A research tradition in lexicology--a
branch of science which can be broadly described as the linguistic study
of words--has developed at the Department of Swedish Language in Göteborg.
In the research to date, the stylistic dimension has had a subordinate
role.
Information about stylistics is scanty and far from systematic in our dictionaries.
None of the larger, up-to-date Swedish dictionaries or thesauruses provides
a systematic description of either style level or the stylistic value of
words.
This lack stems from the neglect of stylistics in linguistic science. Although
stylistics has been an important part of the curriculum for Nordic language
studies, the subject of stylistics has always had a status subordinate
to that of other areas of linguistic research; those who have dedicated
themselves to research in stylistics have been at a disadvantage in the
competition for professorships.
Stylistics research has by preference worked with texts. Most major investigations
of style (even within the field of Nordic languages) have used fictional
texts as material; only a few studies have used words and phrases as a
point of departure. The relationship between the parts that make up a text
(the words) and the text as a whole is extraordinarily complex. The question
of how words influence each other in texts has often been left unresolved
and unexplained; the importance of context is invoked without making the
important distinction between the meaning and the function of a word--a
prerequisite for determining which variables have an influence on the whole.
We still have only a limited understanding of the mechanism which makes
single words influence the whole, and the whole influence the content of
single words. In my opinion, this question cannot be resolved without investigating
the lexical stylistic value of individual words.
Do words have a lexical, i.e. inherent, stylistic value? Researchers have
taken diametrically opposed views on the question of whether words have
a semantic and stylistic value of their own or if they obtain their meaning
and value from and through their context. In my dissertation, Deskriptiv
stilistik ('Descriptive Stylistics) (1970), I quote, among others,
Bengt Hesselman, who categorically states: "Words and sounds have
no existence outside the phrase and do not have meaning in themselves"
(Huvudlinjer, p. 18). In principle, today's semanticists seem to
follow this reasoning: Jean Aitchison gives an insightful characterisation
of the situation in Words in the Mind (1987). Nevertheless, I have
my doubts about Hesselman's use of the word "existence" in the
quotation above. While I thoroughly agree with the hermeneutic paradox
which states that the whole is composed of parts which obtain their meaning
from the whole, language still has definable units in the form of words
which can be organized in dictionaries--in all their incompleteness.
The ideal dictionary would list not only all of the prototypical potential
meanings of each word but also the ways in which a the meaning and function
of a word may be influenced by its context. It occurs to me that one possible
feature would be the word's potency for meaning or, roughly speaking,
the extent to which a word can be predicted to influence or be influenced
by its context.
I believe that it is possible to establish general rules for the meaning
potential, and that one of the decisive variables for this is the stylistic
value of a word. A word with strong connotative associations in the stylistic
dimension should influence context rather than be influenced by it. Needless
to say, this hypothesis cannot be put to the test until such lexical connotations
have been established. (1)
Stylistics and semantics are closely interwoven. In my attempt to define
the elusive concept of style in the report mentioned above, the full significance
of Stephen Ullman's remarks on this subject was brought home: "Every
major problem of semantics has stylistic implications, and in some cases,
as for example in the study of emotive overtones, the two approaches are
inextricably intertwined" (Semantics, 1967, p 9).
The "emotive overtones" constitute, in my opinion, one of the
two dimensions which, together, determine a word's stylistic value or style
value. This feature, which will be examined below, is controversial and
less self-evident than the categorization of words according to style level.
The style level of a word is determined by the word's contextual environment
and subject; information about level of diction is the most usual, sometimes
the only, stylistic information in Swedish dictionaries, which describe
style level partly in terms of specified subject categories or technical
language (technology, medicine, poetry, law, etc.) and partly in terms
of degree of formality in spoken language or in written language. One reason
that this information is given in dictionaries is probably that the grounds
for this information is relatively easily established, whereas the line
between the stylistic value and the semantic dimensions of a word is more
difficult to draw. (More on this below). The notation of style level in
dictionaries seems to me to indicate that style level is deemed lexicalised:
a word which appears frequently in a particular context, linguistic or
otherwise, is characterised by this context in such a way that the style
level becomes closely bound to the meaning of the word, regardless of whether
the word is used in an unusual context or not -- the style level is, in
other words, lexicalised. Examples are hardly needed; neither is it controversial
to allot the words of the language a certain level of style.
In my attempt to define style I found that stylistics and semantics clearly
have much in common, but that semantics could hardly offer an ideal foundation
for a theory of stylistics; linguists have frequently doubted their ability
to develop a consistent theory of semantics. Chomsky once maintained that
"modern semantics gives us no way of determining whether synonymy
holds between oculist and eye-doctor, or between oculist
and horse" and even writers of major works in this field tend
to include introductory comments on these theoretical difficulties, generally
formulated in strong metaphors--as when Miller & Johnson-Laird in Perception
and Language (1976) characterize the relationship between words and
concepts as "a morass of complexity and ignorance" (p 212).
Explanations are many but they will not be elaborated upon here. One possible
explanation is that signification itself differs between different types
of words; the difference between content words and function words is the
most obvious example. However, it may also be possible to distinguish between
different types of signification in a single content word: in General
Semantics (1970), D. Lewis observes that "meanings may turn out
to be complicated, infinite entities built up out of elements belonging
to various ontological categories" (p 19).
The complexity and diversity of semantic research support Lewis' observation.
To mention only a few of the theoretical approaches, we find "traditional"
semantics (represented by Gustav Stern and Stephen Ullman), semiotic, i.e.
communicatively directed semantics according to Pierce, structural semantics
according to Greimas or applied to field theory according to Coseriu or
Nida, transformationally conditioned component analysis according to Katz
and Fodor, generative semantics of Fillmore and MacCawley or the pragmatic
communication and speech-act analysis by Grice and Searle, to psycholinguistic
theory (as exemplified by Jean Aitchison among many others) to pragmatic-logical
phrase semantics formulated by theorists such as Bar-Hillel, Montague and
Lewis.
The differences in points of departure, angles of approach and objectives
among these schools of thought appear to confirm Lewis' argument. In fact,
it seems as though the differences between word- and phrase-semantics as
well as the internal differences in theory and models particular to each
school of thought are so important that no single analytical theory of
semantics could account for all the aspects relevant to our research. Analysis
of components may be used primarily for concrete nouns, the case theory
is strongly linked to verbs, truth value analysis theory and other logical
systems of analysis are applied to phrases, the Osgood differential (see
below) is linked to socalled non-cognitive (or, perhaps better, non-referential)
factors. This may be one of many possible explanations for the failure
to construct a comprehensive, homogenic, universal theory of semantics.
In the last decades, however, developments in cognitive and semantic-syntactic
theory have been substantial. A combination of analysis of components of
meaning and valences according to Fillmore was integrated in a model which
yielded tangible results in the project, "The Swedes and their Words."
The objective of the study was to determine what makes a word easy or difficult.
In this analysis, however, I did not examine non-referential factors -
which I should, perhaps, have done!
All semanticists seem to agree--in spite of other differences of opinion--that
there is a clear line of demarcation between cognitive and non-cognitive
factors of signification. The "connotative," "affective"
and "emotive" factors in the meaning of words remain, however,
inadequately researched; it seems as if the majority of semanticists would
rather do without them. Just as only a few decades ago linguists shunned
semantics as " a messy, largely unstructured intellectual no-man's
land on the fringes of linguistics" (Geoffrey Leech, Semantics,
1974 p x), the same may be said today about those elements of meaning which,
for lack of a better term, are here referred to as non-referential. (The
terminological diversity in labeling of affective, connotative and associative
meaning is already proof that the subject has not been satisfactorily researched
and specified). Although Leech, in the aforementioned work, attempts to
remedy this imbalance between cognitive and non-referential meaning (p
xi), he devotes relatively little attention to the latter. Non-referential
phenomena are discussed in fewer than ten pages of the introductory chapters
of this 400-page work and, though "affective" meaning is treated
in other contexts, the focus is never explicitly on stylistic concerns.
The reluctance to make non-referential elements of signification the subject
of thorough investigation is also evident in John Lyon's major work, Semantics
(1977), a fact which is all the more regrettable since he is considered
one of the late twentieth century authorities on the subject. As early
as in his thesis, Structural Semantics: Analysis of Part of the Vocabulary
of Plato (1963), Lyon expresses the hope that, in consequence of the
method of analysis he applies, "we are no longer tempted to invent
rather nebulous 'emotive' differences of meaning that can be given no operational
significance" (p 77). It is therefore not really surprising that,
in Semantics, Lyons introduces his discussion of style by referring
to Martin Joos; in Joos' terminology, style is reduced to style level,
which is defined according to degrees of formality. Lyons refers (in regard
to Joos) to five levels of formality: frozen, formal, consultative, casual,
and intimate (Semantics, p 580). In this context it is remarkable
that in "Linguistics and Poetics," Roman Jakobson's renowned
closing statement for the 1958 conference on stylistics in Bloomington,
Indiana, Jakobson characterized Joos' "emphatic demands for a 'purging'
of the emotive elements from linguistic science" as a radical experiment
in reduction--in reductio ad absurdum" (sic).
According to Joos' and Lyon's reductive definition of style, stylistic
variation becomes linguistic variation dependent upon situations (or context).
With this perspective it is understandable that stylistic variation falls
into the category of socio-linguistics. However, this is only one of a
number of possible categorizations and it primarily concerns the stylistic
level, not the stylistic value, of words. In my opinion,
moreover, stylistic value is the subject most urgently demanding research;
stylistic value--and its relation to style level--are at the center of
this project.
The research which serves as a foundation for the continued work on the
non-referential properties of words is The Measurement of Meaning
(1957),by Charles Osgood et al., which was introduced in Sweden in the
sixties. Using factor analysis, Osgood could discern three semantic aspects
of meaning: evaluation, potency and activity. In my work
(from my 1970 thesis onwards), I have considered these variables to constitute
a word's stylistic value. These factors undoubtedly correspond to properties
which semantics terms "affective," etc. (The fact that non-referential
factors are precisely those measured by the semantic differential is the
most common objection to their label).
Since Lyons' work does not broach what I would call the core of stylistics,
it is in keeping that Osgood is not mentioned in Semantics. Leech,
on the other hand, does present The Measurement of Meaning by Osgood
et al., but in his Semantics, he criticizes their approach and ends
his explication of the semantic differential in the chapter, "Associative
Meaning: A Summary Term," with a statement which probably contains
an explanation of the reluctance of many semanticists to deal with the
non-referential factors of meaning: "Whereas conceptual meaning is
substantially part of the 'common system' of language shared by members
of a speech community, associative meaning is less stable, and varies with
the individual's experience" (p 22).
One objection to the argument above is that, just as prototypical meanings
on the "cognitive" side of signification can be established (cf.
Aitchison, Words in the Mind, Chapter 5), so can prototypical style
values for the words of a language. Indeed, this appears to be Leech's
intention in presenting a graphic schema of the proportions of "conceptual"
(unmarked white field) and "affective" meaning (hatched field)
of three words, assuming "a fairly typical use of the cited words."
(These figures are reproduced at the end of this paper.)
FIG 1
Leech, in other words, considers that words are loaded with prototypical
value which, in my terminology, constitutes part of their prototypical
stylistic value. It should also be noted that, in this chapter, Leech discusses
"use"--the word in operation. When "fascist" is used
as invective, its referential meaning is reduced to a very small, though
by no means negligible, part of the whole. In the functional terminology
of Bühler, the symbol function, i.e., the referential/conceptual meaning,
is reduced once the word is used as invective and, to a corresponding degree,
the symptom and signal functions become dominant: the word becomes a manifestation
of censure as much as a direct signal of this same censure. It occurs to
me that the degree of referentiality could be tested by denial: the more
meaningful a denial is, the more of its referential meaning will remain.
It hardly makes sense to reply to someone who has called you a Jew, meaning
you are an usurer, "I am no Jew," while it is possible and meaningful
to deny being a fascist.
Both words may have exclusive or almost exclusive cognitive and referential
meanings, for instance, in historical description: the proportions of referential
versus non-referential meaning obviously depend upon the function
of the expression. Bühler expressly calls his representation an "organon
model" (Sprachpsychologie, Chapter 1, § 2). There
appears to be a fascinating resemblance between the non-referential factors
of words in a functional sense and presuppositions since negation does
not affect the presupposition. This makes them, like the affective meaning
described by Leech, "covert, implicit and potentially insidous,"
rendering them difficult to handle in debate (Leech p 50). The relationship
of presuppositions to stylistics is, in all probability, a very rewarding
field of inquiry.
Bühler's theory is a theory about language in operation. As already
noted, it is important in a preliminary stage to distinguish between the
meaning of an isolated word (which may turn out to be pure abstraction)
and the meaning of a word in operation, in other words, in a text. Evading
this distincton gives rise to the difficulties Rolf Hedquist meets in his
discussion of the Swedish word risk in his doctoral thesis, Emotivt
språk (1977, p 34ff). Hedquist proposes that risk only
has an "emotive meaning" in expressions where the sender, the
receiver or both are affected by the risk (p 36). This example illustrates
the necessity of keeping the functional and the lexical aspects of meaning
apart. Lexically, risk clearly contains a negative element. This
is made clear by the definition in Svensk ordbok ('Swedish Dictionary')
(p 992): "möjlighet till negativ utveckling eller negativt resultat.
. . mots. chans. . ." (possibility of a negative development
or result . . . ant. chance . . ."). The fact that riskvilligt
kapital (risk capital) with the negative definition "särsk.
i fråga om [påbörjande av] handling som kan få negativt
resultat/i vissa uttr./: . ..spec. även i ekon. sammanhang: riskvilligt
kapital" (esp. regarding [start of] action which may have a negative
result/ in certain expressions/: . . . esp. in an economic context: risk
capital) confirms one of Osgood's hypotheses: the combination of the
negative connotations of the word risk in Swedish with the positive
connotations of the word villigt (willingly) produces an overall
positive effect; most Swedish speakers would understand riskvilligt
as positive). The conceptual pair, risk and chans, provide
a clear example of a set of semantic components identical except in regard
to factors of value. The Swedish words lova (promise) and hota
(threaten) have a similar relationship; these words could be analyzed as
meaning "to assure someone of something that will bring positive or
negative consequences, respectively, to the person concerned." This
does not imply that the whole of humanity will experience all promises
as positive and all threats as negative: When the allied UN forces threatened
Iraq with military intervention a year ago, Kuwait naturally understood
the threat as a promise.
In Studier över ordförståelse (Studies in Word
Comprehension), mentioned abvove, different ways of graphically illustrating
the relationship between cognitive and non-referential factors in signification
are discussed. The traditional image of meaning may be represented as a
core meaning of a word surrounded by "emotive" components. Outside
of these are "sub-meanings" and "associations" even
more loosely connected to the word. (The quotation marks imply that these
terms are seldom distinctly defined). This view of meaning is illustrated
below.
FIG 2
In my opinion, on the other hand, the stylistic variables are lexicalized
in language. The positive value of härlig, chans and lova
('wonderful, chance, and promise') and the negative value of risk, hota
and sadist ('risk, threaten, and sadist') are not individually or
arbitrarily ascribed, regardless of whether these factors are to be assigned
to the semantic or the stylistic dimension of the words. One who does not
grasp these value factors (and in the word sadist, the factor POTENCY
as well) cannot be said to fully understand these words. The Svenska Akademins
Ordbok (Dictionary of the Swedish Academy ) defines sadism as, among other
things, unnatural cruelty of a high degree ("höggradig och onaturlig
grymhet,", p 62). According to the analytical model I applied in Studier
över ordförståelse (Studies of Word Comprehension),
sadist would be analyzed as follows.
The word itself would be labeled as "agent" or "actor"
and the "semantic roles" would be marked
AGENT human being
PURPOSE (sexual) pleasure
AFFECTED OBJECT (other) human being
INSTRUMENT optional: instruments . . .
RESULT pain (in the Object) / pleasure (in the Agent)
LOCUS/DIRECTION agent-object
The stylistic dimension must now account for the word sadist as
a highly negative word. From the semantic analysis of the comprehensive
category for the word (which was suggested to be agent or actor), it is
clear that the word is active and, consequently, agency exists. Whether
there is need for a non-referential variable such as "activity"
strikes me as doubtful. In words with strong activity I consider the variable
"strength" to operate on the semantic factor which expresses
action or movement. In my opinion, the factors "strong" and "negative"
are inextricably linked to the word's meaning; therefore, its stylistic
value should be included in the word's lexical "nucleus."
FIG 3
The three circles are probably not of equal value in all respects: stylistic
values are, for instance, in all probability more dependent on individual
and social variables such as age and education than are cognitive variables.
Our research in this regard, as anticipated, shows that there are great
differences between older and younger people's appreciations of the stylistic
value of words--a phenomena which is directly linked to developments in
society.
Osgood's classification of the three fundamental factors (strength, value
and activity, or what I call stylistic value) in the semantic differential
as "associative meaning" is probably justified inasmuch as the
stylistic charge is developed through associations rather than through
concept formation. Nevertheless, I would prefer to reserve the term associations
for the private sphere--a sphere which generally cannot be "lexicalised."
No linguistic theory should be held responsible for my own idiosyncracies.
The fact that words may become loaded while the stylistic value of words
may be described in general terms provides a point of departure for a number
of plausible parametres. Leech's own account in Semantics (p 50
ff) is of interest in this context; there he notes, among other things,
that particularly political designations seem to result in such strong
charges that the cognitive part of the meaning may disappear completely.
Figure 3 above is probably misleading in that it may give the impression
that cognitive and non-referential elements of meaning, on the one hand,
and different kinds of non-referential elements, on the other, can be clearly
separable and discreet phenomena. Adding to the difficulty of satisfactorily
resolving the problems in this whole area is the fact that the phenomena
as we describe as emotive meaning, sub-meaning, associative meaning and
perhaps even cognitive meaning are complexly interdependent. It might therefore
be more accurate to illustrate the lexical meaning using the following
figure:
FIG 4
The fact that we are approaching fundamental epistemological issues in
our work with these problems might explain the reluctance of certain linguists
to concern themselves with stylistic variables; in his review of Measurement
of Meaning, Uriel Weinreich goes so far as to call the factors measured
by Osgood et al. "extra-linguistic" (Semantic Differential
Technique: A Source Book. Snider & Osgood 1969, p 124). Although
the concepts of emotive and associative factors do bring us closer to issues
which are also handled in psychology, this is no reason to evade a discussion
of these issues: competence from different disciplines must be brought
to bear on them. The considerable number of linguists who feel that, in
examining emotive or affective meaning, they distance themselves from their
own domain only to end up in what might, for lack of a better word, be
called psychology, only lends justification to the idea that the part of
the meaning of a word which is non-referential is of a different character
than the referential/cognitive.
Could it be that the aspect of signification referred to as "associative"
or "connotative" or "non-referential" is more direct
than the referential meaning since it is not primarily linked to conceptualizations
but to associations? Once the stylistic value appears to be tied to the
word itself to a greater degree than to the referent (even if this relationship
is extremely complex), the affective phenomenon is not--as is the referential--a
sign for anything else and is thus no longer arbitrary. The emotive factor
does not denote anything other than what it expresses, that is to say,
an attitude towards the represented or towards the partner in dialogue.
It therefore seems reasonable to designate these elements of signification
as stylistic; this is a view which is clearly in accordance with a definition
of style which, in spite of its vagueness, is, in my opinion, the most
accurate definition to have ever been given: "Style is detail of meaning
or small-scale meaning. [. . .] Where there is either no difference in
meaning at all, or else a gross difference, we do not say there is a difference
in style; where the difference in meaning is relatively subtle and is present
along with some basic similarity on the primary level, we call the difference
in meaning a difference in style" (Monroe Beardsley, Aesthetics:
Problems in the Philosophy of Criticism, 1958, p 223).
The project, Lexical Stylistics, must begin from the research tradition
founded by Osgood in Measurement of Meaning and further developed
in the major work, Semantic Differential Technique: A Sourcebook
(Snider & Osgood 1969), where the most promising (though, in general,
highly critical!) reviews of the original work are reproduced along with
Osgood's responses. One of the critical contributions is Uriel Weinreich's
"Travels through semantic space," originally published in Word,
in which Weinreich, despite many critical objections, expresses the hope
that the results of measurements based on the semantic differential will
be able to result in practical applications: "The lexicographer must
be grateful for the semantic differential, as he must appreciate studies
of word association, because of the possibilities it has opened for the
systematic description of the "affective" capabilities of words.
The day may come when the location of a word in the "affective"
space of a language will be included in descriptive dictionairies"
(Semantic Differential Technique: A Sourcebook, Snider & Osgood,
1969, p. 127).
As I have already emphasized in this overview, the "affective"
field has not been particularly well researched by linguists, nor have
they developed the semantic differential to any degree. The semantic differential
has primarily been used in behavioral research: in his Focus on Meaning,
Osgood notes that he knows of about three thousand publications of varying
scope and range which discuss or apply the technique. In this context it
is important to stress that the semantic differential measures attitudes
towards concepts (inasmuch as we are able to differentiate them from meaning)
and is thus not primarily an instrument for linguistic measurement. Nevertheless,
I hold that the semantic differential may be used as a point of departure
in stylistics and, as I stated at the beginning of this paper, there is,
to my knowledge, no empirical research on the stylistic value of words
based on Swedish material.
SUMMARY
The main, initial task for a study in lexical stylistics is to establish
the fundamental stylistic values with which isolated words enter into a
text, i.e., to determine the stylistic values of words, but also the relationship
between style level and stylistic value. In a second stage, results from
research on individual words should be used as a basis for working out
parameters for evaluating the way they interact. To this end, we need to
develop a methodology which, in turn, is dependent on theory. Of particular
interest would be to investigate whether Osgood's stylistic factor, ACTIVITY,
corresponds to a factor, POTENCY, operating on a variable of activity in
the word's semantic position: Osgood's method requires further theoretical
development from a linguistic point of view. It is my hope that this development
will take place in cooperation with representatives from other departments
and disciplines. The line of demarcation between semantic and stylistic
variables is, on the whole, an important area for research, though it would
hardly be productive to erect artificial walls around these concepts. Furthermore,
a discussion of the relationship between presuppositions and stylistic
value would certainly prove fruitful.
The results of the project will establish methods for stylistic analysis
and descriptions of individual words and will also establish data that
can be used as a foundation for deliberations on the function of words
in combination, i.e. in a text. It is reasonable to expect that the results
can provide a foundation for systematic notation of style factors in dictionaries.
Such notations will undoubtedly have a great value for users with a less
developed linguistic and stylistic competence, e.g. young people in school
and students of Swedish as a second language. I view the potential for
practical application of the project results as a highly desirable outcome.
Summary of the main hypotheses of the project and their research areas.
1 The hypothesis that it is possible to isolate prototypical ("lexicalised),
non-referential elements in the semantic position of words is tested.
1.1 Osgood's theory of the factors evaluation, potency and activity constitutes
the point of departure.
1.1.1. The hypothesis that the factor of ACTIVITY may, in fact, be eliminated
and that in its place the factor of POTENCY, which operates as a functor
on an element of meaning with the implication of MOVEMENT, is tested.
1.2. The relationship between non-referential elements of meaning and presuppositions
is investigated.
2. The relationship between style level and stylistic value is clarified.
2.1. With the aid of a reference group and panel of experts, variations
in understanding of style dependent on social variables, primarily age,
is researched.
3. Considerations of the effects of word combinations and the "meaning
potential" of words based upon project results provide the foundation
for a stylistic theory of the text.
FIGURES 
1 A brief
discussion of my argument concerning meaning potential may be found in
my report, Studier över ordförståelse. Rapport från
projektet SVENSKARNA OCH DERAS ORD (Studies in Word Comprehension.
Report from the project THE SWEDES AND THEIR WORDS), 1977, p 86