JANUA LINGUARUM

STUDIA MEMORIAE
NICOLAI VAN WIJK DEDICATA

Series Crìtica 5

edenda curai

WERNER WINTER
LINGUISTIC
STYLISTICS
by

NILS ERIK ENKVIST

1973

MOUTON
THE HAGUE-PARIS
 © Copyright 1973, Mouton & Co., Publishers, The Hague,
The Netherlands.
No part of this book may be translated or reproduced in any form,
by print, photoprint, microfilm, or any other means, without written
permission from the publishers.

L I B R A R Y OF CONGRESS CATALOG CARD NUMBER: 72-88182

Printed in Hungary
For Kjell, Kristian, and Elisabeth
 PREFACE

My aim was to attempt a concise, introductory, and ge-

neral inventory of current problems in linguistic stylistics, all
within some 160 pages. Such a project suggests a number of
decisions on policy of which the reader ought to be warned in
advance.
To discuss all the various concepts and theories of stylistics
that have been proposed, even in the past decade or two, would
at once explode the limits of a small book. I have therefore
frankly adopted a basic view of style as a differential between
a t e x t and a contextually related norm. Within this general
frame I have, however, tried to mention a number of different,
and sometimes conflicting, approaches. I have avoided voicing
my own opinions about a host of controversies where such opi-
nions did not seem necessary for my immediate purpose. For
instance, I suggest that the description of style is not tied to
any single grammatical model, though many of these models
have virtues t h a t other models lack.
Another problem is exemplification. The illustration and
description of actual and potential style markers in different
languages would have needed many times the space of the
present essay. For concrete examples readers will therefore
have to consult works listed in the references and in the bib-
liography. Nor have I tried to give anything like a full review
of the enormous body of stylistic research available in t h e
major European languages alone. My bias has been frankly
in favour of English.
6 PREFACE

I doubt that anybody can honestly claim to master all the

research into stylistics in all the world's languages. My own
references have been picked from among the works I have
come across during several years of fairly comprehensive
reading. At several points I touch upon matters general
enough to warrant very long lists of references. Such lists
were omitted as impracticable and not very helpful. The
bibliography does, however, include a number of works not
referred to in the text. Some are general and bear on many
relevant questions; others are specific, and their relevance
should appear from their titles.
My debts of gratitude are numerous enough to defy brief
listing. Most of the arguments have been tried on generations
of students at Abo Akademi as well as at several other univer-
sities in various countries. Many queries and discussions have
helped to define problems, though not always to solve them.
My most concrete debts are to Professor I. R. Galperin, Dr. Ha-
kan Ringbom, Lector Geoffrey Phillips, and Lector Kurt Jo-
hansson. To Professor Werner Winter and Miss Swantje Koch
I owe thanks for friendly editing, and to my family for pati-
ence and encouragement while I was writing beside a heavy
load of teaching and administration. Travel grants from the
Academy of Finland and from the H. W. Donner Fund and
Swedish Jubilee Fund of Abo Akademi enabled me to look
up references abroad.

The manuscript was completed in February, 1971.

Abo Akademi
20500 Abo 50, Finland
February, 1971 N.E.E.
 CONTENTS

Preface 5

1 Introduction 11
1.1 Does style exist? 11
1.2 Style as a notational term.
Style as departure, addition, and connotation . 14
1.3 Types of linguistic variation 16
1.4 Overlap of stylistics and other branches
of linguistics 17
1.5 Style, genre, function 20
1.6 Style as comparison and prediction 21
1.7 Summary 25

2 Style and literary study 27

2.1 Relations between linguistics, stylistics
and literary study 27
2.2 Close reading. Yossler, Croce, Spitzer 28
2.3 Russian formalism 29
2.4 Literary structuralism 31
2.5 Linguistic versus literary context 33
2.6 Summary 34

3 Style, langue and parole, competence and performance 36

3.1 Stylistic variation in linguistic theory 36
3.2 Saussure: langue, parole 36
3.3 Chomsky: competence, performance 42
8 CONTENTS

3.4 Ways of describing styles within grammar . 46

3.5 Choice of delicacy levels 49
3.6 Stylistic rules 50

4 Context parameters 51
4.1 Need for context classification 51
4.2 Relationships of context and language . . . . 52
4.3 Approaches to, and taxonomies of, context . . 54
4.4 Linguistic division of text into portions . . . . 62
4.5 Group contexts. Slang 64
4.6 Dynamism in contextual categorization . . . . 65

5 Grammatical models in the description of style markers 67

5.1 Style as a differential 67
5.2 Requirements for grammatical models 68
5.3 Style in some grammatical models 71
5.4 Transformational grammar in stylistics . . . . 79
5.5 Systemic grammar in stylistics 81
5.6 From linguistic feature to grammatical model.
Words, rhetorical figures 82
5.7 Styles across languages 86
5.8 Style and rhetoric. Allegory, irony, parody . . 87
5.9 Textual complexity and readability 89
5.10 Stylistics and content analysis 90
5.11 From stylolinguistics to literary analysis . . . 91
5.12 Summary 96

6 Deviance 98
6.1 Deviance 98
6.2 Grammatically, acceptability, tolerance . . . . 98
6.3 Transformationalist and behaviourist views
of well-formedness 101
6.4 Grammaticality scales 103
6.5 Strategies in the description of deviant texts . 104
6.6 Closed texts, open texts, and choices
of grammatical models 106
 CONTENTS 9

6.7 Generative-transformational methods

in the comparison of language variants . . . . 1 0 7
6.8 Psycholinguistic approaches to deviance . . . . 108

7 Linguistic style markers beyond the sentence . . . 110

7.1 Style as a quality of texts 110
7.2 Textual well-formedness, textual .acceptability,
textual tolerance 112
7.3 Textual deviance as a style marker 114
7.4 Theme and rheme. Theme dynamics.
Cohesion devices 115

8 Stylostatistics 127
8.1 Role of statistics in stylistics 127
8.2 Some early stylostatisticians. Zipf and Yule . . 128
8.3 Guiraud. Theme words and key words.
Josephine Miles 132
8.4 Approaches and stylostatistical characteristics . 133
8.5 Sample and norm 135
8.6 Attribution studies 136
8.7 Summary 140
8.8 The role of the computer 140
8.9 Perspectives 143

9 Conclusion and Summary 145

10 Bibliography 149
10.1 Stylistic bibliographies 149
10.2 Books and articles 150

11 Index to references 177

 1

INTRODUCTION

1.1 DOES STYLE EXIST?

STYLE is a concept as common as it is elusive: most of us

speak about it, even lovingly, though few of us are willing to
say precisely what it means. Those who write vaguely, sub-
jectively, and impressionistically about it remain open to
charges of conceptual looseness, however elegantly they may
express their prejudices. Yet those who spend their energies
on rigorous definitions of style and who support their state-
ments with exact facts and figures fare no better. They have
often been ridiculed for tortuous pedantry and chastised for
breaking butterflies on the wheel.
Far fewer are those who deny the very existence of style.
Indeed most people have accepted a tacit, ontological argum-
ent: the fact that the idea of style is so widespread and so
useful proves that there is an objective entity underlying
this idea. 'So many people cannot be wrong' has been one of
the silent tenets of many a stylistician's creed.
Ontological proofs can, however, be refuted just as readily
as they can be upheld. The most recent attempt at refutation
comes from Bennison GRAY, and his argument should be
briefly rehearsed to start us off (GRAY 1969). GRAY'S central
question is, Does style in fact exist at all? His answer is a
vigorous negative. Style, he says, is like the emperor's clothes
or the ether of old physics. It exists merely because people
want to see it because everybody else does, or perhaps because
it buries a number of mysteries under a handy, respectable
12 INTRODUCTION

term. In GRAY'S opinion it hides a vacuity that more thorough

research would be bound to reveal.
GRAY supports his contention by trying to refute one
approach to style after another. If we study style as behaviour
in the manner of the psychologist, he says, we in fact study
character, personality, or individuality. Therefore we should
say so and not pretend to be working with style. If, in the
rhetorician's manner, we identify style with the speaker, we
assume that a man's language has a physiognomic relation to
the man himself. But this must be proved, not assumed. If we
turn philologists and study style as the 'latent' in the manner
of Leo SPITZER, we are in fact studying subject matter:
The lesson of Spitzer's fifty years of application t o the problem of
style is that he could not succeed in distinguishing style from the
work, or, if you will, structure from content, ( Q U A Y 1969: 62)

But — alas — those who define style as the individual, and

thus become literary critics, fare no better. For individuality,
says GRAY, is not a matter of style but of language, subject
matter, content, theme, and referent. Indeed the collocation
individual style is a needless tautology: if we can recognise a
given writer's work, we do so not by 'style' but by a total
pattern. This view of Btyle as the individual element is but
another aspect of an exaltation of romantic originality. Nor
do we solve our problem if we define style as an 'implicit
speaker'. GRAY reminds us how Professor WIMSATT once de-
fined stylistic judgments as comparisons of what a writer had
said with what he ought to have said, but nevertheless later
subscribed to the theory of the intentional fallacy (WIMSATT
1941, 1954). One might, however, reply that comparing a text
with an imaginary norm does not involve any reference to
the author's intentions. The norm is rather set up by the
critic.
So far, a linguist might be tempted to nod in at least
partial agreement. But GRAY uses the same Occamist razor to
slash at linguists as well. To begin with, he says, linguists live in
 INTRODUCTION 13

the realm of the about-to-be-established. Nor is style as choice

— a favourite linguistic notion — a workable concept, because
we can never know what choices in fact were available to a
particular author. And how can we define choice in linguistic
terms? To hedge behind 'synonymy' by defining style as
choice between items t h a t mean the same is not a satisfactory
solution. If two words are synonyms, there is no difference
between them. If they are not, the difference is one of mean-
ing, and the term style is once again superfluous. GRAY sums
up:

In every case of the use of the word 'style' which we have examined,
the user has found it necessary to go outside the work to establish
the existence of style, and in every case he has had to go to something
for which there exists no evidence but the particular work whose
style he wishes to discuss. (GKAY 1969: 107)

GRAY is a diehard organicist: to him a literary work is one

and indivisible, without dichotomies between content and
form, style and expression, process and result. His scepticism
is bent on reducing terms and concepts to a minimum. If we
can do without the concept of 'style' we must discard it a t
once, he insists. B u t does this necessarily follow? One might
agree with GRAY that it is our duty to define precisely what
we mean by style, and still insist t h a t the term is a convenient
abbreviation, precisely as 'horse-power' is a useful concept
for 'a unit measuring the work of a prime motor taken as
equalling 550 foot-pounds per second', or as 'yellow' is handier
than 'the most luminous primary colour occurring in the
spectrum between green and orange'. Philosophers of science
have dwelt on this difference between SUBSTANTIVE TERMS
and NOTATIONAL TERMS. Substantive terms are irreplaceable
without loss of conceptual content, whereas notational terms
are, basically, abbreviations and can thus be paraphrased
(KAPLAN 1964: 4 9 ) . All the same, even notational terms are
useful and may even be necessary. There is no reason to ban
the term style even if it were notational, not substantive, as
14 INTRODUCTION

long as we remember that notational terms must always be

defined with the aid of more basic concepts.
As to the organic view, one may indeed admit that both
meaning and aesthetic effect may be regarded as the sum total
of everything that goes into a text, including the "style" of
that text, in whatever terms it be defined. But one may still
insist that texts must be analysed and taken apart into
components by various methods, and that the quest for corre-
lations between the effect of the whole work and the individual
stimuli and their combinations in the text that caused this
response is a highly legitimate procedure. Indeed the business
of literary criticism might be defined as such a quest.

1.2 STYLE AS A NOTATIONAL TERM.

STYLE AS DEPARTURE, ADDITION, A N D CONNOTATION

Many scholars have felt that style is a notational term and

therefore tried to give us stringent definitions. Some have even
tried to botanize in the flora of such paraphrases (DEVITO
1967, ENKVIST - SPENCER - GREGORY 1964, SAYCE 1962).
A taxonomy of definitions might be based on a number of
principles. Thus some students of style have concentrated on
the relations between the speaker/writer and the text, and
thus found clues to style mainly in the personality and en-
vironment of the people who have generated the text. Others
have focussed their interest on the relations between the text
and the listener/reader, often pointing out that the receiver's
reactions to textual stimuli are more readily accessible to
study than are the generative impulses that motivated the
sender of the message. I f so, the study of style now becomes
based on the recipient's responses to certain features in the
text. A third group of investigators have tried to objectify the
approach and to eliminate references to the communicants at
either end of the communication process. They look for clues
to style in descriptions of the text, not in appeals to personal-
 INTRODUCTION 15

ities. This necessitates the use of objective methods, not

least to distinguish the stylistically significant features of a
text from the non-stylistic ones.
Along another dimension it has been suggested that all
linguistic views of style tend to be based on one of three
fundamentally different views. First, style can be seen as a
DEPARTURE from a set of patterns which have been labelled
as a NORM (style comme écart). According to this principle,
stylistic analysis turns out to consist of comparisons between
features in the text whose style we wish to analyse, and
features in the body of text that we have defined as a norm
and therefore regard as a relevant background. Secondly,
style has been viewed as an ADDITION of certain stylistic
traits to a neutral, styleless, or prestylistic expression. If so,
stylistic analysis becomes a stripping process in which we
peel off, isolate, and describe the stylistic skin and meat that
surround the stylistically neutral or unmarked core. Con-
versely, in the generation of texts, the speaker/writer is sup-
posed to start from a prestylistic or stylistically neutral or
unmarked core of meaning, which he then surrounds with a
halo of style. Thirdly, style has been viewed as CONNOTATION,
whereby each linguistic feature acquires its stylistic value
from the textual and situational environment. Stylistic
analysis therefore becomes a study of the relations between
specific linguistic units and their environment.
But these three approaches may also be seen as complement-
ary rather than as contradictory or mutually exclusive. If we
define a relevant norm against which we match our text, and
if we thus succeed in isolating those departures from the
norm that have stylistic significance, we seem to be applying
the first of the three methods. If, however, the norm is regarded
as STYLISTICALLY NEUTRAL or UNMARKED, the same compar-
ison also satisfies the requirements of the second definition:
the features by which the text differs from the norm are then
identical with those by which a stylistically marked text dif-
fers from a stylistically unmarked, neutral norm. And if we
16 INTRODUCTION

define the norm, not on grounds of general relevance or of

stylistio neutrality but expressly with the aid of definite CON-
T E X T U A L RELATIONSHIPS which justify the comparison of text
and norm, we are in fact also viewing style as connotation.
We are defining stylistically significant features in terms of
contextual, that is, textual and situational, environment.

1.3 TYPES OF LINGUISTIC VARIATION

Linguists are, of course, only one of the groups that are

interested in the styles of language. Linguistics might be
defined as that branch of learning which builds models of
texts and languages on the basis of theories of language, and
which evaluates the success of such models with the aid of
explicit criteria. It is the task of LINGUISTIC STYLISTICS, or
ST YLOLENGUISTICS, to set up inventories and descriptions of
stylistic stimuli with the aid of linguistic concepts. One might
add at once that practical work in stylolinguistics may be, but
need not be, an exercise in its own right. Stylolinguistic
analyses may be directed towards goals beyond linguistics
proper. They may, for instance, be a first step in a wider,
structural, literary, and historical study of a text or a
language.
B y definition, linguists should be interested in all kinds of
LINGUISTIC VARIATION. And style is only one of many types
of such variation. Other types are TEMPORAL, REGIONAL, and
SOCIAL DIALECT, as well as IDIOLECT and REGISTER. A linguistic
form is temporal if it correlates with a given period. The
English of the Authorized Version of the Bible, for instance,
represents an earlier stage in the history of the language.
Linguistic forms whose occurrences correlate with areas on a
map are definable as regional dialects, or dialects for short.
And those whose occurrences correlate with the social class
of their users are social dialects, or SOCIOLECTS. A S we saw
above, one way of defining styles is to regard them as varia-
 INTRODUCTION 17

tions that correlate with contexts and situations. Some lin-

guists prefer the term REGISTER for types of language that
correlate with situation, and use the term style to indicate
individual variation within each register. Others have reserved
register for the different subtypes of language t h a t people
use when acting in different social roles: thus a doctor uses
one register in the operating theatre, another with a patient,
and a third when playing with his children. The term IDIO-
LECT is often used to indicate the language of one individual,
usually in its totality. Actually the terminology varies among
different schools of linguists and even with individual linguists
within each school. One useful term is LINGUISTIC DIATYPE,
which covers all the different subvarieties of language men-
tioned above.

1.4 OVERLAP OF STYLISTICS A N D OTHER BRANCHES

OF LINGUISTICS

Unless stated otherwise, the term style will here be reserved

for t h a t type of linguistic variation which correlates with con-
text in a wide sense of the term, including both textual con-
text and situational context. B u t as style is a notational
term rather than a linguistic prime, it should be emphasized
t h a t other definitions and terminologies are possible and
perhaps even plausible. There is nothing to prevent readers
who so desire from translating the substance and models of
this book's argument into the terminology of their own pre-
ference.
Such translation may be motivated not only by the nota-
tional character of the term style, b u t also by the obvious
fact t h a t the different categories of linguistic variation over-
lap. The English of the 1611 Bible is a temporal variant of
English, but at the same time a living style of twentieth-
century English which must be included in a full inventory of
current stylistic variants. Innumerable examples could be
18 INTRODUCTION

cited of situations in which regional dialect correlates not

only with region but also with social class: for instance, upper-
class people may speak a supraregional standard language,
whereas lower-class people use regionally definable dialects.
And if, as often happens, there are people who speak standard
language in one type of situation and a patois or regional
dialect in another type of situation, the use of regional
dialect thus comes to correlate with situation and the dialect
becomes a style. In a similar manner, sociolects may assume
the function of styles, for instance if a speaker uses upper-class
language in one type of situation and lower-class language in
another. There are many societies in which situational corre-
lations exist even for the choice of language. Thus languages
such as Latin, French, English, Russian, and Hindi have often
ousted various vernaculars from certain contexts such as
scientific or administrative communication, and in societies
characterized by DIGLOSSIA the choice of language is con-
textually conditioned. Thus higher variants such as Classical
Arabic and katharevusa are used in sermons, formal speeches,
and lectures, whereas the lower variants such as colloquial
Arabic and dhimotiki occur in instructions to workmen and
in familiar conversation. (GUMPERZ 1 9 6 6 , FERGUSON 1 9 6 6 ,
DENISON 1 9 6 9 ) In societies with many immigrants, varying
patterns of language loyalty may also lead to a bilingualism, or
even multilingualism, in which languages assume the context-
determined role of styles (FISHMAN 1 9 6 6 ) . Where such
situations exist, we may view the use of a language in the
function of a style as another of the many instances of rank
shift that occur in language.
In this light, our terminology of linguistic subtypes must be
determined by our aims and perspectives. I t is relative rather
than absolute. Of course the examination of all these overlaps
is relevant to stylistics as well as to other branches of language
study. In our description of styles it is important to note for
example that a person's formal style consists of, say, a supra-
dialectal, non-regional standard, whereas his familiar style is
 INTRODUCTION 19

a regional or social variant. We must accept the fact t h a t

stylistics often intersects other areas of linguistics: historical
linguistics, dialectology, and sociolinguistics. But the choice
of conceptual frame and terminology must fit the purpose and
approach of each investigation. Therefore, different branches
of language study ought to be regarded not as rivals but as
different sets of tools, each of which may have its own ad-
vantages in a particular job. We should choose the one t h a t
works best. Thus it becomes idle to argue whether a given
type of language is a temporal, regional, social, or stylistic
variant. I t should perhaps be given more labels than one. To
the historian of English, thou lovest is an older form than you
love; to the student of contemporary styles, it is a feature of
à style that one might label as 'Biblical' or 'archaic'. To
those who find you ain't characteristic of a social class it is a
CLASS MARKER, b u t to those in whose studies it correlates
with a certain range of situations it becomes a STYLE MARKER.
In a full study of linguistic variation, both observations may
be equally relevant. (Cf. ELLIS - U R E 1 9 6 8 , GREGORY 1 9 6 7 ,
LABOV 1 9 6 6 , LOMAN 1 9 7 0 , U R E 1968.)
Such overlaps can be very complicated, not least in assess-
ments of the style of old texts. One example of such a compli-
cation was cited by RIFFATERRE in connection with the
contrast between French réussite and heureux succès. Today,
réussite is stylistically more neutral, heureux succès archaic.
I n the seventeenth century, however, réussite was a bold
neologism readily spotted as a borrowing of Italian riuscita
(RIFFATERRE 1 9 5 9 ) . Such changes in, and crossovers of,
stylistic values pose a caveat to all readers of old texts, while
also showing the relevance of historical, diachronic perspec-
tives even in apparently synchronic stylistics.
Many sociolinguistic studies are of great importance for
students of style as well. For instance in William LABOV'S
sophisticated study of the social strata of English in New York
City (LABOV 1 9 6 6 ) , a number of observations concern corre-
lations between language and situation. Thus LABOV found
20 INTRODUCTION

t h a t many New Yorkers tend to use postvocalic r more often

in formal than in informal contexts, while at the same time
the frequency of postvocalic r could be used to rank people
on a social scale. For example,
A professor of sociology born and raised in N e w Y o r k City began a
lecture with a n (E) index of 50 t o 60; as h e proceeded and warmed
t o his subject, t h e index dropped precipitately, as low as (r) — 0 5 ;
t h e n a s he began t o make his final points, t h e (r) i n d e x began t o rise
again, though it never quite reached its initial value. A Negro w o m a n ,
living on welfare in a bare t e n e m e n t apartment, used a carefully
articulated style o f speech w i t h (r) — 1 9 ; n o w and t h e n she interrupted
herself t o scold her children, using a radically different style of s p e e c h
w i t h ( r ) — 0 0 . (LABOV 1966: 56-7)

This suggests t h a t both the professor and the Negro woman

had two styles, one more formal and one less formal, and each
consisted of a sociolect marked by a certain frequency of post-
vocalic r. Another of the categories of sociolinguistic investi-
gation that have obvious stylistic importance is that dealing
with the choice of pronouns and modes of address (e.g.
LAMBERT 1967).

1.5 STYLE, G E N R E , FUNCTION

Altogether, one scholar's style may be another scholar's

dialect, historical form, sociolect, or even language. As
Werner W I N T E R has noted, stylistics may profit from such
overlaps by borrowing methods from well-developed areas of
linguistics such as dialectology (WINTER 1 9 6 4 ) . In literary
study, too, certain concepts are highly useful in stylolinguistic
description. Thus there is a very close relationship between
style and GENRE, if genre is defined as 'a culturally definable,
traditional type of communication'. A genre could then be
regarded as a culturally definable stable context category —
or stable cluster of contextual features — which usually
correlates to some extent with a certain style, that is, with
 INTRODUCTION 21

a certain type of language. Also, the connection between genre

and function of language is a close one. If we identify genre
with linguistic function, GENRE STYLES such as the styles of
poetry, scientific communication, journalism, and colloquial
conversation become FUNCTIONAL STYLES. This presupposes
the setting-up of correlations between traditional genres and
those constellations of contextual features that we like to
define as FUNCTIONS OF LANGUAGE. I n fact, terms such as
genre and function also turn out to be notation al terms
rather than linguistic primes. They are shorthand labels for
various combinations of contextual primes. For example, in a
certain social setting, both poetry and journalism may share
the feature 'written', whereas the feature 'metrical' may be
characteristic of poetry alone. This is worth noting because if
we can develop a frame for the description and classification
of contextual features, we may also use it for the classification
of genres and functions in the sense of culturally definable,,
traditional uses of language.

1.6 S T Y L E A S C O M P A R I S O N A N D PREDICTION

Even if style is defined as that variety of language which

correlates with context, including situation, the recognition
and analysis of styles are squarely based on comparison. The
essence of variation, and thus of style, is difference, a n d
differences cannot be analysed and described without com-
parison.
Comparisons are also necessary to show ranges of linguistic
variation in specific situations. Situational correlations do
not always operate with hundred-per-cent certainty. In fact,
strong stylistic effects are often created precisely by departing
from the usage that is customary in the situation involved.
The effect becomes strong because it thwarts the recipient's
expectations. If, for instance, a clergyman indulges in slang
and in four-letter words in a high-church sermon, his audience
is likely to be shocked. And if a corporal addresses a squad of
22 INTRODUCTION

recruits on the barrack square in the language usually asso-

ciated with sermons, the effect will also be remarkable. I n
each situation, the recipient of a message — provided t h a t he
knows the language and is familiar with the situation — will
expect t h a t message to be couched in a certain type of
situation-bound language. When the message emerges, the
expectations are fulfilled or disappointed to varying degrees.
If the expectations are fulfilled, the message has conformed
to the style usually associated with that situation: the message
has a high stylistic predictability, a HIGH STYLISTICITY. If
they are disappointed, the message departs from the usual
style and has LOW STYLISTICITY.
There are several corollaries to this simple principle. One
has t o do with the range of variation characteristic of each
situation. There are contexts in which language is very highly
formalized: military commands, greetings, legal formulae,
and ritual language in general. Such "FROZEN" LANGUAGE
(Joos 1962) can also occur in everyday situations, for instance
around the breakfast table. There are other contexts which
are characterized by very wide ranges of variation, like
modern poetry. In fact the RANGE OF LINGUISTIC VARIATION
in a given situation may itself turn into a characteristic of
style: certain styles allow far wider variation than others.
Another basic aspect of stylistic comparison emerges out of
the well-known fact that different people may react differently
to the style of a given text. Their reactions depend on what
they expect of it, that is, what norm they choose to compare
it with. Now the choice of norm is conditioned by past ex-
perience of language in context. Even a native speaker can-
not be conversant with all the styles of his language. Therefore
people with different backgrounds will often react differently
to one and the same text: their stores of past experience and
their norms are different. And even people with a similar
background may, consciously or subconsciously, choose to
match a given text against different norms. One of the
traditional aims of literary education has been to make
 INTRODUCTION 23

students conversant with a body of recommended texts

embodying a set of styles, and thus stylistic norms, against
which they can match the texts they meet. In foreign-
language teaching, one of the tasks necessary before the
student achieves near-native command of the foreign language
is to expose him to a suitably selected range of language
variants, including situational ones, with which he can com-
pare other texts and which he can use as a diagnostic frame
for the recognition of styles. As we shall see below in section
8.1.2, these problems also appear in stylostatistical com-
parisons of texts.
When we compare two texts, we may find t h a t a linguistic
feature occurs in one text but not in the other. I t may be more
frequent in one than in the other. Or it may have roughly the
same frequency in both. In each of these three possible cases
we are involved with FREQUENCIES, greater or smaller, in-
cluding zero frequency (which is the same as non-occurrence).
And if we reckon with the length of the text when assessing
the frequencies, we should more properly speak about DEN-
SITIES of linguistic features, the density of a given feature
being definable as its number of occurrences divided with a
measure for the length of the text, such as the number of
running words.
If the densities of certain features are appreciably different
in the two texts — which we might label as TEXT and NORM,
the t e x t being the text we study and the norm the body of
texts against which we match our text, or the body of expec-
tations based on past experience — these features are STYLE
MARKERS, that is, stylistically significant features. If, again,
the density of a given feature is roughly the same in text and
norm, that feature is, in terms of this comparison, stylistically
unmarked or neutral — that is, equally characteristic of both
text and norm. I t should perhaps be repeated once again
t h a t the results of such comparisons are entirely dependent
on the choice of norm: if two people match the same text
against different norms, their impressions of its style will be
24 INTRODUCTION

different. In the evaluation of literary texts, very often new

norms are brought to bear on a classic text.
As I have argued elsewhere (ENKVIST - SPENCER - GREGORY
1964), if this reasoning is pursued further, it will lead to a
definition of style based on comparison between text and
norm. In ordinary communication, people who know the
language do not choose just any norm for comparison with a
text: the norm chosen must have a definite CONTEXTUAL
RELATIONSHIP with the text. If we read a sonnet, we are more
likely to compare it with other sonnets than with, say, a
telephone book or a newspaper article. The impression of
style, then, arises out of a comparison of the densities of
linguistic features in the text with the densities of the corres-
ponding linguistic features in a contextually related norm.
If we view the same process, not as a comparison between a
static text and a static norm but as a continuous matching of
a linear, emerging text with a set of expectations conditioned
by past experience, we may rephrase the definition in terms
of expectations, or, better, probabilities. The style of a text
thus becomes the aggregate of the contextual probabilities of
linguistic features. Such probabilities may be studied either
with the aid of informants — REFFATERRE (1959) suggested
the concept 'AR' or 'average reader' — or through compu-
tation of densities of linguistic features in text and norm.
Those who desire concrete illustrations of stylistic analysis
with the aid of density comparisons and contextual probabi-
lities may, for instance, consult statistical studies of author
identification such as those by ELLEGARD (1962a, 1962b) or
MOSTELLER — WALLACE (1963, 1964). Such studies are
based precisely on computations of significant differences in
the densities of linguistic features in a text and in a con-
textually related norm.
If we agree to define style as the result of a comparison of
densities of linguistic features in a text and a contextually
related though suitably contrasting norm, we have at the
same time covered a wide range of problems sometimes re-
 INTRODUCTION 25

garded as separate branches of stylistics that are hard to fit

under a single definition. The style of an individual will
emerge out of a comparison of that individual's texts with a
norm consisting of comparable texts but by other people.
A study of the special characteristics of poetic styles would
naturally begin by matching poems against non-poetic texts.
The study of 'expressive' or 'emotive' features should start by
matching texts containing such features against non-expres-
sive or non-emotive texts, and so forth. If this argument is
correct, all stylistic analysis will consist of matching a text
against a norm. Different varieties of stylistics arise through
variations in types of text and types of norm. And different
results, including contradictory opinions of the style of one
and the same text, may arise if different analysts match that
text against different norms, all of which may be justified by
definite contextual relationships with the text at hand.

1.7 S U M M A R Y

In this introduction, I began by noting that style is a

notational term rather than a linguistic prime. That is, style
is a term that can be defined in terms of other, more basic
notions, some of which will be discussed in the following
chapters. If styles are defined as those varieties of language
that correlate with contexts, including both textual and
situational envelopes, there will always be considerable over-
lap between them and a number of concepts of historical,
geographical, and sociological linguistics. Stylolinguistics is
thus one of the ways in which we may view language. A stylo-
linguist is concerned with the same linguistic features as
others may wish to consider in other perspectives and describe
with different terminologies. I have also argued once again
that impressions of style always arise out of comparisons. We
match the text against another body of texts which we might
label as norm, this norm being chosen because it is con-
26 INTRODUCTION

textually relevant as a background for the text. And this

matching process will result in a,n assessment of the differences
in the densities of linguistic features that make the text dif-
ferent from the norm. Features whose densities are significantly
different in the text and in the norm are style markers for the
text in relation to the norm used. A change of norm may re-
sult in a different inventory of style markers.
The norm may be chosen from a wide field. One portion of
a text may be matched against other portions or the whole of
the same text. One text may be compared to other texts. Or
the text may be set against an imaginary norm that only ex-
ists in a critic's mind.
 2

STYLE AND LITERARY STUDY

2.1 RELATIONS BETWEEN LINGUISTICS, STYLISTICS,

AND LITERARY STUDY

I have already spoken for a view which frankly admits that

stylolinguistics is merely one of several possible ways of
looking at language. The features t h a t one scholar may regard
as stylistic may be labelled as historical, regional, or social
variants by those who prefer to start out from different
linguistic premises. But this is not all. Indeed there are many
devotees of other disciplines who are vitally concerned with
stylistic features and who wish to consider them from angles
very different from that of the stylolinguist. Prominent
among them are the students of literature. Literary tra-
ditions also perforce enter into stylolinguistics because they
are part of the context that helps us to define the norm with
which we compare our text. I t is therefore relevant to survey
the vexing question of the relationship between linguistics,
stylistics, and t h a t comprehensive Literaturwissenschaft which
I shall here, for convenience, simply call LITERARY STUDY.
The relationship between these three disciplines can be set
in different ways. We may, for instance, regard stylistics as a
subdepartment of linguistics, and give it a special subsection
dealing with the peculiarities of literary texts. We may choose
to make stylistics a subdepartment of literary study which
may on occasion draw on linguistic methods. Or we may re-
gard stylistics as an autonomous discipline which draws
freely, and eclectically, on methods both from linguistics and
from literary study.
28 S T Y L E AND L I T E R A R Y STUDY

Each approach has its own virtues, and arguing for the
general supremacy of one over another would be a futile
exercise. Such arguments only become meaningful in specific
instances: for a given task, one approach may be better than
another. It is one thing to study styles as types of linguistic
variation, and a very different thing to describe the style of
one particular text for a literary purpose such as finding out
where its author drew his inspiration. All the same, literary
methods have so dominated stylistics that a brief list of some
important schools of literary stylistics is indicated even in a
book whose burden is on the linguistic side. Indeed the
catalogue of literary schools that have contributed germinal
ideas relevant to stylolinguistics is a long one. I t comprises at
least French explication de texte, the approaches of V O S S L E R
and CROCE and S P I T Z E R , Russian formalism, the literary
tenets of the Prague school, Anglo-American New Criticism
with its various offshoots, and, most recently, French Neo-
Structuralism.

2.2 CLOSE READING. VOSSLER, CROCE, SPITZER

Explication de texte has nineteenth-century roots. Its aim was

a close reading which correlated historical and linguistic in-
formation and sought connections between aesthetic responses
and specific stimuli in the text; its refusal to congeal into set
patterns was both a strength and weakness. The New Criticism
that began in Britain before World War I I and became a
dominant movement in British and American criticism in the
post-war years shared this interest in the text. The important
basis of criticism was the text, not the biography of the author
or the history of his times. At best, the New Critics threw
sharp light on poetic details, on technique and structure; at
worst they worked in a cultural vacuum and ignored such
features as could be understood only in elaborate reconstruc-
tions of their original, cultural, and linguistic setting. Ob-
 S T Y L E AND L I T E R A R Y STUDY 29

viously, the increase in attention to textual detail was apt to

bring literary criticism closer to linguistics. Curiously enough,
the gap between these two disciplines was not bridged, how-
ever — perhaps because contemporary linguists were pre-
occupied with small units such as phonemes and morphemes,
which were particularly amenable to the methods of taxonom-
ic structuralism. Some persons associated with New Criticism,
notably I . A . R I C H A R D S , have made their mark in linguistics
as well.
V O S S L E R , C R O C E , S P I T Z E R , and other "idealists" tried dif-
ferent avenues to find the ideas underlying the surface of
linguistic expression. V O S S L E R was particularly interested in
clues to national cultures behind linguistic details. CROCE
regarded language as creation and therefore made linguistics
a subdepartment of aesthetics. And S P I T Z E R wrote his stim-
ulating essays by allowing his famous literary sensibility
to react to different types of features in a range of text and
by giving his vast erudition free rein in tracing parallels
between culture and expression. Linguists, who are brought
up to use stringent and explicit methods, are tempted to find
such studies lacking in methodological rigour. Indeed S P I T Z E R
himself was the first to emphasize that he had no method
that could be precisely described. His approach — the
S P I T Z E R I A N C I R C L E — was first to spot stylistic stimuli by
intuition, and then to offer such explanations as brought
back the argument to its starting-point. S P I T Z E R ' S own bril-
liance guaranteed a stock of norms against which he could
meaningfully match his texts. But those who lack his insight
are likely to be better served by more pedestrian and explicit
methods.

2.3 RUSSIAN FORMALISM

The Russian formalists have not yet exhausted their role,

partly because of the language barriers. In a sense they have
been rediscovered in the past several years, not least thanks
30 STYLE A N D LITERARY STUDY

to new translations (TODOROV 1 9 6 5 , STRIEDTER 1 9 6 9 ) , and

their impact for instance on the French Neo-Structuralists is-
obvious. The origins of Russian formalism can be sought in a
protest against the academic preoccupation with linguistic-
and literary history and against the symbolist movement.
During the early years of World War I, groups of brilliant,
young scholars at Moscow and Petrograd crystallized the
formalist principles into a programme and began applying it
to a wide range of concrete problems. Their focus was on the
devices of artistry, not on content: the student of literature
was to concentrate on the 'how' rather than on the 'what'.
In stylistics, the formalists were the opposite pole from the
idealists. Though formalism remained a collection of indi-
vidual ideas with a common background, not a monolithic
doctrine, and though stylistics meant different things to
different formalists, the centre of attention was always the
formal manifestation of styles. From CHRISTIANSEN'S Philo-
sophic der Kunst ( 1 9 0 9 ) the formalists borrowed the principle
of Differenzqicalitat, through which qualities originate in a di-
vergence from a norm. Roman JAKOBSON began formulating
the distinction between artistic prose based on metonymy and
poetry based on metaphor; T Y N J A N O V studied relation be-
tween genres of poetry and the corresponding levels of speech;
¿IRMTTNSKIJ and EICHENBATTM adopted more inclusive views
of style which included composition as well as theme; and
VINOGRADOV analysed verbal reflections of motifs in key
words and word clusters, as well as relations between context
and language in texts such as Avvakum's seventeenth-century
autobiography (ERLICH 1 9 6 9 , L E O N T ' E V 1 9 6 8 ) .
The Russian formalists thus commented on the relation-
ship between style and context as well as on the character of
style as deviance from a norm. The movement, however,
declined, its very name became synonymous with official
opprobrium, and by about 1930 it had lost its first impetus.
Many of its principles have best survived in the works of the
Prague School. Still, even Soviet stylistics was continually
 STYLE AND LITERARY STUDY 31

inspired by these beginnings. V . V . VINOGRADOV was one of

those instrumental in defining the subsequent positions of
Soviet stylistics. Stylistic analysis, he said, should view the
literary text as a dialectical combination of elements at dif-
ferent levels, and thus not only at the formal level: ideas,
themes, and literary structure are the forces shaping the
linguistic surface of a text. Style concerns itself with organi-
zation at all levels, and linguistic stylistics is the subdepart-
ment which studies such organization at the linguistic level.
Therefore formal, linguistic stylistics is only one part of
stylistics, and the study of literature is not merely a sub-
department of linguistics. VINOGRADOV has divided stylistics
into three departments: linguistic or structural stylistics
which operates with linguistic concepts; the stylistics of
speech or rec' which deals with the characteristics of genres,
groups of texts, and individual texts; and the stylistics of
artistic literature whose highest form is the discovery of the
aesthetic effects of structural features in a literary work.
VINOGRADOV thus professed a certain antiformalism:

in its method of approach to the analysis of structures of verbal art,

literary stylistics — a branch of the general stylistics of art — is
dominated by the categories and concepts of philosophical aesthetics
and literary theory (VINOGRADOV 1963: 205).

2.4 L I T E R A R Y STRUCTURALISM

Another line of influence runs from modern stylistics back to

Vladimir PROPP'S studies of the morphology of the folktale,
also an important outgrowth of Russian formalism. PROPP
noted that the plots of Russian folktales could be described
in terms of combinations of discrete elements. He listed thirty-
one thematic categories such as the theme of absence ('one of
the members of a family is absent from home'), departure
('the hero leaves home'), interdiction ('the hero is forbidden
something'), provision ('the hero is given a magic agent'), and
32 STYLE AND LITERARY STUDY

so on. Thus each folktale is no longer seen as a unique object

sui generis, but rather as a selection from, and combination
of, these universal themes or "functions". When the analytic
principle is extended to cover character, Russian folktales are
found to have seven types of dramatis personae: a villain, a
donor, a helper, a sought-for person, a dispatcher, a hero, and
a false hero. (PROPP 1958)
Such categories are more closely concerned with the struc-
ture of the literary work than directly with style. But if style
is defined as that kind of linguistic variation that correlates
with context, and if elements of literary structures such as
those of PROPP are viewed as contextual categories, they
become relevant to stylolinguistics as well. They give us a
more sensitive taxonomy of literary contexts than, say, rough
distinctions between dialogue and description.
The quest for such universals of narrative structure has
been more recently continued by BARTHES (1966), TODOROV
and others, and applied to some English themes by CHATMAN
(1969). According to one version of neo-structuralist theory,
narrative consists of story and discourse. The story is formed
by the action and by the characters, whereas the discourse
consists of the relations between narrator and reader such as
the time scale, the aspect, and the mode of the narrative. The
story can be split up into minimal narrative units called
functions. These functions combine into larger units or actions,
which together form the narrative. Functions are either ker-
nels, that is branching points in the story in which choices
between alternative courses of behaviour take place, or
catalysts which merely elaborate a path of behaviour chosen
in a kernel. A third type of functions are the indices which do
not link a function to the events proper in the narrative but
which refer to character or atmosphere. Thus a description of
a knife in a person's belt is a catalyst if the knife is actually
used in the story, but remains an index if it merely illustrates
the appearance and character of its owner. Proportions
between kernels, catalysts, and indices can be regarded as
 STYLE AND LITERARY STUDY 33

characteristics of a writer's narrative structure, and in t h a t

sense of his narrative style. But such universals might also
be viewed as stylolinguistic context categories. They make it
theoretically possible to study and contrast for instance the
language of kernels and catalysts with the language of indices,
and to find out whether a given writer might be said to have
a different style for the expression of each of these functions.

2.5 LINGUISTIC VERSUS LITERARY CONTEXT

Thus various, sometimes very dogmatic, attitudes have been

voiced about the relations between linguistics, stylistics, and
literary study, and they have even acquired political overtones.
I n practical work, such problems tend to solve themselves,
pragmatically, as long as each investigator allows himself t h e
freedom of choosing and shaping his methods as they best
help him to achieve his own particular goals. In some studies,
stylistics may be an auxiliary brought in to elucidate narrative
structure; in others, categories of narrative structure provide
contexts for stylistic analysis.
An example (ENKVIST 1 9 6 4 , CASSIRER 1 9 7 0 ) : In The Doll's
House, Nora says: " I leave the keys here." As long as we limit
ourselves to the linguistic garb of this utterance, we may
content ourselves by noting t h a t it is dressed up in everyday-
language which perfectly harmonizes with its domestic con-
text. We must go into the dramatic structure of the play t o
notice t h a t Nora's line in fact is a major kernel, as it signals
Nora's determination to break with her past. To understand
the full impact of an utterance, we must simultaneously
analyse both its linguistic surface in terms of a linguistic con-
text such as 'everyday middleclass conversation', and its
meaning in terms of the context of narrative and action.
Only thus can we note that an expression which, against one
contextual background, seems trivial and highly predictable
may carry a very great amount of information and thus be
highly significant when seen in the light of another con-
34 STYLE AND LITERARY STUDY

textual background (here the structure of The Doll's House).

How far we wish to go in our discussion of an utterance such
as this will of course depend on our purpose. If we study
Ibsen's Norwegian style in general, we may dismiss Nora's
line as a trivial example of everyday dialogue. If, on the
contrary, we study The Doll's House or the way in which
Ibsen builds u p a dramatic climax, we should carefully note
the tension between a major narrative kernel and its un-
dramatic expression. The apparatus developed by P R O P P ,
BARTHES, TODOROV, and others is likely to be of use for those
who wish to study correlations between narrative elements
and their linguistic expression.

2.6 SUMMARY

Altogether: we may approach styles in more than one way.

There is a basic distinction between mere linguistic description
of stimuli t h a t we have defined as stylistic, and a description
not only of the stimuli themselves but also of their full
narrative, semantic, and aesthetic effects. I n the words of
Amado ALONSO, stylistics may study a text both as ergon and
as energeia (ALONSO 1 9 4 2 ) . Elsewhere (ENKVIST 1 9 6 4 ) and in
a somewhat different context I have suggested the terms
STYLOLINQTJISTICS or SL for the linguistic description of
stylistically significant features, and STYLOBEHAVIOTJRISTICS
or SB for attempts at correlating such stimuli with their
responses. If we wish to venture into SB, we may frankly use
ourselves as informants and try to verbalize and classify our
responses and to correlate them with the stimuli that caused
them. We may set up ideal responses such as those of R I F F A -
TERRE'S "Average Reader" (RIFFATERRE 1 9 5 9 ) , we may ob-
serve informant behaviour, and we may try to elicit descrip-
tions of responses from informants or informant groups. The
wells of literary criticism may of course be tapped for a
wealth of responses to classic literary texts.
 STYLE A N D LITERARY STUDY 35

In the strict sense, stylolinguistics is concerned with the

linguistic description of stylistic stimuli, as well as with the
methods by which such stylistic stimuli may be defined and
identified. To what extent the linguist should be concerned
with people's responses to stylistic stimuli depends on where
he wishes to place the borders of his own discipline. Responses
to stylistic stimuli are important clues to the meaning of
such stimuli, and meaning is certainly a central concern of all
study of language.. And in foreign-language teaching, every
teacher is compelled to concern himself with giving his pupils
the proper patterns of response to stylistic features. Thus,
however alien the study of stylobehaviouristics may seem to
some linguists, it is all the same an area t h a t many of us
must be concerned with, and an area that all linguists will
gnore at their own peril.
 3

STYLE, LANGUE AND PAROLE,

COMPETENCE AND PERFORMANCE

3.1 S T Y L I S T I C VARIATION I N L I N G U I S T I C T H E O R Y

In addition to devising linguistic methods for the identification

and description of stylistic stimuli, stylolinguists must try
to define the place of style in linguistic theories. Let us
therefore look at some of the ways in which linguists have
tried to fit stylistic variation into their theories.

3.2 S A U S S U R E : LANGUE, PAROLE

One of the major topics in the discussions around the theory

of stylolinguistics has been the question whether style should
be regarded as part of langue or of parole in the famous
SAtrssuBKan dichotomy. The definition is best given in the
words of SATTSSTTRE'S text, remembering that langue and pa-
role together form the total of langage;

L a langue existe dans la collectivité sous la forme d'une somme d'em-

preintes déposées dans chaque cerveau, a peu près comme un diction-
naire dont tous les exemplaires, identiques, seraient répartis entre
les individus [. . .] C'est donc quelque chose qui est dans chacun d'eux,
tout en étant commun a tous et placé en dehors de la volonté des
dépositaires [. . .] De quelle manière la parole est-elle présente dans
cette même collectivité? Elle est la somme de ce que les gens disent,
et elle comprend: a) des combinaisons individuelles, dépendant de la
volonté de ceux qui parlent, b) des actes de phonation également
volontaires, nécessaires pour l'exécution de ces combinaisons. H n'y a
donc rien de collectif dans la parole; les manifestations en sont indi-
viduelles et momentanées [. . .] (SAUSSURE 1955: 38)
 STYLE, "LANGUE" AND "PAROLE" 37

Such dichotomies cannot be ignored by those interested in

linguistic variation. B u t students of style have not been un-
animous in their attitudes to, and comments on, the S A U S -
S U R E a n dichotomy. There are at least four different ways in
which theorists of style have reacted to the distinction be-
tween langue and parole. One solution is to find stylistic sub-
sections under each of these two concepts. Another is to
equate stylistics with the linguistics of parole. Yet another is
to declare t h a t though S A U S S U R E ' S distinction is too valuable
to throw away altogether, it needs supplementing if it is to be
salvaged for stylolinguistic use. Finally, some scholars discard
or ignore S A U S S U R E ' S theory, implying t h a t it is poorly suited
for the theory and study of style.
Those who wish to maintain langue and parole might, then,
identify stylistics with the study of parole. But such views (e.g.
N A E R T 1949) lead to difficulties, however well they may work
in the analysis of single texts by one individual. Some of the
difficulties are methodological. If langue is only observable as
an abstraction from parole, and if styles are only observable as
results of comparison between one sample of parole and
another, how can these two samples be compared without
recourse to langue % That is, each sample supposedly reflects
the same, underlying langue, which directs them and makes
them commensurable. And if langue must be drawn into such
comparisons, then style must be related to langue and not
only to parole. Also, if parole is defined as non-collective, in-
dividual, and momentaneous, it does not cover all of the
language variants t h a t we may wish to label as styles. This is
true of non-individual, collective, group styles regulated not
only by demands of a single, individual speech act but by the
wider norms of groups or communities. Indeed there are
important categories of texts such as laws, statutes, and
certain types of scientific communication in which writers
often take great pains to repress all traces of individualism
from their expression. Thus, even if individual styles fit
nicely under parole, group styles seem to contain the inter-
38 STYLE, "LANGUE" AND "PAROLE"

individual and normative element which places t h e m under

langue.
Presumably these are the reasons w h y some scholars have
maintained t h a t each of the two spheres has its own stylistics.
I t has been suggested (e.g. VINOGRADOV 1963b) t h a t the
stylistics of language operates on the paradigmatic level,
assigning stylistic values t o each grammatical rule and
lexical item. The stylistics of speech operates on the syntag-
matic level and shows how the stylistic system of language has
been used to form the style of an individual text. Thus the
stylistic potentialities of the paradigmatic system are eclectic-
ally used b y those generating texts: t h e y choose features from
the paradigmatic system a n d p u t t h e m into linear, syntag-
matic strings.
Another solution is to divide styles into two categories:
group styles belonging t o langue, and individual styles belong-
ing to parole. Lubomir DOLEZEL is one of the Czech scholars
who have emphasized the distinction between the style of a
single utterance, and the style of a category or type of
utterance (DOLEÈEL 1 9 6 0 ) . The former is close to SAUSSURE'S
parole, as it implies the capacity of an individual to order
certain features in a single utterance. B u t to study this
aspect of utterances, we shall need a special theory of dis-
course which is not the same as stylistics:

La différence entre la théorie de l'énonciation et la stylistique réside

dans le fait que celle-la e x a m i n e l'aspect général de l'énonciation et
les lois générales de sa structure linguistique, tandis que celle-ci étudie
l'aspect spécifique des énonciations et les lois spécifiques de la structure
linguistique des énonciations e t des t y p e s d'énonciations. (DOLEZEL
1960: 196)

The type of stylistics devoted to categories, not to single

texts, must define and group its corpora by individuals,
epochs, functions, or genres, all of which can be studied
synchronically as well as diachronically. Particularly import-
ant are t h e functional styles defined by "le b u t général du
 STYLE, "LANGUE" AND "PAROLE" 39

complexe normalisé des moyens linguistiques" (DOLEÎEL

1960: 197). Four such functional categories used by Czech
scholars are conversation, publicity, science, and artistic
communication. In such frames, artistic style becomes b u t
another functional style, not a special language system op-
posed to t h a t of non-artistic communication as suggested by
those Russian formalists who maintained t h a t artistic language
(and, most notably, poetry) is language oriented towards
itself.
A similar principle of divorcing individual styles from group
styles appears in Josef VACHEK'S Dictionnaire de linguistique
de VÉcole de Prague, where a distinction is drawn between
SPECIAL LANGUAGES a n d FUNCTIONAL STYLES:

L a différence e n t r e l a l a n g u e spéciale et le s t y l e f o n c t i o n n e l est la

s u i v a n t e : le s t y l e f o n c t i o n n e l est d é t e r m i n é p a r le b u t c o n c r e t d e t o u t e
m a n i f e s t a t i o n l i n g u i s t i q u e ; il s ' a g i t d e la f o n c t i o n d e l a m a n i f e s t a t i o n
l i n g u i s t i q u e (c.-à.-d. d e la ' p a r o l e ' ) , t a n d i s q u e l a l a n g u e spéciale est
d é t e r m i n é e p a r le b u t g é n é r a l d e l ' e n s e m b l e n o r m a l i s é des m o y e n s
l i n g u i s t i q u e s , elle est la f o n c t i o n de l a ' l a n g u e ' . (VACHEK 1966: 44)

This elegant placing of special languages under langue and

of functional styles under parole brings with it some difficulties,
however. I t tends to separate descriptions of individualistic
styles from those of impersonal styles, the type of commu-
nication where individuals struggle to escape from the norm
from the type where communicants do their utmost to efface
all traces of themselves. For some purposes the latter might
justifiably be regarded as 'special languages' or group styles
even if they have a but concret like 'functional styles'.
Problems of these kinds have led some scholars to posit
supplements to SAUSSURE'S dichotomy. One well-known
attempt in this direction is t h a t of Eugenio COSERIU. In his
essay "Sistema, norma y habla" (COSERIU 1962) he has re-
hearsed the arguments agains the languejparole distinction
and illustrated the need for an additional, third level, labelled
40 STYLE, " L A N G U E " AND "PAROLE"

as norm or usage, as an intermediary between the two.

Another suggestion towards supplementing SAUSSURE comes
from Luigi ROSIELLO, who adds two levels, usage and norm,
and places them between langue and parole (ROSIELLO 1 9 6 5 ) .
The structure of language both acts on, and is acted on, by
usage; usage is linked to norm by the forces of standardi-
zation. The norm has both a grammatical and a stylistic
component, and these two components of the norm join both
the collective and the individual elements of usage finally to
determine parole. Some Prague linguists have also developed
a three-level approach: between the concrete speech event
and the abstract sentence pattern there intervenes an utter-
ance level which includes features such as functional sentence
perspective ( D A N E § 1 9 6 4 ) .
In stylostatistics, too, there have been efforts to supplement
SAUSSURE. Gustav H E R D AN repeatedly pointed out that
statistics had proved the soundness of two of SAUSSURE'S
fundamental principles: the arbitrariness of the linguistic
sign, which is revealed by the stability of phoneme distr-
ibutions in different samples of language, and the linearity of
the signifiant, which appears in the harmony between voca-
bulary connectivit and the theory of random partitions.
Further, the relation between langue and parole has a stat-
istical equivalent in the relation between the statistical uni-
verse and the sample. As SAUSSURE was so obviously right in
matters such as these, his basic ideas ought to be salvaged.
On such grounds, H E R D A N argued for the use of SATJSSURE'S
third, often ignored, concept, langage, which could be reinter-
preted statistically as

the total of samples (parole) that can be withdrawn from the statistical
universe, represented by la langue plus probabilities. ( H E B D A N 1964;
74)

Langage thus becomes the level of stylistic norms. H E R D A N ' S

writings are, however, sometimes diffident at this very point:
 STYLE, "LANGUE" AND "PAROLE" 41

W h e t h e r le langage o u g h t t o be regarded a s s o m e t h i n g different f r o m

la langue, or w h e t h e r it is preferable t o regard la langue as already
incorporating these probabilities, this is really more a m a t t e r of t a s t e
and definition, and I personally would n o t p u t m u c h w e i g h t u p o n
one in preference t o t h e other conception. (HERDAN 1964: 74)

W h a t is clear is t h a t we need a level of norms incorporating

not only structures b u t also their probabilities of occurrence
— a point we shall return to, and one whose relevance has
already appeared. If such a level can be incorporated into
langue, well and good; if not, it must be given a zone of its
own in linguistic theory. H E R D A N was right in emphasizing
t h a t the existence of styles compels us t o have, somewhere in
our theory, a level with structures as well as their probabilities.
I t is not enough to have two levels only, one of structures
without probabilities (langue) and another of features bound
to a given text {parole).
I n his studies of linguistic aspects of literary communica-
tion, Archibald A. TTTT.T. has also reminded us t h a t SAUSSURE'S
two levels do not suffice. HILL'S addition is a third element
labelled as interpretation ( H I L L 1964). When hearing succes-
sive features of a text, says H I L L , the receiver interprets
them, apparently with the aid of projection rules which bring
previous signals to bear on later ones and narrow down t h e
range of possible interpretations. In the words of Professor
HILL,

T h e s e successive interpretations are like h y p o t h e s e s . T h e y are n o t

t h e s a m e t h i n g as parole, since t h e y are conclusions a b o u t parole.
T h e y are n o t t h e s a m e t h i n g a s langue, since if t h e y were w e should
be forced t o suppose t h a t langue changes w i t h each change in inter-
p r e t a t i o n ( H r t i 1964: 44).

Interpretation is relevant t o stylistics because style can be

regarded as the aggregate of contextual probabilities. A n d
this aggregate of probabilities is based on the receiver's p a s t
experience, which determines and guides his interpretation. If
I have understood H I L L correctly, his "interpretation" is
42 STYLE, " L A N G U E " AND "PAROLE"

closely related to H E R D A N ' S langage and to the stock of norms

and norm-conditioned expectations against which we match
the texts we are exposed to.
In effect, H I L L ' S interpretation, H E R D A N ' S probabilities,
and COSERIU'S and ROSIELLO'S levels of norm and usage all
serve one basic requirement. They insert a probabilistic level
between a more or less deterministic grammar, in which rules
are given without the frequencies in which they are used in
different contexts, and the taxonomic description of actual
occurrences of grammatical features in a limited text. This is
important in more than one respect. First it suggests that
dichotomies such as the SAUSSUREan may not give an adequate
base for linguistic theories that have expressly to cater for
style. Secondly, in a wider connection it shows that the human
element of linguistic experience must be reckoned with at
some place of a complete theory of language. For, if probab-
ilistic levels are necessary, and if our sense of linguistic pro-
babilities is determined by our past experience of language,
this past experience becomes a very major force in shaping
our ability to generate and to interpret linguistic texts. We
carry with us not only a deterministic, all-or-nothing gram-
mar, but also a body of statistical data which we extrapolate
from past experience into current probabilities and expec-
tations.

3.3 CHOMSKY: COMPETENCE, PERFORMANCE

In transformational grammar, the counterpart of the SAUS-

SUREan dichotomy is the distinction between competence and
performance. To cite a paper by Noam CHOMSKY from the
mid-sixties:

A distinction m u s t be made between w h a t the speaker of a language

knows implicitly (what we m a y call his competence) and what he does
(his performance). A grammar, in the traditional view, is an account
of competence. I t describes and a t t e m p t s t o account for the ability
 STYLE, "LANGTJE" AND "PAROLE" 43

of a speaker to understand an arbitrary sentence of his language and

to produce an appropriate sentence on a given occasion. (CHOMSKY
1966: 3)

I take it this statement, and other similar ones, were not

composed especially to cater to theories of style, which
were not a focus of interest in early transformational gram-
mar. In fact the notion of performance has more often been
evoked to explain those deviations from grammaticality that
are characteristic of ordinary speech. Nor should we here
spend much time in trying to analyse the precise differences
between competence and SAUSSTTRE'S langue. I t must suffice
to note that CHOMSKY'S wordings suggest a greater emphasis
on an individual's internalized set of language rules, whereas
SAUSSURE brought in the social, interindividual aspect as
basic for langue (LEVXN 1965). Therefore it may be easier to
build an individual's past experience into competence than
into langue.
We might, however, ask, where should styles go, into
competence or into performance ? If a person knows implicitly
that a text he hears or reads is in, say, legal style, and if he is
also capable of generating texts in this style, then one might
argue that the characteristics of legal style should be re-
garded as part of his competence. If so, a full analysis of com-
petence should include an apparatus describing stylistic
variation. And if grammar must be capable of accounting for
a speaker's ability to understand an arbitrary sentence, and
if understanding the impact of the style of an arbitrary
sentence is part of this ability, it follows that a full grammar
must be concerned with style as well. Indeed CHOMSKY'S very
words "to produce an appropriate sentence on a given occa-
sion" (CHOMSKY 1966: 3) echo the sentiments of generations
of normative rhetoricians and teachers of style, who have in-
sisted that a sentence is not appropriate if it is not dressed in
the style proper for the occasion. In this light it is interesting
to recall once again that early transformational grammar was
but little concerned with style. In fact the term stylistic
44 STYLE, "LANGTJE" A N D "PAROLE"

variation was often used to indicate that such variation

need not be exhaustively analysed within the basic grammar
or even within competence, but could be relegated to some
other, less fundamental part of linguistics, and perhaps to
performance. Another difficulty was caused by the lack of
interest in, and even hostility to, statistics in early transfo-
rmational grammar. Historically, this is easy to explain as one
of the reactions against behaviourist doctrine. But if styles
remain intimately linked to frequencies, we must find some
way of quantifying our generative-transformational rules if
we wish to include a stylistic component in our accounts of
linguistic competence.
In the description of texts, it is of course perfectly possible
to analyse their grammatical characteristics and styles with
the aid of a generative-transformational model. We may
readily state how many times a given writer used a certain
rule when producing a given text (assuming that we have an
adequate transformational grammar at our disposal). And we
can do this even though styles and quantifications were not
built into the grammar itself. (This is the reason why style
has been regarded as part of performance, not competence.)
But if we agree that linguistic variation must be explained in
the rules themselves, we might try to provide our rules with
an additional apparatus of quantification. This need has been
met by K L E I N ( 1 9 6 5 ) and by L A B O V in his study of the copula
in American ghetto English (LABOV 1 9 6 9 ) . L A B O V found that
only a statistical apparatus could adequately describe the
type of free variation that occurred in his materials. That is,
some variation was found to remain "free" in spite of the se-
arch for new parameters which could have determined when
each variant was used. L A B O V was therefore dissatisfied with
the traditional type of transformational grammar, in which
every rule either operates or does not operate depending on
the presence or absence of explicit, rule-activating features
or feature combinations. He needed a grammar whose rules-
were activated in a certain percentage of the instances but
 STYLE, " L A N G U E " AND "PAROLE" 45

which remained inactive in the rest. A s LABOV'S method is of

potential interest also to students of style, precisely because
it shows us an example of how probabilistic quantification
can be built into a generative grammar, it should be briefly
presented here.
LABOV starts out by noting that current rules in generative
grammar have the general form X -* Y/A — B. That is,
every time X occurs in the environment A — B, it is rewritten
as F ; otherwise the rule does not operate. Such rules are
CATEGORICAL INSTRUCTIONS. I f we introduce OPTIONAL RULES
into our grammar, as CHOMSKY had done in his original 1 9 5 7
format, we sidestep the crucial question as to when a rule is
actually used and when it is not. Therefore the solution is to
introduce VARIABLE RULES with a

specific quantity 99 which denotes the proportion of cases in which

the rule applies as part of the rule structure itself. This proportion is
the ratio of cases in which the rule actually does apply to the total
population of utterances in which the rule can possibly apply, as
defined by the specified environment, if it were a categorical rule
[. . . ] (LABOV 1 9 6 9 : 738)

The values for 99 range from 0 to 1. For categorical rules, the

value for <p is always 1, and the rule always operates if trig-
gered off by the environment.
It becomes convenient to define q> = 1 — k0, where k0 is
the VARIABLE INPUT TO THE RULE which limits its application.
The greater the value of k0, the more rarely the rule operates.
If we need a more elaborate formula to compute <p not in terms
of a single factor but in terms of several variables, we may use

q> = 1 — (k0 — — . . . — vkn) ,

where k0 ... kn are empirically determinable constants, and

x — v their weights. Here-the formula was designed so that
the presence of a positive constant in a given subset of sen-
tences will diminish k0 and thus increase the application of
46 STYLE, "LANGUE" AND "PAROLE"

the rule by decreasing constraints and increasing cp. I n

principle one might think in terms also of negative constants
or constraints, and write another formula with pluses instead
of minuses within the parentheses.
When the values of the various constants have been de-
termined through empirical study of the data, they may be
arranged into a single hierarchy. The criterion for this ordering
is that each constraint should outweigh the effects of all con-
straints below it in the hierarchy. This ordering helps us to
p u t related constraints into a single system and thus to get a
more accurate idea of their connections. Hierarchization is
therefore an improvement over putting each constraint into a
separate rule. I t will also help to spot those constraints that
fail to fit into the system and t h a t will need rules of their own.

3.4 WAYS OF DESCRIBING STYLES WITHIN GRAMMAR

The dichotomies betwen langue and parole or between com-

petence and performance serve to reconcile two views: lan-
guage as a rule-bound system, and language as a wide range
of behaviour. As the study of linguistic variation goes on, the
discovery of new regularities may warrant moving certain
areas of language from parole into langue.
Assuming t h a t we have opted for describing stylistic
variation in grammar proper, and not in parole or perform-
ance or some other extragrammatical limbo, how should we
proceed? Obviously, we must mark our rules for stylistic
applicability. Some types of stylogrammatical rules are
categorical within each style. Thus we may wish to generate
a religious text which always uses thou lovest, never you love.
Other stylogrammatical rules are variable in LABOV'S sense.
We might, for instance, conceive of a scientific style in which
the passive transformation operates in a given percentage of
all sentences with certain specified underlying subjects (such
as I , we, the author(s) of the present paper, etc.). We shall
 STYLE, "LANGTJE" AND "PABOLE" 47

therefore need two kinds of stylogrammatical rules, categorical

and variable, in addition to a battery of marks explaining the
contextual range of application of each rule.
But there are some other basic decisions that have to be
made first. One decision concerns the attribution of a given
text to a given language. Should, say, Shakespeare's style,
the style of Dr Johnson, and the style of Dylan Thomas all be
regarded as subvarieties of one single language, English, or
should they be regarded as separate languages meriting
grammars of their own ? Such decisions are made on cultural
and widely sociolinguistic, not on grammatical, grounds.
There is no linguistic definition of 'language' and 'dialect'
that agrees with the traditional use of such terms. For in-
stance, there are dialects traditionally labelled as "Norwegian"
that are more readily intelligible to speakers of some Swedish
dialects than to speakers of some other "Norwegian" dialects.
Nevertheless such decisions are important. If we decide that
Shakespeare, Dr Johnson, and Dylan Thomas all wrote the
same language, though in different styles, we must design our
description of that language to incorporate descriptions of
these three styles within the same general frame. If not, we
are free to describe each style in its own terms, without
glancing — or "squinting", as linguists once used to say — at
the others.
In this particular instance, the answer may seem obvious
enough. If we write separate grammars for Shakespeare, Dr
Johnson, and Dylan Thomas, we lose a lot in economy,
because a very large number of statements will have to be
repeated in each grammar. Also we run the risk of making the
three grammars incommensurable. We may design them so
differently that their comparison becomes awkward and
laborious, though we know that such comparison is motivated
by the fact that a vast number of people agree in calling them
part of one and the same language, English. There was, how-
ever, a period when linguists used to insist that each text had
to be described exclusively in terms of its own, immanent
48 STYLE, "LANGUE" AND "PAROLE"

categories t h a t were actually a n d overtly manifested on the

surface of t h a t text. Importing categories from other texts —
not t o speak of categories f r o m other languages — was for-
bidden. Absurdly rigid adherence to such principles would,
of course, make stylistic analysis difficult or even impossible.
To guarantee easy comparison, texts t h a t are commonly and
justifiably regarded as p a r t of the same language should be
described within the same basic grammatical frame.
Once we have decided t h a t there is sufficient cultural justi-
fication for regarding a set of texts as being in the same
language, we must decide how the description of t h a t language
is t o incorporate the description of its variants. We m a y
scatter stylistic information throughout the grammar, ex-
plaining stylistic variation in connection with each rule.
Rules would then contain a contextual matrix showing in
what constellations of contextual features they operate, a n d
how often (categorically and always, or with w h a t probability).
Or we m a y write a basic, stylistically neutral grammar incor-
porating only those features t h a t appear in all styles — a
grammar of the degré zero — and then a d d a set of sections
t h a t explain how the stylistically u n m a r k e d language de-
scribed in the basic p a r t of t h e grammar can be further devel-
oped into stylistically m a r k e d language. The former method
harmonizes with the view t h a t every t e x t has a style and t h a t
there is no styleless or prestylistic expression. The latter
assumes t h a t there is a stylistically unmarked core out of
which stylistically marked t e x t s can be generated through the
addition, subtraction, or transformation of certain specifiable
features.
I n theory, both methods have advantages a n d disad-
vantages. Writing a " n e u t r a l " grammar with stylistic appen-
dices would most probably be the more difficult of the two
approaches. Granted, methods such as factor analysis a n d
numerical taxonomy ( K R A U S - P O L Â K 1 9 6 7 , CARVELL - SVART-
VIK 1969) might be used as aids t o reveal the structures t h a t
are shared by a sufficient number of t e x t s t o qualify as stylis-
 STYLE, " L A N G U E " AND "PAROLE" 49

tically unmarked and neutral in the statistical sense. But the

assignation of stylistically marked or unmarked status to
grammatical features is likely to involve more than mere
statistical weights. If such a grammar could be written, it
would, however, be of considerable interest. I t would also
have practical applications for instance in foreign-language
teaching, for it would focus attention on the contextually
most neutral, most widespread, and in this sense basic, fea-
tures of the language.
All the same there are strong reasons for agreeing with
David C R Y S T A L and Derek D A V Y , who prefer to scatter
stylistic information into the grammar. This, they say,

has the advantage of allowing easy comparative statement [. . .] Thus

when the grammar defines the notion of sentence, all stylistically
interesting points about sentence-types, distribution, etc. are described;
when the grammar goes on to discuss adjectives in nominal groups,
all significant stylistic points about adjectives are described, and so on.
Such an approach means that in order to obtain a complete description
of any one variety a description has to be pieced together by working
through the grammar in some predetermined way, and noting points
about a variety as they arise; but this is no objection to the approach,
as it would in any case be necessary to work through the grammar in
this way in order to specify the common-core information. (CRYS-
TAL—DAVY 1 9 6 9 : 4 2 - 3 )

CRYSTAL and D A V Y were working with practical goals in

mind. Their worry is not only theory b u t also the convenience
with which the user of a grammar may extract its stylistic
information.

3.5 C H O I C E O F D E L I C A C Y LEVELS

One of the remaining problems is ubiquitous. I t concerns the

choice of the optimal L E V E L O F ABSTRACTION for description
of styles, or, to use M. A. K . H A L L I D A Y ' S excellent term, the
choice of D E L I C A C Y L E V E L S . In practice, if we write a grammar
50 STYLE, "LANGUE" AND "PABOLE"

of a language, we must decide not only how and where we are

going to put stylistic information, but also how much of such
information we should be concerned with. Should we describe
a few of the most important, major stylistic categories, or
should we aim at greater delicacy and try to describe a large
number of styles ? We should note that if style is regarded as
the result of density comparisons between text and norm,
the number of possible styles is unlimited and depends on the
number of possible combinations of texts and norms. In this
sense, all 'styles' cannot possibly be described within one
grammar, and the investigator is bound to decide where lies
the point of diminishing returns for his particular purpose.

3.6 STYLISTIC RULES

Finally, what should our rules look like? Their form will of
course depend on what model we choose for our grammar.
Some requirements are, however, general enough to apply to
all grammars, and they were already touched upon above.
Thus, if a grammar is to cater for stylistic variation, con-
textual matrices and indications of rule applicability in each
constellation of contextual features must accompany each
rule to mark its stylistic applicability and thus its stylistic
value. More will be said about such requirements below in
section 5.2.
 4

CONTEXT PARAMETERS

4.1 NEED FOR CONTEXT CLASSIFICATION

In the previous chapter I tried to list some general problems

concerning the ways in which descriptions of styles could be
fitted into linguistic description and into grammar. One of the
statements most frequently emphasized in the course of the
argument so far has been the role of contexts as determinants
of styles. Indeed styles have been defined as those variants
of language that correlate with contexts. Before going on I
should therefore discuss some questions of the classification of
contexts as a background for the definition and determination
of styles.
If style is defined as contextually restricted linguistic varia-
tion, stylolinguistics must be capable of defining contextual
ranges and restrictions. Many linguists have in fact tried to
ignore, or even hedge, this task. To set up classifications and
taxonomies for all the situations in which language may occur
is admittedly a difficult operation. I t has been doubly uncon-
genial to those linguists who are used to strict limitations of
their subject and to such stringent methods as are possible if
one works with narrow problems. Indeed there are linguists
who confess defeat when facing the need for bringing taxo-
nomic law and order into the welter of situations and contexts
offered by the world's endlessly varied sociophysical settings.
Still, linguistics has on the whole succeeded well in extracting
relevant features out of very complex chains of events. For
instance, phonologists have not been entirely defeated by the
great complexity of the acoustic speech stream: their efforts.
52 CONTEXT PARAMETERS

to distinguish between distinctive and redundant features have

yielded fairly good results. And the classification of linguistic
contexts is still likely to be simpler than the classification of
chains of events in, say, history or even the history of art.
However awkward context classification may be, there is no
way around it. We can neither ignore the fact t h a t language
occurs in speech situations, nor can we wait for non-linguists
to present context taxonomies tailored expressly for stylo-
linguistic use. A full study of language must recognize the fact
that linguistic systems are used by living people in complex
environments.

4.2 RELATIONSHIPS OF CONTEXT AND LANGUAGE

Connections between variants of language and contexts can

be established in two ways. One of the approaches is from con-
textual categories: we may start out from a contextually de-
finable body of text and see what types of language occur
within this corpus. We may thus study the language of an
individual, of a genre such as scientific communication, of a
period such as the eighteenth century, and so on, and compare
this language with that of a relevant norm to pinpoint its
own distinctive characteristics. B u t we may also approach
the problem in reverse and study the contextual spread of
certain linguistic features. We might, for instance, go through
a corpus of contemporary English texts looking for the forms
thou lovest and he loveth, and define all the passages dominated
b y these forms as 'archaic'. I n practice, all of us use both
methods. When we are exposed to texts in context, we norm-
ally start from the contexts and learn what linguistic features
tend t o occur in them. When we see a text without a context
— say, the first paragraph of an unknown typescript without
a headline — we apply our knowledge in reverse and conjure
forth -a probable context to fit the language. Once we can
establish firm connections between a certain range of contexts
 CONTEXT PARAMETERS 53

and a linguistic form (for instance between thou lovest and

Biblical or archaic style), we may use this linguistic form
(thou lovest) as a criterion for context classification, and label
every context in which it occurs with the proper label. In
some instances, such a procedure works very well. In others,
where the style markers are more complex and styles more
varied and open-ended, firm connections are harder to estab-
lish. Still, in principle it is possible to proceed in both direc-
tions: to define sets of linguistic forms with the aid of the
contexts in which they occur, or to define contexts with the
aid of the linguistic forms that occur in them.
In cultures with stable relationships between context and
language, such correlations are firm and therefore comparat-
ively easy to establish. Historians of stylistics have duly
noted that fixed correlations between genre — that is, tradi-
tional literary context — and style go back a long way. In
book III, chapter xii of the Rhetoric, A R I S T O T L E noted that
each kind of rhetoric has its own appropriate style, and that
the styles of speech and writing are different. The text
usually given pride of place in histories of stylistics is, how-
ever, the Bhetorica ad Herennium, which had this to say about
genre and style:
Sunt igitur tiia genera, quae genera nos figuras appellamus, in quibus
omnis oratio non vitiosa consumitur: unam gravem, alteram medio-
crem, tertiam extenuatam vocamus. Gravis est quae constat ex
verborum gravium levi et ornata constructione. Mediocris est quae
constat ex huxniliore neque tamen ex infima et pervulgatissima ver-
borum dignitate. Adtenuata est quae demissa est usque ad usitatissi-
mam puri consuetudinem sermonis. (CAPLLAN 1954: 252)

The grave style, says the author, expresses ideas through the
most ornate words, and should be used in the discussion of
serious and grave matters. It is illustrated by a grand speech
before a jury. The middle style is illustrated by a more relaxed
speech for a jury, and the low style with a colloquial passage
on an incident at the baths. Out of such beginnings grew the
rigid theory of levels of style, which worked excellently as long
54 CONTEXT PARAMETERS

as literature was closely tied to the corresponding genre

categories (QUADLBAUER 1962). But when borders between
genres became diffuse, the distinction between their styles was
blurred, too. We may conclude t h a t there tends to be a con-
nection between the fixity of genres and the fixity of styles: a
stable inventory of contexts suggests a stable inventory of
styles. A society with very complex context parameters and
context constellations could be expected to have very complex
sets of styles as well.

4.3 APPROACHES TO, AND TAXONOMIES OF, CONTEXT

Contexts have been variously defined by different scholars.

An example of a very inclusive definition is t h a t of SLAMA-
CAZACTJ ( 1 9 6 1 : 2 0 9 ) : here context is a function of the INTEN-
TION of the communication, the MEANING of the text, and the
recipient's possibilities of INTERPRETATION. RIFFATERRE
( 1 9 6 0 ) distinguishes between a MICROCONTEXT, which forms
the stylistically unmarked set of constituents against which a
stylistic device stands in contrast, and a MACROCONTEXT or
that p a r t of the message which precedes the stylistic device
and is exterior to it. He has also defined context as " a linguistic
pattern suddenly broken by an element which was unpredict-
able" (RIFFATERRE 1 9 5 9 : 1 7 1 ) . RIFFATERRE'S context t h u s
emphasizes the role of the textual environment, whereas
SLAMA-CAZACU brought in a number of factors from beyond
the text.
Once again, we should admit that the range and detail of
contextual features we should reckon with are determined b y
our problem and by our materials. I n classifying contexts,
the first task is to decide what features in them are STYLISTIC-
ALLY RELEVANT and what features are STYLISTICALLY IRRE-
LEVANT or REDUNDANT. As this distinction may be of great
importance, in some types of work it may be wise to use dif-
ferent terms for the total, "etic" textual and situational en-
vironment, and for our "emic" selection of stylistically
 CONTEXT PARAMETERS 55

significant features from among the endlessly large number of

characteristics of a given sociophysical setting. One could, for
instance, use the term ENVELOPE for the totality of features,
and the term CONTEXT for the aggregate of stylistically
significant features.
Every situation contains an infinite number of constituent
features. The totality of such features, which also includes the
time and place of the communication act, will in practice make
the sociophysical envelope of every text unique. Precisely as
in phonemic or distinctive-feature analysis of the speech
stream, or in the kinesic analysis of gesture (BIRDWHISTELL
1 9 5 2 , SEBEOK 1 9 6 4 ) , our task is to extract the significant
contexts from among the welter of features in the envelope.
In other words, we should find the proper level of abstraction
and of delicacy. And the method is, in principle, precisely the
same as in phonemics orkinesics: we must isolate those re-
current features that prove to have an invariant relation to
specific linguistic features. We must, in other words, eliminate
redundant features that have no stylistic significance. When
stated thus, the problem of context analysis should at once
look more familiar and less forbidding. Nor should this
approach be taken to suggest that all redundant features are
necessarily meaningless. In phonology, allophonic and para-
linguistic features also have meanings at certain levels:
they may help us to identify speakers by voice colour or
idiolectal mannerisms, and so forth. Similarly, stylistically
redundant features of the sociophysical envelope may be of
great interest in many ways, though we may decide to call
them redundant in our quest for parameters of style.
We should, then, try to isolate those features from the
sociophysical envelope whose presence or absence correlates
with the frequency, presence, or absence of specific linguistic
features. Let us take an example. John Smith is writing. He is
in fact typing in his study at midnight. I t rains. Now the rain
may be irrelevant to this choice of language: we are not likely
to find that Smith's expression will change with every squall.
56 CONTEXT PARAMETERS

The fact t h a t he is writing at midnight may be relevant at a

very delicate level of analysis: he may be tired, he may ex-
press himself less tersely and grammatically than he does
when he is fresh in the morning, he may be less able to avoid
undesirable individual idiosyncrasies. His typing might also
affect his language: had he dictated the text, his sentences
might have been more complex and rambling and contained
more embeddings, as happened to Henry James. Such features
are relevant at a very delicate level of analysis; yet such
delicate analyses may be called for in certain studies of
literary style. All the same, John Smith's choice of language
is much more likely to be decisively affected by the purpose
and recipient of the text he is composing. I t will make a great
deal of difference whether he is writing a love letter, a letter
to the Editor of The Times, a legal brief, or a note for his son
who has overspent his allowance. And the very fact t h a t he is
writing will also have a profound effect on his language: he
would put things differently were he speaking on the telephone.
One reason why it would be overambitious to begin by
attempting inventories of contextual universals is that a
taxonomy of contextual features is closely bound to each
culture and to each sociolinguistic situation. The contextual
spectrum of Eskimo is different from that of Yoruba, and
both differ from t h a t of English. Japanese is often cited for its
complicated and subtle systems of keigo or 'honorificlanguage',
whose distinctions readily strike Europeans as complicated.
Its various inflexional forms must be analysed in terms of
three-dimensional, or perhaps even multidimensional, system,
two of whose axes are the axis of attitudes to the conversation
partner (plain to polite) and the axis of attitudes to the sub-
ject matter (humble to exalted). To translate keigo styles into
European languages is far from easy. At a level of greater
delicacy, even the languages of Europe will be found to have
different contextual spectra. I t is therefore wise to start b y
assuming t h a t contextual taxonomies are culturally restricted
rather than universal. This also suggests that the categories of
 CONTEXT PARAMETERS 57

style of one language do not necessarily have exact equi-

valents in another language, and raises a host of questions
related, among other things, to the theory of translation.
In the extraction of stylistically significant contextual
features from the sociophysical envelope we shall, however,
need a conceptual frame and a terminology. And once this
frame can be made universal, general, and inclusive, the
features actually used in individual languages and styles can
be regarded as selections from it. So far, different scholars
have arrived at very different terminologies, though some-
times their terminologies disguise considerable similarities of
thought. For instance, I have here used the short and therefore
convenient term CONTEXT as a blanket both for the textual
and for the situational features in the envelope surrounding a
given linguistic unit. Others prefer to make a clear-cut dis-
tinction between SITUATION and CONTEXT, reserving the latter
f o r INTRATEXTTTAL CONTEXT o r CO-TEXT.
Some schools of linguists, including the Prague school and
many Soviet scholars, have written about functional styles in
a way which suggests that the "functions" in fact represent
major constellations of contextual features. The difficulty
with the term functional style has been that — even apart
from its mathematical senses — the term function has been
understood to mean very different things. First, one well-
known set of linguistic functions is that defined by Roman
JAKOBSON in his famous paper "Linguistics and poetics"
(JAKOBSON 1 9 6 0 ) . Here JAKOBSON listed six factors: addresser,
message, addressee, context, contact, and code, which are in-
alienably present in all verbal communication. Each factor
has a corresponding function. Though each message is likely
to fulfill more than one function at once, it is still characterized
by a hierarchic ordering of the six functions. Thus each text is
dominated by one of six universal functions: the COGNITIVE
or denotative or referential function related to context; the
EMOTIVE function focussed on the addresser; the CONATIVE
function oriented towards the addressee; the PHATIC function
58 CONTEXT PARAMETERS

with emphasis on contact; the METALINGUISTIC function

centred upon the code; and the POETIC function linked with
the message itself. Secondly, other adherents of the term
functional style have equated their functions, not with
JAKOBSON'S universals but with culturally conditioned, signi-
ficant contextual groupings very much akin to genres. Among
functional styles they list the styles of scientific, journalistic,
artistic, poetic, colloquial, and perhaps other types of commu-
nication. Thirdly, the meaning of the term function has been
diluted by yet another sense perhaps best describable as
"functional efficacy": a style is functional if it works efficiently
in a given situation.
I n another book (ENKVIST - SPENCER - GREGORY 1 9 6 4 ) I
tried to set up a relatively weakly ordered list of the contextual
features that may deserve consideration in stylistic analysis
and from among which the significant features could be ex-
tracted. The inventory looked as follows:

textual context
linguistic frame
phonetic context (voice quality, speech rate, etc.)
phonemic context
morphemic context (he sings / he singeth)
syntactic context (including sentence length and
complexity)
lexical context
punctuation, capitalization
compositional frame
beginning, middle, or end of utterance, paragraph,
poem, play, etc.
relationship of text to surrounding textual portions
metre, literary form, typographical arrangement
extratextual context
period
type of speech, literary genre, subject matter
speaker/writer
 CONTEXT PARAMETERS 59

listener/reader
relationship between speaker/writer and listener/reader in
terms of sex, age, familiarity, education, social class
and status, common stock of experience, etc.
context of situation and environment
gesture, physical action
dialect and language
I further noted that if items from this list are regarded as
stylistic characteristics, they must be omitted from the con-
text and transferred to the category of potential style markers.
Such unordered lists are, however, merely a starting-point
for context taxonomies. I n British linguistics, for instance,
some significant contributions have appeared in the past
several years. SPENCER and GREGORY are among those who
have emphasized the need for "placing" a text, first into its
proper historical and dialectal setting, and then by three con-
textual parameters called FIELD, MODE, and TENOR of discourse
(ENKVIST - SPENCER - GREGORY 1964: 85-91). FIELD re-
lates the discourse to its subject matter. Thus for instance an
article on nuclear physics and a love letter differ in field. In
long texts, the field may shift: in novels, for instance, a
novelist is likely to move from one field to another. By MODE,
SPENCER and GREGORY mean

the dimension which accounts for the linguistic differences which

result from the distinction between spoken and written discourse.

They carefully note t h a t in some texts such as poems, authors

may be very conscious of the spoken mode though they are in
fact writing, and t h a t the devices used b y novelists and
dramatists to provide an illusion of natural speech are p a r t of
the modal dimension. The difference between scripted and un-
scripted speech should also be noted here. The third parameter,
TENOR, reflects the relationship between the speaker/writer
and the listener/reader, chiefly in terms of degrees of formality
on a continuous scale between extreme formality and extreme
60 CONTEXT PARAMETERS

informality. These five dimensions — period, dialect, field,

mode, and tenor — of course also interact. For instance, field
is likely to affect tenor. An article on physics is likely to use
more formal tenors than a love-letter.
In their discussion of dimensions of situational constraint,
C R Y S T A L and D A V Y attempted a further refinement of similar
variables. Their system looks like this ( C R Y S T A L — D A V Y
1969: 66):

A. Individuality
Dialect
Time
B. Discourse
(a) [Simple/Complex] medium (speech, writing)
(b) [Simple/Complex] participation (monologue, dia-
logue )
C. Province
Status
Modality
Singularity

The features under A and B should explain themselves.

Under C, P R O V I N C E reflects occupational or professional
activity: the languages of public worship, advertising, science,
or law each have their own province. S T A T U S is the term for
the relative social standing of the communicants in terms of
formality, respect, politeness, intimacy, kinship, business re-
lations, and the like. M O D A L I T Y covers differences in the form
and medium of communication such as those between a letter
postcard, note, telegram, memo, lecture, report, essay, mono-
graph, or textbook. S I N G U L A R I T Y is a term for occasional,
personal idiosyncrasies which are said to differ from those
under "individuality" in that

the former are typically short, temporary, and manipulable, usually

being deliberately introduced into a situation to make a specific lingu-
istic contrast, whereas the latter are relatively continuous, permanent,
 CONTEXT PARAMETERS 61
and n o t able t o be manipulated in this w a y — in short, non-linguistic.
(CRYSTAL - DAVY 1 9 6 9 : 76)

To mention yet another attempt at an inventory of context

parameters: in a paper on the linguistic theory underlying the
teaching of Russian to foreigners, V I N O G R A D O V and K O S T O -
MAROV abandoned the usual inventory of functional styles
and set up a system of five, mainly binary, categories ( V I N O -
GRADOV - KOSTOMAROV 1 9 6 7 ) . They consisted of ( 1 ) means of
communication: speech, writing, gesture, (2) presence versus
absence of partner, (3) one-way or two-way flow of informa-
tion, (4) individual versus mass communication, and (5)
communication in contact or at a distance. With five binary
features we get a theoretical number of 25 = 32 possible com-
binations.
An area in need of further research is the ordering and
hierarchization of context parameters. For the rough group-
ings in examples such as the above should not be taken to
imply hierarchic ordering. As C R Y S T A L and D A V Y have
emphasized, certain patterns of co-occurrence, and thus of
redundancy, can readily be found. Legal language is usually
formal, conversational language is probably informal, and
legal language is very improbably ever colloquial. Observa-
tions of this kind suggest that a certain amount of hierarchi-
zation might be extracted out of large empirical materials on
the co-occurrence of contextual parameters, and t h a t certain
redundancy rules could be set up to indicate for instance t h a t
the feature [ + legal] makes the feature [ + formal] unneces-
sary to repeat. Still, we shall be likely to find t h a t studies of
context hierarchies will first of all necessitate a very sophisti-
cated set of definitions of the parameters (thus two lawyers
talking shop over a cup of coffee should not be said to use
legal language if legal language is to be formal, not colloquial).
And we may also find a large amount of overlap and fluctua-
tion, less in cultures with stable context categories and more
in societies with complex and fluctuating contexts and in
62 CONTEXT PARAMETERS

situations where individual idiosyncrasies are permitted or

even encouraged.
Some schools of linguists like to illustrate hierarchies with
numbered brackets or tree diagrams, and sets of unordered
features with matrices. The overlaps of stylistic categories
seem to be easier to handle in matrices than in trees.

4.4 LINGUISTIC DIVISION OF TEXT

INTO PORTIONS

It was noted above that the student of style is often able to

start out by comparing a well-defined text with a well-defined
norm. If so, his task is to find the linguistic features that make
the text significantly different from the norm. But in some
situations it is necessary to reverse the process, to start out
from an undivided and massive body of text, and to see
whether it can be divided into different portions by analysing
the distribution of its linguistic features. If these linguistically
distinct portions prove to correlate with contextual cate-
gories, their languages qualify as styles. Such investigations
will profit from operations such as factor analysis (Rratjs -
P o l A k 1967) a n d numerical t a x o n o m y ( C a r v e l l - S v a r t v i k
1969). A representative example of this approach is a paper
by A. Ja. Sajkevic. He asked the basic question, What
allows us actually to classify a given variant of language as a
style ? Is there an objective method for distinguishing styles
that makes it possible for us to assert that in a given body of
text (or, why not, in an entire language) there are, say, five
rather than six or ten or twenty "functional" styles? He then
went on to analyse a corpus of one million words out of 307
texts by Elizabethan and early Stuart authors: 161 texts
from plays (about half the corpus), 76 poetic texts (about one-
fifth of the corpus), the Bible (8 per cent), prose fiction by
Lyly and Deloney (5 texts; 4.5 per cent), Bacon (6 per cent),
travel descriptions (4 per cent), letters, Royal edicts, and so
on. By factor analysis by computer of the patterns of noun
 CONTEXT PARAMETERS 63

modification, SAJKEVIC obtained four major style categories:

a poetic group, a colloquial group, a factual group, and a re-
ligious or archaic group. The proportion of style markers
characteristic of each category could also be used to measure
the poetic, colloquial, factual, and archaic ingredients in
texts of any of the four major categories. This result was ob-
tained through purely statistical methods, and it may be
said to correlate well with rough-and-ready intuitive catego-
rizations of Elizabethan and Early Stuart prose (SAJKEVIC
1968). I t would, however, be interesting to know whether
criteria other than noun modification might have suggested a
different taxonomy of major styles. Another example of a
somewhat similar approach, though with less emphasis on
formal statistics, is the study of Plato's styles b y Holger
THESLEFF. By using close to one hundred grammatical and
rhetorical style markers and by studying the distribution of
a few hundred words, THESLEFF (1967) arrived at ten classes
or shades of style, as he called them. All of Plato's dialogues
contained more than one style, and some correlations could
be suggested between each style and contextual categories
such as speeches, myths, visions, mimetic play, parody, and
the like.
I t might be added in passing that the identification of con-
texts by their style markers is by no means merely an exercise
for linguists or stylostatisticians. On the contrary, it charac-
terizes many types of communication. When A speaks to B,
B will readily draw conclusions from A's classification of the
context: he will, for instance, note what level of formality
and politeness A regards suitable for speaking with him. And
this leads to a mirror effect: B's ideas of A's ideas of B will
affect B's views of A, and also influence B's classification of
the context and hence his style.
There are thus occasions where contexts can be consciously
defined and even manipulated through choices of style, which
may range on a scale between extreme flattery and extreme
rudeness. The relation between context and style thus works
64 CONTEXT PARAMETERS

both ways: the use of language may influence context, and

not only vice versa. A simple example. In Swedish, the recip-
rocal use of the familiar pronoun du has come to mark not
only familiarity but also contexts in which the communicants
wish to appear equal to emphasize in-group solidarity. From
use between members of the same profession, du has now
spread to more heterogeneous groups such as those working
in universities and in large offices. There are instances in
which heads of organizations have decreed the use of the
familiar pronoun obligatory in all oral communication within
the organization. That such decrees have been made indicates
a belief in the power of style markers to define contexts: the
universal use of a familiar pronoun within an organization
presumably makes people happy because it marks contexts
in which communicants have equal status. How long the effect
of such reforms-by-decree will last is, unfortunately, another
matter. If the use of, say, a familiar pronoun becomes uni-
versal, this pronoun will inevitably lose its power to function
as a style marker and thus to identify contexts. In such in-
stances, languages are likely to develop new means to mark
those differences in context that a tacit consensus — or
"usage" — finds in need of marking.

4.5 GROUP CONTEXTS. SLANG

This kind of linguistic dynamism in the realm of style leads us

to the origin and formation of SLANG as well. Here, words play
a decisive role as markers. Slang might be said to arise out
of a given group's need to form a style of its own, again
to mark solidarity within the group. If suitable markers for
such an IN-GROUP STYLE do not already exist, they have
to be created. The means are the same as in word forma-
tion. Old words can be used in new meanings and new con-
texts and collocations. New words arise through affixation.
Loans can be brought in from other regional or social dialects,
from neighbouring languages, and from other languages
 CONTEXT PARAMETERS 65

familiar to the group. Even root-creation may occur in slang.

Particularly common are phonological neologisms such
as abbreviations and clippings. I n favourable situations,
where the in-group is large and has a strong need for de-
monstrations of solidarity, the slang may spread very rapidly.
B u t if a style-marker of slang is so successful t h a t it passes
into the language of other groups, or into general use, it loses
its original function. I t s originators m a y therefore have to
adopt new slang terms. Another reason why slang m a y be
ephemeral is t h a t many groups are short-lived: more lasting
groups tend to have more lasting slang.
SLANG is a relative term, not an absolute: the distinction
between concepts such as 'slang', 'professional language', and
'sociolect' depends on our classification of groups of people
(in-group, profession, social class) r a t h e r t h a n on linguistic
criteria. If a certain subvariety of language which was original-
ly used by members of a restricted in-group when communicat-
ing within itself spreads t o a whole social class, it stops being
slang and becomes a sociolect. I n turn, a sociolect may acquire
more prestige t h a n other sociolects and begin spreading, and
even become a standard language. Once again, such con-
siderations show the futility of setting u p rigid borders
between styles, slang, sociolects, and other linguistic diatypes.
Among those who must closely follow the dynamics of these
processes and keenly watch linguistic fashions are advertisers,
especially those who appeal to groups such as teen-agers.
Thus b y the summer of 1970, former high-frequency expres-
sions such as jet set, the adjective swinging, and to be with it
were obsolescent in advertisements of charter-flight holidays,
for young British customers.

4.6 DYNAMISM IN CONTEXTUAL CATEGORIZATION

The lesson to be drawn from such considerations is t h a t rela-

tions between contexts and styles must not be regarded as.
66 CONTEXT PARAMETERS

constant. In fact both the sociophysical envelope and the

language used in á given contextual constellation are always
subject to change. This means that the contextual probabilities
of linguistic features are affected both by changes in contexts
and by changes in language. All those interested in styles,
whether linguists, literary scholars, advertisers, or others,
should remember that in stylistics, too, strict synchrony and
static views remain simplifications of a dynamic reality.
Looking at styles as constants is a very useful working hypo-
thesis which helps us to isolate features from their complex
environment and thus to see them more clearly. But in styl-
istics, as in all linguistics, the complete view should be a pan-
chronic synthesis of synchronic and diachronic considerations.
Special problems can also be caused by SHIFTS OF STYLE
within a text. Some such shifts can be readily correlated with
changes in context. In a novel, for instance, we are often able
to distinguish between dialogue and non-dialogue, or between
passages dominated by action and passages dominated by
background description. Such contextually delimited portions
of the text may have styles of their own. But sometimes
writers and speakers achieve stylistic effects by changing
their style without overt changes in context. It then becomes
the reader's and listener's task to note that the relations
between context and language have changed. Thus a writer
may rise to a vigorously polemic climax after a calmly factual
and deliberative passage. This may be reflected in an increase
of certain syntactic features such as rhetorical questions and
imperatives, and a higher incidence of evaluative or emotion-
ally charged words. Another kind of example is available in
Joyce's TJlysses in that famous passage where Joyce imitates
a number of English literary styles from Old English to
modern times. To grasp the effect of this passage, the reader
must be capable of associating Joyce's variants of language
with their proper diachronic contexts.
 5

GRAMMATICAL MODELS
I N T H E D E S C R I P T I O N OF S T Y L E M A R K E R S

5.1 STYLE AS A DIFFERENTIAL

Though styles are varieties of language that correlate with

specific constellations of contextual features, stylistic analysis
must always be based on comparison, tacit or explicit. I f we
define a body of text by contextual criteria — say, a set of
political leaders in The Daily Mirror — we can, of course,
study its language and describe it in any manner we may wish.
But such a description remains an account of the language,
not the style, of this corpus. I f we wish to spot those linguistic
features that nlake political leaders in The Daily Mirror dif-
ferent from other kinds of texts — such as other types of
articles in the same paper, political leaders in The Times, or
any other texts that we define as relevant and sensible norms
of comparison — we must match the densities of the linguistic
features in our text against the densities of the corresponding
features in the norm. Comparison is the only key to stylistic
differentials, that is, to the style markers that characterize our
text as different from other texts. And all stylistic descriptions
must begin with an inventory of style markers.
This point has sometimes been obscured because so many
students of style have operated with tacit rather than with
explicit comparisons. Their norm has not been explicitly
circumscribed and defined. I t has remained a product of past
experience of the uses of language in comparable contexts,
and the comparison itself has been performed intuitively in
the sense that some or all of the materials on which it was
based are irrecoverable. Such tacit comparisons are very often
68 GRAMMATICAL MODELS

expressed in impressionistic or metaphoric terms. An ex-

perienced reader may thus spot the style of a given text as
" h e a v y " . But he may not be able at once to describe what
makes the text "heavy", whether the "heaviness" owes to a
high density of difficult or long words, to long and complex
sentences, to numerous embeddings, or to something else.
A " h e a v y " text may also be "lighter" than many other texts,
b u t still be "heavy" for its context: a sentence which strikes
us as relatively short in a legal contract may appear very long
on a postcard. As always, the norm, too, enters into the picture.

5.2 REQUIREMENTS FOR GRAMMATICAL MODELS

Such considerations help us to formulate the requirements

t h a t a model of language has to satisfy if it is to give us a
good basis for stylolinguistic description. We might posit
four basic requirements. First, the model has to allow for
linguistic VARIATION and admit systematic description and
classification of contextual categories. The second requirement
demands CONSISTENCY: both text and norm have to be
described in terms sufficiently similar and consistent to permit
comparison. The third requirement has to do with ADEQUACY:
the linguistic model chosen must be capable of describing all
relevant style markers. The fourth requirement was dealt
with in sections 3.2 and 3.3: the model should admit both
categorical and probabilistic rules.
We have already seen t h a t linguists have made some be-
ginnings towards the setting-up of taxonomic systems of con-
t e x t parameters. The purpose of such systems is to make pos-
sible the extraction of the stylistically significant contextual
features out of the endlessly varied cultural and sociophysical
envelope that surrounds language. Granted, the systems
available today may well be characterized as sketchy, rough,
and provisional. Still they are adequate for a number of
practical tasks in the analysis and description of styles. These
systems are not a prerogative of one or another linguistic
 GRAMMATICAL MODELS 69

movement or grammatical school. They may be used together

with various grammatical models, though some models may
well prove more hospitable to them than others.
The requirements of consistency and adequacy are even
more directly relevant to the choice of linguistic models in the
description of style markers and styles. Consistency can be
dismissed briefly. As style is based on comparison, the de-
scriptions of text and norm must be commensurable. Both
must be described in terms sufficiently similar to allow com-
parison. I have already touched on this point above in section
3.4. I f the text and norm are described separately according
to rigid principles of immanence, and if the two descriptions
are to result in a maximally economic presentation of features
overtly and manifestly present on the surface of the text and
the norm, respectively, there is no guarantee that the two de-
scriptions are fully consistent. Accidental differences of sur-
face structure may lead to great differences in basic levels of
the description, because certain structures did not happen to
be overtly present in the text or in the norm. Text and norm,
then, have to be described in sufficient depth and at optimal lev-
els of abstraction and delicacy to allow for maximally meaning-
ful comparison. Of course no levels of maximal meaningfulness
can be set universally and a priori. For one task, one level is
best; for another task, another level may be more rewarding.
The requirement of adequacy demands that the model be
capable of describing all potential style markers. All linguistic
features, as well as all combinations of linguistic features, in
fact qualify as stylistic discriminators. In practice, however,
we cannot sit back to wait for the ultimate perfection of
grammar before we venture to describe styles. On the contrary,
from the point of view of stylistics alone, any grammar that
does the job is by definition adequate. This does not of course
contradict the obvious fact that some grammatical models are
more satisfactory than others. It only means that a number
of style markers are linguistic features of a kind that can be
described even with the aid of relatively simple, crude, and
70 GRAMMATICAL MODELS

unsophisticated grammars. We do not, for instance, need

particularly advanced grammars to study sentence length in
written texts, numbers of finite verbs per sentence or per
word, the ratio of adjectives per substantive (always assuming
t h a t our grammar gives exact, formal definitions of 'finite
verb', 'sentence', 'adjective', 'substantive', and other such
concepts), the density of passives, certain basic varieties of
word-order patterns, and the like. Those who need an excuse
for the crudeness of their grammatical apparatus may define
their stylistics as a branch of applied linguistics, where the
eclectic choice of simple approaches is a virtue and not a vice.
On the other hand, those who analyse styles may not wish to
stop at stylistic analysis. They may also wish to relate their
findings to certain linguistic theories or to certain full de-
scriptions of the language. If so, they should choose a model
which makes this as easy as possible, even when this may
complicate the definition of some individual style markers.
For some tasks, simple grammars will thus prove adequate.
In other tasks, we may at once come u p against the frontiers
of current grammar and even of linguistics proper. If, for in-
stance, the text and the norm are suspected to differ because
one might have different metaphors from the other, we shall
need a linguistic model capable of describing different types
of metaphor. If the text seems to be characterized by a
certain pattern of thematic progression, the grammar adopted
must offer a system for the description of thematic movements
from one sentence to the next. And what about irony? Ex-
perience tells us that a high density of ironical statements may
well be a characteristic of a given style. B u t to what extent
is irony — always including those irony signals which warn
us to interpret a passage ironically rather than literally —
definable by strictly linguistic methods? We should be duly
encouraged by the fact t h a t in recent years, linguistics has
expanded rapidly and become increasingly concerned with
problems of these kinds. Still, many features of great interest
to students of style lie well beyond the current borders of
 GRAMMATICAL MODELS 71

rigorous linguistics. Meanwhile, work must go on. If the

student of style cannot find what he needs in the linguistic
models that are available, he has no choice but to look for
help elsewhere, for instance in traditional rhetoric and seman-
tics. And if this fails, he will have to devise concepts and
methods of his own.

5.3 S T Y L E I N SOME G R A M M A T I C A L MODELS

Such liberal views have occasionally seemed vague, unscholar-

ly, and therefore repugnant to those linguists who are de-
dicated to one grammatical model to the exclusion of all
others. But even today, the world of grammar remains a
pluralistic one. Linguists have not universally and unanimous-
ly agreed to regard any single grammatical model as superior
to all others. On the contrary, several schools all speak for
their own models. Even the much-reviled traditional grammar
has been partially rehabilitated by some recent developments,
though linguists are in reasonable agreement about its weak-
nesses and shortcomings. Tagmemic grammar is one of the
major movements continuing the traditions of behaviour-
ist structuralism, to which linguistics owes more than many
tend t o admit today. H j E L M S L E v i a n glossematics has contri-
buted not only valuable theoretical foundations b u t also some
concrete descriptions of French and Danish. In Britain, the
systemic grammar of M. A. K . HAT,T,TDAY and others has
gone through various stages of development in the past
decade or so. Generative-transformational grammar has con-
tinued its triumphs ever since C H O M S K Y ' S Syntactic Structures
of 1957, but also given rise to new offshoots such as the deep-
semantic transformationalism of L A K O F F , M C C A W L E Y , R O S S ,
and others, and the case grammar of FILLMORE.1 I t also has a
1
S e e LAKOFF ( 1 9 6 9 ) , MCCAWLEY ( 1 9 6 8 ) , R o s s (1970), FILLMORE
(1969), a n d t h e readers BACH-HARMS (1968), FODOB-KATZ (1964),
JACOBS - ROSENBATTM ( 1 9 7 0 ) , KIEFER ( 1 9 6 9 ) , a n d REIBEL - SCHANE
(1969).
72 GRAMMATICAL, MODELS

rival in LAMB'Sstratificational grammar. In the Soviet Union,

SATJMJAN and his followers advocate their own variety of
generative grammar which they call "the Applicational
Model". To such lists, others might wish to add yet others
of their favourite models of grammar.
I t would take a library, not a book, to run through all these
models, one after the other, and try to assess all of their
relative advantages and disadvantages in solving a wide
spectrum of different types of linguistic problems. All I wish
to attempt here is a very brief discussion of a few points
particularly relevant to stylistics.
In traditional grammar, the border towards rhetoric, and
thus also towards stylistics, was left open. I t was up to each
grammarian to decide how much stylistics and rhetoric he
wanted to p u t into his grammar. And, the other way round,
rhetoricians were free t o make what forays they liked into
grammatical territory. This freedom was bought at the cost
of weak systematization. Grammarians dealt with those
features of a language t h a t seemed important or interesting.
They were at liberty to ignore the rest. Then, one of the
common aims of traditional grammarians was to provide
norms for writers and speakers. This, too, led to an intimacy
between grammar and stylistics. In fact many of our tradi-
tional grammars, including a host of old school grammars,
are treatises on style in so far as they mark certain structu-
res as appropriate to certain occasions, and brand others as
unsuitable (often in terms of 'right' and 'wrong').
Behaviourist-structuralist grammars were usually preoc-
cupied with an objective description of the language actually
occurring in a definite corpus. The corpus could consist of
spoken or colloquial language, as in the grammars of Charles
Carpenter FRIES. The restrictions placed on the corpus
readily turned the grammars into descriptions of one style
rather than into comparative analyses of whole ranges of dif-
ferent styles. Other circumstances contributed to the com-
parative neglect of stylistic considerations. Structuralists
 GRAMMATICAL MODELS 73

were anti-normative. They insisted on immanent descriptions,

which did not encourage the comparison of different texts.
A n d t h e y were preoccupied with the surface of language, and
even there with the smaller units such as phonemes and mor-
phemes. This also led them away f r o m style, which resides
in the larger units as well as in t h e small ones.
Though the behaviourist-structuralist model was poorly
adaptable to stylolinguistic description, this did not mean t h a t
its adherents wholly neglected stylistics. Though BLOOMFIELD
himself avoided the word style in his classic Language (BLOOM-
FIELD 1933), m a n y of his pupils a n d followers wrote about it.
B e r n a r d BLOCH elegantly a n d suggestively defined style as

t h e m e s s a g e carried b y t h e frequency distributions a n d transition

probabilities of its linguistic features, especially as t h e y differ f r o m
those of t h e same features in language as a whole, (BLOCH 1953: 40)

I n Zellig HARRIS'S papers, including those on discourse

analysis, keys to style are sought in distribution. (HARRIS
1952a, 1952b). Martin J o o s ' s Five Clocks (Joos 1962) contains
a wealth of acute observations as well as an a t t e m p t t o s e t u p
five m a j o r style categories for English: frozen, formal, consulta-
tive, casual, a n d intimate. Archibald A. H I L L has written wide-
ly a n d incisively about ways of analysing literature with a
linguistic apparatus, and also about the theoretical principles
of stylistics. And K e n n e t h P I K E has been concerned with t h e
place of language in a wider frame of patterns of human be-
haviour, which led him to set u p concepts and categories
useful to context analysts (PIKE 1967). B u t in spite of efforts
such as these, it would be a distortion t o say t h a t the focus of
behaviourist structuralism lay on stylolinguistics.
I have already noted t h a t the early formats of transforma-
tional grammar were hardly more hospitable to style. Trans-
formationalists tended to react against the behaviourists' pre-
occupation with the surface analysis of limited corpora, a n d
this also prejudiced them against statistics. B u t I have also
74 GRAMMATICAL MODELS

emphasized that we may count the rules and transformations

that went into a given text, and that LABOV'S variable-input
rules in fact are an attempt to turn the originally categorical
model into a probabilistic one better suited to the description
of linguistic, including stylistic, variation. And transforma-
tionalists have yet other strings to their bow. For instance,
they have helped to modify our attitudes to the problem of
choice, which has bedevilled stylistics for a long time.
Style, many experts have maintained, arises as a result of
CHOICE as a preference of one feature to another. However,
there are other types of choice in language, in addition to
stylistic choice. There is, first of all, PRAGMATIC CHOICE,
which is based on the preference of one utterance to another
because of its truth value. Thus a person may prefer to say it
is snowing, not it is raining, because the former is true and
the latter untrue (or the other way round, if he wants to mis-
lead his hearers). Another type of linguistic choice is GRAM-
MATICAL CHOICE: English people prefer to say he is here, not
he am here or he are here, choosing is because they wish to con-
form to the patterns of well-formed English. Neither of these
types is identical with STYLISTIC CHOICE. Therefore those
who wish to define style as the result of choice must also face
the task of distinguishing stylistic choice from pragmatic and
grammatical choice. This is far from easy. One criterion is
given by the study of correlations: choices that correlate only
with context are stylistic, whereas choices that correlate at
least with expression content, and perhaps also with context,
are pragmatic. As features with different expression content
mean different things, this might be rephrased by saying that
pragmatic choice takes place between features that have dif-
ferent meanings, whereas stylistic choice takes place between
features which mean the same. "Mean the same", that is, apart
from their stylistic meaning, which must still be regarded as
part of the total meaning of an utterance. An example of
stylistic choice would be the use of he is a fine man rather
than he is a nice chap.
 GRAMMATICAL MODELS 75

The view of style as choice has a lot to commend it. Among

other things it agrees nicely with our intuition: we often feel
how the choice of a certain word instead of another gives a
definite stylistic texture and flavour to a sentence or passage.
And when struggling for stylistic effect we often consciously
seek le mot juste from among a set of words which all satisfy
the pragmatic and grammatical requirements we wish t o
satisfy. There are two difficulties, however. First, with our
present equipment it is hard to say which formal features —
words or structures — mean sufficiently the same — or which
correlations exist between formal features and their content
— t o qualify as stylistic, not pragmatic, variants. Our
machinery of semantic analysis and description is rarely
accurate and sensitive enough to provide us with firm infor-
mation of this kind. In fact, the flow of information has so
far been from stylistics and from the study of items in context
to semantics rather than the other way round. Secondly,
those grammatical models t h a t dealt exclusively with the
surface of a text were incapable of mapping out the processes
of choice t h a t underlay that surface. When the linguist saw
the text, all choices had already been made. Choice, said the
linguist, was beyond conjecture, and speculating about it was
vacuous, rank mentalism.
If we ever have a model of language which allows us to
chart all processes of choice and to list all the alternatives
rejected and paths not taken, this objection will lose its
main point. I n principle, generative grammars ought to be
capable of giving us a generative programme in which a large
number of choices, except those of the deepest and most fun-
damental pragmatic meanings, are made explicit in the
rules. I t is another matter to say which type of generative
grammar will best realize this formidable requirement. One
might, for instance, indulge in speculations about deep
semantic components generating prestylistic meanings, which
form the input into a syntactic and lexical component sensi-
tive to the textual and situational context of the utterance.
76 GRAMMATICAL MODELS

Deep meanings would t h e n be m a r k e d with the a p p r o p r i a t e

matrix of constellations of stylistically relevant contextual
features, which would either trigger off t h e generative pro-
cesses characteristic of t h e appropriate style, or define a set
of constraints preventing stylistically inappropriate transfor-
mations f r o m taking place. I f such inappropriate t r a n s f o r m a -
tions actually do take place, the result will be a breach of
style, a venture beyond t h e margin of tolerance. As I have
noted before, style-sensitive rules m a y be of two kinds: some
m a y be categorical and s t a t e t h a t in a given style a given rule
must operate either always or never, whereas others m u s t b e
probabilistic a n d have a variable input which activates t h e
rule in a certain percentage of the instances in which it could
be activated.
In the light of such r a t h e r fanciful speculation it is interest-
ing to note a nascent addition to t h e handling of some, t h o u g h
strictly limited, contextual features in transformational
grammar. W e owe this addition to J o h n R o s s . Briefly, in
transformation grammar it has long been customary t o r e g a r d
certain constructions as results of deletion. An imperative
such as Go! is t h u s a deletion of an underlying You go. Bor-
rowing the concept of performatives f r o m A U S T I N ( 1 9 6 2 ) ,
R o s s t u r n e d these deletions into a system by positing t h e
existence of a set of ABSTRACT P E R F O R M A T I V E V E R B S . T h u s t h e
sentence Prices slumped is a surface representation of an u n d e r -
lying structure with an a b s t r a c t performative verb. (Cf. p. 77)
If this abstract structure were t o be verbalized as such, i t
would read something like I tell you: prices slumped. The
performative superstructure I tell you is then deleted, a n d
the o u t p u t will be Prices slumped.
I n his paper on declarative sentences R o s s gives f o u r t e e n
reasons in support of t h i s particular b r a n d of abstract syn-
t a x (Ross 1970). These reasons are not primarily stylistic
in the sense of having been devised expressly to bring in
features f r o m the sociophysical envelope or f r o m the stylistic
context into grammar. They have t o do with syntactic f e a t u r e s
 GRAMMATICAL MODELS 77

such as the use of reflexive pronouns. Nevertheless it is clear

t h a t the performative superstructure refers to certain areas
of situational context b y referring to the speaker/writer, the
listener/reader, and the "modality" of the sentence (its charac-

NP VP

V NP NP
-f- verb I I
+ performative you S
+ communicative
+ linguistic N p y p
declarative
prices V
I
slumped

ter of statement, question, c o m m a n d , etc.). Ross's performa-

tives might therefore be said to bring in some features of con-
text into a deep layer of generative-transformational grammar.
Ross himself has emphasized that one of the alternatives to
P E R F O R M A T I V E ANALYSIS W O u l d b e PRAGMATIC A N A L Y S I S . I n
the latter, the underlying form of a s e n t e n c e such as Prices
slumped w o u l d be, simply, Prices slumped, plus — instead o f
the performative — a list of the contextual features present
in the speech act. This list must have a form t h a t is capable
of triggering off or constraining syntactic processes. Its fea-
tures must also be isomorphic with the system of performatives
because they must be capable of doing everything perform-
atives can do.
A precise [pragmatic] theory would have to specify formally what
features of the infinite set of possible contexts can be of linguistic
relevance. Furthermore, these features would have to be described
with the same primes which are used for the description of syntactic
elements, so that rules which range over syntactic elements will also
range over them. While such a theory can be envisioned, and may even
78 GRAMMATICAL MODELS

eventually prove to be necessary, it is obvious that it does not exist

at present [. . .] If the pragmatic analysis is to be carried through,
contexts must be assumed to have the structure of clauses: they must
have elements which share properties with subject NP's, elements
which share properties with indirect object NP's, and elements which
share properties with verbs of saying. Furthermore, if Lakoff and I
are correct in our claim that questions are to be derived from structures
roughly paraphrasable by I request of you that you tell me . .] then
contexts also exhibit properties of syntactic constructions with em-
bedded clauses. However, while such observations may be interesting,
they only serve to illustrate the enormous gap between what can now
be said about contexts in fairly precise terms and what would have
to be said in any theory which could provide a detailed understanding
of language use. (Ross 1970: 257-8)

The first sentence of the quotation will sound familiar enough

to those who have read this book from the beginning. The re-
mainder is concerned with the form t h a t contextual features
must assume if they are to fit Ross's type of grammar. With-
out minimizing the gap between syntax and stylistics one
may still find it both interesting and promising t h a t certain
syntactic reasons have motivated the introduction of some
contextual features into a deep component of syntax. Optim-
istic students of style might even be tempted to interpret per-
formative analysis as another tiny step from grammatical
theory towards stylistics. Certainly this type of inclusion of
contextual features in a deep, possibly semantic component
of generative grammar is more elegant and economical than
a device which would, say, generate all possible stylistic
variants of a given deep sentence, and then eliminate those
that do not fit a given contextual matrix by straining the
lot through a stylistic filter.
Transformational grammar has suggested further refine-
ments t h a t are potentially useful in the description of styles.
We shall see in another chapter how it has modified our atti-
tudes to deviant expressions. In the analysis of a m b i g u i t i e s ,
transformational grammar has suggested a clear theory: a
given sentence is ambiguous n ways if it can be generated in
 GRAMMATICAL MODELS 79

n different ways. We should recall t h a t a t least since William

Empson, ambiguities have been an important concern for
students of literary styles. ELLIPSIS is another of the pheno-
mena t h a t can now be studied with more stringent methods
t h a n before. Structuralists could do little with ellipsis because
they insisted t h a t description be restricted to features actually
present on the surface of the t e x t : they failed to extend their
use of the zero to the word level or above, though they used it
freely in morphemics. And in the study of METAPHOR, which
was one of the most recalcitrant problems of stylolinguistics,
transformational grammar has also been beneficial. I shall
r e t u r n to this below.

5.4 TRANSFORMATIONAL GRAMMAR IN STYLISTICS

A pioneering a t t e m p t a t using transformational grammar in

t h e analysis of styles was t h a t of Richard OHMANN (OHMANN
1964). He found certain uses of transformations character-
istic of certain writers. His method was t o go down to deeper
structures f r o m the surface of his texts b y reversing the pro-
cesses through which these texts h a d been generated. I n the
terminology of the early 'sixties, OHMANN was reconstructing
the kernel sentences, a n d listing and counting the optional
transformations between the kernel sentences and the t e x t u a l
surface. He found t h a t Faulkner's apparent complexity was
the restflt of surprisingly few — in fact mainly three — dif-
ferent transformations, all of which were additive in t h a t they
combined such kernel sentences as shared a t least one mor-
pheme. High frequencies of relativization, conjunction, a n d
comparison were thus markers of Faulkner's style. The
crucial features of a passage of Hemingway consisted of trans-
formations t h a t turn kernels into style indirect libre or erlebte
'Rede. I n H e n r y James, t h e embedded elements outweighed
the main sentence, and his style seemed t o build on the
positioning of structures rather t h a n on the qualities of the
80 GRAMMATICAL MODELS

structures as such. D. H . Lawrence was OHMANN'S example

of a writer whose style built largely on deletion. To repeat:
OHMANN had thus demonstrated the applicability of gene-
rative-transformational grammar to the description of styles,
irrespective of the fact t h a t neither style nor frequencies had
occupied places of their own in transformational theory. The
patterns of frequency with which each writer had used certain
optional transformations and types of transformations proved
to be explicit, concretely describable style markers capable of
distinguishing between Faulkner, Hemingway, Henry James,
and D. H. Lawrence.
Since OHMANN wrote his paper, there have been several
other applications of transformational grammar to the de-
scription of styles (e.g. H A Y E S 1 9 6 8 ) . Also, as we have noted,
the transformational model has gone a long way since the late
'fifties and early 'sixties. Still the comparatively simple prin-
ciples of the early format (CHOMSKY 1 9 5 7 ) , however unsatis-
factory they may seem in the light of later revisions, were quite
adequate for the analysis of a number of stylistic problems.
Distinctions such as those between kernels and transforms,
and between obligatory and optional transformations seemed
to harmonize nicely with some stylistic concepts and methods
of analysis. Thus a text could be reduced to its underlying
kernel sentences, which gave a reasonably well-defined
starting-point for descriptions of what a given speaker/writer
had done to produce the final text, even if one might hesitate
before equating the string of kernel sentences directly with
stylistically neutral or prestylistic expression. And the
contrast between obligatory and optional transformations
sounded familiar, and perhaps even congenial, to those who
liked to define style as choice from among the optional
features of language. Those students of style who wish to use
transformational models to describe a strictly limited range
of relatively simple syntactic structures may, in their practical
work, come to regret some of the later developments in trans-
formational formats. They might even wish to return to
 GRAMMATICAL MODELS 81

simpler models by omitting a number of such complications as

are irrelevant to their particular problem. Here, too, the
student of style may wish to feel free to use the simplest gram-
matical model t h a t satisfies the requirements of his particular
job.

5.5 SYSTEMIC GRAMMAR IN STYLISTICS

I n recent British linguistics, the concern with style and

register as well as with the linguistics of utterances has been
more explicit, as one can expect of a school t h a t acknowledges
J . R . FIRTH as a major stimulus. I have already cited a number
of examples of British discussions of contexts and context
parameters, styles, and registers, and of fields, modes, and
tenors of discourse. Some British achievements are particular-
ly relevant to all students of stylistics. One is the theory
developed by M. A. K . HALLIDAY and his collaborators. Here,
languages are said to operate with four basic categories: unit,
structure, class, and system. Units are arranged into struc-
tures, which are descriptions of syntagmatic strings. The
category of class contains arrangements of units according to
the way they operate. And the category of system is an in-
ventory of those limited possibilities of choice that a speaker/
writer has at a given place of the structure of language, the
whole of a language being thus conceived of as a system of
systems. HALLIDAY'S model contains a number of other con-
cepts vital to students of style as well, among them scale,
cline, and delicacy. I t is also strong on choice. I n transfor-
mation grammar, all choices are usually built into a single
system, which is split up into a very few components. I n
HALLIDAY'S scale-and-category or systemic grammar, related
choices are built into limited subsystems of their own. I n
practical application, this makes it easier to extract the sub-
systems one happens to need out of a systems model than out
of a transformational model: one can use the relevant sub-
system as an entity without having to involve oneself in more
82 GRAMMATICAL MODELS

extensive and perhaps tenuous considerations. HAT.TJDAY'S

theory has also integrated certain essential areas of language
t h a t other models have found hard to include, among them
transitivity and theme (HALLIDAY 1 9 6 7 - 6 8 ) . Its concepts have
stimulated work in some fields t h a t seem increasingly relevant
to stylolinguistics, among them the cohesion of sentences in
texts (HASAST 1 9 6 8 ) .
Another major feature of today's British linguistics is the
rigorous processing of large masses of text. Professor Randolph
QUIRK'S Survey of English Usage at University College,
London, has already resulted in a number of observations on
usage that all students of English styles will find interesting.
The methods of the Survey also contain relevant elements,
among them the study of serial relationship which helps us
to grasp the full complexities of linguistic subcategorization
and gives us a means of contrasting actual occurrences with
the range of theoretical choices. Problems of grammati-
cality and acceptability have been studied by means of
elicitation experiments rather than through introspection
(QUIRK - SVARTVIK 1966, GREENBAUM - QUIRK 1970), and
the surveyors have also been concerned with prosodic and
paralinguistie features in spoken texts (CRYSTAL — QUIRK
1 9 6 4 ) . These studies are worth the attention of all who in-
vestigate styles of English, partly because of the light they
throw directly on subvarieties of English and their use, and
partly because they offer concepts and methods that might be
found suggestive in the analysis and description of styles.

6.6 FROM LINGUISTIC FEATURE TO GRAMMATICAL

MODEL. WORDS, RHETORICAL FIGURES

I have now briefly glanced at some prominent models of

grammar, not because they are the only ones t h a t should be
considered or the best, but because they show t h a t different
 GRAMMATICAL MODELS 83

models may well prove to have different advantages in the

linguistic description of style markers. I n one model, a given
area of grammar may have been more fully developed than in
another; in one model, a set of stylistically relevant, and t h u s
related, problems may have been kept together within one
subsystem, whereas another model may have scattered t h e m
into different parts of the grammar and thus made their con-
centrated treatment more difficult.
There is, of course, another approach to the same problem:
instead of taking a grammatical model and discussing how
well suited it is to the description of certain types of potential
style markers, we may take a number of style markers and
then examine a number of different grammatical models to
see how they might describe them. I n fact, potential style
markers can be found throughout the whole spectrum of
language. They range from paralinguistic, phonetic, and
phonemic features all the way through the lexis to the study
of sentences, including word order and sentence length and
complexity, and of textual as well as narrative units. A full
discussion of such problems would therefore lead to a recapi-
tulation of all of today's linguistics in the widest and most
comprehensive sense. Once again, the following remarks
amount to no more than a few hints at a few of many per-
tinent issues, beginning with a glance at lexicography.
A branch of stylolinguistic description conspicuous even t o
the non-linguist is the lexicographic categorization of words
according to their stylistic value. Users of dictionaries are
familiar with markings such as colloq., hist, or rhet. Usually,
however, such markings point out special terms — military,
scientific, technical, historical, rhetorical, and so on — and the
marking of stylistic values of more general words is less
specific, though certain labels such as 'colloquial' or 'slang' are
often used. I know of no dictionary t h a t would have a t t e m p t e d
an elaborate classification of contextual categories and a con-
comitant system of stylistic markings t h a t would correlate
with specific ranges of contexts. Rather, everyday experience!
84 GRAMMATICAL MODELS

suggests that any educated native speaker will be able to

supplement dictionary entries with a wealth of further obser-
vations on the range of contexts in which a given word is most
likely to occur. Today there is increasing agreement that
dictionaries should not be compiled in a linguistic vacuum,
that is, independently of linguistic theory. But the linguistic
principles of lexicography tend to vary with each school of
linguists — witness for instance the markings for strict sub-
categorization and for selection required by a lexicon in the
model of CHOMSKY ( 1 9 6 5 ) — and with each major theory.
This may suggest that dictionaries marking frequent con-
textual ranges of occurrence should be produced especially
for the needs of applied stylistics, for instance in foreign-
language teaching. Frequency dictionaries — now usually
produced by computer — can be excellent aids. Overall fre-
quencies computed out of large samples tend to permit few
conclusions as to the contextual range of words, but some
frequency dictionaries (e.g. KTTCERA - FRANCIS 1 9 6 7 ) offer data
by genre as well. Concordances, too, invite to the study of
word frequencies in certain well-defined literary contexts. Some
are devoted to restricted genres such as Old English poetry;
more commonly, they cover the works of one great writer,
usually a poet. In many stylistic problems, the great historical
dictionaries with their wealth of examples (such as the Oxford
English Dictionary) also offer some help. Nor should the
student of style forget the special dictionaries of slang or of
special subjects such as medicine or engineering, or the studies
of special word fields (e.g. CROSLAND 1 9 6 2 ; G L A S E R 1 9 6 3 ,
1970a; GUILBERT 1 9 6 5 , 1 9 6 8 ; LEECH 1 9 6 6 ; W E X L E R 1 9 5 5 ) .
Students of older stages of languages might even profit from
the markings of words in old dictionaries, which often reveal
stylistic values and attitudes (OSSELTON 1 9 5 8 ) .
SIMILES, METONYMIES, METAPHORS, and other figures usually
counted as rhetorical must not be left out from any list of
potential style markers. Sometimes it suffices to reckon with
their overall densities; often, however, a given style is char-
 GRAMMATICAL MODELS 85

acterized not merely by frequent metaphors but by a high

density of certain specific types of metaphors. METAPHOR
CLASSIFICATION will then become necessary (cf. BICKERTON
1969, P A S I N I 1968). Two main avenues are open here. Either
we set up a syntactic taxonomy of metaphors in the manner of
Christine BROOKE-ROSE, staying close to the syntactic surface
(BROOKE-ROSE 1958). Or we operate with semantic matrices,
trying to show what components of meaning the LITERAL
TERM and the METAPHORIC TERM have in common. For in-
stance, in John is a lion, both the literal John and the meta-
phoric lion are [ + noun], [-f animate], and, apparently,
[ + brave]. But John is [ + human], lion is [— human]. We
may proceed to label the set of semantic components that are
shared by the literal and the metaphoric term (or, in terms of
set theory, the intersection) as METAPHORIC CONSTRAINTS, and
those different (the difference set) as METAPHORIC DIFFEREN-
TIALS. In these terms we may find, for instance, that the style
of a certain poet is characterized by a high density of certain
specific metaphoric differentials. This high density becomes
a style marker if it is significantly higher than the correspond-
ing density in the norm against which we wish to view our
poet's works. A frequent metaphoric differential in Emily
Dickinson's poems, to take just one example, is divine].
Of course the student of style-marking metaphors will need a
stringent apparatus of basic definitions. He must concern
himself with points such as: if an expression of the type x is y
is to qualify as a metaphor, the literal term x and the meta-
phoric term y must not be subsets one of the other. John is a
crusader is thus not a metaphor if John belongs to the set of
crusaders. But it becomes a metaphor if we presuppose that
John is not a crusader, but for instance a pugnacious professor:
pugnaciousness will then satisfy the requirement of meta-
phoric constraint, the difference between a real crusader and
a pugnacious professor satisfying the requirement of meta-
phoric differential. This example was, by the way, chosen de-
liberately to illustrate how difficult some metaphoric con-
86 GRAMMATICAL MODELS

Btraints and differentials may be to verbalize in simple terms

of a matrix. Again, a student of style is likely to profit more
from fairly abstract categories such as [ ± divine] than from
very detailed, delicate ones t h a t characterize single metaphors
rather than metaphor classes. Here, too, finding suitable
levels of delicacy and abstraction will be a crucial problem.
Nor should we forget t h a t cultural and literary traditions play
an important role in specifying what kinds of metaphoric
constraints and differentials are permissible in a given culture,
language, and style. Some metaphors overlap with idioms in
refusing direct translation from one language into another.
Traditions also enter into specifying where runs the fine border
between a metaphor and a lie.

5.7 STYLES ACROSS LANGUAGES

I have here touched upon the question of styles across lan-

guages, which has theoretical as well as practical importance.
If we wish to study the similarities and the differences
between, say, legal style in English, French, and Russian, we
must have at our disposal a linguistic apparatus capable of
deciding what features in these different languages we must
regard as equivalent, and what features we must regard as
different. If, for instance, we wish to know whether legal Eng-
lish uses longer sentences than legal French, we must first know
whether all kinds of English perhaps use longer sentences
than all kinds of French. We shall, in other words, need a
comparison of English and French sentence length in more
general terms, and this comparison gives us the background
against which to view the results of our comparison of legal
English and legal French. Only if the results of the latter
comparison prove significantly different from the former are
we justified in regarding them as stylistic. Such comparisons
ought to be based on a theory of their own; otherwise they
readily become a series of unsystematic, scattered — though
perhaps interesting and suggestive — observations. Such
 GRAMMATICAL MODELS 87

questions are highly relevant to the theory of translation

(see e.g. CATFORD 1 9 6 5 , MOUNIN 1 9 6 3 , N I D A 1 9 6 4 ) , and also
to practical translating. In fact, comparative stylistic studies
of the languages involved would give us a welcome means of
checking the stylistic accuracy of translations. Style is p a r t
of total meaning, and if a translator's aim is to make the text
read as if it had been originally written in the target language,
it must consistently follow the patterns of the style the original
author would have used had he written in this, target,
language, and lived within its culture.

5.8 STYLE AND RHETORIC.

ALLEGORY, IRONY, PARODY

Some potential style markers have been studied mainly b y

rhetoricians, and they sometimes need more stringent linguist-
ic definition. Many examples could be cited of rhetorical de-
vices that are potential style markers. One of them is the con-
cept of the LOOSE SENTENCE, which can be interrupted before
its end so t h a t the part, or parts, before the interruption form
a grammatical, acceptable sentence. The density of loose
sentences might in fact be relatively easy to compute even out
of relatively long texts. As Karl BOOST has shown, the co-
hesive tension within sentences differs, and is upheld by dif-
ferent means, in different languages (BOOST 1 9 5 9 ) . The
English Yesterday I bought a book in the shop is thus inherently
more loose — it can be interrupted after book — than the
German Oestern hdbe ich ein Buch in dem Laden gekauft, in
which the final verb functions as a signal upholding the tension
and preventing earlier interruption. I have quoted this in-
stance of contrastive English-German grammar to show
another example of the risks of uncontrolled bilingual styl-
istics: figures on densities of loose sentences in English a n d
German are not directly comparable as criteria of style,
because these two languages have different methods of up-
holding cohesive tension within the sentence. Sentence
88 GRAMMATICAL MODELS

looseness is also connected with embedding and with the left

or right-branching structures described by Y N G V E ( 1 9 6 0 ) ,
whose proportions may contribute to giving a text a con-
spicuous stylistic flavour.
So far, rhetoric has been more successful than strict gram-
mar in dealing with textual units larger than the sentence. To
what extent such units are amenable to strict linguistic defi-
nition is a very topical question whose dissection cannot be
attempted here. In writing, the PARAGRAPH is a typographical-
ly marked unit, and therefore objectively verifiable. And in
practice it is by no means impossible to use paragraph
structure as a potential style marker. We may, for instance,
proceed to count densities of paragraphs whose controlling
idea is expressed in an initial topic sentence, paragraphs in
which the controlling idea is implied rather than summarized
in any one sentence, and paragraphs in which the controlling
idea comes last. (See rhetorics such as B R E W S T E R 1 9 1 2 and
GENUNG 1900.)
When pushing upward towards larger textual units we are
also approaching problems traditionally regarded as literary,
not linguistic. Features such as allegory and irony will require
a very sophisticated semantic apparatus before they become
accessible to stringent, mechanical stylolinguistic description.
A L L E G O R Y builds on the isomorphism between two simultan-
eous meanings of the text: the literal, surface meaning, and
the allegorical, deep meaning. I R O N Y is based on a tension
between a surface meaning and its opposite. In terms of the
communication chain, irony might be described as a con-
spiracy between speaker/writer and listener/reader, both of
whom know t h a t a given textual unit should not be under-
stood literally. This has to be signalled through special I R O N Y
SIGNALS ( W E I N R I C H 1 9 6 6 ) , which warn the receiver not to
interpret certain sets of textual stimuli in their literal sense.
Such irony signals need not be overtly present in a text.
Some irony signals consist of contradictions of presupposed
facts known to both communication partners, as in Hitler was
 GRAMMATICAL MODELS 89

the kindest of men. Others involve an inconsistency: if we are

told that Chaucer's prioress only gave her dogs the finest of
foods, and if we assume t h a t many people were starving all
around her and recall t h a t it was the duty of nuns to help
those in distress, we are predisposed to interpret Chaucer's
praise of her gentility as irony. Irony signals may also be
"exophoric" in that they refer to actions outside the text
itself. If, in a play, we are shown that a character behaves
horribly and then told that he is noble, the inconsistency
between action and verbal characterization signals the irony.
In speech, irony may be conveyed through specific patterns
of stress and intonation and by paralinguistic means: rhythm,
tempo, tone of voice, loudness, and the like. I n writing, inver-
ted commas are sometimes used as a crude kind of irony signal.
PARODIES are very often built around stylistic features. I n
fact one common kind of parody consists of a text with a very
high, and perhaps exaggerated, density of the parodee's most
conspicuous style markers. Non-discrete style markers, such
as sentence length, may be given exaggerated values, too.
The parodist's method is the verbal counterpart of caricature,
where the artist seizes upon a few prominent features of his
subject — a large nose and big ears, for instance — and ex-
aggerates them so that they come to dominate his portrait.
Similarly, the parodist's first task is to find a set of character-
istic style markers, and then to compose a text with an ex-
aggerated concentration of these very features. If the density
of such imitated style markers is roughly the same as in the
original, the result may be labelled as IMITATION or PASTICHE.
( c f . IKEGAMI 1969)

5.9 TEXTUAL COMPLEXITY AND READABILITY

Some scholars have investigated characteristics of styles t h a t

consist, not of sets of single style markers but of complex
resultants of many simultaneous style markers, sometimes in
ways in which the single vectors are hard or impossible to dis-
90 GRAMMATICAL MODELS

entangle. Examples of such complex style markers are

T E X T U A L COMPLEXITY a n d R E A D A B I L I T Y ( W n O 1 9 6 8 , M L L L E R
- COLEMAN 1 9 6 7 ) . Readability and textual complexity are
influenced by many factors: the ease or difficulty of the vocab-
ulary (which may roughly correlate with average word
length); the ease or difficulty of clause and sentence structure
(which should be measured not only in terms of sentence
length but also of sentence complexity: ADMONI 1 9 6 6 , L E S S -
K I S 1 9 6 4 ) ; patterns of intersentence cohesion and connectivity
in texts; the character and density of allusions and rhetorical
figures; and the like. Textual complexity has been calibrated
in its totality, without reference to the individual stimuli in
the text, by methods such as cloze scores (obtained by having
informants guess at words omitted from the text), answering
questions on a passage, calculating distributional constraints
and redundancy, and the like.

5.10 S T Y L I S T I C S A N D C O N T E N T ANALYSIS

Stylistics also borders on CONTENT ANALYSIS (POOL 1959,

ROSENGREN 1968, STONE 1966), especially those methods by
which content analysts have investigated the densities of
words and collocations in well-defined contexts. It should be
noted that content analysis is not a single, hard-and-fast
method but rather an approach in three steps. First, the in-
vestigator must limit his object of study. Secondly he must
choose a sample according to explicit criteria. And, thirdly,
he studies the frequency of certain objectively — and thus in
practice linguistically — definable features in these materials
through systematic counts and statistical controls. Both the
student of style and the content analyst are thus interested in
contextual densities of linguistic features. The difference is
often one of purpose rather than method: content analysts
choose their linguistic features, not because of their linguistic
interest but because of their semantic content. Their basis is
 GRAMMATICAL MODELS 91

not linguistic but epistemological; the ultimate focus is on

content, not on form, and forms are regarded not as entities
in their own right but merely as reflections of content. The
similarities between content analysis and the study of style
are, however, sufficient to suggest t h a t some of the content
analysts' methods are potentially useful to students of style
as well. Thus the co-occurrence of words that has often been
studied b y content analysts is also a potential style marker.

5.11 FROM STYLOLINGUISTICS TO LITERARY ANALYSIS

Above I listed four requirements for a linguistic model in

stylolinguistic description. The model had to allow for varia-
tion and context. I t had to be consistent. I t had to be adequate.
And it had to admit both categorical and probabilistic rules.
In case it were not regarded as self-evident, we might now add
a fifth requirement. Stylolinguistic methods have to be ob-
jective in the sense t h a t anybody who repeats the analysis of
the same materials by the same methods must arrive at the
same result. The concepts and definitions must be operation-
ally concrete and unambiguous. As long as we use modern,
strict grammatical models as the basis of our stylolinguistic
description, this requirement is easy to satisfy. B u t if we
venture into those areas t h a t are important for stylistics but
which so far lie in the margin of, or even beyond, linguistics
proper, we must watch the integrity of our concepts and
methods. When we are dealing with rhetorical or literary
features, with metaphor, irony, and the like, the temptation
to relax the rigour of the apparatus will be overwhelming. To
yield to such temptations is, of course, not a sin. Actually, it
may be the only way in which some jobs will get done a t all.
Indeed we may, at the present stage of linguistics, be compel-
led to give u p the demands that a strict methodology puts on
operational concreteness and rigorous conceptual frames. But
we should then be aware t h a t our investigation stops being
stylolinguistic and becomes something else, for instance,
92 GRAMMATICAL MODELS

literary criticism, where brilliant intuitions and elegant, often

metaphoric, verbalizations of subjective responses are at a
premium. In comparison, linguistic stylistics of the stricter
kind often seems a highly pedestrian subject.
To illustrate some of the grammatical categories t h a t stylo-
linguists have counted I shall, finally, refer to a few examples
of quantitative studies. These examples will also support the
observation that, though modern linguistics offers many
highly sophisticated and subtle concepts and approaches, in
stylistics a relatively simple and traditional grammar can go
a long way.
An excellent example is Louis T. M t l i c ' s computer study of
Swift's style ( M i l i c 1 9 6 7 ) . M i l i c studied Swift's seriation, in-
cluding percentages of doublets and triplets, sequences of
words and groups of words, normal and asyndetic and poly-
syndetic series, series with a continuator such as etc., and so
on. His report contains another section on Swift's connectives:
sentence-initial co-ordinating conjunctions, subordinating
conjunctions, and conjunctive adverbs. The most elaborate
part of M x l i c ' s analyses was concerned with word-classes and
word-class distribution. Among the statistical characteristics
are the proportion of function words; the "Stable Style
Characteristic" computed with the aid of values for four
groups of frequencies, namely those of verbals, finite verbs
and auxiliaries, modals, and connectives; ratios between
nominals and verbals, adjectives, and verbs; and the occur-
rence of different three-word patterns within the sentence
(which a computer can readily be programmed to identify and
count). As M i l i c concludes, these features, all of which can
be easily defined with the aid of a relatively elementary
grammar, suffice to show a whole set of style markers by
which Swift's style differs from the styles of Addison, Johnson,
Gibbon, and Macaulay. The burden lies on the statistics. All
the linguistics M i l i c really needed was a basic grammar
capable of adequate definitions of series, conjunctions, and
word classes.
 GRAMMATICAL MODELS 93

Richard OHMANN'S study of Shaw (OHMANN 1962) is

different in appearance: it shows how a primarily literary
analysis can gain support from statistical counts, and how a
body of statistical d a t a can be presented without fanfare in
a slender appendix. Those who prefer readability to statistics
should seek comfort in OHMANN'S combination of a sensitive,
elegant essay and a sound method. The essay concentrates on
connections between Shaw's literary qualities and his language.
The first appendix gives comparative counts of samples of
some 2,500 words out of Shaw, Yeats, Bertrand Russell, t h e
Webbs, Chesterton, and Oscar Wilde — in other words, from
a contextually related norm. The features OHMANN counts
are long series, comparatives, would, all-or-nothing determin-
ers a n d degree words as well as degree intensity, extent,
limits, and quantity; appositions, negatives, direction shifts,
requests, dependent clauses, quotations, introductory that, re-
ferences to beliefs and to statements and propositions, proper
names, personal pronouns, the proportion of person words
among grammatical subjects, mental causation, infinitives,
abstract nouns, a n d adjectives. The list is remarkable for its
blend of very different types of units. As OHMANN himself has
emphasized, some of the categories were defined more by feel
t h a n b y exact linguistic rule; he m a y thus have overstepped
t h e bounds of strict linguistics. Also the sample, though care-
fully chosen, was too small to permit confident statistical
generalizations. I n the respects considered, all ( t h e other
writers still proved more like one another t h a n any one of
t h e m was like Shaw. OHMANN'S conclusion is t h a t

a small but carefully selected group of counts is more useful in isolating

the style of a writer than would be a much larger but randomly chosen
group (OHMANN 1962: 185).

The second appendix is devoted t o a study of revisions in

some of Shaw's typescripts, and it shows t h a t revisions are
particularly frequent precisely within those areas t h a t are of
particular relevance to Shaw's style.
94 GRAMMATICAL MODELS

Walker GIBSON is one of the students of style who have

attempted setting up statistical frames for the diagnosis of
styles, not only descriptions of single styles. (GIBSON 1966)
Though written in a light vein with a somewhat cavalier
approach to statistics, GIBSON'S book — appropriately label-
led Tough, Sweet and Stuffy — suggests a battery of statistical
tests for the distinction of three major styles in modern
American prose. GIBSON began by classifying a number of
texts under each of his three impressionistic headings,
'tough', 'sweet', and 'stuffy'. He then studied each group to
see what its texts had in common. To give just a few examples
of GIBSON'S criteria: a'tough', text has mope than 70 per cent
monosyllables and less than 10 per cent words of 3 syllables or
more; a 'stuffy' text has a maximum of 60 per cent mono-
syllables and 20 per cent or more of words of at least 3 syl-
lables. Texts with values between 'tough' and 'stuffy' are
'sweet'. A text with more than one passive per 5 verbs is stuffy,
one with less than one passive per 20 verbs is tough, and one
with no passives at all is sweet. GIBSON finally gives a set of
ten rules how to avoid being stuffy, which — characteristically
— ends with a tenth commandment: "Don't obey all these
rules at once !" GIBSON is no doubt right in claiming t h a t the
profiles t h a t emerge out of measuring a text with his quanti-
tative tests do indicate some general stylistic characteristics
of t h a t text. The recognition of three major styles was, of
course, based entirely on GIBSON'S initial intuition. But it
would be interesting to study a corpus of texts more thorough-
ly and see how far GIBSON'S intuitions are supportable by
factor analysis and numerical taxonomy.
I n a study of over 60,000 sentences brought from 30 German
texts out of four genres (plays, fictional prose, factual prose,
scientific prose), Werner W I N T E R similarly restricted himself
to three comparatively simple, rapidly countable criteria:
sentence-initial parts of speech, clause length, and sentence
complexity (WINTER 1961). There was a clear correlation
between context and the frequencies of some of these features.
 GRAMMATICAL MODELS 95

The density of sentences t h a t begin with adverbials was

higher in written, and particularly in scientific, prose than in
plays, which were here taken to reflect patterns of spoken
German. The density of finite verbs per total number of
words was low in scientific prose and higher in plays and in
fiction, with a great deal of scatter in factual prose. These
findings can be reinterpreted as contextual probabilities,
which can be read both from context to linguistic density, and
from linguistic density to context. In written German, about
35 per cent of all sentences began with adverbials; in primarily
spoken German, only 17.5 per cent did. Thus if we take a
scientific text in German, we are likely to find t h a t some 35
per cent of its sentences begin with adverbials, whereas, the
other way round, if we find t h a t a text has 35 per cent sentence-
initial adverbials, we may predict t h a t this will contribute to
giving it a scientific flavour.
The comparison between text and norm need not be syn-
chronic as in the studies of MTLIC, OHMASTN, GIBSON, and
W I N T E R . An example of a comparison between texts from the
same contextual sphere but from different periods is Lars
GRAHN'S pilot study of Swedish newspaper styles (GRAHN
1965). A study of 85 articles from Dagens Nyheter in 1890
(41,753 words) and of 124 articles from the same newspaper
in 1964 (83,760 words) showed t h a t the average length of
sentences had shrunk from 24.39 words in 1890 to 17.49
words in 1964. The greatest difference, a shift from 35.45 to
20.38 words per sentence, was found in the leaders; the smal-
lest, from 13.80 to 13.72 words per sentence, in causeries. The
number of embedded subordinate clauses per 1,000 running
words of text had similarly diminished from 6.61 in 1890 to
3.96 in 1964. The relations between different contextual genres
had also changed. I n 1890, cultural articles had the greatest
density of embedded subordinate clauses; by 1964, they had
fewer embedded subordinate clauses than leaders and foreign
news reports. Comparable differences, both between the styles
of 1890 and 1964 and between the styles of different genres
96 GRAMMATICAL MODELS

first in 1890 and then in 1964, were also found in the densities
of attributive participles, participles embedded between
genitive attribute and head, and some other syntactic struc-
tures. Altogether, GRAHN'S figures offer ample statistical
reasons why Swedish newspapers of 1890, and especially some
subcategories of their text, strike modern readers as 'heavy'.

5.12 SUMMARY

Many, no less successful and suggestive reports could be cited,

and many more will no doubt be forthcoming in the near
future. But these must suffice. They show concrete examples
of a number of statements made in the course of my essay on
defining style (ENXVTST - SPENCER - GREGORY 1 9 6 4 ) and re-
peated in this book. They all relate texts to contexts. Then
they compare the densities of selected linguistic features in
the text with the corresponding densities in a contextually
definable, related, and suitably contrastive norm — suitably
contrastive, that is, to bring out the desired stylistic different-
ial. For Swift, Mn.ro found a suitable norm in Addison,
Johnson, Gibbon, and Macaulay; for Shaw, OHMANN chose a
norm from Yeats, Bertrand Russell, the Webbs, Chesterton,
and Wilde; GIBSON compared three categories of texts;
W I N T E R contrasted primarily written and primarily spoken
texts; and GRAHN worked with the same newspaper from 1 8 9 0
and 1964. In every instance, the norm was carefully chosen
for its definite contextual relationship with the text. And
those features whose density in the text was found to be
significantly different from the density of the corresponding
features in the norm were regarded as stylistic characteristics,
or style markers, for that text. Once the correspondence
between style markers and contextual ranges had been
established, it could be made use of both ways: if we know the
style markers we may draw conclusions about the context,
and if we know the context we may predict what types of
 GRAMMATICAL MODELS 97

style markers are likely to occur in it. The former, diagnostic

procedure was used by GIBSON to classify texts, and it forms
the basis of stylistic author determination which I shall dis-
cuss in a later chapter. The latter is the approach we use when
characterizing the language that occurs in given contexts, for
instance when discussing the styles of authors, genres, or
periods. Finally, all the five studies cited made good use of
relatively simple grammatical features (which OHMANN
supplemented with a few complex and subjective ones without
attempting complex and stringent linguistic definitions).
Even such simple features proved operationally adequate to
bring out inventories of style markers.
 6

DEVIANCE

6.1 DEVIANCE

One of the areas in which linguists have gained in sophisti-

cation is the study of GRAMMATICALITY , or GRAMMATICAL
WELL-FORMEDNESS. This area is relevant to students of style
on two counts. First, certain texts — modern poems, for in-
stance — achieve some of their stylistic effects precisely by
departing from the ordinary norms of well-formedness. If we
are to describe their style in linguistic terms, we need a theory
and method of description of the ways in which a text can
depart from normal usage. Secondly, all styles must, by de-
finition, be regarded as "deviant" — not necessarily from
normal patterns of well-formedness, but from contextually
definable norms and thus from other styles. The former type
of deviance, which is the focus of this chapter, results from
actual tinkering with the rules of normal grammar. The latter
results from the fact t h a t the features of normal language
and the rules of normal grammar are used with different
densities in the generation of different texts. As we have seen,
such "well-formed" styles can be explained by the comparison
of densities of well-formed features in the text with the cor-
responding densities in a suitable, contextually related norm.

6.2 GRAMMATICALITY, ACCEPTABILITY, TOLERANCE

Another very important type of deviance is the normal and

acceptable ungrammatically of spontaneous, unscripted
speech. I n fact, this kind of deviance was one of the motives
 DEVIANCE 99

for the introduction into linguistics of concepts such as parole

and performance. Studies such as BOWMAN (1966) and L o -
MAN - JORGENSEN (1971) have pinpointed a number of
problems that analysts of deviant patterns in spoken styles
have to face. One of them concerns the definition of the basic
unit of spoken language. In speech, people often express
themselves in units which do not satisfy the requirements of
grammatical sentences, for instance in sentence fragments.
Therefore the analyst of spoken language will have to develop
an apparatus capable of segmenting a text into suitable syn-
tactic units, some of which do not qualify as well-formed
sentences. BOWMAN has divided utterances into sentences,
which are marked by intonation contours and terminal junc-
ture, and fragments, which are interrupted; sentences are
further subdivided into major sentences containing both
subject and predicate, and minor sentences (BOWMAN 1966).
LOMAN and his co-workers operate with macrosyntagms,
which can be subclassified as follows:

macrosyntagms

interjectional sentence vocative

macrosyntagms macrosyntagms macrosyntagms

sentence sentences
fragments ^ | ^

incomplete complete
sentences sentences

ill-formed well-formed
sentences sentences

Fragments are defined as sentences lacking a finite verb in

the main clause, whereas incomplete sentences lack a contex-
tually relevant constituent at the end of the sentence (aposio-
pesis). Ill-formed sentences can be subclassified by type of
100 DEVIANCE

deviance. Such types are anacoluthon, self-correction on a

syntactic level, omission of a syntactically relevant p a r t of
the clause, attempts at continuing an already complete sent-
ence, and interruptions (EINARSSON 1 9 7 1 : 3 2 , 1 1 8 ; J O R -
GENSEN 1 9 7 1 ) .
As terms such as grammaticality and well-formedness have
been used in different senses, their user owes his audience a
definition. I shall here use grammaticality simply to indicate
the extent to which a given sentence has been formed accord-
ing to the rules of a given grammar. If the grammar G gene-
rates the sentence S, S is GRAMMATICAL, or, synonymously,
W E L L - F O R M E D with respect to grammar G. If grammar G
has been set u p hierarchically so t h a t it has a set of ordered
rules, some of which are more fundamental, or deeper, than
others, we can set up a scale of grammaticality. Sentences
that violate more fundamental, deeper rules are less grammatic-
al than sentences that violate less basic rules that lie closer
to the surface. ACCEPTABILITY, on the other hand, is a con-
cept relative not to a given grammar but to the opinion of
an informant or group of informants. Within this terminology
it is therefore perfectly possible to write a grammar of a langu-
age which generates unacceptable sentences: in terms of the
grammar, they are grammatical and well-formed, even if a
group of informants rejects them as unacceptable. Con-
versely, informants may accept sentences not generated by
the grammar. Such sentences are acceptable but, in terms of
our grammar, ungrammatical. Like grammaticality, accepta-
bility can be scaled: if all of one hundred suitably selected
informants agree on accepting a given sentence, this sentence
has a higher acceptability than a sentence which is only accep-
ted by half of the informants. I n practice, groups of informants
can be far from unanimous when judging the acceptability
of certain types of structures. This suggests that grammat-
icality, which is an absolute and can be defined in terms of a
given set of rules, and acceptability, which is relative to in-
formant opinion, will never fully agree ( Q U I R K - SVARTVIK
 DEVIANCE 101

1966, GREENBAUM - QUIRK 1970). We therefore need both

concepts (Cf. CHOMSKY 1 9 6 5 ) .
Acceptability does not, however, exist in a situational
vacuum. Informants may accept a given sentence in one
context (including situation) and reject the same sentence
in another. I shall use the term tolerance for 'acceptability in
a given context'. Thus in the context of modern poetry,
anyone lived in a pretty how town is tolerated, though in the
context of factual prose it would be unacceptable and though
it is ungrammatical in the light of ordinary grammars of
English. Each context is in fact characterized by a TOLE-
RANCE RANGE or MARGIN or TOLERANCE. Certain rituals in
armies, the Church, and other tradition-bound establish-
ments may have no tolerance margin in the sense t h a t their
language is "frozen" (Joos 1962) to certain fixed expressions.
I n other contexts, the range of tolerance may be very wide,
even wide enough to permit verbal play. I already noted this
above in section 1.6 in connection with context parameters.
Finally, I shall use the term DEVIANCE to indicate the differ-
ence between a text and the overall grammatical norm of the
language. Deviance is thus the sum of nongrammaticality
and nonacceptability, and a blanket term covering two poten-
tially different factors. However, here deviance does not in-
clude those non-deviant styles t h a t arise through differences
in densities of well-formed features in a text and a contextu-
ally related norm; deviance thus does not have the second of
the senses referred to in the first paragraph of this chapter.

6.3 TRANSFORMATIONALIST
AND BEHAVIOURIST VIEWS OF WELL-FORMEDNESS

I t should perhaps be noted t h a t the need for distinctions not

only between "right" and "wrong" but between degrees of
grammaticality and acceptability haa arisen both in trans-
formational grammar and in large studies of actual usage.
102 DEVIANCE

In all work with generative grammars we are compelled to

test the output. When writing the grammar we set up a ten-
tative sequence of rules t h a t generate a set of sentences. We
must then decide whether these sentences are desirable or
not. If they are, our rules may stand; if t h e y are not, we must
revise the rules. In practice, writing a generative grammar
involves very large numbers of such decisions. Processes of
this kind have contributed greatly to sharpening our views of
deviant expressions and evaluations of their degrees of gram-
maticality and acceptability. Obviously, acceptability and
tolerance are, in a sense, more fundamental than grammatica-
lity. Grammars are written by linguists, who can manipulate
their grammaticality scales; acceptability and tolerance
reside in a collective Sprachgefühl and can only be influenced
indirectly through changes in usage and style. Of course a
grammarian should all the same test the output of his grammar
against some strict norms of acceptability and tolerance,
even if these are merely the ones of his own idiolect and
of his own subjective sense of linguistic and stylistic pro-
priety.
I n general, behaviourist structuralists found deviance an
awkward concept. The reasons seem clear enough. They based
their linguistic descriptions on explicitly defined corpora
of texts, and they insisted on the use of immanent categories
actually present on the surface of the text, without permitt-
ing judgments as to degrees of correctness or grammaticality
in the text itself. Thus if the corpus contained a deviant ex-
pression, t h a t expression had to be incorporated into the
description precisely in the same way as a non-deviant ex-
pression. Now if, say, a grammarian decided to include the
poems of E. E. Cummings, their structures became p a r t of
modern English and thus non-deviant. If, again, the gramma-
rian decided to exclude them and p u t them into a separate
grammar of Cummingsese, t h a t grammar had to be immanent,
t h a t is, based not on categories of non-Cummingsian English
but on Cummingsese alone. Now if Cummingsese is described
 DEVIANCE 103

as a new and separate language, its categories may well be-

come very different from those of normal English, especially
if the grammarian uses criteria of maximal descriptive eco-
nomy in his description of Cummingsese. The grammar of
Cummingsese runs the risk of becoming hard to compare with
t h a t of non-deviant English. Yet we well know that Cumm-
ings's poems are not effective in a vacuum. Their effect, our
intuition tells us, is founded not on their form as such, b u t
rather on our knowing how each of Cummings's expressions
differs from a comparable set of related expressions in non-
deviant English. Altogether, the effect of deviant poetry is
based on comparisons between deviant and normal expres-
sions.

6.4 GRAMMATICALITY SCALES

The density of deviant structures may be a style marker in

its own right, like the density of metaphors. B u t just as one
poet's style may differ from t h a t of another, not by an overall
density of metaphor but by densities of different types of
metaphors, so one deviant text may differ from another de-
viant text because of different kinds of deviance. Cummings's
deviance is not the same as Lewis Carroll's. We therefore
need a classification of different types of deviance.
One starting-point for such a classification is the ranking
of deviant expressions on a scale of grammaticality, which
becomes possible as soon as we have a generative grammar
with ordered rules. I n terms of generative-transformational
grammar, a string such as *a who the of today would be found
to violate some very basic rules and thus to be greatly un-
grammatical. In *I thermometer you, the deviance results
from use of a substantive where normal grammar requires a
verb. I n * / elapsed the book to her we are using a verb, elapse,
in a structure reserved for certain other verbs such as give,
and thus breaking what one transformational model (CHOM-
104 DEVIANCE

SKY 1965) called "strict subcategorization rules". And in

*The stone snores, the ungrammatically consists of a breach
of a "selection rule" which says that animate verbs such as
snore require animate subjects, not inanimate ones such as
stone. (Breaches of selection rules may result in metaphor.)
There will, however, be a residue of sentences such as the
famous Both of John's parents are married to aunts of mine,
which are hard to diagnose as deviant in grammatical terms
unless we incorporate into our grammars a semantic compon-
ent capable of blocking these logical absurdities. How this
should be done is a controversial question.

6.5 STRATEGIES IN THE DESCRIPTION OF DEVIANT

TEXTS

Devices such as those exemplified above will suggest ways

t o describe different patterns of deviance in more precise
terms. We should, however, still face the question of what
method we ought to prefer when describing a deviant text.
At least three alternative strategies seem to be available
( L E V I N 1963, 1965a; THORNE 1965, 1969, 1970; HENDRICKS
1969).
First, if we have a full generative grammar of the language
a t our disposal, we may note at what point the deviant
structure departs from the normal sequence of generative
rules. In terms of the above examples, * / elapse the book
would be less grammatical than *The stone snores. Secondly,
we may regard the deviant t e x t as having been written in a
different language. We may then proceed to write a grammar
of this new language. If we do so, our new grammar is likely
to generate not only the particular sentences of our original,
deviant text, but a large — and perhaps infinite — number of
other texts as well. Such overgeneration is not, however,
necessarily an evil: these other, "overgenerated" texts may
 DEVIANCE 105

afford interesting examples of the type of deviation we are

concerned with. They are, in a sense, imitations or pastiches
o f the original. If, for instance, we write a grammar for Cumm-
ings, a generative rule allowing he danced his did is likely to
generate not only those sentences of this type that Cummings
actually used, but also a number he did not use, such as he
ran his ate or he wrote his slept. And a generative rule permitt-
ing Dylan Thomas's a grief ago might also generate a ciga-
rette ago and, perhaps, three husbands ago. The more basic
the deviant rule, the larger the number of overgenerated
structures. The risk of this second approach is that the
grammar written for the deviant text may depart too
far from the grammar of non-deviant language. We must
therefore take care to formulate constraints compelling us
to make the deviant grammar conform sufficiently closely
to that of non-deviant language. W e shall need a theory for
such restraints.
The third strategy is, in a sense, the most interesting be-
cause it tries to incorporate the actual process of interpreta-
tion of deviant structures. Here we again take for granted that
we have a full grammar of the non-deviant language at our
disposal. When faced with a deviant structure, we interpret
it by associating it with that structure, or perhaps those
structures, in the non-deviant language which we feel to
be its nearest equivalents. T o cite KATZ:

The speaker's understanding of the [deviant] semi-sentence is nothing

other than his understanding of the sentences in the [non-deviant]
set with which the semi-sentence is associated (FODOR - KATZ 1964:
411).

Our description of the deviant sentence will thus come to

consist of a set of normal sentences that we might call its
COMPREHENSION SET, and a set of rules which explain precisely
how the deviant sentence is associated with each of the sen-
tences of the comprehension set. Such TRANSFER RULES will
106 DEVIANCE

also throw light on the reasons why a given comprehension

set, and not another, was associated with a given deviant
sentence. This model in fact suggests active experiments with
informants to show with what degree of agreement different
people can offer comprehension sets for specific instances and
types of deviance.

6.6 CLOSED TEXTS, OPEN TEXTS,

AND CHOICES OF GRAMMATICAL MODELS

William S . CHISHOLM has posed a question t h a t we should

now t r y to answer. "Is it," he asked "necessary to have a
reserve of all sentences in order to describe a few? Or, to
put it differently, is a generative grammar the one t h a t stylist-
ics needs?" (CHISHOLM 1 9 6 7 : 2 6 )
If style is a result of a comparison of frequencies in two
closed texts — a CLOSED TEXT being a text which is available
in its entirety and an OPEN TEXT being a text which may be
added to — we may in principle choose between a taxonomic
and a generative model. For taxonomic models are also ca-
able of describing closed texts. If, however, the norm, or
both text and norm, are open, we must use a grammar which
is adequate for the description of open texts and which can
predict what occurs even beyond the actually available sample
or samples. And if we suspect t h a t our text may use deviant
structures as style markers, the requirement of adequacy
speaks strongly for the use of a generative model. For here,
adequacy requires a description of generative processes rather
than of surface structures: the densities of deviant structures
as such may not suffice. We may need subclassifications of
different types of deviant structures t h a t are based on depth
of deviance, and this presupposes a generative grammar with
ordered rules. But whatever model we use, it must be capable
of allowing comparisons of densities of linguistic features.
Such arguments may be modified b y working hypotheses,
however. Not infrequently, linguists have tacitly assumed
 DEVIANCE 107

t h a t an open corpus such as modern English may be described

in terms of a closed text, which is then regarded as a repre-
sentative sample of the open corpus. And even in their work
with closed texts, linguists are usually interested in those
underlying general patterns t h a t allow predictions beyond
the text itself. If a linguist studies Old English poetry, for
instance, he is not only interested in the surface of the poems
b u t also, and perhaps even more, in the set of rules that Old
English poets used when composing their poems. He thus
operates with the hypothesis t h a t his corpus is in fact open
a n d comprises all of the actual and potential poems in Old
English, even though he only has a sample of actual, surviv-
ing poems at his disposal. This is one of the basic differences
between most linguists and most literary scholars. The latter
generally study closed texts for their uniqueness, whereas
linguists are often committed to regarding them as samples
t h a t help them to discover the underlying patterns and rule
systems that were applied to their generation. I n linguistic
stylistics, decisions as to whether text and norm should be
regarded as closed or as open follow from the choice of text
and norm. In work with closed texts, densities may take the
form of actual numbers of occurrence; in work with open texts,
we must reckon with probabilities computed from t h a t closed
body of text which we regard as our sample.

6.7 GENERATIVE-TRANSFORMATIONAL METHODS

IN THE COMPARISON OF LANGUAGE VARIANTS

Yet another question should be mentioned. If we compare

two variants of a language in terms of generative grammar,
in what form should we express our results? I empha-
sized above t h a t two variants of a language may differ in
two ways. First, if both use the same set of rules, but
with different patterns of frequency, we must somehow
attach to each rule a measure of the frequency with which
108 DEVIANCE

it is used in each variant. OHMANN did this by listing

the transformations of closed texts together with their
frequencies, whereas LABOV extended the method to open
texts by setting up probabilistic rules with variable input
(see above, sections 3.3 and 5.3). Secondly, if the two variants
use different rules, we may proceed to classify such rule differ-
ences into a few definite subtypes. Thus one or the other vari-
ant may have a rule which is missing from the other grammar.
Two otherwise identical rules may be differently ordered in
the two grammars. Two rules in the two grammars may apply
to the same feature but rewrite it differently: one grammar
rewrites x -»• y / A — B, the other x ->- z/A — B. Finally,
two grammars may have rules with the same structural
change b u t their sphere of application can be different. Such
differences between related grammars have so far been studied
mainly by those interested in historical syntax (KIPARSKY
1 9 6 8 , TRATJGOTT 1 9 6 5 ) , but their methods are potentially
applicable to the comparison of any related grammars, in-
cluding those of stylistic variants of the same language.

6.8 PSYCHOLINGUISTIC APPROACHES TO DEVIANCE

J u s t as a number of style markers can be packaged and stu-

died collectively under labels such as "complexity" or "read-
ability", total patterns of deviance have also been analysed
by psycholinguistic methods (CHAPMAN 1 9 6 8 , D A N K S 1 9 6 9 ) .
At least in some situations, informants are able to scale de-
viant sentences in accord with formal criteria of grammatical-
ity, and the ease of learning and recall increase with increas-
ing grammaticality. Meaningfulness, however, can sometime»
be scaled independently of grammaticality. Such findings are
interesting, b u t the experiments have so far used compara-
tively rough methods when making up the deviant sentences
for the experiments. Thus words of grammatical sentences
have been scrambled randomly, or deviant sentences have
 DEVIANCE 109

been created by writing a number of sentences with simi-

lar surface structures under each other and then reading
the resulting matrix diagonally. The experiments might well
gain in linguistic and stylistic relevance if methods could be
devised to set up deviant sentences with an improved control
not only of the rough level of grammaticality, but also of the
number and type of transfer rules needed to make each test
sentence grammatical. For it may well be rash to assume
t h a t depth of deviance is the only factor affecting the meaning-
fulness of the deviant sentence. Meaningfulness may also be
affected by other factors such as the complexity of the trans-
fer from deviant to normal: a transfer close to the surface,
but requiring many transfer rules, may perhaps rank high
on a scale of transfer complexity, and make the deviant
sentence difficult to understand.
 7

LINGUISTIC STYLE MARKERS

BEYOND THE SENTENCE

7.1 STYLE AS A QUALITY OF TEXTS

In their quest for simple definitions that could be hung on the

pegs of well-established linguistic concepts, some scholars
tried to equate style with the linguistics of parole. J u s t as
simple are the definitions of style as the linguistics of units
larger than the sentence ( H i l l 1958: 406). Such views have,
however, also been contradicted by style theorists empha-
sizing the role of the sentence as a style carrier (Ohmann
1967).
If definitions of style were rigidly restricted to features
involving spans larger than the sentence, they would lead
to difficulties. First, even single sentences have a style, and
stylistic incongruities such as the use of a colloquial word in
an otherwise solemn, high-style frame may occur within the
bounds of one sentence. And the other way round: quite a
few features of textual cohesion between sentences can be
regarded as grammatical rather than as stylistic. Pronominal
references, concord, and certain other grammatical pheno-
mena do not stop at sentence borders. Indeed, in some langu-
ages, for instance Czech and Russian, the structure makes the
word order of sentences strongly sensitive to their textual
environment (Adamec 1966; B u t t k e 1963, 1969; Firbas
1964a; Kovtunova 1969). In fact, the old, tacit assumption
that grammars deal with sentences has been honoured in the
breach as well as in the observance. In traditional grammar,
the apparatus was so formless and flexible that problems of
text linguistics could intrude freely into a discussion osten-
 LINGUISTIC S T Y L E MARKERS BEYOND THE SENTENCE 111

sibly devoted to sentences rather than to texts. In more highly

formalized grammatical models such as generative-transfor-
mational grammar, the introduction of textual features be-
comes explicit and therefore conspicuous. I t may take the
form of PRESUPPOSITIONS, or of special INTERSENTENCE FOR-
MATIVE S such as the mentioned] which is attachable to
noun phrases to govern the choice of syntactic function (such
as theme or subject), of the choice of article (indefinite or
definite), and of certain phonological patterns ( N I C K E L
1970).
The need for text linguistics has thus become obvidus
within sentence grammar: certain features of sentences cannot
be described, or correctly generated, without reference to
intersentence features and to portions of the text beyond the
sentence under consideration. At the same time, the study
of discourses and texts has led to a new branch of linguistics
which expressly devotes itself to text spans larger than the
sentence. ( B E N S E 1962, 1969; VAN D I J K 1970a, 1970b; H A R -
WEG 1 9 6 8 ; HENDRICKS 1 9 6 8 ; K A R L S E N 1 9 5 9 ; K I N N A N D E R 1 9 5 9 ;
K L O E P F E R - O O M E N 1 9 7 0 ; KOCH 1 9 6 6 , 1 9 7 0 ; M E L ' 6 U K - 2OL-
KOVSKIJ 1 9 7 0 ; P E T O F I 1 9 7 0 ; SGALL 1 9 6 9 ; STEMPEL 1 9 7 1 ; e t c . )
There is an obvious area of overlap between studies in text lin-
guistics, which approach single sentences as parts of larger
units, and that part of grammar which studies cohesion devices
in terms of sentences (e.g. HASAN 1968). Those who wish to
keep these overlapping areas apart may give them separate
labels and, for instance, distinguish between T E X T LINGUISTICS
and INTERSENTENCE GRAMMAR. The former is mainly pre-
occupied with texts, the latter with those features of sentences
that require reference beyond the sentence itself. Text lingu-
istics is thus a discipline in its own right, whereas intersen-
tence grammar is a necessary part of "ordinary" sentence
grammar.
112 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

7.2 TEXTUAL WELL-FORMEDNESS,

TEXTUAL ACCEPTABILITY, TEXTUAL TOLERANCE

Some of the most fundamental problems of these branches of

linguistics emerge from a contemplation of a RANDOM SEN-
TENCE STRING:

John is a boy. N o t yet, but you soon will. If we examine

the wave form in figure 4.6, we can see that, while the sound
lasts, peaks of pressure occur every thousandth of a second.
Good morning. At the beginning of Queen Victoria's reign,
many Londoners were very poor. Its invention is commonly
ascribed to the late Nicholas Longworth. He will be an ideal
companion for a night at Las Vegas.

Here every sentence is grammatical as well as acceptable when

taken in isolation. Still anybody who knows English will
classify the passage as textually ill-formed: the succession
of these individual sentences seems odd, and the sentences
will not be tolerated in this particular textual environment,
though they would be perfectly all right in certain other
environments. One might note in passing that the study of
such random sentence strings is by no means uninteresting.
By trying to spot the features t h a t make them textually ill-
formed we may gather clues as to what characteristics are
necessary if a sentence string is to qualify as a well-formed
text. This should also remind us of the well-known fact that
in many situations, people go to great lengths to make sense
of sentence strings that seem queer at first sight. And this
is true not only of strings of sentences but of strings of other
elements as well: many metaphors, for instance, require the
receiver to devote considerable effort to their interpretation.
Only if no hypotheses as to possible cohesion t a n be maintained
are we willing to give up, and either to dismiss the text as
deviant beyond reason, or, more humbly, to acknowledge our
own shortcomings as interpreters.
There does, then, seem to be a quality of TEXTUAL WELL-
FORMEDNESS or TEXTTJALITY which corresponds to the gram-
 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE 113

matical well-formedness or grammaticality of sentence lin-

guistics. In this sense, the ability of distinguishing between
textually well-formed and textually ill-formed sentence strings
seems to be part of each speaker's internalization of linguistic
patterns, and thus part of his total linguistic competence.
Again, if we adopt the view that all features of competence
must be placed within the domain of grammar, it seems hard
to escape the conclusion that a complete grammar ought to
offer explicit criteria not only for the grammaticality of single
sentences, but also for textual well-formedness. We may
similarly extend our definitions not only of well-formedness,
that is, of conformity to the explicit rules of our grammar, but
also of acceptability and of tolerance, to the textual domain.
In addition to textuality we should then speak of TEXTUAL
ACCEPTABILITY and of TEXTUAL TOLERANCE, as well as of
TEXTUAL DEVIANCE. In analogy with grammatical acceptabi-
lity, textual acceptability should be studied through elicita-
tion experiments with informants, though at the textual level
such experiments will inevitably turn much more complex,
perhaps forbiddingly so.
Textual tolerance can exist at two levels. First, a sentence
acceptable in one place — for instance in the middle — of a
text may not be tolerated in another place, for instance as
a text-initial sentence: the place of a sentence in a string
of sentences may require that sentence to satisfy certain
textual requirements. Secondly, a string of sentences may
satisfy the demands of textual acceptability in one context
but not in another. For instance, the string

The girl was liked since she was pretty. Since she was pretty
the girl was liked. The girl, since she was pretty, was liked.
The girl was poor if pretty. The girl, if pretty, was poor. If
pretty, the girl was poor.

is presumably within the tolerance margin of its actual con-

text, a grammar-book offering illustrations for the rule "If
and since convert one subject-predicate group into a movable
114 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

modifier of another". I t might perhaps be tolerated in some

literary contexts — one thinks of Gertrude Stein — b u t its
repetitions and permutations would probably make people
regard it as textually ill-formed if it occurred for instance in
a legal context, where repetitions obey a different set of
intersentential and intrasentential rules. Thus tolerance in
the sense of 'context-bound acceptability' seems relevant to
text linguistics as well.

7.3 T E X T U A L D E V I A N C E A S A S T Y L E M A R K E R

J u s t as the use of deviant patterns of sentence grammar is a

frequent device with poets such as Cummings or, in Swedish,
Gunnar Bjorling, so prose writers and dramatists occasionally
use odd, and perhaps deviant, textual patterns for literary
effect. In Hamlet Il.ii, Polonius is rightly puzzled by thematic
jumps in Hamlet's speech, though he suspects that "though
this be madness, yet there is method i n ' t " . In Ionesco, brea-
ches of textual tolerance presumably symbolize the absurdi-
ties of life; in La chantatrice chauve, some of them result from
the contextual transfer of sets of sentences from a linguistic
phrase-book into real-life situations. And in some of Harold
Pinter's dialogues, the relations between successive lines seem
contrived to illustrate how difficult it is for one human being
to achieve contact with another. Pinter's means are, however,
more subtle and not as patently deviant as those of Hamlet's
mad phase, Ionesco, or experimentators such as Gertrude
Stein. Interior monologue and stream-of-consciousness prose
are also examples of texts in which the patterns used to link
sentences often differ from those t h a t occur in expository
prose.
The use of textually deviant sentence strings in certain
contexts supports the view t h a t the manner in which senten-
ces are strung together into texts may also function as a style
marker. I n linguistic stylistics we should therefore try to
 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE 115

develop concepts and frames for the description of different

types of sentence sequences, whose range of occurrence could
be studied in terms of context, categories. I f certain patterns
of sentence sequence are significantly more frequent in a
given text than in a norm chosen for its contextual relation-
ship with that text, they qualify as style markers precisely
like any other linguistic features.

7.4 THEME AND RHEME. THEME DYNAMICS.

COHESION DEVICES

T o develop an apparatus for the description of patterns of

sentence sequence we shall do wisely in beginning not only
from ordinary syntax but also from the studies of theme that
have been developed notably by the Prague linguists (MATHE-
SIUS, FIRBAS, DANE§, ADAMEC), by HALLIDAY (1967-68),
and some others (e.g. DAHL 1969). The basic observations
that suggested such studies concerned word order as a result
of various system of organization. In English, the basic lin-
guistic task of word order is to mark the function of groups
in the sentence: in John kicked Jack and Jack kicked John,
we identify subject and object by position before and after
the verb. In languages where these functions are marked b y
case endings, word-order patterns may be free to express
other functions such as those of THEME or TOPIC or, roughly,
what is talked about, and RHEME or COMMENT or what is said
about it. An additional distinction is that between GIVEN or
DATUM, which has been mentioned before in the text, and
NEW or NOVUM, which has not. Often, but not always, the
theme consists of what is given, and the rheme of what is new.
I n Russian and Czech, for instance, these criteria of FUNCTIO-
NAL SENTENCE PERSPECTIVE are fundamental factors in deter-
mining word order. In English, where word order is partially
tied to functions such as subject and object, certain other
devices are used to indicate functional sentence perspective.
116 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

Passives may thus make themes out of objects of transitive

verbs by lifting them into the sentence-initial position which
usually identifies themes. Other structures such as introduct-
ory there, the cleft construction of the type It was x that did
y, and the inversion that allows sentences to begin with the-
matic adverbials (In London once lived a witch) may be viewed
as part of the machinery of functional sentence perspective
in English.
It would take another book to discuss the present state of
knowledge of this intricate subject. We cannot here analyse
connections between THEMATIZATION and EMPHASIS, the
marking of themes and of textual patterns in spoken language
by means of stress and intonation and other features, the
correlation between theme and the use of the definite article,
the various instances in which non-initial themes may occur,
and the like. A t present the discussion is further complicated
by the fact that different scholars do not always agree about
the definitions of basic concepts and terms such as theme and
rheme, topic and comment, and given and new. All the same the
problems of thematization are now beginning to assume their
rightful place in linguistics as one of the systems necessary
to explain the form of sentences and of texts.
In intersentence grammar as well as in text linguistics, we
cannot rest satisfied with a linguistic apparatus only capable
of discussing the statics of themes and rhemes within single
sentences. On the contrary, we shall need a THEME DYNAMICS
expressly designed for descriptions of patterns of thematic
cohesion in a string of sentences. In other words, theme dy-
namics charts the patterns by which themes recur in a text
and by which they run through a text, weaving their way
from clause to clause and from sentence to sentence.
Theme dynamics must consist of three parts. The first part
is a THEME STATICS, that is, a theory of theme in a clause and
sentence. The second part must consist of a theory and method
of THEMATIC IDENTIFICATION — some linguists might prefer
to speak about IDENTIFICATION OF GIVEN — which enables
 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE 117

us to compare thematically definable parts of different sent-

ences and to decide whether we wish to regard them as the
same or as different. And the third part should give a taxo-
nomy for PATTERNS OF THEME MOVEMENT through the suc-
cessive sentences of a text.
Approaches to theme statics are already available. Much
less work has been done on theme dynamics (cf. D A N E § 1 9 7 0 ) ,
though many of its problems have been touched upon in
analyses of style and narrative structure (e.g. K O C H 1 9 7 0 ) .
The basic problems of theme dynamics may therefore be
worth illustrating here in terms of some examples deliberately
restricted to written texts. In spoken ones, a number of
complex features of stress, intonation and prosody would
have to be reckoned with as well.
The second part of theme dynamics must, then, compare
themes of different sentences and give methods of deciding
whether two themes are the same or different, irrespective
of whether they are expressed with the same words or not.
At present, there is no sufficiently rigorous semantic theory
of synonymy, and in practical stylistic analysis we must
therefore content ourselves with some very rough-and-ready
systems of theme identification. Themes may thus be regarded
as the same if they fit into certain patterns of semantic rela-
tionship such as

repetition: The process of charging a capacitor consists of

transferring a charge from the plate at lower potential to
the plate at higher potential. The charging process there-
fore requires the expenditure of energy,
reference: On the station platform were Negro soldiers. They
wore brown uniforms and were tall and their faces shone,
synonymy: Rome was still the capital of the Pope. As if she
knew t h a t her doom was upon her, the Eternal City
arrayed herself to meet it in all her glory,
antonymy: Wise men should speak. Fools are much less inter-
esting to listen to.
118 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

comparison: John was hurt by all these accusations. Even more

painful were the suspicions of his wife,
contracting hyponymy: People got on and off. A t the news-
stand Frenchmen, returning to Paris, bought that day's
papers.
expanding hyponymy: Tulips are cheap even in January. But
then flowers seem to be necessary to Scandinavians during
the darkest season,
co-membership of the same word field: Tulips are cheap.
Roses are expensive,
sustained metaphor: The sun sagged yellow over the grass
plots and bruised itself on the clotted cotton fields. The
fertile countryside that grew things in other seasons
spread flat from the roads and lay prone in ribbed fans of
broken discouragement.

As even these few examples show, sentences are often themati-

cally linked by the simultaneous use of more than one device
of theme identification. In the example given above under
"synonymy", the sentences were thus not only kept together
by synonyms of Rome, but also by the reference Rome —
she — her — her. If necessary, these categories can be further
subdivided for greater delicacy. For instance in

Y o u could not get away from the sun. The only shadows were
made b y rocks.

the relationship between sun and shadows might be labelled

with the semiotic term INDEXAL to indicate a special type of
word-field relationship. And in passages like

She had a strong sense of her own insignificance; of her life's

slipping by while June bugs covered the moist fruit in the
fig trees with the motionless activity of clustering flies upon
an open sore. The bareness of the dry Bermuda grass about
the pecan trees crawled imperceptibly with tawny cater-
pillars. The matlike vines dried in the autumn heat and hung
 LINGUISTIC STYLE MARKERS B E Y O N D THE SENTENCE 119

like empty locust shells from the burned thickets about the
pillars of the house.

it is hard to find simple instances of repetition, reference, or

synonymy, though the choice of words is homogeneous
enough. Here, one device holding the sentences together is
the repeated juxtaposition of botanical and zoological terms,
the way in which plants suggest insects and locust shells. Of
course such thematic looseness, such a lack of more stringently
definable patterns of reference, synonymy, and the like can be
a style marker in its own right.
Even if we succeed in thus developing a provisional appa-
ratus for the identification of themes enabling us to trace
sets of themes through a text, we still need the third compon-
ent of theme dynamics: the taxonomy of the ways in which
themes move through a text. To escape the difficulties of the
terms theme and rheme, which some, but not all, investigators
identify with initial and non-initial sentence elements, we
might strictly and operationally discuss thematic movement
in terms of two positions, I(nitial) and N(on-initial). If so,
we shall arrive at four possible patterns of thematic move-
ment (assuming that I and N are precisely defined):

I to I : The fields outside the villages were full of vines. The

fields were brown. (Identification device = repetition)
I to N: A lady stood in the midst of the hail of bullets. I t
was obviously impossible to frighten her. (Pronominal
anaphoric reference)
N to I: The ratio of the velocity of light in a vacuum to the
phase velocity of light of a particular wavelength in any
substance is called the index of refraction of the substance
for light of t h a t particular wavelength. The index of
refraction will be designated by n. (Repetition)
N to N: That afternoon Jack came to London. Peter was
also there. (Anaphoric reference by adverbial of place)
1 2 0 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

If we have operationally unambiguous definitions for theme

and rheme or topic and comment, we may of course substitute
them for the distinction between I and N. Several other
principles of classifying thematic movements could be de-
vised. One criterion is syntactic function. Thus a theme may
move from the subject of one sentence to the subject of another,
from subject to object, from object to subject, and so on.
Another is syntactic structure. Thematic features may thus
move from noun phrase to verb phrase, from substantive to
verb, and so forth. One principle of classification is based on
the distance of sentences with related themes. Some texts
make frequent use of thematic movements from one sentence
to the next, that is, from sentence n to sentence n -(- 1»
whereas in other texts, movements from sentence n to
sentence n + 2, n + 3, and so on may be comparatively
common.
A terminology for what I have here called theme dynamics
was published by Frantisek DANES after this chapter had
been written (DaneS 1970a, 1970b). To identify themes,
DANE§ suggests the use of questions. If, for instance, we
wish to find the theme in Er bekam das Buch von einem Kol-
legen, we should ask Von wem bekam er das Buch ?, the answer,
Von einem Kollegen, being the rheme, the rest — Er bekam
das Buch — the theme. (Why we should not ask Wer bekam
das Buch von einem Kollegen? or Was bekam er von einem
Kollegen ? is, however, not quite clear.) Further, DANES classi-
fies patterns of theme movement as

(a) simple linear progression, in which the rheme of one sent-

ence becomes the theme of the next,
(b) passages with run-through themes (a sequence of senten-
ces with the same theme but different rhemes),
(c) progression of derived themes (there is one Hyperthema
and several hyponymic Teilthemen), and
(d) the development of a split rheme (the themes of successive
 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE 121

sentences are co-members of a concept forming the rheme

of the initial sentence, as in

Die Widerstandsfähigkeit [. . .] ist bei verschiedenen Arten

pathogener Viren sehr unterschiedlich. Poliomyelitisviren
sterben in trockene Luft sofort ab [ . . .] Bei Grippeviren
ist es hingegen umgekehrt [. . .]) (DÄNES 1970b: 77).

Here, then, theme identification and theme movement have

been fused into one single taxonomy. Such packaging of
different features is no doubt practical in those analyses in
which a group of co-occurring elements is likely to have
stylistic significance. In actual practice, a complete consider-
ation of all the factors of theme dynamics listed above will make
the actual analysis of texts very slow and intricate. Indeed
the work readily becomes so laborious that the processing
of large bodies of text consumes time and energy beyond the
point of diminishing returns. In theme-dynamic analysis as
in all practical analyses of texts, the purpose must be allowed
to determine the choice of levels of abstraction and delicacy.
As long as stylistic analysis is viewed as a practical business,
we shall do wisely in registering and counting only those
features that are likely to function as style markers. The in-
vestigator will also very soon come to realize, and to learn
the hard way, the need for simple methods of surveying and
presenting his complex findings. One type of theme-dynamic
display is the COHESION CHART, in which the clauses and sen-
tences of a text are numbered and plotted against the various
cohesion devices. Another is the STYLISTIC PROFILE, in which
relative numbers of cohesion devices are given in staple
diagrams or histograms.
Theme-dynamic patterns are not, however, the only means
by which strings of sentences satisfy the requirements of
textual well-formedness. In fact, textual well-formedness (or
"TEXTUALITY") is a function of three major types of factors.
First it depends on the grammatical well-formedness of the
1 2 2 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

individual sentences. A text consisting of ill-formed sentences

is ill-formed. Secondly, it depends on the way in which the
sentences are strung together. And thirdly, as we noted above,
it may also depend on the context: in a grammar-book we may
tolerate a string of grammatical examples which make little or
no sense outside their grammatical context.
In addition to anaphoric and cataphoric reference, pronomi-
nalization, the use of referential do or one, and other cohesion
devices traditionally discussed in sentence grammar, there
are yet other cohesion features t h a t are amenable to linguistic
analysis and description. Some of them may be readily listed:
CONTEXTUAL COHESION keeps together passages occurring
in the same matrix of contextual features. For instance, a
piece of dialogue in a novel has a contextual matrix different
from a descriptive passage in the same novel; in a play,
stage directions have a contextual matrix different from t h a t
of the dialogue; and so on. Sentences having the same con-
textual matrix are felt to belong together.
LEXICAL COHESION is a term suggesting that coherent texts
often have a homogeneous vocabulary, which contributes
to their unity. The homogeneity of the vocabulary may be
affected by the subject matter of a text. An article on nuclear
physics is likely to contain a high density of terms related to
nuclear physics. I t is also affected by other contextual fea-
tures, including style: a colloquial text is likely to use a stylis-
tically homogeneous, colloquial vocabulary.
CLAUSAL LINKAGE provides us with an arsenal of formal
means marking the ways in which clauses cohere within
sentences and sentences cohere within texts. Grammarians
have traditionally paid attention to the ways in which clauses,
join into sentences, but though many of the devices of inter-
sentence linkage are much the same as those of clausal link-
age within the sentence, ways of linking sentences into texts
still deserve special study (LONGACRE 1 9 7 0 ) . MILIC ( 1 9 6 9 : 2 1 )
has suggested a system of eight basically logical, rather than,
formal, relations between sentences. They are:
 L I N G U I S T I C STYLE MARKERS B E Y O N D T H E S E N T E N C E 123

Additive, a proposition which has no organic relation with

its predecessor (and).
Initial, the first sentence of a paragraph.
Adversative, a proposition which changes the direction of
the argument (but).
Alternative, a proposition which may be substituted for the
previous one (or).
Explanatory, a restatement, definition or expansion of the
previous proposition (that is).
Illustrative, an instance or illustration (for example).
Illative, a conclusion (therefore).
Causal, the cause for a preceding conclusion (for).

Density patterns of types of sentence linkage may offer us a

battery of additional style markers.
By ICONIC LINKAGE — another term borrowed from semio-
tics — I mean those situations in which two or more sentences
cohere because they are, at some level of abstraction, iso-
morphic (or, more popularly, "pictures of each other"). For in-
stance, one line of Pope is highly likely to be metrically iso-
morphic with another line of Pope. If so, these two lines cohere
iconically at the metrical level. To use the term iconic linkage
meaningfully, however, the investigator has to decide at what
level of abstraction the isomorphism is significant as an
iconic link. As a rule, such isomorphisms have to lie close to
the surface: it would be meaningless in this connection to
regard all sentences as pictures of each other merely because
they all share some deep structure such as N P + VP. In-
stances of rhythmic and metrical regularities, rhyme, allite-
ration, and assonance all qualify as instances of iconic linkage.
Iconic links may also be syntactic: they link The old gentle-
man elegantly kissed the young lady with The striped tiger cruelly
bit the innocent lamb. They also connect CHOMSKY'S classic
examples He is eager to please and He is easy to please, Revo-
124 LINGUISTIC STYLE MARKERS B E Y O N D T H E SENTENCE

lutionary new inventions appear infrequently and Colourless

green ideas sleep furiously. Such examples support the view
t h a t stylistically meaningful iconic links are based on simi-
larities on, or close to, the surface.
There are yet other linguistic features t h a t may add cohe-
sion to a text, for instance the consistent use of certain tenses
( W E I N R I C H 1 9 6 4 ) or the consistent use of such aspects of point
of view as can be linguistically defined ( S I N C L A I R 1 9 6 8 : 2 2 3 - 4 ) .
We are here approaching the border between text linguistics
in the strict sense, and poetic or narrative analysis of the
kinds developed by the Russian formalists, New Critics, and
French Neo-Structuralists. Today, the wisest policy may well
be to leave this border open and to welcome any translations
of literary concepts and devices into linguistic terms. To men-
tion just one instance: in some contexts, the first sentence of
a text is marked by certain characteristics some of which
are amenable to reasonably stringent linguistic description.
This does not apply only to the Once upon the time of fairy-
tales. In a body of scholarly articles in philological journals,
there seems to be a restricted number of ways and means
that writers use to provide their first sentences with apparent
'givens': existential thematization with there is, thematiza-
tion with presuppositions expressed by it is, quotations, uses of
what Sir George Rostrevor H A M I L T O N ( 1 9 4 9 ) has called the
"tell-tale article" (that is, uses of the in instances where the
following noun is neither universally known nor given in a
previous sentence), nominalizations (as in The esteem in which
scholars hold Bacon is . . ., rather than Scholars esteem Bacon),
general quantifiers such as every or all, and certain other struc-
tures seem to be particularly characteristic of text-initial
sentences. Sometimes, too, writers use part of their title as
'given' and turn it into the theme of the first sentence. In a
wider, less linguistic, description one might say t h a t text-
initial sentences reveal one of two possible strategies. Either
the first sentence is frankly acknowledged as lacking a 'given'
by the use of a narrative formula or existential phrase. Or
 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE 125

the writer tries to hide this lack by appealing to undisputed

facts, authorities, general statements, unquestioned nomi-
nalizations and the like, often marking them with the definite
article to show that he assumes them to be known not only
to himself but also to his reader. Beginning with a quotation
is another way in which writers conjure forth a "PSEUDO-
GIVEN".
What, then, is the significance of all these patterns of inter-
sentence grammar and text linguistics for stylistic analysis ?
First, they show what kinds of conceptual frames we must
use if we agree that style is not merely a quality of sentences
but also of texts. If so, we must also devise means of describ-
ing styles which reckon with textual, intersentential features
and not only with terms that refer to phenomena within the
confines of single sentences. The first step from a sentence
stylistics to a text stylistics in this, more rigorous, sense
where "text stylistics" is not merely the sum total of the
stylistic analysis of individual sentences but also a conside-
ration of the way in which sentences form texts, must be a
close study of intersentence grammar. This must concern
both spoken and written texts, for spoken language has its
own ways of marking cohesion, theme, and forms. Here we
are approaching text stylistics from the linguistic end. But
text stylistics also has other aspects. For instance, the study
of narrative patterns may be regarded as part of textual
study. Here we are once again straining against the fence
between stylolinguistics and literary stylistics. Some narra-
tive patterns may be amenable to reasonably stringent, even
linguistic, definition; others may not. Therefore the investi-
gator who moves in this frontier area may find that some
problems need a shift of method from linguistic to literary,
or vice versa.
. Secondly, patterns of intersentence grammar and text
linguistics provide us with a vast arsenal of additional style
markers. We may try to express the stylistic differential
between text and norm with the aid of densities of cohesion
1 2 6 LINGUISTIC STYLE MARKERS BEYOND THE SENTENCE

devices. We may test hypotheses such as: " X ' s scientific

style is characterized by a comparatively high density of
thematic movements from rheme in sentence n to theme in
sentence n + 1". Observations of textual cohesion patterns
and of devices of theme dynamics may also yield materials
for practical tasks such as the teaching of composition and of
normative stylistics.
Altogether, LINGUISTIC TEXT STYLISTICS is an area which is
gradually opening up for new types of inquiry. One possible
rough classification of textual style markers is into three major
fields: theme dynamics, including anaphoric and*cataphoric
reference; cohesion devices between sentences and larger
textual units; and linkage which overtly marks relations
between sentences.
 8

STYLOSTATISTICS

8.1 ROLE OF STATISTICS IN STYLISTICS

Many linguists are likely to agree that style is affected by

frequencies. In this book I have argued that the impression
we have of the style of a text is caused b y significant differ-
ences in the densities of linguistic features in this text, and
in a norm consisting of another, contextually related text or
body of text. In quantitative studies of such stylistic differen-
tials, statistical methods may come to play an important
role, not least in testing probable significance levels of the
differences between text and norm.
In fact students of style have often made pseudo-quanti-
tative statements of the type "N.N.'s style is characterized
by the frequent use of construction x." Note that such state-
ments in themselves imply comparison. If feature x is frequent
in the style of N.N., and if this frequency is characteristic of
N.N., it follows t h a t others — and in fact those whose writings
we regard as worth comparing with N.N.'s — use feature x
less often than N.N. One possible source of disagreement lies
in the choice of norm. If one critic finds N.N.'s use of feature
x very frequent, and another finds N.N.'s use of the same
feature very rare, both may still be right though they dis-
agree. For as long as the norm remains undefined, one critic
may have tacitly compared N.N. to A.A., the other to B.B.
As soon as the text and the norm are explicitly defined, how-
ever, it becomes possible to move the discussion of such
disagreements from the realms of opinion to the realms of
verifiable fact. I t becomes possible to set up inventories of
128 STYLOSTATISTICS

the stylistic differentials between text and norm, and it be-

comes possible to test their statistical significance. Of course
this need not end the disagreement and argument between
our two critics. One of them may still argue that the impor-
t a n t and therefore significant comparison of N.N. should be
with A. A. and not with B.B., and perhaps also accuse his
opponent of having neglected to pay attention to those lin-
guistic features that are really significant and therefore worth
counting. And the stylobehavioristic description of each
critic's responses to the same set of stimuli in a text will al-
ways remain subjective. Such subjectivity is the critic's
most precious privilege.
In the stylolinguistic description of texts, however, sta-
tistics has its given place. Style has often been viewed as a
statistical and probabilistic, not a deterministic, branch of
linguistics. In an earlier chapter, the capacity of making
statistical and probabilistic, and not only categorical, state-
ments was regarded as one of the characteristics required of
those grammatical models t h a t we wish to apply to stylo-
linguistic description. To those who see an opposition be-
tween a fundamentally deterministic linguistics and a funda-
mentally probabilistic stylistics, different methods are open
to resolve the conflict. One extreme solution would be t o
divorce stylistics from linguistics. But as we have seen, neat
dichotomies between a deterministic langue and a probabilistic
parole have been opposed by those who find t h a t certain
probabilities are likely to be inherent not only in texts b u t
also in our internalized linguistic rules. Another solution is
to build frequencies into one's theories of linguistics and
methods of language description.

8.2 SOME EARLY STYLOSTATISTICIANS. ZIPF AND YULE

I n spite of frequent scepticism and many justifiable caveats

(e.g. POSNER 1 9 6 3 ) , the list of attempts to apply statistics to
 STYLOSTATISTICS 129

the study of style is a long one. Those who wish to review

statistical studies of language with style in mind will do
wisely in noting that statisticians have in fact studied language
with two very different aims. Some of them have tried t o
find those statistical patterns that are common to large
samples of text, and perhaps even to all texts in all languages.
They have, in other words, been looking for statistical uni-
v e r s a l . On the contrary, others have concentrated on extract-
ing those features that make one text different from other
texts. They have sought statistical differentials, not uni-
v e r s a l . For obvious reasons, students of style are primarily
interested in the differential approach. They want to know
how one text differs from another. But both approaches are
nevertheless related and relevant. If we know what features
can be regarded as statistical universals and thus as inde-
pendent of the text and perhaps of the language, we may a t
once disregard them in our quest for stylistically significant
statistical parameters. We may, in other words, set u p an
inventory of those statistical features t h a t are potential style
markers by subtracting the statistical universals from the
total inventory of statistical characteristics of the text we
are studying.
As long as we bypass early work such as the Massoretic
studies of Biblical Hebrew, the writings of various kinds of
linguistic numerologists, and the endeavours of cryptanalysts
and other forerunners of linguistic statistics, the origins of
stylostatistics can be found in the mid-nineteenth century.
Augustus D E M O R G A N , professor of mathematics in the
University of London, showed his interest in the subject in a
letter to the Reverend W. H E A L D at Cambridge in 1851. H E
suggested t h a t H E A L D count word length in various Greek
texts to prove

that one man writing on two different subjects agrees more closely
with himself than two different men writing on the same subject.
(WILLIAMS 1 9 7 0 : 6)
130 STYLOSTATISTICS

Such data could be used for author identification. A number

of critics gradually developed the STYLOMETRIC APPROACH to
literature. They studied both ancients and moderns b y count-
ing averages and percentages of lines with metrical variations,
rhyme, run-on lines, extra syllables and the like, though with-
out those statistical controls of significance levels and standard
deviation that modern stylostatisticians regard as basic. One
important figure in this development was T. C. MENDENHALL,
an American physicist working in the last decades of the nine-
teenth century. He was apparently stimulated by a memoir
of D E MORGAN published by his widow in 1 8 8 2 ; he studied a
number of texts — Dickens, Thackeray, John Stuart Mill,
Caesar, Shakespeare, Bacon, Marlowe, and others — in terms
of word-length distribution. One of the sophisticated features
of MENDENHALL'S work was his reckoning with the influence
of sample size on the conclusions.
The importance of George Kingsley Z I P F ( 1 9 0 2 — 1 9 5 0 ) lies
in his having stimulated the statistical study of language,
especially its statistical universals. One of the focal points of
his interest was the relationship between word frequency and
word length: he found t h a t in languages as different as Chinese,
Latin, and English, the length of words tends to have an
inverse relationship to their frequency. Common words are
shorter, rare words tend to be longer. Another of ZIPF'S
contributions was an attempt at stating a massive, universal
relationship between WORD RANK and WORD FREQUENCY: if we
multiply the rank of a word (that is, its place on a frequency
list, so t h a t the most frequent word has rank 1, the second
most frequent word rank 2, and so on) with its frequency, the
result is a constant. This relation holds best in the middle
region of the vocabulary, but is less accurate at both ends,
t h a t is, with the most common words and with rare words.
ZIPF'S critics have, however, suggested t h a t the inverse re-
lationship between rank and frequency is a statistical necessity
inherent in these two concepts, and not a characteristic of
language. ZIPF went on to speculate voluminously about the
 STYLOSTATISTICS 131

all-important role of the principle of least effort in all human

behaviour, and about the character of language as a compro-
mise between elaboration and reduction, dull over-articulation
and incomprehensible under-articulation.
Very different in temperament was the Cambridge statisti-
cian, G. Udny Y U L E , whose achievement lies within differen-
tial rather than universal linguistic statistics. Y U L E had a close
look at " Z I P F ' S Law" but could not agree that it held a satis-
factory degree of precision. One of his investigations concer-
ned De imitatione Christi with the purpose of judging between
the claims of Thomas á Kempis and Jean Charlier de Gerson
as to its authorship. His main problem was to devise a
statistical measure relatively independent of sample size, and
for this purpose he developed a formula for "Yule's Character-
istic". Y U L E ' S K can be regarded as a measure of the chance
that any two nouns selected at random in a given text will
be identical. Y U L E also used sentence length as a stylo-
statistical criterion.
Among those who have refined, and added to, the pioneering
studies of ZRPE and Y U L E , three theorists of linguistic statistics
should be briefly mentioned. Benoit M A N D E L B R O T modified
Z I P F ' S rank-frequency formula by adding to it factors reflect-
ing certain features of individual texts. M A N D E L B R O T ' S
formula gives better approximations than Z I P F ' S , though it
too has been criticised on the same grounds. Wilhelm F U C K S
studied particularly the number of syllables per word, not
only as a characteristic of various languages but also as a
feature of individual styles and of the diachronic development
of styles within a language. One of his additions was the con-
cern with entropy which links stylostatistics with information
theory (see e.g. C H E R R Y 1 9 5 7 ) . And in numerous books and
articles, Gustav H E R D AN presented a number of approaches
both to the study of linguistic universals and of stylistic
particulars, including many analyses of data both his own
and cited from other studies. H E R D A N tried, among many
other things, to improve Y U L E ' S characteristic, and one of
132 STYLOSTATISTICS

his achievements is a stringent formulation of a vocabulary

connectivity test measuring the lexical homogeneity between
different partitions of a text or between different texts. Such
tests give us chances of contrasting one portion of a text
against a norm taken from the same text.

8.3 GUIRAUD. THEME WORDS AND KEY WORDS.

JOSEPHINE MILES

I n the works of Pierre GUIRAUD, we are introduced to the

concepts of THEME WORDS, and KEY WORDS which are interes-
ting both to stylolinguists and to content analysts. GUIRAUD
begins by noting that in any text, a small number of words will
make up the major part of the text: the 100 most frequent words
account for 60 per cent, the 1,000 most frequent words for 85
per cent, and the 4,000 most frequent words for 97,5 percent
of the text (GUIRAUD 1954). Having emphasized the impor-
tance of proper definitions of concepts such as 'word' — for de-
finitions must be explicit and rigorous if we are to base
statistical counts on them — GUIRATTD goes on to discuss
characteristics of word distribution patterns in terms of rank
versus frequency, and number versus frequency. One important
type of calculation attempts computation of the size and
structure of a writer's TOTAL LEXICON, of which the vocabulary
actually appearing in a given text is a sample. With the aid of
empirical data, GTTIRATJD arrives at a way of estimating the
total lexicon as a function of text length, the number of dif-
ferent words, and the structure of the vocabulary. He also
sets up formulae for the calculation of measures for VOCA-
BULARY CONCENTRATION, VOCABULARY DISPERSION, a n d VOCA-
BULARY RICHNESS. The setting-up of norms for these functions
will then permit comparison of individual texts with the
norms. GUIRAUD'S theme words, mots-themes, are those words
t h a t appear most frequently in a given writer's text (excluding
function words). Key words, mots-clés, are those full words in
the text whose frequencies significantly differ from their
 STYLOSTATISTICS 133

frequencies in the norm (GUIRAUD used VANDER BBKE'S

word-frequency list as the basis for the norm). The method
permits conclusions such as that Rimbaud and Claudel have
very much larger lexicons than Valéry, but that Rimbaud's
vocabulary dispersion is greater and vocabulary concentration
lower than Claudel's, so that Rimbaud can be said to put his
large lexicon to fuller use.
Still more akin to content analysis are the studies of poetic
vocabulary of Josephine MILES. Her materials consisted of
1,000 lines from each of 130 poets from different periods, and
she focussed her series of investigations partly on high-
frequency words and partly on proportions between adjectives,
substantives, and verbs. The favourite words of poets and
periods reflect on content, but MILES also defined certain
major types of style in English poetry. One is a PHRASAL
STYLE characterized by nominalizations and participles as
well as by metrical irregularities. Another is a CLAUSAL STYLE
marked by active use of verbs and a more complex sentence
structure. A n d the third is a BALANCED, CLASSICAL STYLE
which is intermediate between the first two. When matching
the texts of different periods against these norms, each
century from 1500 to 1900 is in fact found to fall into three
parts, each of which is characterized by the dominance of one
of the three stylistic norms.

8.4 APPROACHES AND STYLOSTATISTICAL

CHARACTERISTICS
The student of style can thus easily find a number of studies
that teach him the principles and methods of stylostatistics,
partly in the works of the scholars mentioned above, partly
in special textbooks (e.g. MTTLLER 1 9 6 8 ) and anthologies
(HALLBERG et alii. 1 9 6 6 , DOLEÈEL - BAILEY 1 9 6 9 ) . Statisticians
cannot, however, tell students of style what features will be
worth counting for stylistic significance. Statisticians operate
with figures and take no responsibility for the correlations
134 STYLOSTATISTICS

between the figures and the entities they stand for. The
student of style must decide on his own what linguistic
features he wishes to count in order to arrive at the figures
that can be tested by statistical operations. In practice, no
linguist can ever count all the features of a text of any length:
even short texts include vast numbers of features and feature
combinations that may be of stylistic significance. Therefore
it is usually wise to begin by limiting the counting of features
to those that the investigator suspects will support his obser-
vations and intuitions as to stylistic differentials between text
and norm. Some hypotheses are reasonably safe: thus we
may expect that the vocabulary patterns in the works of a
given poet will contrast sufficiently against a general norm
provided by a frequency dictionary to yield a set of what
G u i r a u d called hey words. Other hypotheses are not, and
weeks of diligent counting may fail to produce significant
results.
The distributions of sounds, phonemes, and letters appear
to be language-bound and even to show certain universal dis-
tribution patterns. Therefore they are comparatively ill
suited to the differentiation between one style and another,
except in such areas as in metrics and in sound patterns where
iconic relationships play a major role. F o c k s and some others
have found that measures of word length such as the number
of syllables per word can discriminate between individual
styles and certain period styles within a given language. Most
stylostatisticians have, however, worked at the lexical level.
Here stylistic characteristics can be defined either in terms of
general distributional characteristics of a text such as
G u i r a i t d ' s measures of richness, concentration, and dis-
persion, or in terms of frequencies of individual lexical items
such as theme words and key words, or even certain function
words. Proportions between native and borrowed words have
also been used. A t the syntactic level, sentence length and
sentence complexity are among those features that are easiest
to quantify, and they have therefore been another focus of
 STYLOSTATISTICS 135

statistical study. There are, so far, relatively few attempts a t

using intersentence patterns for stylostatistical purposes
(SWIECZKOWSKI 1961, WlNBURNE 1964).

8.5 SAMPLE AND NORM

Stylolinguistic criteria must, of course, always affect the de-

cision whether one should embark on statistical analyses of
one's materials or not. But the statistical operations them-
selves also place some limitations on stylostatistical studies.
For instance, statistical formulae are apt to penalize those
who use small samples. In some situations, a linguist cannot
manipulate the sizes of his text and his norm; he may for in-
stance be working with a dead language with few surviving
texts, or with other kinds of closed texts. I n other instances,
the linguist can himself decide how large his samples are to be.
If so, he may do wisely in consulting a statistician to find out
what minimum texts he should analyse to get samples
adequate for a statistical control of the significance of stylistic
differentials. In addition to linguistic and to statistical
factors, students of style must always reckon with the cost
of labour and other economic realities. Elaborate analyses of
large materials readily become impracticable because they
involve vast amounts of text processing. In practice, nobody
can count everything. The number of features one can count
must in practice be limited to what the investigator can
handle in his particular situation.
One way of saving labour is to use a norm which is already
available and against which one's text can be matched. I n
many types of study, such ready-made norms do not exist. I n
other studies, a suitable hypothesis may make it possible to
use already processed materials for norm. In word studies, for
instance, we may define our norm as t h a t large sample of
language from different contexts that compilers of frequency
dictionaries used for their frequency counts. For some studies,
136 STYLOSTATISTICS

such a norm may be too general; for others it may prove

adequate. Thus VANDER BEKE'S word list was sufficient to
bring out the differences between the key words of several
French poets. Nowadays, large masses of text have been pro-
cessed by compilers of frequency dictionaries, students of
usage, concordance-makers, and others; a student of style
with access to such materials may use them to great advantage,
for instance by extracting new information out of existing
data banks by means of computer programmes of his own.
There are many types of stylistic studies in which it would be
most unwise for the student to closet himself to count manual-
ly thousands or millions of features which might, with proper
permission and knowhow, be extracted from materials already
on tape for computer treatment. The periodical Computers in
the Humanities is one clearing-house of relevant information.

8.6 ATTRIBUTION STUDIES

I t is one thing to start out from the observation that two

styles are different, and to apply statistics to check the
validity of this observation. I t is another matter to apply
statistics to the testing of the hypotheses that two texts may,
or may not, be stylistically different. The statistical methods
used in both situations are, of course, very much the same.
But the latter type of question — testing whether text and
norm are significantly different or not — occurs in one parti-
cular variant of stylostatistics, namely author identification.
Attribution problems have, of course, been studied without
strict stylolinguistic quantification (ERDMAN - VOGEL 1 9 6 6 ) ,
but they have gained in stringency with the introduction of
explicit statistical tests evaluating the probability levels at
which we may regard two texts as having been written by
the same person.
Once again, available statistical data on author identifica-
tion support our definition of style as the aggregate of signi-
 STYLOSTATISTICS 137

ficant differences between a text and a contextually related

norm. For what is done to identify the author of a text is a
comparison of the densities of selected linguistic features in
the problem text with the corresponding densities in a set of
comparable texts by the suspected author. In other words:
attribution studies are based on matching the densities in the
text with those in a norm, or in a set of norms. (Cf. HALLBERG
1 9 6 2 , 1 9 6 3 , 1 9 6 8 ; M C K I N N O N - WEBSTER 1 9 6 9 ; MORTON 1 9 6 5 ;
MTJNDT 1969).
To exemplify these procedures, let us briefly glance at just
two of the best-known stylostatistical author attributions in
English. A classic instance is Alvar ELLEGARD'S study of the
Junius Letters, the celebrated political pamphlets from the
eighteenth century whose authorship has been in dispute ever
since they burst upon the world. The letters themselves
comprise some 82,200 words. ELLEGARD matched them against
a norm comprising about one million words of political literat-
ure from the same period. His growing familiarity with this
genre suggested t h a t some words were more common in the
Junius Letters than in the norm, and that others were less
common in the text than in the norm. ELLEGARD then went
on to count these "Junian plus" and "Junian minus" words.
As the likeliest identification of Junius was with Sir Philip
Francis, he also used a 231,300-word sample known to be b y
Francis for an additional norm. The conclusions have to do
both with the authorship of the Junius Letters and with the
method. Sir Philip Francis indeed proved to be the most
likely author. As to method, ELLEGARD states t h a t his
statistical test is sufficiently sensitive to identify the author
of a work, provided that the work itself amounts to 10,000
words or more (though considerable discrimination can be
obtained even with texts of 2,000 words), and t h a t there is a
sample of about 100,000 words by the suspected author as
well as a comparable, contemporary sample of some 1,000,000
words of contextually related text. ELLEGARD'S method also
included fusion of certain groups of features and texts with
138 STYLOSTATISTICS

the purpose of increasing their size a n d thus adding to the

effectiveness of the test by minimizing the statistical penali-
zation of small samples.
ELLEGARD'S list contained 4 5 8 features, of which the
majority were substantives, adjectives, verbs, and adverbs,
and some were prepositional phrases, prepositions, and con-
junctions. There are, however, situations where content words
such as substantives, adjectives and verbs fail to discriminate
between texts that may have been written by different au-
thors. If no other criteria such as syntactic structures, sentence
length, or word length can be used, the investigator is compel-
led to base his tests on densities of function words such as
articles, conjunctions, pronouns, and auxiliary verbs. This was
the method by which Frederick MOSTELLER and David L.
WALLACE analysed the authorship of the disputed items in
The Federalist Papers (MOSTELLER - WALLACE 1 9 6 3 , 1 9 6 4 ;
L E E D 1 9 6 6 ) . By comparing the 1 8 papers known to be by
Hamilton with the 14 papers certainly written by Madison,
they set u p a first list of 165 discriminator words, which was
later reduced to those 30 words that had been found best to
separate Hamilton's and Madison's works. Madison turned
out to be the likelier author of the disputed papers. Apart from
their statistical method which was based on Bayes's theorem
and the Poisson distribution, the most interesting and sur-
prising contribution of MOSTELLER and WALLACE to the
methodology of stylostatistics was their demonstration of the
discriminatory power of function words. These "utterly
mundane high-frequency function words" such as wpon, also,
and, by, of, on, there, this, to, and although had often been dis-
missed as useless by students of author-identification tech-
niques. I t is in the nature of such words to appear in a wide
spectrum of contexts, and it had therefore been assumed
that their density pattern is relatively unaffected by style.
But once they are shown to be capable of discriminating
between different authors, as MOSTELLER and WALLACE did
in their work, a study of their densities is in some respects
 STYLOSTATISTICS 139

very favourable in stylolinguistic investigations. To begin

with, they are frequent; they are also less affected by content,
genre, and subject matter than are many other types of
linguistic features.
Stylistic tests and author-identification techniques have
also been applied to forensic purposes, not only by scholars
b u t also by others. Two of the cases freely publicized have
been those of Bishop Helander in Sweden, whose career was
ruined after a lawsuit centering on the authorship of a series
of anonymous letters (three polemical publications are
MOD£ER 1 9 5 2 , RTTNQUIST 1 9 5 8 , a n d T E N O W 1963), and of
Timothy John Evans in England, who was hanged in 1950
and posthumously pardoned in 1 9 6 6 (SVARTVIK 1 9 6 8 ) . In the
Helander case, the suspect was a highly literate professional
man, and the comparison could be based on a number of
books and articles he had written. The linguistic analysis of
the Evans statements by J a n SVARTVIK had to be founded on
slender materials consisting of policemen's written records of
the oral statements of an illiterate. Though the materials fail
to permit firm conclusions, those paragraphs t h a t Evans later
claimed were untrue and owed to his being frightened and
upset proved to be linguistically somewhat different from
those paragraphs which Evans persisted in admitting as true
evidence. Even the Helander case was, of course, not decided
on the strength of linguistic evidence alone. Opinions differ as
t o what weight stylostatistics and author identification can,
and should, carry as judicial evidence. Even those who be-
lieve that a person's style is one of his stable individual
characteristics must reckon with the possibility of stylistic
imitation and forgery. I t is another matter that stylostatistics
may, under suitable circumstances, claim considerable dis-
criminating power. Thus chi-square testing of 19 stylistic
variables in samples of 1,000 words sufficed to distinguish
adequately between 19 Greek and Roman writers ( M O E R K
1 9 7 0 ) . Such findings should not, however, be rashly translated
into forensic terms.
140 STYLOSTATISTICS

8.7 SUMMARY

Altogether, stylostatistics is capable of two types of operation.

I t can devise statistical parameters of various kinds, such as
those listed by G u i r a u d , H e r d a n , and M i s t r i k (1967). And
it can test the fit and significance level of various hypotheses
and data, and thus estimate the likelihood t h a t certain dif-
ferences and distribution patterns owe to choice and not to
chance. Any textbook on statistics will contain relevant in-
formation on these approaches to the testing of significance.
They range from simple standard-deviation and chi-square
analysis to very elaborate and sophisticated devices.

8.8 T H E ROLE OF T H E COMPUTER

Before we leave stylostatistics, a glance at the role of t h e

computer in studies of language is in order. I n fact it already
entered into our argument in section 8.5.2. In processing large
masses of text, the computer is often useful, and sometimes an
indispensable aid: there are certain types of investigation
t h a t cannot in practice be carried out without computer as-
sistance. Some of these investigations involve too much
mechanical drudgery, others too much meticulous precision,
for any investigator, assistant, or team working manually.
Still, as any investigator will very quickly find, computer
studies cannot be undertaken lightly. First, the problems have
to be defined with very great precision. Often a student of
style is groping in a jungle of hypotheses too vague for exact
programming. Then there are the costs of programming, in-
put (card or tape punching), and actual computer time. Once
the problem has been given a formulation sufficiently exact
for programming, the decision becomes largely economic —
not a problem of what can be done, but one of what is worth
doing. If the text is already available in punched format or on
tape — either in a text bank or, as in Sture A l l e n ' s frequency
 STYLOSTATISTICS 141

dictionaries, on tapes used for the typesetting of newspapers

— one major expense, t h a t of input, has been eliminated.
What, then, can the computer do ? The basic operations are
few in number. A computer can sort out large numbers of
items and thus produce suitably ordered lists such as concor-
dances or, in reverse, rhyme indices. If the input consists of
coded representations, for instance of syntactic structures, the
computer can sort out these in any desired order by any
criterion or combination of criteria. The computer can also
count the items it lists. And it can perform any numerical
calculations, for instance of all those controls of significance
t h a t enter into stylostatistics. Out of combinations of simple
operations, we may build very complex programmes for
specific purposes. Computers can collate texts, list colloca-
tions and co-occurrences of features, list those sentences out of
different texts that share certain features or feature combina-
tions, and the like. (Examples are available in B E S S I N G E R
1964, LEED 1966, BOWLES 1967, and A L L E N - THAVENITJS
1 9 7 0 . ) B u t even if the computer can perform the desired
operations, the crucial question for the investigator is likely
to be economical. Are the yields likely to stand in a reasonable
relation to the expense? In practice, manual pilot studies
simulating computer operations as well as preruns of limited
materials will often save the investigator from costly mistakes.
So far, relatively few linguists have fully mastered the intri-
cacies of programming and computer technology and been
able to keep their knowledge and skill u p to date. Therefore
most students of style who wish to embark on computer studies
will have to find and to employ programmers willing to im-
merse themselves in the special problems of the stylolinguistic
processing of texts.
Quite a few aids for the study of styles have already been
produced with the aid of computers. The number of COM-
P U T E R CONCORDANCES is steadily increasing. Students of
English are by now familiar with the Cornell concordances of
Matthew Arnold, Blake, Emily Dickinson, and Yeats. Marvin
142 STYLOSTATISTICS

SPEVACK'S Shakespeare concordance (SPEVACK 1 9 6 8 - 7 0 ) is

the best example so far of a CONTEXT-SENSITIVE WORD LIST.
I n fact SPEVACK'S concordance is not one but several, inter-
locking concordances: one for each character, one for each
play, one for each group of plays, and one for the whole of
Shakespeare. Such contextual partition makes possible a rapid
matching of a large number of Shakespearean subtexts
against any other Shakespearean subtext or combination of
subtexts. We may, for instance, rapidly analyse the individual
vocabularies of Shakespeare's characters and match them
against the vocabularies of other characters or character
groups. Another advantage of computer concordances for
the student of style is their ease of giving full lists of gram-
matical form-words. They can be stylistically significant, b u t
were often omitted from manually compiled concordances
because of the labour involved.
Another category of aid nowadays provided by computer
i s t h e FREQUENCY DICTIONARY. (JOSSELSON 1 9 5 3 , JELINEK -
BECKA -TE&ITELOVA 1 9 6 1 , STEJNFEL'DT 1963, JUILLAND -
CHANG-RODRIGUEZ 1964, JUILLAND - EDWARDS - JUILLAND
1 9 6 5 , KUCERA - FRANCIS 1 9 6 7 , MISTRIK 1 9 6 9 , ALLEN 1971).
As noted above, the very general character of such overall
frequency dictionaries may make their use awkward for those
stylistic purposes in which we need a contextually more re-
stricted norm. Some frequency dictionaries do, however, offer
data on contextually restricted subsections of the corpus. And
even the general lists can be used for the basis of indirect com-
parisons. Thus if we wish to compare a text T with a norm N,
we may t r y to compare both, T and N, separately with the
frequency dictionary, and then base our comparison of T and
N on the results of the two differentials, t h a t between T and
the frequency dictionary and t h a t between N and the fre-
quency dictionary. Such comparisons will give us a measure
of the similarities and differences between T and N in the
light of a large, contextually more varied sample of language.
Of course the older, manually compiled frequency dictionaries
 STYLOSTATISTICS 143

(KAEDING 1 8 9 8 , M E I E R 1 9 6 7 , V A N D E E B E K E 1929, THOEN-

etc.) are still useful, too. And special word
DIKE - L O E G E 1 9 4 4 ,
lists such as those of the Enquête du français élémentaire of St.
Cloud (GOTJGENHEIM 1 9 5 9 ) and the Finnish dictionary of school-
children's vocabularies (KARVONEN et alii 1970) have a more
direct concern with contextually restricted styles.
In addition to computer programmes devised for content
analysts but also applicable to work with style (STONE et alii
1966), there are descriptions of computer programmes design-
ed especially for stylistic analysis. Thus Sally A. SEDELOW'S
V I A programme is essentially a thesaurus establishing collo-
cations for stylistic comparison, her M A P T E X T being a code
adaptable to the study of syntagmatic patterns of various
k i n d s (SEDELOW 1 9 6 5 - 6 7 , 1 9 6 6 ) .

8.9 PERSPECTIVES

Stylostatistics, then, is an area where different cultures meet.

In some quarters there has been a curious reluctance to admit
statistics, and particularly computer-produced ones, into
linguistic and literary study. There is a fear of technology in
general and the statistical trivialization of the humanities in
particular. Indeed it is true that the computer now makes it
possible for us to produce enormous quantities of meaningless
figures more rapidly than ever before. A t the same time, it
gives us a chance of obtaining meaningful data that used to
be denied us b y the mechanical labour involved in their ex-
traction. The responsibility, the praise or the blame, should
not, however, be placed on the computers but on the
men and women who use them. A computer only obeys
orders; the scholars and students are the ones who ought to
guarantee that its capacities are not misused and that the
statistical figures represent meaningful data. Those who still
fear that the humanities will be inundated by vast masses of
computer-produced statistics, containing plenty of sound and
144 STYLOSTATISTICS

fury but signifying nothing, can also seek comfort in the fact
that computerized research is neither cheap nor particularly
convenient or soothing to the investigator. In practice, con-
siderations of economy and of personal convenience will often
combine to keep investigators from using computers until
they really must.
 9

CONCLUSION AND SUMMARY

In this book, style has been viewed as one of the responses

that people have to texts. Irrespective of how they try to
verbalize their impressions of this response, the response itself
is based on the tacit or explicit comparison of the text with an
imagined or explicit norm which the recipient regards as a
relevant background for the text. The norm against which
the text is matched is chosen on the basis of contextual relev-
ance, context being definable as the aggregate of stylistically
relevant features in the textual and situational envelopes sur-
rounding the text. Our impressions of styles result from com-
parisons of densities of linguistic features in the text and in
the contextually related norm: features common in their
context give a different impression from features that are
rare in that context. The choice of norm is, however, depend-
ent on the recipient. Different people may compare the same
text with different norms and may accordingly view the style
of that text in a different light. The comparison of text and norm
will result in observations of similarities as well as of differences.
From a point of view starting from the generation of texts,
stylistically significant linguistic features have also been
viewed as results of stylistic choice. However, the range of
choice available to the speaker/writer does not appear from
the text itself. If style is to be analysed in terms of choice, we
must, first of all, be capable of distinguishing stylistic choice
from other types such as pragmatic and grammatical choice.
Secondly, we shall need a linguistic apparatus which ex-
plicitly maps out all those alternatives from among which the
one actually present in the text was chosen. In the past
146 CONCLUSION A N D SUMMARY

fifteen years, linguistic models have become much more ex-

plicit in their description of the sequence of choices t h a t lead
to the generation of a given sentence. No model presently
available is, however, capable of offering detailed mechanisms
showing precisely how a generative grammar could react to
matrices of contextual features and choose stylistic variants
appropriate to each situation.
When we hear or read a text in its linear, temporal progres-
sion, we analyse it simultaneously in a number of ways. One
of these analyses takes place in terms of stylistic probabilities.
These probabilities are determined by the context and by our
past experience of the densities of linguistic features in con-
texts that we regard as comparable. Our past experience of
contextual frequencies thus turns into a set of present pro-
babilities, against which we match the features t h a t actually
emerge during the progression of the text. In certain societies,
epochs, and styles, high conformity to contextual probabilities
has been at a premium. Individual departures from the norm
have been frowned upon; surprises are regarded as undesirable,
ugly, vulgar, socially dangerous, or even morally reprehensible.
In other societies, epochs, and genres, communicants have been
praised for their novel departures from stylistic norms. That
poet has been said to be the greatest who has most departed
from the language customarily used in the same kind of poetry.
Thus some periods favour a classic harmony, others a romantic
disorder; some groups have been conditioned to love fulfilled ex-
pectations, whereas others have preferred the shocks of surprise.
Those linguistic features whose densities in the text are
significantly different from those in the norm are called style
markers. Style markers are not absolutes: their inventory
depends not only on the text b u t also on the choice of norm.
If the norm changes, the inventory of style markers may also
change. Style markers may be features whose frequency in
the text is greater or smaller than in the norm. Thus the
absence of features that are usually present in similar contexts
can be a style marker, too.
 CONCLUSION AND STJMMABY 147

Any linguistic features may function as style markers.

Some may be paralinguistic and include an individual's voice
or gestures. Others may be phonetic: in English, they might
involve greater or less use of "weak" or "reduced" forms in
solemn contexts. Rhythmic patterns, including cursus, may
mark both poetic and prose styles. Examples of morphological
style markers would be the use of thou lovest and he loveth in
English, and I aren in Swedish, religious texts. Some style
markers are lexical, as in nice chap and fine man; others are
syntactic and may comprise not only individual syntactic
features within clauses and sentences, but also word order
and sentence complexity and length. Nor do style markers
stop at the borders of sentences. Intersentence cohesion de-
vices, patterns of theme dynamics, ways of opening texts or
sections of text, and other features of intersentence grammar
and text linguistics will provide us with a large stock of ad-
ditional style markers. Orthographic and typographical de-
vices of different types may also help to mark written styles.
And as new areas of language become accessible to stringent
linguistic description, we shall gain new potential style
markers which are today counted among rhetorical and
literary, not linguistic, features.
As my attempts at definitions have shown, style is a nota-
tional term, an abbreviation for a concept t h a t can be de-
fined in terms of other, more basic, concepts. Style is not a
linguistic prime. This means that each stylolinguist owes t o
his readers an explicit report of precisely what he means by
style, and by what methods he has arrived a t his results and
conclusions. The virtue of linguistic stylistics, as opposed t o
other kinds of stylistics, rests squarely on its ability to make
precise, objectively verifiable statements.
Linguistic stylistics applies certain linguistic concepts t o
the study of texts. B u t the same texts will most likely be
worth studying from other points of view and with other
methods as well. Of all people, students of style ought to be
least parochial: they work in a busy border zone, where
148 CONCLUSION A N D SUMMARY

courtesy and tact should count as virtues. When the methods

of stylolinguistics prove inadequate, the needs of textual
analysis may in fact suggest borrowing from neighbouring
areas such as rhetoric and literary study. There is also another
branch of linguistic stylis tics, that trying to provide for
stylistic variation in linguistic theory. Here, different linguistic
theories must be evaluated on the basis of their capacity to
include styles, and developed so that they can better incor-
porate varieties of language such as styles.
A number of studies of linguistic stylistics begin by stating
that stylolinguistics has come of age, and I feel bound to re-
peat this true, though perhaps commonplace, statement as
my final remark. The surge of interest was marked by events
such as the Bloomington conference in 1958 and the stylo-
linguistic debate in Voprosy Jazykoznanija in the middle
'fifties. Those wishing to substantiate the continuous and
growing preoccupation with style may count articles on
stylistic subjects in linguistic journals, list new periodicals
devoted largely or entirely to stylistic questions, and cite the
development of sections on stylistics at linguistic congresses
in support of their views. Also, thanks to an increasingly
lively debate, a number of fundamental questions of stylo-
linguistics have been brought out to the fore and given their
due perspective. On some of them, there is increasing agree-
ment. Many others remain unsolved. Some will be solved dif-
ferently by different schools of linguists. And many answers
are bound to change with the changes and developments in
linguistic perspectives. A number of different movements and
specialities have already arisen within linguistic stylistics, and
many more are bound to come. Different areas of specializ-
ation exist, and no single person can hope to master, or even
follow, all the lines of development that are topical today. If
an introduction of the size of mine succeeds in giving its
readers some of the basic perspectives they need for further
reading, it has achieved its goal.
 10

BIBLIOGRAPHY

A very large number of articles bearing on stylistics and on

styles has been, and is being, published in a very wide range
of linguistic and literary periodicals, some of which — such as
Language, and Style, Lingua e stile, and Style — are expressly
devoted to stylistics. Standard bibliographies such as the
Bibliographie linguistique, Language and Language Behavior
Abstracts (LLBA), the annual bibliography of the Modern
Language Association of America, the Annual Bibliography of
English Studies, the Year's Work in English Studies, and the
Tear's Work in Modern Language Studies, as well as more
specialized bibliographies contain numerous references to rel-
evant books and articles. Many of the literary and period
bibliographies often scatter relevant references and therefore
need close perusal.

10.1 STYLISTIC BIBLIOGRAPHIES

Bailey, Richard W. - Dolores M. Burton

1968 English Stylistics; a Bibliography (Cambridge, Mass. —
London: The M.I.T. Press).
Bailey, Richard - Lubomir Doleiel
1968 An Annotated Bibliography of Statistical Stylistics (= Michi-
gan Slavic Series) (Ann Arbor, Mich.: University of Michi-
gan Press).
Guiraud, Pierre
1954 Bibliographie critique de la statistique linguistique (Utrecht —
An vers: Spectrum Publishers).
Hatzfeld, Helmut
1952 A Critical Bibliography of the New Stylistics Applied to the
Romance Literatures, 1900—1952 (— University of North
150 BIBLIOGRAPHY

Carolina Studies in Comparative Literature 6) (Chapel Hill,

N.C.: U n i v e r s i t y of Carolina).
1956 Bibliografia critica de la nueva estilistica applicada a las litera-
turas romanicas (Madrid: Gredos).
Hatzfeld, H e l m u t - Y v e s L e H i r
1961 Essai de bibliographie critique de stylistique française et romane
(1955—1960) (Paris: Presses Universitaires de F r a n c e ) .
Milic, Louis T .
1967 Style and Stylistics; an Analytical Bibliography (New Y o r k :
T h e F r e e Press).
Todorov, T z v e t a n
1970 " L e s É t u d e s d u style; bibliographie sélective", Poétique
2: 224-32.
Tyl, Zdenëk (ed.)
1970 A Tentative Bibliography of Studies in Functional Sentence
Perspective ( P r a h a : Ceskoslovenska Akademie Ved).

Many of the books and articles listed below contain ex-

tensive bibliographies.

10.2 B O O K S A N D A R T I C L E S

The following list contains the items referred to in the text, as

well as a small selection of other representative works. Many-
important older studies have been omitted because references
to them are available in the works listed here. Collections of
papers have been marked withe, and only those articles that
are expressly referred to in the text have been re-listed
separately under their authors' names.
Adamec, P .
1966 Porjadok slov v sovremennom russkom jazyke [Word order
in c o n t e m p o r a r y Russian] ( - Rozpravy ÔSAV-RSP
76.15) ( P r a h a : Ceskoslovenska Akademie Vëd).
Admoni, V . G.
1966 " R a z m e r predlozenija i slovosoôetanija k a k javlenie sintaksi-
ôeskogo s t r o j a " [Length of sentences a n d of word combina-
tions as p h e n o m e n a of syntactic s t r u c t u r e ] , Voprosy jazy-
koznanija 15.4: 111—8.
A k h m a n o v a , O.S. et alii
c l 9 6 3 Exact Methods in Linguistic Research, t r a n s l a t e d f r o m t h e
 BIBLIOGRAPHY 151
Russian by David G. Hays and Dolores V. Möhr (Berkeley —
Los Angeles: University of California Press).
cl966 0 principax i metodax lingvostilistiSeskogo issledovanija [On
principles and methods of lingùostylistic research] (Moskva:
Izdatel'stvo Moskovskogo Universiteta).
Âkermalm, Âke
1966 Eubriksvenska [The Swedish of headlines] ( — Skrifter uigivna
av Modersmälslärarnas förening 100) (Lund: Gleerups).
Allén, Sture
1970a Nusvensk frekvenaordbok 1 (Göteborg: Almqvist & Wiksell).
1970b "Vocabulary Data Processing" in Benedektsson (1970:
235-61).
Allén, Sture — Jan Thavenius (eds.)
cl970 Sprâklig databehandling. Datamaskinen i sprdk- och litteratur-
forskning [Language data processing. The computer in
linguistic and literary research] (Lund: Studentlitteratur).
Alonso, Amado
1942 "The Stylistic Interpretation of Literary Texts", Modem
Language Notes 57: 489—96.
Anttila, Raimo
1963 "Loanwords and Statistical Measures of Style in the Towneley
Plays", Statistical Methods in Linguistics 2: 73—93.
Austin, J . L.
1962 How to Do Things with Words (Oxford: The Clarendon Press).
Bach, Emmon - Robert T. Harms (eds.)
cl968 Universals in Linguistic Theory (New York, etc.: Holt, Rine-
hart and Winston, Inc.).
Bally, Charles
1951 Traité de stylistique française3, 2 vol. (Geneva: Georg; Paris:
Klincksieck).
Berthes, Roland
1966 Le Degré zéro de l'écriture (Paris: Éditions du Seuil).
Barthes, Roland (ed.)
cl966 L'Analyse structurale du récit (= Communications 8) (Paris:
Éditions du Seuil).
Benediktsson, Hreinn (ed.)
cl970 The Nordic Languages and Modem Linguistics; Proceedings
of the International Conference of Nordic and General
Linguistics, University of Iceland, Reykjavik, July 6—11,
1969 (Reykjavik: Visindafélag islendinga).
BeneS, Eduard
1968 "On Two Aspects of Functional Sentence Perspective", Tra-
vaux Linguistiques de Prague 3: 267-74.
152 BIBLIOGRAPHY

Bense, Max
1962 Theorie der Texte; eine Einführung in neuere Auffassungen
und Methoden (Köln: Kiepenheuer & Witsch).
1969 Einführung in die informationstheoretische Ästhetik; Grund-
legung und Anwendung in der Texttheorie (Reinbek-Ham-
burg: Rowohlt).
Bezzel, Chris
1969 " A G r a m m a r of Modern German Poetry", Foundations of
Language 5: 470-87.
Bickerton, Derek
1969 "Prolegomena to a Linguistic Theory of Metaphor", Founda-
tions of Language 5: 34—52.
Bessinger, Jess B., J r . , et alii
cl964 IBM Literary Data Processing Conference Proceedings (New
York: Modern Language Association, Materials Center).
Birdwhistell, R a y L.
1952 Introduction to Kinesics; an Annotation System for Analysis
of Body Motion and Gesture (Washington, D. C.: Department
of State Foreign Service Institute).
Bloomfield, Leonard
1933 Language (New York: H e n r y Holt Company).
Bloomfield, Morton
1967 "The Syncategorematic in Poetry: F r o m Semantics t o
Syntactics", in: To Honor Roman Jakobson; essays on the
Occasion of His Seventieth Birthday 11 October 1966 (— Janua
Linguarum, series maior 31) (The Hague — Paris: Mouton),
pp. 309-17.
Boost, K a r l
1959 Neue Untersuchungen zum Wesen und zur Struktur des deut-
schen Satzes; der Satz als Spannungsfeld (= Deutsche Aka-
demie der Wissenschaften zu Berlin, Veröffentlichungen des
Instituts für deutsche Sprache und Literatur 4) (Berlin: Aka-
demie-Verlag).
Bloch, Bernard
1953 "Linguistic Structure a n d Linguistic Analyses", in: Report
of the Fourth Annual Round Table Meeting on Linguistics and
Language Teaching (Ed.: Archibald A . H i l l ) (= Monograph
Series on Language and Linguistics 4) (Washington, D. C.:
Georgetown University Press), p p . 40—4.
Bowles, E. A. (ed.)
cl967 Computers in Humanistic Research (Englewood Cliffs, N. J . :
Prentice-Hall Inc.)
 BIBLIOGRAPHY 153
Bowman, Elizabeth
1966 The Minor and Fragmentary Sentences of a Corpus of Spo-
ken English ( = Publications of the Indiana University Research
Center in Anthropology, Folklore, and Linguistics 42) (The
H a g u e : Mouton).
Brewster, William T. (ed.)
1913a English Composition and Style; a Handbook for College
Students (New York: The Macmillan Company).
cl913b Representative Essays on the Theory of Style (New York: The
Macmillan Company).
Brook, G. L.
1970 The Language of Charles Dickens (London: Andre Deutsch).
Brooke-Rose, Christine
1958 A Grammar of Metaphor (London: Seoker a n d Warburg).
Brown, H u n t i n g t o n
1966 Prose Styles; five Primary Types (— Minnesota Monographs
in the Humanities 1) (Minneapolis, Minn.: University of
Minnesota Press).
Buttke, K u r t
1963 "Besonderheiten der Wortfolge in Stilen der russischen
Literatursprache der Gegenwart", Zeitschrift für Slawistik
8: 785-92.
1969 Gesetzmässigkeiten der Wortfolge im Russischen (= Lingui-
stische Studien) (Halle/Saale: Max Niemeyer Verlag).
Budagov, R . A.
1967 Literatumye jazyki i jazykovye stili [Literary languages and
language styles] (= Biblioteka filologa) (Moskva: Vysäaja
Skola).
Caplan, H a r r y (ed.)
1964 [Cicero] ad C. Herennium De ratione dicendi - Rhetorica ad
Herennium (—Loeb Classical Library) (London: William
Heinemann Ltd.; Cambridge, Mass.: H a r v a r d University
Press).
Carvell, H . T. — J a n Svartvik.
1969 Computational Experiments in Grammatical Classification
(= Janua Linguarum, series minor 61) (The Hague: Mou-
ton).
Cassirer, P e t e r
1970 Deskriptiv stilistik; en begrepps- och metoddiskussion [Des-
criptive stylistics; a discussion of concepts and methods]
(= Acta Universitatis Gothoburgensis, Nordistica Gothobur-
gensia 4) (Stockholm: Almqvist and Wiksell).
154 BIBLIOGRAPHY

Catford, I. C.
1965 A Linguistic Theory of Translation (= Language and Lan-
guage Learning 8) (London: Oxford University Press).
Chapman, R . A.
1967 The Interpretation of Deviant Sentences, Dissertation, Uni-
versity of California, Berkeley. (Ann. Arbor, Mich.: U n i -
versity Microfilms No. 68-6702).
Chatman, Seymour
1962 "Linguistic Style, Literary Style and Performance: Some
Distinctions", Georgetown Monographs on Language and
Linguistics 13: 73—81.
1966 "On t h e Theory of Literary Style", Linguistics 27: 13-25.
1967 "Style: A Narrow View", College Composition and Communi-
cation 18: 72-6.
1969 "New Ways of Analyzing Narrative Structure, with an E x -
ample f r o m Joyce's Dubliners", Language and Style 2: 3—36.
Chatman, Seymour (ed.)
cl971 Literary Style: A Symposium (New York: Oxford University
Press).
Chatman, Seymour - Samuel R . Levin (eds.)
cl967 Essays on the Language of Literature (Boston: The Houghton
Mifflin Company).
Cherry, Colin
1966 On Human Communication1 (Cambridge, Mass. — London::
The M. I. T. Press).
Chisholm, William S., Jr.
1967 "An Exercise in Syntactic Stylistics", Linguistics 33: 24—36-
Chomsky, Noam
1957 Syntactic Structures (= Janua Linguarum, series minor 4)-
(The Hague: Mouton).
1965 Aspects of the Theory of Syntax (Cambridge, Mass.: M. I. T.
Press).
1966 "Topics in the Theory of Generative Grammar", in: Current
Trends in Linguistics (Ed.: Thomas A. Sebeok) 3 (The H a -
gue, — Paris: Mouton).
Christiansen, Broder
1909 Philosophie der Kunst ( H a n a u : Clauss & Feddersen).
Closs, see Traugott
Cooper, Lane (ed.)
cl907 Theories of Style (New York: The Macmillan Company).
Coseriu, Eugenio
1962 Teoria del lenguaje y lingüistica general (Madrid: Gredos)..
 BIBLIOGRAPHY 155

Craddock, Sister Clare Eileen

1952 Style Theories as Found in Stylistic Studies of Romance
Scholars (1900—1950) (= Catholic University of America,
Studies in Romance Languages and Literatures 43) (Cambridge:
University Microfilms), [typescript]
Cressot, Maurice
1947 Le style et ses techniques (Paris: Presses Universitaires d e
France).
Croce, Benedetto
1904 Estetica como scienza dell'espressione e linguistica generate2
(Palermo: Sandron).
Croll, Morris W .
1966 Style, Rhetoric and Rhythm; essays b y Morris W . Croll edited
b y J . Max Patrick, Robert O. Evans et alii (Princeton, N. J . :
Princeton University Press).
Crosland, M. P .
1962 Historical Studies in the Language of Chemistry (London, etc.:
Heinemann).
Crystal, David - Randolph Quirk
1964 Systems of Prosodic and Paralinguistic Features in English
(= Janua Linguarum, series minor 39) (The Hague: Mouton).
Crystal, David — Derek D a v y
1969 Investigating English Style (London — Harlow: Longmans,
Green and Co. Ltd.).
Cunningham, J . V. (ed.)
1966 The Problem of Style (Greenwich, Conn.: Fawcett WorldLibrary).
Dahl, Liisa
1970 Linguistic Features of the Stream-of-Consciousness Techniques
of James Joyce, Virginia Woolf and Eugene O'Neill (= Anna-
tes Universitatis Turkuensis B. 116) (Turku: T u r u n Yliopisto).
Dahl, Osten
1969 Topic and Comment; a Study in Russian and General Trans-
formational Grammar (= Acta Universitatis Gothoburgensis,
Slavica Gothoburgensia 4) (Stockholm: Almquist & Wiksell).
DaneS, FrantiSek
1964 " A Three-Level Approach to Syntax", Travaux linguistiques
de Prague 1: 225-40.
1970a "One Instance of Prague School Methodology: Functional
Analysis of Utterance and Text", in: Method and Theory in
Linguistics (Ed.: Paul L. Garvin) ( = Janua Linguarum, series
maior 40) (The Hague - Paris: Mouton).
1970b "Zur linguistischen Analyse der T e x t s t r u k t u r " , Folia Lin-
guistica 4: 72—8.
156 BIBLIOGRAPHY

Danks, Joseph H .
1969 "Grammaticalness and Meaningfulness in the Comprehension
of Sentences", Journal of Verbal Learning and Verbal Be-
havior 8: 687-96.
Davie, Donald
1952 Purity of Diction in English Verse (London: Chatto & Win-
dus).
1955 Articulate Energy (London: Routledge - Kegan Paul).
1963 The Language of Science and the Language of Literature,
1700-1740 ( = Newman History and Philosophy of Science
Series 13) (London - New York: S h e e d - W a r d ) .
Delbouille, P.
1960 "Definition du fait de style", Gahiers d'analyse textuelle
2: 94-104.
Denison, Norman
1969 "Sociolinguistics and Plurilingualism" in GRATO (1970 1:
551-9).
DeVito, Joseph A.
1967 "Style and Stylistics; an Attempt at Definition", Quarterly
Journal of Speech 53: 248-55.
Devoto, Giacomo
1950 Studi di stilistica (Firenze: Le Mounier).
1962 Nuovi studi di stilistica ( — Biblioteca di litteratura e dell'arte)
(Firenze: Le Monnier).
1969 "Criticism of Style in Italy", Style 3: 17-26.
Dijk, T. A. van
1970a "Nogle aspekter af en generativ-transformationel tekst-
teori" [Some aspects of a generative transformational text
theory], Poetik 3: 155-77.
1970b "Sémantique générative et théorie des textes", Linguistics
62: 66-95.
Dolczel, Lubomir
1960 O stylu moderni ceské prózy [On the style of modern Czech
prose] (Praha: Öeskoslovenska Akademie Ved).
1967 "The Prague School and the Statistical Theory of Poetic
Language," Prague Studies in Mathematical Linguistics
2: 97-104.
Doleäel, Lubomir - Richard W. Bailey (eds.)
cl969 Statistics and Style (New York: American Elsevier Publishing
Company Inc.).
Doleäel, Lubomir - Karel Hausenblas
1961 "O sootnoSenii poetiki i stilistiki" [On the interaction of
poetics and stylistics], in: Poetics. Poetyka. Poètika 1 (Eds.:
 BIBLIOGRAPHY 157

D. Davie, I. Fonagy, R . Jakobson et alii) (Warszawa: P a n -

stwowe Wydawnictwo Naukowe; 's-Gravenhage: Mouton
& Co.), pp. 39-52.
Drijkoningen, F . F . J .
1963 "Stilistiek en het Onderzoek der Periodestijlen" [Stylistios and
t h e investigation of period styles], Forum der Letteren 4:207-14.
Dubois, J . ef alii
1970 Rhétorique générale (—Langue et langage) (Paris: Librairie
Larousse).
Einarsson, J a n
1971 Sammanträdet som talsituation ( = Lundastudier i nordisk
sprdkvetenskap C.4) (Lund: Studentlitteratur).
Ellegârd, Alvar
1962a A Statistical Method for Determining Authorship; the Junius
Letters, 1769-1771 ( = Gothenburg Studies in English 13)
(Göteborg: Elanders Boktryckeri Aktiebolag).
1962b Who Was Junius? (Stockholm etc.: A l m q v i s t - W i k s e l l ) .
Ellis, Jeffrey
1966 " O n Contextual Meaning", in: In Memory of J. E. Firth
(Ed.: C. E . Bazell et alii) (London: Longmans), pp. 79-96.
Ellis, Jeffrey - J e a n N. Ure
1968 "Language Varieties, Register" in MEETHAM - HUDSON
(1969: 251-9).
Enkvist, Nils Erik
1969 "Stylistios in Sweden and Finland; a Historical Survey",
Style 3: 27-43.
1971 " O n t h e Place of Style in Some Linguistic Theories" in
CHATMAN ( 1 9 7 1 ) .
Enkvist, Nils Erik — J o h n Spencer — Michael Gregory
1964 Linguistics and Style (= Language and Language Learning 6)
(London: Oxford University Press).
E r d m a n , David V. - E p h i m G. Vogel (eds.)
cl966 Evidence for Authorship; Essays on Problems of Attribution
(Ithaca, N. Y.: Cornell University Press).
Erlich, Viktor
1969 Russian Formalism; history — Doctrine (= Slavic Printings
and Reprintings 4) (The Hague: Mouton).
Faccani, R e m o - Umberto Eco (eds.)
cl969 I sistemi di segni e lo strutturalismo Sovietico (Milano: Valen-
tino Bompiani).
Fasold, Ralph W .
1970 "Two Models of Socially Significant Linguistic Variation",
Language 46: 551-63.
158 BIBLIOGRAPHY

Ferguson, Charles A.
1964 " D i g l o s s i a " in H y m e s (1964: 4 2 9 - 3 9 ) .
Fillmore, Charles J .
1969 " T y p e s of Lexical I n f o r m a t i o n " in K i e f e r (1969: 109-37).
Firbas, J a n
1964a " F r o m Comparative Word-Order Studies", Brno Studies in
English 4: 111-28.
1964b " O n Defining the Theme in Functional Sentence Analysis",
Travaux linguistiques de Prague 1: 267-80.
1966 "Non-Thematic Subjects in Contemporary English", Tra-
vaux linguistiques de Prague 2: 239—66.
1968 "On t h e Prosodie Features of t h e Modern English Finite
Verb as a Means of Functional Sentence Perspective", Brno
Studies in English 7: 11-48.
Firth, J . R .
1957 Papers in Linguistics, 1934-1951 (London: Oxford University
Press).
Fishman, Joshua et alii
cl966 Language Loyalty in the United States, the Maintenance and
Perpetuation of Non-English Mother Tongues by American
Ethnic and religious Groups (= Janua Linguarum, series
maior 21) (London — The H a g u e — Paris: Mouton & Co.).
Flydal, Leif
1962 " R e m a r q u e s sur certains r a p p o r t s entre le style et l'état de
langue", Norsk tidsskrift for sprogvidenskap 16: 241-58.
Fodor, J e r r y A. - Jerrold J . K a t z (eds.)
cl964 The Structure of Language; Readings in the Philosophy of
Language (Englewood Cliffs, N. J . : Prentice-Hall Inc.).
Fowler, Roger (ed.)
cl966 Essays on Style and Language (London: Routledge — Kegan
Paul).
Fowler, Roger - Peter Mercer
1969 "Criticism and t h e Language of Literature: Some Traditions
and Trends in Great Britain", Style 3: 45-72.
Freeman, Donald C. (ed.)
cl970 Linguistics and Literary Style (New York, etc.: Holt, Rinehart
a n d Winston Inc.).
Fries, Charles Carpenter
1940 American English Grammar (New York: Appleton-Century-
Crofts).
1952 The Structure of English; an Introduction to the Construction
of English Sentences (New York: Harcourt, Brace & World).
 BIBLIOGRAPHY 159

Fucks, Wilhelm
1955 Mathematische Analyse von Sprachelementen, Sprachstil und
Sprachen (= Arbeitsgemeinschaft für Forschung des Landes
Nordrhein—Westfalen 34 A) (Köln-Opladen: Westdeutscher
Verlag).
Gal'perin, I. R .
1958 OSerki po stilistike anglijskogo jazyha [Notes on t h e stylistics
of English] (= Biblioteka füologa) (Moskva: Izdatel'stvo
literatury n a inostrannyx jazykax).
1967 " J a v l j a e t s j a Ii stilistika urovnem j a z y k a ? " [Is stylistics a
level of language?] in: Problemy jazykoznanija. Doklady i soob-
s&enija sovetskix u5enyx na X Mezdunarodnom kongresse ling-
vistov (Bukarest, 28.8.-2.9. 1967) (Reds.: F. P. Filin, B. A.
Serebrermikov, et alii) (Moskva: Izdatel'stvo Nauka), pp.
198-202.
1968 An Essay in Stylistic Analysis (Moskva: Izdatel'stvo Nauka).
1971 "Some Principal Issues of Style and Stylistics as Viewed by
Russian Linguists", Style 5: 1-20.
Gal'perin, I. R . (ed.)
cl969 NauSnaja konferencija problemy lingvisti&eskoj stilisliki [A
scientific conference on problems of linguistic stylistics]
(Moskva: Ministerstvo vysäego i srednego special'nogo obra-
zovanija SSSR).
Garvin, Paul L. (ed.)
cl964 A Prague School Reader on Esthetics, Literary Structure, and
Style (Washington, D. C.: Georgetown University Press).
Genung, J o h n Franklin
1900 The Working Principles of Rhetoric (Boston etc.: Ginn and
Company).
Gibson, Walker
1966 Tough, Sweet and Stuffy; an Essay on Modern American Pro-
se Styles (Bloomington - London: Indiana University Press).
Ginzburg, R .
1969 "On Some Trends in Stylistic Research", Style 3: 73-90.
Gläser, Rosemarie
1963 "Neuwörter im politischen Englisch", Zeitschrift für Angli-
stik und Amerikanistik 11: 229-47.
1970a "Sprache und Pragmatik der englisch-amerikanischen kom-
merziellen Werbung", Zeitschrift für Anglistik und Amerika-
nistik 18: 314-23.
1970b Thesen zur Habilitationsschrift 'Linguistische Kriterien der
Stilbeschreibung' (University of Leipzig), [typescript]
1970c " E x t r a t e x t u a l e Faktoren der Stilbeschreibung", Wissen-
160 BIBLIOGRAPHY

schaftliche Zeitschrift der Pädagogischen Hochschule "Dr.

Theodor Neubauer" Erfurt/Mühlkausen, Gesellschafts- und
sprachwissenschaftliche Reihe 7.2: 89-91.
Gopnik, Irwin
1970 A Theory of Style and Richardson's Clarissa ( — De propietati-
bus litterarum, series practica 10) (The Hague— Paris: Mouton).
Gordon, I a n A.
1966 The Movement of English Prose (London: Longmans, Green
a n d Co. Ltd.).
Gorrell, R o b e r t M. (ed.)
cl967 Rhetoric: Theories for Application (Champaign, 111.: National
Council of Teachers of English).
Gougenheim, G.
1969 " L a statistique de vocabulaire", Revue de l'enseignement
supérieur 1959.1: 137-44.
Grahn, Lars
1965 " D e t moderna t i d n i n g s s p r â k e t " [Modem newspaper lan-
guage], Sprâkvàrd 4: 3 - 1 1 .
Graur, A. et alii (eds.)
cl970 Actes du 10e Congrès International des Linguistes, 4 vol. (Buca-
rest: Editions de l'Academie de la ¡République Socialiste de
Roumanie).
Gray, Bennison
1969 Style; the Problem and Its Solution ( = De proprietatiòus
litterarum, series maior 3) (The Hague: Mouton).
Greenbaum, Sidney - Randolph Quirk
1970 Elicitation Experiments in English; Linguistic Studies in Use
and Attitude (London: Longman Group Ltd.).
Gregory, Michael
1967 "Aspects of Varieties Differentiation", Journal of Linguistics
3: 177-98.
Greimas, A. J .
1966 Sémantique structurale; recherche de méthode (— Langue et
langage) (Paris: Librairie Larousse).
Guilbert, Louis
1965 La formation du vocabulaire d'aviation (1861-1891) (Paris:
Larousse).
1967 Le vocabulaire de l'astronautique ( = Publications de l'Uni-
versité de Rotten, Faculté des Lettres et Sciences Humaines)
(Paris: Larousse).
Guiraud, Pierre
1954 Les caractères statistiques du vocabulaire (Paris: Presses Uni-
versitaires de France).
 BIBLIOGRAPHY 161
1959 Problèmes et methodes de la statistique linguistique (Dordrecht:
D. Reidel Publishing Company).
1963 La stylistique ( — Que sais-je i 646) (Paris: Presses Universi-
taires de France).
Guiraud, Pierre - Paul Zumthor - A. Kibedi Varga — J . A. G. T a n s
1962 Style et littérature (La Haye: Van Goor Zonen).
Gumperz, J o h n J .
1964 "Speech Variation and t h e Study of Indian Civilization"
i n HYMES ( 1 9 6 4 : 4 1 6 - 2 8 ) .
Hallberg, Peter
1962 Snorri Sturluson och Egils Saga SkaUagrimssonar; ett försök
till sprdklig författarbestämning [Snorri Sturluson and t h e
saga of Egil Skallagrimsson; a n a t t e m p t a t linguistic deter-
mination of authorship] ( — Studia Islandica 20) (Reykja-
vik: Heimspekideild Häsköla Islands og Bôkaûtgâfa Men-
ningarsjôds).
1963 ôlafr ßördarson Hvitaskàld, Knytlinga Saga och Laxdœla Saga;
ett försök till sprdklig författarbestämning [Ölafr J)ördarson
Hvitaskàld, Knytlinga Saga, and Laxdaela Saga; a n a t t e m p t
a t linguistic determinatiorl of authorship] ( = Studia Islandi-
ca 22) (Reykjavik: Heimspekideild Hàskâla Islands og Böka-
ûtgâfa Menningarsjôds).
1968 Stilsignalement och författarskap i norrön sagalüteratur; syn-
punkter och exempel [Stylistic features and authorship in
Nordic saga literature; viewpoints and examples] ( — Acta
Universitatis Qothoburgensis, Nordica Gothoburgensia 3)-
(Stockholm, etc.: Almqvist and Wiksell).
Hallberg, Peter et alii
cl966 Liiteraturvetenskap; nya mal och metoder [The study of litera-
ture; new goals and methods] (Stockholm: N a t u r och.
Kultur).
Halliday, M . A . K .
1967-68 "Notes on Transitivity and Theme in English", Journal of
Linguistics 3: 37-81, 199-244; 4: 179-215.
H a m i l t o n , Sir George Rostrevor
1949 The TM-Tale Article; a Critical Approach to Modem Poetry
(London, etc.: Heinemann).
Harris, Zellig
1952a "Discourse Analysis", Language 28: 1-30.
1952b "Discourse Analysis: A Sample T e x t " , Language 28: 474-94.
Harweg, Roland
1968 Pronomina und Text/constitution ( = Beiheft zu Poetica
2) (München: Wilhelm Fink Verlag).
162 BIBLIOGBAPHY

Hasan, R u q a i y a
1968 Grammatical Cohesion in Spoken and Written English 1
(= Programme in Linguistics and English Teaching 7) (Lon-
don: Longmans).
Hayes, Curtis W .
1968 " A Transformational-Generative Approach t o Style: Samuel
Johnson and Edward Gibbon", Language and Style 1: 39-48.
Hendricks, W . O.
1968 " O n t h e Notion 'Beyond t h e Sentence'", Linguistics 37:
12-51.
1969 "Three Models for t h e Description of Poetry", Journal of
Linguistics 5: 1-22.
Herdan, Gustav
1966 Language as Choice and Chance (Groningen: Noordhoff).
1960 Type-Token Mathematics (= Janua Linguarum, series maior
4) (The Hague: Mouton).
1962 The Calculus of Linguistic Observations (— Janua Linguarum,
series maior 9) (The Hague: Mouton).
1964a Quantitative Linguistics (London: Butterworths).
1964b " O n Communication between Linguists", Linguistics 9: 71—6.
1966 The Advanced Theory of Language as Choice and Chance (Ber-
lin, etc.: Springer Verlag).
1969 "The Jig-Saw Puzzle of Saussurian and Quantitative Lin-
guistics", Lingua e stile 4: 69-76.
Hill, Archibald A.
1968 Introduction to Linguistic Structures; from Sound to Sentence
in English (New Y o r k : H a r c o u r t Brace & Co.).
1964 "The Locus of the Literary W o r k " , in: English Studies Today,
Third Series (Ed.: G. J . Duthie) (Edinburgh: University
Press), pp. 41-50.
1965 Essays in Literary Analysis (Austin, Tex.), [a collection of
fouiteen mimeographed or xeroxed, previously published
articles]
Hill, Archibald A. (ed.)
cl953 Report on the Fourth Annual Round Table Meeting on Lingu-
istics and Language Teaching (Washington, D. C.: Georgetown
University Press).
Hough, Graham
1969 Style and Stylistics (London: Routledge - Kegan P a u l ; New
York: Humanities Press).
Hymes, Dell (ed.)
cl964 Language in Culture and Society; a Reader in Linguistics and
Anthropology (New York: H a r p e r - Row).
 BIBLIOGRAPHY 163
Ikegami, Yoshihiko
1969 "A Linguistic Essay on Parody", Linguistics 55: 13-31.
Jacobs, Roderick A. - Peter S. Rosenbaum
cl970 Readings in English Transformational Grammar (Waltham,
Mass. — Toronto — London: Ginn Co.).
Jakobson, Roman
1960 " L i n g u i s t i c s a n d P o e t i c s " i n SEBEOK ( 1 9 6 0 : 3 5 0 - 7 7 ) .
Jelinek, J . — J . Y. Vecka — M. Teäitelova
1961 Frekvence slov, slovnich druhù a tvarù v cesMm jazyce [The
frequency of words, word combinations and idioms in Czech]
(Praha: Öeskoslovenska Akademie Vëd).
Johannet, José
1963 "Le style nominal en russe dans la langue administrative
du 18e siècle", Bernte des Études Slaves 42: 97-107.
Joos, Martin
1962 The Five Clocks ( = Publications of Indiana University
Research Center in Anthropology, Folklore, and Linguistics 22,
= International Journal of American Linguistics 28.2.)
(Bloomington, Ind.: Indiana University; The Hague:
Mouton & Co.).
Jörgensen, Nils
1971 Om makrosyntagmer i informell och formell stïl ( = Lwndastudi-
er i nordisk sprâkvetenskap C.3) (Lund: Studentlitteratur).
Josselson, H. H.
1953 The Russian Word Count and Frequency Analysis of Gramma-
tic Categories of Standard Literary Russian (Detroit: Wayne
State University Press).
Juilland, A . - E . Chang-Rodriguez
1964 Frequency Dictionary of Spanish Words ( = The Romance
Languages and Their Structures SI) (The Hague: Mouton).
Juilland, A. - P. M. H . Edwards - I. Juilland
1965 Frequency Dictionary of Rumanian Words ( = The Romance
Languages and Their Structures R l ) (The Hague: Mouton).
Kaeding, F. W.
1898 Häufigkeitswörterbuch der deutschen Sprache, festgestellt durch
einen Arbeitsausschuss der deutschen Stenographiesysteme
(Steglitz: Mittler), [reprint: (New York: The Macmillan
Company 1928)]
Kaluga, Irena
1967 The Functioning of Sentence Structure in the Stream-of-Con-
sciousness Technique of William Faulkner's 'The Sound and
the Fury'; a Study in Linguistic Stylistics ( = Prace Historycz-
164 BIBLIOGRAPHY

no-literackie 13) (Kraköw: Zeszytu Naukowe Universytetu

Jagiellonskiego CLVI).
Kaplan, Abraham
1964 The Conduct, of Inquiry; Methodology for Behavioral Science
(San Francisco: Chandler Publishing Co.).
Karlsen, Rolf
1959 Studies in the Connection of Clauses in Current English (Ber-
gen: J . W. Eides Boktrykkeri A. S.).
Karvonen, Juhani et cdii
1970 Opettajan sanastokirja [A teacher's word-book] (Jyväskylä:
K. J . Gummerus Osakeyhtiö).
Kayser, Wolfgang
1948 Das sprachliche Kunstwerk (Bern: Francke).
Kiefer, Ferenc
1967 On Emphasis and Word Order in Hungarian (= Indiana
University Publications, Uralic and Altaic Series 76) ( B l o o -
mington: Indiana University; The Hague: Mouton).
Kiefer, Ferenc (ed.)
cl969 Studies in Syntax and Semantics (= Foundations of Lan-
guage, Supplementary Series 10) ( D o r d r e c h t : D . R e i d e l
Publishing Company).
Kinnander, Bengt
1969 Sammanhangsanalys; studier i sprdkets struktur och rytm
[Context analysis; studies in the structure and rhythm of
l a n g u a g e ] ( = Skrifter utgivna av Institutionen for nordisJca
sprdkvid Uppsala Universitet 5 ) ( U p p s a l a : A l m q v i s t & W i k s e l l ) .
Kiparsky, Paul
1968 "Linguistic Universals and Linguistic Change" in Bach -
Harms (1968: 171-202).
Klein, Sheldon
1966 "Control of Style with a Generative Grammar", Language
41: 619-31.
Kloepfer, Rolf —Ursula Oomen
1970 Sprachliche Konstituenten moderner Dichtung; Entwurf einer de-
skriptiven Poetik. Rimbaud (Bad Homburg: Athenäum Verlag).
Koch, Walter A.
1966 Recurrence and a Three-Modal Approach to Poetry (— De
proprietatibus litterarum, series minor 2) (The Hague:
Mouton).
1970 "Zur Ablesbarkeit von Mustern von Textanalysen" in Graub
(1970 3: 389-93).
Kondraäov, N. A. (ed.)
cl967 Praiskij lingvistiSeskij kruiok; sbornik statej [ T h e L i n g u i s t i c
 BIBLIOGRAPHY 165
Circle of Prague; a collection of articles] (Moskva: Izdatel'stvo
Progress).
Kovtunova, I. I.
1969 Porjadok slov v russkom literaturnom jazyke XVIII —• pervoj
treti XIX v. Puti stanovlenija sovremennoj normy [Word
order in the Russian literary language of t h e 18th and t h e
first third of t h e 19th century; t h e development of t h e
modern n o r m ] (Moskva: Izdatel'stvo Nauka).
Kraus, J i f i - Josef Poläk
1967 "Text Factors and Characteristics", Prague Studies in
Mathematical Linguistics 2: 155—71.
Kreuzer, H e l m u t - R u l Gunzenhäuser (eds.)
c1967 Mathematik und Dichtung ( — Sammlung Dialog 3 ) (München:
Nymphenburger Verlagshandlung).
Kußera, Henry — W . Nelson Francis
1967 Computational Analysis of Present-Day American English
(Providence, R . I.: Brown University Press).
Labov, William
1966 The Social Stratification of English in New York City (Wash-
ington, D. C.: Center for Applied Linguistics).
1969 "Contraction, Deletion, and Inherent Variability of t h e
English Copula", Language 15: 715-62.
Labov, William - Joshua Waletzky
1967 "Narrative Analysis: Oral Versions of Personal Experience",
in: Essays on the Verbal and Visual Arts: Proceedings of the
1966 Annual Spring Meeting of the American Ethnological*
Society (Ed.: J . Holm) (Seattle, Wash.: University of Washing-
ton Press).
Lakoff, George P .
1969 "On Generative Semantics", in: Semantics; an Interdisciplinary
Reader (Ed.: D . D . Steinberg — L. A. Jakobovits). (London:
Cambridge University Press).
L a m b , Sidney M.
1966 Outline of Stratificational Grammar (Washington, D. C.:
Georgetown University Press).
Lambert, W. E.
1967 "The Use of Tu and Vous as Forms of Address in French
Canada: A Pilot Study", Journal of Verbal Learning and
Verbal Behavior 6: 614-7.
Leech, Geoffrey N.
1966 English in Advertising; a Linguistic Study of Advertising
in Great Britain (London: Longmans, Green & Co., Ltd.).
166 BIBLIOGRAPHY

1969 A Linguistic Guide to English Poetry (London: Longmans,

Green & Co, Ltd.).
Leed, Jacob (ed.)
cl966 The Computer and Literary Style (Kent, Ohio: K e n t S t a t e
University Press).
Leont'ev A. A.
1968 "Issledovanija poetiSeskoj ro5i" [The study of poetic speech],
in: T eoretiSeskie problemy sovetskogo jazykoznanija (Moskva:
Izdatel'stvo Nauka).
Lesskis, G. A.
1964 " O zavisimosti mezdu razmerom predloienija i ego struk-
t u r o j v raznyx vidax t e k s t a " [The relationship between
sentence length a n d sentence structure in various types of
t e x t ] , Voprosy jazykoznanija 3.3: 99-123.
Levin, Samuel R .
1963 "Deviation — Statistical a n d Determinate — in Poetic
Language", Lingvo, 12: 276-90.
1965a " I n t e r n a l and E x t e r n a l Deviation in Poetry", Word 21:
225-37.
1965b "Langage and Parole in American Linguistics", Founda-
tions of Language 1: 83-94.
Lodge, David
cl966 Language of Fiction; Essays in Criticism and Verbal Analysis
of the English Novel (London: Routledge & Kegan Paul; New
York: Columbia University Press).
Loman, Bengt
1970 "Social Variation in t h e Syntax of Spoken Swedish" in
B e n e d i k t s s o n (1970: 211-34).
Loman, Bengt — Nils Jorgensen
1971 Manual for analys och beskrivning av makrosyntagmer ( =
Lundaetudier i nordisk sprdkvetenskap C.l) (Lund: Student-
litteratur).
Longacre, R o b e r t E .
1970 "Sentence Structure as a Statement Calculus", Language
46: 783-815.
L o t t , Bernard
1960 Style and Linguistics; an Inaugural Lecture (Djakarta: Pener-
bit Djambatan).
Love, Glen A. - Michael P a y n e (eds.)
cl969 Contemporary Essays on Style (Glenview, 111.: Scott-Fores-
m a n & Company).
L u n t , Horace G. (ed.)
cl964 Proceedings of the Ninth International Congress of Linguists,
 BIBLIOGRAPHY 167
Cambridge, Mass., August 27-31, 1962 ( = Janua Linguarum,
series maior 12) (London - The Hague - Paris: Mouton & Co.)
Malblanc, Alfred
1961 Stylistique comparée du français et de l'allemand (Paris —
Bruxelles: Didier).
Mandelbrot, Benoit
1954 "Structure formelle des textes et communication", Word
10: 1-27.
Marcus, Solomon
1970 Poetica matematica (Bucureçti: E d i t u r a Academiei Republicii
Socialiste Romania).
Marouzeau, Jules
1946a Précis de stylistique française (Paris: Masson).
1946b Traité de stylistique latine (Paris: Les belles lettres).
Martin, Harold C. (ed.)
cl959 Style in Prose Fiction; English Institute Essays 1958 (New
York - London: Columbia University Press.).
McCawley, J a m e s D.
1968 "Concerning the Base Component of a Transformational
G r a m m a r " , Foundations of Language 4: 243-69.
Mcintosh, Angus
1965 "Saying", Review of English Literature 6: 9-20.
McKinnon, Alastair — Roger Webster
1969 " A Method of Author Identification", Computer Studies in
the Humanities and Verbal Behaviour 2: 19-23.
Mathesius, Vilém
cl947 Ceétina a obecny jazykozpyt, soubor stati [Czech and General
Linguistics; a collection of papers] (Praha: Melantrich).
Meetham, A. R . - R . A. Hudson (eds.)
cl969 Encyclopaedia of Linguistics, Information, and Control (Ox-
ford, etc.: Pergamon Press).
Meier, H .
1967 Deutsche Sprachstatistik3 (Hildesheim: Olms). [1st edition
(1964)]
Mel'Cuk, I. A. - A. K . Zolkovskij
1970 "Towards a Functioning 'Meaning-Text' Model of Language",
Linguistics 57: 10-47.
Mihailescu-Urechia, Venera
1970 "Are Novelists Free t o Choose Their Own Style ?", Linguistics
59: 37-61.
Miles, Josephine
1960 Renaissance, Eighteenth-Century, and Modern Language in
168 BIBLIOGBAPHY

English Poetry; a Tabular View (Berkeley, Cal.: University

of California Press).
1964 Eras and Modes in English Poetry (Berkeley, Cal.: University
of California Press).
1965 The Continuity of Poetic Language, 3 vol. (New York: Octagon
Books Inc.).
1967 Style and Proportion; the Language of Prose and Poetry
(Boston: Little - Brown).
Milic, Louis T.
1967 A Quantitative Approach to the Style of Jonathan Swift ( = Stu-
dies in English Literature 23) (The Hague: Mouton).
cl969 Stylists on Style; a Handbook with Selections for Analysis
(New York: Charles Seribner's Sons).
Miller, G. R . - E . B. Coleman
1967 " A Set of Thirty-Six Prose Passages Calibrated for Com-
plexity", Journal of Verbal Learning and Verbal Behavior
6: 851-4.
Mistrik, Josef
1967 "Matematiko-statistiSeskie m e t o d y v stilistike" [Mathe-
matical-statistic methods in stylistics], Voprosy jazykozna-
nija 16.3: 42-52.
1969 Frekvencia slov v slovenëine [Word frequency in Slovak]
(Bratislava: Vydavatelstvo Slovanskej Akadémie Vied).
Modéer, I v a r
1952 Utldtande om professorerna Erik Wettanders och Ture Johan-
nissons utredning rôrande fôrfattarskapet till vissa anonyma
brev [Remarks on t h e statements of professors Erik Wellander
and Ture Johannisson concerning t h e authorship of certain
anonymous letters] (Uppsala: Lundequistskabokhandeln).
Moerk, E r n s t L.
1970 "Quantitative Analysis of Writing Styles", Journal of
Linguistics 6: 223—30.
Morton, A. Q.
1965 "The Authorship of the Pauline Epistles: A Scientific Ap-
proach", Journal of the Royal Statistical Society 128: 169.
Mosteller, Frederick - David L. Wallace
1963 "Inference in an Authorship Problem", Journal of the Ame-
rican Statistical Association 58: 276-309.
1964 Inference and Disputed Authorship: The Federalist (Reading,
Mass.: Addison-Wesley Publishing Company, Inc.).
Mounin, Georges
1963 Les problèmes théoriques de la traduction (Paris: Editions
Gallimard).
 BIBLIOGRAPHY 169
Mourot, J .
1964 "Stylistique des intentions et stylistique des e f f e t s " , O A I E F
16: 71-9.
Muller, Charles
1968 Initiation d la statistique linguistique (—Langue et langage)
(Paris: Librairie Larousse).
Mündt, Marina
1969 Sturla Pördarson und die Laxdaela Saga (Bergen, etc.: Uni-
versitetsforlaget).
Naert, Pierre
1949 Stilen i Vilhelm Ekelunds ajorismer och essäer [The style in
Vilhelm Ekelund's aphorisms and essays] (Lund: Gleerup).
Nickel, Gerhard
1970 "Some Contextual Relations between Sentences in English"
in G r a u b (1970 2: 877-84).
Nida, Eugene A.
1964 Toward a Science of Translating; with Special Reference to
Principles and Procedures Involved in Bible Translating
(Leiden: Brill).
Nikolaev, P . A.
1968 Teorija stilja [A theory of style] (Moskva: Iskusstvo).
N o w o t t n y , Winifred
1962 The Language Poets Use (London: The Athlone Press).
Ohmann, Richard M.
1962 Shaw; the Style and the Man (Middletown, Conn.: Wesleyan
University Press).
1964 "Generative Grammars and t h e Concept of Literary Style",
Word 20: 423-39.
1966 " L i t e r a t u r e a s S e n t e n c e s " in C h a t m a n - L e v i n (1967: 231-8).
Osselton, N. E .
1958 Branded Words in English Dictionaries before Johnson (=
Groningen Studies in English 7) (Groningen: J . B. Wolters).
Pasini, Gian Franco
1968 " L o studio delle metafore", Lingua e stile 3: 71-89.
Petöfi, J a n o s S.
1970 Von der 'Explikation des Begriffes Satz' zu der 'Explikation
der Texte'; zur Frage einer generellen Texttheorie (Göteborg:
Göteborgs Universitet, Forskningsgruppen f ö r modern
svenska, Spräkdata). [mimeographed]
Pike, Kenneth L.
1967 Language in Relation to a Unified Theory of the Structure of
Human Behavior (= Janua Linguarwm, series maior 24)
(The Hague: Mouton).
170 BIBLIOGRAPHY

Posner, Rebecca
1963 "The Use and Abuse of Stylistic Statistics", Archivum Lin-
guisticum 16: 111—39.
Propp, Vladimir
1958 Morphology of the Folktale, translated by Laurence Scott
(The Hague: Mouton).
Quadlbauer, Franz
1962 Die antike Theorie der genera dicendi im, lateinischen Mittel-
alter (— Sitzungsberichte der Oesterreichischen Akademie der
Wissenschaften, philosophisch-historische Klasse 241.2) (Graz
etc.: H. Böhlaus Nachfolger).
Quirk, Randolph
1968 Essays on the English Language, Medieval and Modern (Lon-
don: Longmans).
Quirk, Randolph - Jan Svartvik
1966 Investigating Linguistic Acceptability (— Janua Linguarum,
series minor 64) (The Hague: Mouton).
Reibel, David A. - Sanford A. Schane (eds.)
cl969 Modern Studies in English; Readings in Transformational
Grammar (Englewood Cliffs, N . J . : Prentice-Hall Inc.).
Riffaterre, Michael
1959 "Criteria for Style Analysis", Word 15: 154-74.
1960 "Stylistic Context", Word 16: 207-18.
1961a "Problèmes d'analyse du style littéraire", Romance Philology
14: 216-27.
1961b "Vers la définition du style", Word 17: 318-44.
1964 "The Stylistic Function" in L o t t (1964: 316-23).
Rosengren, Karl Erik
1968 Sociological Aspects of the Literary System (Stockholm: Natur
och Kultur).
Rosiello, Luigi
1965 Struttura, uso e funzioni della lingua (Firenze: Vallecchi).
Ross, John
1970 " O n D e c l a r a t i v e S e n t e n c e s " i n J a c o b s - R o s e n b a t j m (1970:
222-72).
Runquist, C. - G.
1958 Ett falskt kombinationslâs; spräkiUredningen i Helander-
mdlet [A false combination-lock; the linguistic investigation
in the Helander case] (Stockholm: Seelig).
Sajkeviö, A. Ja.
1968 "Opyt statistiSeskogo vydelenija functional'nyx stilej" [An
attempt at the statistical determination of functional styles],
Voprosy jazykoznanija 7.1 : 64-76.
 BIBLIOGRAPHY 171

S a u m j a n , S. K .
1965 StruJcturnaja lingvistika [ S t r u c t u r a l linguistics] (Moskva:
Izdatel'stvo Nauka).
d e Saussure, F e r d i n a n d
1955 Cours de linguistique générale 8 (Ed.: Charles Bally a n d Albert
Sechehaye with t h e collaboration of Albert Riedlinger)
(Paris: P a y o t ) .
Sayce, R . A .
1963 Style in French Prose (Oxford: Basil Blackwell).
1962 " T h e Definition of t h e T e r m Style", in: Actes du 3e Congrès de
l'Association Internationale de Littérature Comparée, Utrecht,
1961 (The H a g u e : Mouton), p p . 156-66.
Sebeok, T h o m a s A. (ed.)
c l 9 6 0 Style in Language (Cambridge, Mass.: The M. I . T . Press).
Sebeok, T h o m a s A. — Alfred S. H a y e s - M a y Catherine Bateson (eds.)
c l 9 6 4 Approaches to Semiotics: Cultural Anthropology, Education,
Linguistics, Psychiatry, Psychology (=Janua Linguarum,
series maior 15) (The H a g u e : Mouton).
Sedelow, Sally Y a t e s
1965-67 Stylistic Analysis; Reports on the First, Second, and Third
Years of Research ( S a n t a Monica, Cal.: Systems Development
Corporation).
Sedelow, Sally Y a t e s — W a l t e r A . Sedelow
1966 " A Preface t o Computational Stylistics" in LEEDS (1966:
1-13).
Segre, Cesare
1968 Lingua, stile e società (Roma: Feltrinelli).
Sgall, P e t r
1969 "L'ordredesmotsetlasemantique"inKiEFER(1969:231-40).
Sil'man, T . I .
1967 Problemy sintaksiSeskoj stilistiki (na materiale nemeckoj
prozy) [Problems of syntactic stylistics, on t h e basis of
G e r m a n prose] ( = Leningradskij ordena trudovogo krasnogo
znameni gosudarstvennyj pedagogiieskij institut imeni A. I .
Qercena, Uéenye zapiski 315) (Leningrad: ProsveSëenie).
Sinclair, J o h n McH.
1968 " A Technique of Stylistic Description", Language and Style
1: 215-42.
Slama-Cazacu, T a t i a n a
1961 Langage et contexte. Le problème du langage dans la conception
de l'expression et de l'interprétation par des organisations
conteoctuelles (=Janua Linguarum, series maior 6) (The
H a g u e : Mouton).
172 BIBLIOGRAPHY

Somers, H . H .
1959 Analyse mathématique du langage (Louvain: Nauwelaerts).
1967 Analyse statistique du style (Louvain - Paris: Nauwelaerts).
Spevack, Marvin
1968-70 A Complete and Systematic Concordance to the Works of
Shakespeare, 6 vol. (Hildesheim: Georg Olms).
Spitzer, Leo
1948 Linguistics and Literary History; Essays in Stylistics (Prince-
ton, N . J . : Princeton University Press).
1961 Stilstudien2, 2 vol. (München: Hueber). [1: Sprachstile ; 2:
Stilsprachen]
Statsrddsberedningen (G. E . , B. M.)
1967 Sprakel i lagar och andra författningar [The language of laws-
and other statutes] (Stockholm: Svenska Reproduktions AB).
Stejnfel'dt, E . A.
1963 Öastotnyj slovar' sovremennogo russkogo literaturnogo jazyka,
[A frequency dictionary of modern literary Russian] (Tal-
linn: Naucno-issledovatel'skij institut pedagogiki Éstonskoj
SSR).
Stempel, Wolf-Dieter (ed.)
1971 Beiträge zur Textlinguistik ( = Internationale Bibliothek für
allgemeine Linguistik 1) (München: Wilhelm Fink Verlag).
Stone, P . J . et alii
1966 The General Inquirer; a Computer Approach to Content Analy-
sis (Cambridge, Mass. — London: The M. I. T. Press).
Straumann, Heinrich
1935 Newspaper Headlines; a Study in Linguistic Method (Lon-
don: G. Allen & Unwin Ltd.).
Striedter, J u r i j , (ed.)
cl969 Texte der russischen Formalisten 1 (München: Wilhelm Fink
Verlag).
Svartvik, J a n
1968 The Evans Statements; a Gase for Forensic Linguistics ( =
Gothenburg Studies in English 20) (Stockholm: Almquist-
— Wiksell).
Swieczkowski, Valerian
1961 " O n the Margin of Syntax and Style", in: Poetics. Poetyka.
Poitika 1 (Eds.: D. Davie, I. Fonagy, R . Jakobson et alii}
(Warszawa: Panstwowe Wydawnictwo Naukowe; 's-Gra-
venhage: Mouton), pp. 463-9.
Tatilon, Claude
1970 "Les grandes options de la stylistique littéraire", Le Français
dans le monde 71: 10-6.
 BIBLIOGRAPHY 173

Teleman, Ulf - Anne Marie Wieseigren

1970 ABC i stüistik [ A B C of stylistics] (Lund: Gleerup).
Tenow, Nore
1963 De anonyma breven i Helandermàlet [The anonymous letters
in the Helander case] (Stockholm: Forum).
Terracini, Benventuno
1968 Analisi stilistici; teoria, storia, problemi (Milano: Feltrinelli).
Thesleff, Holger
1967 Studies in the Styles of Plato ( = Acta Phüosophica Fennica
20) (Helsinki: Societas Philosophica Fennica).
Thorndike, E . L . — I. Lorge
1944 The Teacher's Word Book of 30,000 Words (New Y o r k : Co-
lumbia University Teachers College B u r e a u of Publications).
Thorne, J . P .
1965 "Stylistics and Generative G r a m m a r s " , Journal of Linguistics
1: 49-69.
1969 "Poetry, Stylistics, and Imaginary G r a m m a r s " , Journal of
Linguistics 5: 147-50.
1970 "Generative Grammar and Stylistic Analysis", in: New Hori-
zons in Linguistics ( E d . : J o h n Lyons) (Hannondsworth,
Middlesex: Penguin B o o k s Ltd.), p p . 185-97.
Todd, M . J .
1969 "Newspaper Style; a Practical Investigation", English
Language Teaching 23: 138-41.
Todorov, Tzvetan
c l 9 6 5 Theorie de la litterature; testes des formalistes russes (Paris:
Editions Seuil).
Traugott, Elizabeth Closs
1965 "Diachronic S y n t a x and Generative G r a m m a r " , Language
41: 402-15.
Uitti, K a r l D .
1969 Linguistics and Literary Theory (Englewood Cliffs, N . J . :
Prentice-Hall).
Ullmann, Stephen
1964a Language and Style (Oxford: Basil Blackwell).
1964b Style in the French Novel (London : Cambridge University Press).
Ure, J e a n
1968 "Practical Registers", English Language Teaching 22: 107-14.
Vachek, J o s e f (ed.)
c l 9 6 4 A Prague School Reader in Linguistics (Bloomington, I n d . :
Indiana University Press).
c l 9 6 6 Dictionnaire de linguistique de l'Ecole de Pragtie (Utrecht:
A m vers).
174 BIBLIOGRAPHY

Vander Beke, G. E .
1929 French Word Book (New York: The Macmillan Co.).
Vinay, J e a n - P a u l - Jean-Louis Darbelnet
1964 Stylistique comparée du français et de l'anglais (Paris —
Bruxelles: Didier).
Vinogradov, V. V.
1961 Problema avtorstva i teorija stilej [The problem of authorship
and t h e theory of style] (Moskva: Izdatel'stvo Akademii
N a u k £SSR).
1963a Sjuzet i stiV [Subject and style] (Moskva: Izdatel'stvo Aka-
demii N a u k SSSR).
1963b Stilistika; teorija poetiSeskoj re&i; poètika [Stylisties; theory
of poetic (Izdatel'stvo Akademii N a u k SSSR).
Vinogradov, V. V. - V. G. Kostomarov
1967 "Teorija sovetskogo jazykoznanija i praktika obuienija
russkomu jazyku inostrancev" [The theory of Soviet lin-
guistics and the practice of t h e teaching of Russian t o foreig-
ners], Voprosy jazykoznanija 17.2: 3-17.
Vomperskij, V. P . (ed.)
cl966 Voprosy stüistiki, sbornik statej k 70-letiju so dnja roidenija
[. . .] K. I. Bylinskogo [Problems of stylisties; a collection of
articles on t h e occasion of t h e 70th birthday of K. I. Bylin-
skij] (Moskva: Izdatel'stvo Moskovskogo Universiteta).
Vossler, K a r l
1913 Frankreichs Kultur im Spiegel seiner Sprachentwicklung
(Heidelberg: Carl Winter).
1923 Gesammelte Aufsätze zur Sprachphilosophie (München: M.
Hueber).
Wängler, H . H .
1963 Bangwörterbuch hochdeutscher Umgangssprache (Marburg:
Elwert).
Warburg, J e r e m y
1959 "Some Aspects of Style", in: The Teaching of English ( =
Studies in Communications 3) (Ed.: Randolph Quirk —A. H .
Smith) (London: Seeker - Warburg), pp. 36-59.
1965 "Idiosyncratic Style", Review of English Literature 6: 56-65.
Wartburg, Walther von
1969 Problems and Methods in Linguistics, revised edition with
t h e collaboration of Stephen Ullmann, translated b y Joyce
M. H . Reid (Oxford: Basil Blackwell).
Weinrich, H a r a l d
1964 Tempus; besprochene und erzählte Welt (Stuttgart: Kohl-
hammer) .
 BIBLIOGRAPHY 175

1966 Linguistik der Lüge (Heidelberg: L. Schneider).

Weisgerber, Leo
1953-64 Von der Kräften der deutschen Sprache, 2 vol. (Düsseldorf:
Schwann). [1: Die inhaltsbezogene Grammatik; 2: Die sprach-
liche Erschliessung der Welt~\
Wellek, René - Austin Warren
1949 Theory of Literature (New York: Harcourt, Brace & Co.).
Wexler, Peter J .
1955 La formation du vocabulaire des chemins de fer en France
(1778-1842) (Genève: Droz).
Wiio, Osmo A.
1968 Readability, Comprehension and Readership ( — Acta Uni-
versitatis Tamperensis 22) (Tampere: Tampereen Yliopisto).
Williams, C. E.
1970 Style and Vocabulary. Numerical Studies (London: Griffin).
Wimsatt, W. K., Jr.
1941 The Prose Style of Samuel Johnson (New Haven: Yale Uni-
versity Press).
1954 The Verbal Icon (Lexington, Kent.: University of Kentucky
Press).
Winburne, John Newton
1964 "Sentence Sequence in Discourse" in L t j n t (1964: 1094-9).
Winter, Werner
1961 "Relative Häufigkeit syntaktischer Erscheinungen als Mittel
zur Abgrenzung von Stilarten", Phonetica 7: 193-216.
1964 " S t y l e s as Dialects" in L u n t (1964: 324-30).
Yngve, Victor H.
1960 "A Model and a Hypothesis for Language Structure", Pro-
ceedings of the American Philosophical Society 104: 444-66.
Yule, G. Udny
1938 "On Sentence-Length as a Statistical Characteristic of Style
in Prose", Biometrika 30: 363-90.
1944 The Statistical Study of Literary Vocabulary (Cambridge:
Cambridge University Press).
Ziff, Paul
1964 "On Understanding 'Understanding Utterances' " in Fodor -
K a t z (1964: 390-9).
Zimmermann, Heinz
1965 Zu einer Typologie des spontanen Gesprächs; syntaktische
Studien zur baseldeutschen Umgangssprache ( — Basler Studien
zur deutschen Sprache und Literatur 30) (Bern: Francke).
176 BIBLIOGRAPHY

Zipf, George Kingsley

1949 Human Behavior and the Principle of Least Effort (Cambridge,
Mass.: Addison-Wesley).
1965 The Psycho-Biology of Language; an Introduction to Dynamic
Philology2 (Cambridge, Mass.: The M. I. T. Press).
 11

INDEX TO REFERENCES

Adamec, 110, 115 Chomsky, 42-43, 45, 71, 80,

Admoni, 90 84, 101, 103-109, 123
Allen, 140, 141, 142 Christiansen, 30
Alonso, 34 Coleman, 90
Aristotle, 53 Coseriu, 39, 42
Austin, 76 Croce, 28, 29
Crossland, 84
Bach, 71 Crystal, 49, 60-61, 82
Bailey, 133
Barthes, 32, 34 Dahl, 115
BeSka, 142 DaneS, 40, 115, 117, 120-121
Bense, 111 Danks, 108
Bessinger, 141 Davy, 49, 60 — 61
Bickerton, 85 Denison, 18
Birdwhistell, 55 De Morgan, 129, 130
Bloch, 73 De Vito, 14
Bloomfield, 73 Dolezel, 38, 39, 133
Boost, 87
Bowles, 141 Edwards, 142
Bowman, 99 Eichenbaum, 30
Brewster, 88 Einarsson, 100
Brooke-Rose, 85 Elleg&rd, 24, 137-138
Buttke, 110 Ellis, 19
Enkvist, 14, 24, 33, 34, 58, 59,
Caplan, 53 96
Carvell, 48, 62 Erdman, 36
Cassirer, 33 Erlich, 30
Catford, 87
Chang-Rodriguez, 142 Ferguson, 18
Chapman, 108 Fillmore, 71
Chatman, 32 Firbas, 110, 115
Cherry, 131 Fishman, 18
Chisholm. 106 Firth, 81
178 I N D E X TO REFERENCES

Fodor, 71, 105 Karlsen, 111

Francis, 84, 142 Karvonen, 143
Fries, 72 K a t z , 71, 105
Fucks, 131, 134 Kiefer, 71
Kinnander, 111
Genung, 88 Kiparsky, 108
Gibson, 94, 95, 96, 97 Klein, 44
Gläser, 84 Kloepfer, 111
Gougenheim, 143 Koch, 111, 117
Grahn, 95-96 Kostomarov, 61
Gray, 11-13 K o v t u n o v a , 110
Greenbaum, 82, 101 Kraus, 48, 62
Gregory, 14, 19, 24, ¡ 5 8 - 5 9 , 96 Kuöera, 84, 142
Guilbert, 84
Guiraud, 132, 133, 134, 140 Labov, 19, 20, 44-46, 74, 108
Gumperz, 18 Lakoff, 71
L a m b , 72
Hallberg, 133, 137 L a m b e r t , 20
Halliday, 49, 71, 81 — 82, 115 Leech, 84
Hamilton, 124 Leed, 138, 141
H a r m s , 71 Leont'ev, 30
Harris, 73 Lesskis, 90
Harweg, 111 Levin, 43, 104
H a s a n , 82, 111 Loman, 19, 99
Hayes, 80 Longacre, 122
Heald, 129 Lorge, 143
Hendricks, 104, 111
Herdan, 40-42, 131, 140 Mandelbrot, 131
Hill, 41-42, 73, 110 Mathesius, 115
Hjelmslev, 71 McCawley, 71
McKinnon, 137
Ikegami, 89 Meyer, 143
Mel'cuk, 111
Jacobs, 71 Mendenhall, 130
Jakobson, 30, 57, 58 Miles, 132
Jelinek, 142 Milic, 92, 95, 96, 122
Joos, 22, 73, 101 Miller, 90
Jörgensen, 99, 100 Mistrik, 140, 142
Josselson, 142 Modeer, 139
Juilland, A., 142 Moerk, 139
Juilland, I., 142 Morton, 137
Kaeding, 143 Mosteller, 24, 138
Kaplan, 13 Mounin, 87
 I N D E X TO R E F E R E N C E S 179

Muller, 133 Spitzer, 12, 28, 29

Mundt, 137 Steinfeldt, 142
Stempel, 111
Naert, 37 Stone, 90, 143
Nickel, 111 Striedter, 30
Nida, 86 Svartvik, 48, 62, 82, 100, 139
Swieczkowski, 135
Ohmann, 79-80, 93, 95, 96, 97,
108, 110 Tenow, 139
Oomen, 111 TeSitelova, 142
Osselton, 84 Thavenius, 141
Thesleff, 63
Pasini, 85 Thorndike, 143
Pet6fi, 111 Thorne, 104
Pike, 73 Todorov, 30, 32, 34
Polak, 48, 62 Traugott, 108
Pool, 90 Tynjanov, 30
Posner, 128
Propp, 31, 32, 34 Ure, 19

Quadlbauer, 54 Vachek, 39
Quirk, 82, 100, 101
Vander Beke, 133, 136, 143
Reibel, 71 Van Dijk, 111
Richards, 29 Vinogradov, 30, 31, 38, 61
Riffaterre, 19, 24, 34, 54 Vogel, 136
Rosenbaum, 71 Vossler, 28, 29
Rosengren, 90
Rosiello, 40, 42 Wallace, 24, 138
Ross, 71, 76-78 Webster, 137
Runquist, 139 Weinrich, 88, 124
Wexler, 84
SajkeviS, 62-63 Wiio, 90
Saumjan, 72 Williams, 129
Saussure, 36-42, 43 Wimsatt, 12
Sayce, 14 Winburne, 135
Schane, 71 Winter, 20, 94-95, 96
Sebeok, 55
Sedelow, 143 Yngve, 88
Sgall, 111 Yule, 131
Sinclair, 124
Slama-Cazacu, 54 Zipf, 130-131
Spencer, 14, 24, 58-59, 96 Zirmunskij, 30
Spevack, 142 Zolkovskij, 111