
attempt: not only was the division back in full force two generations later, but
in our own time the dialogue continues to find readers who insist on putting
Plato on one or another side of the debate. Anyone who wants to pass beyond a
conception of the world in terms of the natural and the conventional could do
no better than to study this dialogue together with its history. It is in this sense
that the Cratylus, and the history of language theory generally, have as much to
teach us about the future of Western linguistic thought as about its past.
REFERENCES
Aarsleff, Hans 1982. From Locke to Saussure: Essays on the study of language
and intellectual history. Minneapolis: University of Minnesota Press; Lon
don: Athlone.
Amsler, Mark. 1989. Etymology and Grammatical Discourse in Late Antiquity
and the Middle Ages. Amsterdam & Philadelphia: John Benjamins.
Anagnostopoulos, G. 1971/72. "Plato's Cratylus: The two theories of the cor
rectness of names". Review of Metaphysics 25.691-736.
Archangeli, Diana. 1997. "Optimality Theory: An introduction to linguistics in
the 1990s". Archangeli & Langendoen 1997: 1-32.
Archangeli, Diana & D. Terrence Langendoen, eds. 1997. Optimality Theory:
An overview. Malden, Mass. & Oxford: Blackwell.
Arens, Hans. 1975. La lingüística: Sus textos y su evolución desde la an-
tigüedad hasta nuestros días. Transl, by José María Díaz-Regañón López.
Madrid: Gredos.
Arens, Hans. 1980. "Verbum cordis: Zur Sprachphilosophie des Mittelalters".
Historiographia Lingustica 7.13-27.
Aristotle. 1960. Metaphysics. Transl, by Richard Hope. Ann Arbor: University
of Michigan Press.
Auroux, Sylvain, ed. 1990. Histoire des idées linguistiques, vol. 1. La nais-
sance des métalangages en Orient et en Occident. Bruxelles: Pierre
Mardaga.
Bailey, Cyril, ed. & transi. 1926. Epicurus: The extant remains. Oxford: Clar
endon Press.
Baker, C. L. 1991. "The Syntax of English not: The limits of core grammar".
Linguistic Inquiry 22.387-429.
Baratin, Marc. 1982. "L'identité de la pensée et de la parole de l'ancien
stoïcisme". Langages 16, no. 65, 9-21.
Baratin, Marc & Françoise Desbordes. 1981. L'analyse linguistique dans
l'antiquité classique, I: Les théories. Paris: Klinksieck.
Barbosa, Pilar, Danny Fox, Paul Hagstrom, Martha McGinnis & David Peset-
sky. 1998. Is the Best Good Enough? Optimality and competition in syntax.
Cambridge, Mass.: MIT Press.
Barney, Rachel (1998) "Socrates Agonistes: The case of the Cratylus etymolo
gies". Oxford Studies in Ancient Philosophy 16.63-98.
Barsky, Robert. 1998. Noam Chomsky: A life of dissent. Cambridge, Mass.:
MIT Press.
Baxter, Timothy M. S. 1992. The Cratylus: Plato's critique of naming. Leiden:
206 LIMITING THE ARBITRARY
E. J. Brill.
Bentham, Jeremy. 1843. The Works of Jeremy Bentham. Ed. by John Bowring.
11 vols. Edinburgh: William Tait; London: Simpkin, Marshall & Co.
Bloomfield, Leonard. 1933. Language. New York: Henry Holt & Co.
Bloomfield, Leonard. 1939. "Menomini Morphophonemics". Études pho-
nologiques dédiées à la mémoire de N. S. Trubetzkoy, 105-115. Prague:
Cercle Linguistique de Prague. (Repr. in A Leonard Bloomfield Anthology
ed. by Charles F. Hockett, 351-362. Bloomington: Indiana University
Press, 1970.)
Bopp, Franz. 1816. Über das Conjugationssystem der Sanskritsprache [...].
Frankfurt am Main: Andreäische Buchhandlung.
Botha, Rudolf P. 1989. Challenging Chomsky: The generative garden game.
Oxford: Blackwell.
Buridant, Claude. 1984. "Saussure et la diachronie". Travaux de Linguistique
et de Litterature 22.19-51.
Burkert, Walter. 1970. "La genèse des choses et des mots: Le papyrus de Der-
veni entre Anaxagore et Cratyle". Les Études philosophiques 25.443-455.
Burnouf, Eugène. 1825a. Review of Franz Bopp, "Vergleichende Zer
gliederung der Sanskrita-Sprache und der mit ihm verwandten Sprachen.
Erste Abhandlung [...]". Abhandlungen der Königlichen Akademie der
Wissenschaften zu Berlin, Philosophisch-historische Klasse 1825, 117-148.
Journal Asiatique 6.52-62, 113-124.
Burnouf, Eugène. 1825b. Review of Franz Bopp, Ausfürliches Lehrgebunde
der Sanskrit-sprache (Berlin: F. Dümmler, 1824). Journal Asiatique 6.298-
314,359-371.
Cavazza, Franco. 1981. Studio su Varrone etimologo e grammatico: La lingua
latina come modello di struttura linguistica. Firenze: La Nuova Italia.
Chase, Stuart. 1938. The Tyranny of Words. New York: Harcourt, Brace &
Co.; London: Methuen.
Chevalier, Jean-Claude. 1997. "Trubetzkoy, Jakobson et la France, 1919—
1939". Jakobson entre l'Est et l'Ouest, 1915-1939: Un épisode de
l'histoire de la culture européenne, ed. by Françoise Gadet & Patrick
Sériot. (= Cahiers de L'Institut de Linguistique et des Sciences du Langage
de l'Université de Lausanne, no. 9), 33-46.
Chilton, C. W. 1962. "The Epicurean Theory of the Origin of Language: A
study of Diogenes of Oenoanda, fragments X and XI (W)". American
Journal of Philosophy 83.159-167.
Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton.
Chomsky, Noam. 1959. Review of B. F. Skinner, Verbal Behavior (New York:
Appleton, 1957). Language 35.26-58.
Chomsky, Noam. 1966. Cartesian Linguistics: A chapter in the history of ra-
tionalist thought. New York & London: Harper & Row.
REFERENCES 207
Herder, Johann Gottfried. 1782. Vom Geist des Ebräischen Poesie. Erster Teil.
Dessau: auf Kosten der Verlags-Kasse. (Repr. in Herders Sämmtliche
Werke ed. by Bernhard Suphan, vols. 11-12. Berlin: Weidman, 1879. Engl,
transl., The Spirit of Hebrew Poetry, by James Marsh, 2 vols., Burlington,
Vermont: Edward Smith, 1833.)
Hicks, R. D., transl. & intro. 1925. Diogenes Laertius, Lives of eminent phi-
losophers. (Loeb Classical Library.) London: Heinemann; New York: G. P.
Putnam's Sons.
Hooper, Joan B[ybee]. 1979. "Substantive Principles in Natural Generative
Phonology". Current Approaches to Phonological Theory ed. by Daniel A.
Dinnsen, 106-125. Bloomington & London: University of Indiana Press.
Humboldt, Wilhelm von. 1836. Ueber die Kawisprache auf der Insel Java, vol.
1: Ueber die Verschiedenheit des menschlichen Sprachbaues und ihren
Einfluss auf die geistige Entwickelung des Menschengeschlechts, ed. by
Alexander von Humboldt. Berlin: Royal Academy of Sciences of Berlin.
(Engl, transi., On Language: The diversity of human language-structure
and its influence on the mental development of mankind, by Peter Heath,
Cambridge: Cambridge University Press, 1988.)
Hussey, Edward. 1972. The Presocratics. London: Gerald Duckworth.
Jaeger, Jeri J., Alan H. Lockwood, David L. Kemmerer, Robert D. Van Valin,
Jr., Brian W. Murphy & Hanif G. Khalak. 1996. "A Positron Emission
Tomographic Study of Regular and Irregular Verb Morphology in Eng
lish". Language 72.451-497.
Jakobson, Roman. 1928a. "The Concept of the Sound Law and the Teleologi-
cal Criterion". Engl, version first publ. in Jakobson (1962: 1-2). (Orig.
Czech version in Casopis pro moderni filologii 14 [1928], 183-184.)
Jakobson, Roman. 1928b. "Quelles sont les méthodes les mieux appropriées à
un exposé complet et pratique de la phonologie d'une langue quelconque?"
[countersigned by S. Karcevskij & N. Trubetzkoy]. Premier Congrès In-
ternational de Linguistes: Propositions, 36-39. Nijmegen. (Also published
in Actes du Ier Congrès International de Linguistes du 10-15 avril, 1928,
30, Leiden, 1930. Repr. in Jakobson 1962: 3-6.)
Jakobson, Roman. 1929. Remarques sur l'évolution phonologique du russe
comparée à celle des autres langues slaves. (= Travaux du Cercle Linguis-
tique de Prague, 2.) (Repr. in Jakobson 1962: 7-116.)
Jakobson, Roman. 1931. "Die Betonung und ihre Rolle in der Wort- und Syn-
tagmaphonologie". Travaux du Cercle Linguistique de Prague 4.164-183.
(Repr. in Jakobson 1962: 117-136.)
Jakobson, Roman. 1939. "Observations sur le classement phonologique des
consonnes". Proceedings of the Third International Congress of Phonetic
Sciences, Ghent, 1938, 39-41. Ghent: Laboratoire de Phonétique, Univer
sité de Gand. (Repr. in Jakobson 1962: 272-279.)
210 LIMITING THE ARBITRARY
A atheism 14
Aarsleff, Hans 108n, 114n atomism 13-14,58
Abelard, Peter 106 Augustine, Aurelius, St, of Hippo 103
Abel-Rémusat [see Rémusat] Auroux, Sylvain 10
abstraction 97-98, 100, 110, 125, 146, autonomous language faculty 162
151, 154, 156-158, 160, 162, 165-167
adjectives 105-106,118,151,158 B
advertizing 197n Bailey, Cyril 96-97
agglutinating languages 124 Baker, Carl L(eroy, "Lee") 184, 188
Albertus Magnus 107 Bally, Charles 126n
Alexandrians 100, 111, 142 Baratin, Marc 10, 101n, 104n
Alighieri [see Dante] Barbosa, Pilar 199
ambiguity 99 Barney, Rachel 9, 50n, 52, 57
Amsler, Mark 104-105 Barsky, Robert F. 163
Anagnostopoulos, G. 23n Basic English 120, 158, 160
analogy and anomaly 101-102, 111-112, basic word order 3-4, 111, 193
141-143 Baxter, Timothy M. S. 9, 16n, 22, 23n,
anarchy 23 35n
Anaxagoras 53 beauty 80-81
animals 20, 43-44, 60-61, 66, 69, 99-100, behaviourism 132, 137, 163, 183
115, 120, 162 Bentham, Jeremy 106,119-120
anthropology 195 Bible 149
aphasia 181, 184, 197 Bloomfield, Leonard 132, 182-183, 193
applied linguistics 137 Boas, Franz 182
Arabic 4, 117, 138-139, 167 body 94-95, 98-100, 110, 113, 116, 118,
arbitrariness Introd., Chaps 1-3, 103, 105, 120, 125
108-115, 117-120, 124, 126-127, 129- bon usage 112
134, 137, 139, 144, 152-153, 164-165, Bopp, Franz 121
169, 173, 176, 182-183, 198, 200, Af borrowing 106, 186, 190, 192
terword Botha, Rudolf P. 133n
Archangeli, Diana 199 brain 132-134, 137-139, 156, 161, 164-
archiphoneme 174-177, 181-182 165, 182, 187, 189, 195, 196n
Arens, Hans 9, 103 Buridant, Claude 172n
Ariston of Argos 17 Burkert, Walter 10
Ariston of Athens 7 Burnouf, Eugène 121
Aristotle of Stagira 8, 53n, 78, 93-95, 97- Bybee [see Hooper]
98, 108, 118, 145, 150
Arnauld (or Arnault), Antoine 110-111 C
art 60, 62, 65, 67, 69-70, 72, 141, 144- case grammar 132
145, 148, 170 [see also tekhnē] Castilian [see Spanish]
artificiality 141, 144-145, 147-148, 164- Caucasian languages 176-177
167 Cavazza, Franco 101
Aryan [see Indo-European] Charmides 7
218 LIMITING THE ARBITRARY
physis 8-9, 14-16, 19, 28, 31-35, 38, 44, pre-Socratics 13, 52
59, 68, 71-72, 75, 83-85, 87-89, 93, 97- primal words 58, 61, 71-72, 79, 87
98, 165, 192, 198,201 Prince, Alan S. 199
Piaget, Jean 195 principium divisionis 174-177
Piatelli-Palmarini, Massimo 195n Principles and Parameters [see Govern-
Pinborg, Jan 10 ment and Binding theory]
Pindar 15 printing 147-148
Pinker, Steven 4, 102, 130, 137-139 Priscian of Caesaria 104-105
plants 149-150 private language 20
Plato of Athens (original name 'Aristo- privative opposition 178n
cles') 3, 5-6, 7-10, Chaps 1-3, 93-95, Proclus 14
100, 103, 108, 110, 141, 143, 169, 194, Prodicus of Ceos 17n
200, 202-203 pronunciation 152-153
agonistic display 9, 57 propaganda 157-159
Apology 37n Protagoras of Abdera 14, 20n, 22, 24
Cratylus 5-10, Chaps 1-3, 93-94, prototype theory 195-196
101, 118-119, 126, 141, 150, 166, psycholinguistics 137
169,201-203 psychology 165-166, 192, 195
Crito 2 In publishing 148
Euthydemus 66n Pyrilampes 7
Gorgias 8, 27, 28n, 74n, 84n Pythagoreans 49, 169
Ion 8, 43n
Laws 19, 21n, 23n, 32n, 33n, 35, Q
37n, 41n, 45, 70n, 75n, 77n, 85n questione della lingua 154
Meno 44
name change 17, 29 R
Parmenides 23n, 85 race 96,98, 100, 116, 125
Phaedo 37n Rameau, Jean Philippe 117
Phaedrus 8, 28n, 36, 46n, 83n, 85n reality 25, 28n, 34-35, 38, 75, 81, 87-88,
Philebos 54 126, 164-166, 194,200
Protagoras 37n, 43n, 44 religion 103, 105, 115, 125, 144, 148, 169
Republic 19, 20n, 22n, 36, 37, 53, Rémusat, Jean-Pierre Abel 121
60n, 70n, 82n, 85 Renan, Ernest 124-125
Sophist 8, 60, 66n, 70n, 85n rhetoric 8, 15, 39, 46, 55, 62, 76, 81, 85,
Statesman 30n, 33, 75n 94, 103, 106, 112, 141-142, 164, 169
Theaetetus 17n, 20, 20-21n, 52-53 rhetoricians 28n, 55, 57, 63, 89
Timaeus 74n, 85n Richards, I(vor) A(rmstrong) 158
poetics 8, 14, 76, 85 Rickard, Peter 112n
poetry 141, 144 Rijlaarsdam, Jetske C. 9
poets 37, 39-40, 43, 55, 57, 87, 142 Robin, Léon 36, 78
politics 19, 30, 35, 38, 44, 84, 115, 144, Robins, R(obert) H(enry) 8, 10
153, 156-157 Robinson, Richard 8
Pollock, Jean-Yves 135,187-188 Romance 2, 106, 139, 164
polytheism 14 Rosch, Eleanor 195-196
Pomorska, Krystyna 179n, 180 Rousseau, Jean-Jacques 97,115-119
Pompeius 103-104 Russian 4, 176, 197
positivism 126 Russian formalism 170
predestination 115
prescriptivism 35-36n, 119
INDEX 223
s spelling 152
Sanskrit 120-122, 130, 133-134 Speroni, Sperone 149
Sapir, Edward 131-132n, 136 Stalin, Josef 158
Saussure, Ferdinand (Mongin) de 1, 3, 6, Stampe, David 197
9, 48, 74n, 102-103, 126-133, 136-137, standard language [see under language]
141-142, 153-154, 169-173, 176, 181- Starobinski, Jean 48
182, 192, 197n, 198,200 Steinthal, H(eymann = Chajim, also Hein
langue and parole 74n, 103, 127- rich) 6
128, 142 Stoic language theory 100
Saussure, René de 128 Strawperson, Talbot J. 1-2
Schlegel, August Wilhelm (von) 124 structuralism 6-7, 127-128, 132, 137, 170,
Schlegel, (Carl Wilhelm) Friedrich (von) 180, 182-183, 192, 198
121, 124 stupidity 24-25
Schofield, Malcolm 21n Swiggers, Pierre 197
science 60, 65-66, 89, 119, 122, 165-167, synonyms 118
183, 196n syntax 2, 103, 111-112, 122, 124, 128,
Sechehaye, (Charles) Albert 126n 132-133, 136, 150, 152, 189-190, 197-
second language teaching 101 198
Sedley, David 9,49-51,55-56,97 synthëkë [see xynthëke]
semantics 137, 189
semiotics 59n, 119, 197 T
Semitic 125, 139, 176-177 Tagliavini, Carlo 9
sense perception 13, 110, 113, 125, 163, Taylor, Daniel J. viii, 101
182, 194-196 Taylor, Talbot J. viii, 109-110
sensitivity 125 teaching 28, 30, 34, 38, 46, 62, 65, 73, 76,
Servius (Maurus Servius Honoratus) 103- 79-80, 87-88, 152
105 tekhnē 27, 33-34, 55, 58, 63, 65, 69-70,
Shapiro, Michael 198, 200 84-89, 148
Shelden, Michael 155 teleology 171
signs 113-116, 119, 126-128, 145, 173 thesis 15,97
Simone, Raffaele 197 Thirty Tyrants 7
Skinner, B(urrhus) F(rederic) 163 Thomas Aquinas, St 107
slaves [see names of slaves] thought 25, 121, 136, 157, 161
Slavic 172 Thrasybulus 7
Sluiter, Ineke 10 Tolstoy, Lev Nikolajevic, Count 179
Smith, Barbara Herrnstein 82 tool [see language, instrumental nature of]
Smolensky, Paul 199 Tory, Geoffroy 149
society 2, 20 transformations 184, 192
Socrates of Athens 7, 9, Chaps 1-3 translation 150
Solon of Athens 32n Trubetzkoy, Nikolaj S(ergeevic), Prince
Sophists 15, 45, 60, 77, 83-84, 97-98, 6, 170-174, 176-183, 191
197n truth 21-22, 24-25, 32, 39, 55, 66-67, 94-
Soulez, Antonia 9-10 95, 157, 163, 169
sounds 45, 47-48, 60-65, 69-71, 86-89,
141,198 U
Spanish 147-148 unconscious [see consciousness]
speaking 27 understanding 73-76, 88, 110
Speas, Margaret 199 Universal Grammar 2, 121, 132-136, 184,
speculative grammar 107, 110n 190-192, 195, 199
224 LIMITING THE ARBITRARY