Jakobovits, L.A. Rhetoric and stylistics: Some basic issues in the analysis of discourse. College Composition and Communication, 1969, 20, 314-28.


Rhetoric and Stylistics: Some Basic Issues in the
Analysis of Discourse

Leon James (formerly Leon. A. Jakobovits)



This paper addresses itself to some basic issues in the composing process whereby people create sentences, paragraphs, essays, and the like. An inquiry of this kind must be very general: a basic function of speech or writing is the communication of ideas and this process is intimately related to the cognitive abilities of man, his capacity for drawing inferences from past experience. The process of communication always presupposes an audience - real or imagined, and listeners are social-psychological entities: they interpret, they assimilate, they respond, all in terms of their cognitive abilities, their past experiences, their social histories. Thus a theory of the composing process must be at least as general as any psychological theory of the individual or any sociological theory of interacting men.

There is a central issue in all discussions relating to communication via language which must be dealt with before all else. This is the question of whether linguistic expressions are fixed vehicles that carry meaning, like a fountain pen that holds ink, or are they more like the ink itself which spreads on the paper in varying shapes under the will of the hand that holds it. Do words refer to things, do utterances denote propositions, or is it that the speaker himself, using words, refers to this or that, and the writer, composing phrases and sentences, denotes perceptions and cognitions?

There are two extreme views on this question, both having their proponents and detractors. On one account, words have their precise meaning and the problem in composition is to find that linguistic expression which best describes the thought one wishes to communicate. The other view is that linguistic expressions are merely indexical devices for conjuring up in the mind of the recipient some cognitive process that might correspond to the thought that the sender has and wishes to convey.

The consequences of adopting one or the other of these two views have important implications for the kind of theory of language performance one would want to propose. It is important therefore to resolve the issue if at all possible.


Discourse is elliptical. Consider the following examples:

(1a) Hi! How are you?
(1b) Fine, thanks! And you?
(2a) Is Mr. Jones in?
(2b) My husband is out of town.
(3a) I don't seem to have a pencil.
(3b) Why don't you use mine?

The answer in (1b) represents a contextual ellipsis; the full grammatical sentence can be reconstructed on the basis of linguistic cues in the question. In fact, computer programs can decode such elliptical utterances by a mechanical application of a few transformation rules (Holzman, 1968). In (2b), the answer cannot be reconstructed solely on the basis of the linguistic context. It is an instance of "telegraphic ellipsis": inferential reasoning of a non-linguistic sort is required to recognize that (2b) is indeed an answer to the question in (2a). Similarly, the exchange in (3a) and (3b) are instances of telegraphic ellipsis: functionally, both (3a) and (3b) are requests. While syntactically one is a declarative sentence and the other a question.

The juxtaposition of sentences in discourse implies relations between their propositions that are not overtly expressed in linguistic form. Consider:

(4) You can forget about the golf game. It's raining. How about bridge? I'll call John and Mary to see they're interested.

After the first sentence, each subsequent sentence is to be understood in terms of the logical implications of the preceding sentence: "We can't play golf because it's raining. Bridge can be played indoors despite the weather. I'll see if John and Mary want to play since it takes four people to play bridge." The structure of the sequence is demonstrated by the fact that the sentences which compose it cannot be rearranged in any order and still retain the same significance of what is being expressed.

The fact that a word can be looked up in a dictionary seems to suggest that meaning is a characteristic of words which is attached to them and which they can carry around wherever they appear in sentences. However, a more correct view is that dictionary definitions are indexical descriptions of particular words which specify certain essential features of the meaning they may have in actual sentences, but without in fact specifying their meaning. Consider the dictionary entry for "book": "Noun, 1. A number of sheets of paper bound or stitched together; especially, a printed and bound volume. 2. A literary composition or treatise of some length." Now consider three instances of the use of "book":"

(5a) This book is not the book you borrowed.
(5b) There has never been written a book on that subject.

It is clear that in (5a) the two tokens of "book" have different referents, while in (5b) "book" has no referent at all. The three tokens of "book" in (5a) and (5b) share certain common features related to the dictionary entry, but in each of the three instances their referents and their implications-their meanings-are different.

It is the indexical feature of word definitions that allows novel usage. In other words, "dictionaries are not depositories of 'true meanings' ....[They] can only record how a word has been used in the past; [they] cannot predict its use in the future" (Hughes and Duhamel, 1966, p. 192). Thus, while it is possible to define all the words in a language, it is impossible to specify all the meanings of a single word.

One can conclude on the basis of the foregoing considerations that words, phrases, utterances do not stand for things, thoughts, propositions, but rather have the potential of implying such entities provided the listener engages in inferential reasoning that is based on linguistic, non-linguistic, and experiential cues. If this conclusion is accepted, how is one to interpret stylistic characteristics that are often being attributed to discourse as features of the utterances themselves, as opposed to characteristics of the speaker or writer? In what way can a sentence be formal or informal or colloquial? How can a piece of prose be expository rather than loose, a word precise rather than suggestive? These questions deal with the nature of description of stylistic features in language. This issue will now be examined.


Consider the following two sentences:

(6a) Shut up. [meaning: "stop talking."]
(6b) I like this creation of the milliner's art appropriate for the vernal months. [meaning: "I like this spring hat."]1

In qualifying these two sentences, one might say that (6b) is pompous. What justifies the ascription of such qualities to linguistic expression? One hypothesis is that the connotation of coarseness or pompousness is associated with these particular sentences. But this does not seem correct: in making judgments of this type about linguistic expressions we appeal to the taste of "sensitive" or "linguistically sophisticated: people, thereby indicating that the ascribed quality is held to be neither idiosyncratic to the listener nor universal within the cultural speech community. Hence the association hypothesis must be rejected.

A second possibility is to ascribe equalities to linguistic expressions on the argument that people who use these expressions possess these qualitities; vis., people who say (6b) are pompous. Note, however, that the claim that (6a) is coarse is held to be true despite the fact that no information is available as to whether or not a hypothetical speaker that might have uttered it is in fact a caorse person, and the pompousness of the expression in (6b) is not mitigated by any knowledge one might have about the utterer (say, that he is actually a modest and humble person).

Sircello (1967) provides a third hypothesis which avoids the preceding difficulties. He points out that expressions like (6a) and (6b) are "semantic acts": they indicate that the utterer is suggesting, referring, implying, denoting, naming, etc. When a person utters (6a), he is in fact suggesting to the listener that the latter's talk is annoying and that he should stop talking by forcibly keeping his mouth shut as one would a window or door to keep noise out. To suggest such a thing strikes the listener as being a coarse act. To refer to a common object such as a spring hat by the circumlocutions in (6b) is a pompous act. NOte that Sircello's solution to the problem avoids the difficulty whereby one must ascribe a personality trait to the speaker on the basis of the expression he uses. In a number of situations one is forced to distinguish between what a speaker does and what the expressions he is using do: "semantic acts" which are linguistic expressions are not to be confused with "semantic acts" which speakers perform. There are circumstances under which a person may utter an expression (i.e. perform a semantic act) without being aware of the implications of the semantic act (i.e. the linguistic expression). One might excuse the act of the person by saying that "he does not mean what he says." Nevertheless, it is plain that "semantic acts," viz., certain linguistic expressions, are not "acts" in the sense that referring on the part of the speaker who utters the expression is an "act," a performance. One must conclude, therefore, that if a speaker uses a linguistic expression which indicates that he is performing a semantic act that is qualified in a particular way (e.g. coarseness), he is then held responsible for that act, whetheror not he is aware that his act is thus qualified, just as a man in a court of law is held responsible for his prior actions and cannot plead ignorance. It is true, though, that the punishment may be mitigated by special circumstances, and one might excuse a coarse act on the basis of ignorance, emotional disturbance, etc.

It is now time to summarize the argument thus far presented. Words, phrases, sentences, linguistic expressions generally, are indexical devices which a speaker uses to conjure up in the mind of the listener an idea or thiough that he has and wishes to convey. Thus, the communication of ideas makes use of an elliptical process whose success and effectiveness depends on a number of factors: the indexical feature of words whose communicative possibilities are specified by dictionary-type definitions, the inferential capacities of the listener, and the presuppositions (background expectations) which both speaker and listener hold in common about the speech act--the intentions for which a speaker is held responsible (accountable) when performing a particular speech act. In this theoretical context, the nature of stylistic description (variation) pertains to an explicit account of these factors as they have been enumerated.

Definitions of words as given in existing dictionaries represent only one of the many ways one can talk about words. Other metalinguistic possibilities have been explored: giving synonyms and antonyms, listing other words that are associated with a particular word (e.g. Russell and Jenkins, 1954), identifying clusters of words that are stored together in memory (Deese, 1965), arranging them in some relative order such as frequency (Thorndike and Lorge, 1944), similarity (Haagen, 1949), abstractness (Paivio), evaluating them on particular scales of, say, "goodness" (Osgood, Suci, and Tannenbaum, 1957), and so on - the list is very large indeed. An important point to realize in connection with such a taxonomic approach to word description is that it is an open process: the number of ways words can be described is potentially infinite.

A particularly interesting suggestion for the description of words is the notion of identifying "semantic features" that are general enough to be applicable to all words in language (e.g. Katz and Fodor, 1965). by arranging these features into a hierarchy of descending order of inclusiveness one can specify the relations between words in terms of a common conceptual framework. Roget's Thesaurus represents an early version of this approach. To date, attempts of this kind have achieved some measure of success for restricted categories of words (Goodenough, 1967; Osgood, 1968; Steinberg, 1969).

An important aspect of the indexical feature of words is highlighted when one contrasts the description of a word as given in a dictionary-type definition with the description given by semantic feature analysis or, to use the term better known to ethnolinguists. "componential analysis." Consider the dictionary entry for the word "damson": "Noun. An oval purple plum of Syrian origin." A semantic feature analysis might give the following description for the same word: "Concrete object; inanimate; edible; fruit." A componential analysis of the concept category "fuit" might describe a "damson" as follows: "Arboreal; smooth-skinned; juicy; sweet; drupaceous." None of the features or components listed are mentioned directly in the dictionary definition, although they could be discovered by successively looking up the entries for the words used in the definition of "damson" and thereafter. "Plum" yields "edible" and "fruit"; the latter yields "drupaceous," "juicy" and "product"; the latter yields "thing" which in turn leads to "inanimate," etc.). But no such mechanical search could yield information as to the meaning of "damson" in the following sentences:

(7a) He expressed his displeasure rather crudely by an accurately directed damson that hit the speaker in the face.
(7b) He used a damson to keep his desk neat.
(7c) He gave the child a damson for knowing the answer.

(Definitions: "damson": Noun. 1. When very ripe, an effective missile againt unpopular public speakers (as in 7a); 2. When attached to a square wooden base, a temporary paper weight (as in 7b); 3. a reward (as in 7c).) It is obvious that there is no limit to the number of such derived definitions for any one word.

Analogous instances of the indexical feature of words are illustrated by considering certain sets of contrasting words. Thus:

(8a) to stare versus to notice
(8b) fearful versus frightened
(9a) to nag versus to criticize
(9b) wave versus line
(10a) to toss versus to throw
(10b) to nibble versus to eat

In examples (8a) and (8b) there is a contrast between "extended" versus "limited" duration; in (9a) and (9b) there is a distinction of "iteration," and in (10) "nonchalance" marks the difference between the contrasting pairs. "Duration," "iteration," "nonchalance" are some instances of what might be called the "aspectual quality" of words (see Jakobovits, 1968). These distinctions, and others of this type, are made possible by applying inferential reasoning to the indexical possibilities of words.

An important application of this same principle is the derivation of the connotation of words. Consider these examples:

(11a) He is the life of the party.
(11b) He is the life of the company.

In each case, a different connotative aspect of "life" is salient: "activity" in (11a), "importance" and "potency" in (11b). Similarly, the figurative use of words is made possible by their indexical feature, but inferential reasoning is required to isolate the relevant comparison. Consider:

(12a) Juliet is the Sun. (Shakespeare)
(12b) She'll make someone a nice husband. (C.E. Osgood)
(12c) He is a mother to me.

By a literal interpretation of "Sun," "husband," and "mother" should be female. But when certain other features of these words are abstracted, the sentences in (12) become quite apposite: "brightness," "source of heat and life" for "Sun," "masculinity" for "husband," and "with loving care" for "mother."

Thus, as a summary of the discussion in this section, it is to be recalled that the indexical feature of words insures their use in infinitely variable ways, their meaning in any particular instance being specified by inferential reasoning that draws out salient features relevant to the context of the sentence in which they are embedded.

One further aspect of the inferential process remains to be discussed, namely, the background expectations or common understandings which speaker and listener share and without which communication via linguistic expressions is impossible.

To begin with, it is to be recognized that utterances are speech acts that vary in function. the identification of the particular function of the speech act which an utterance has is essential to its being understood in the manner intended by the speaker. Consider some examples from a class of speech acts, "mands," defined as "a type of utterance which the speaker uses when he attempts to induce some action in the listener" (Jakobovits, 1968, p. 12):

(13a) Pass the salt, please.
(13b) Take that coat off! It's hot in here.
(13c) Take that coat off! It's mine.

requests, as in (13a) are appropriate only where the speaker assumes the listener is already disposed to act in the intended manner. When giving advice, as in (13b), the speaker makes the public claim that his purpose is to promote the interest and welfare of the listener. the command in (13c) is effective only when the speaker can claim that lack of compliance may result in retributive action on his part.

A different type of utterance is the "tact" which is defined as a speech act by which the speaker intends to comment on some aspect of the world (e.g. instructing, describing, entertaining). "Autoclitics" are utterances whereby the speaker comments on other utterances in terms of importance, existential status (e.g. possibility vs. certainty vs. probability), relational status to other utterances, and the like (Jakobovits, 1968).

It is important to recognize once again that linguistic information by itself is insufficient for the unambiguous recognition of the function of utterances. As pointed out in the discussion on telegraphic ellipsis, the syntactic form of a sentence is not necessarily related to its function as a speech act: the declarative sentence in (3a) is functionally a question or request, while (3b), having the syntactic form of a question, is functionally a request.

Earlier in this paper, it was stated that a performance theory of language--an account of how people communicate via linguistic expressions --would have to be at least as complex as a psychological theory of the individual and a sociological theory of interacting men. A psychological theory is necessary to account for the inferential reasoning process that makes use of the indexical feature of words and utterances. A sociological theory is required to explicate the role relationship between speaker and listener within the sociocultural context which determines the function of speech acts and conditions the form and selection of elements in the composing process. To engage in social communication via language, the person must be a psychologist and sociologist, and the quality of the psychological and sociological theories that he has developed for himself will determine the quality and effectiveness of his communication act. the psycholinguistic and sociolinguistic enterprise, as formal disciplines are at the present time of their development at a stage far less advanced than the comparable personal subjective enterprise attained by the average individual who has developed communicative competence.


Individual men vary in reasoning ability, in the quality of their thinking, in their creative capacity, in the depth of communicative competence, in the effectiveness of their speech performance. Some people are better psychologists than others and are able to use their practical sociological knowledge more effectively. The discovery of the etiology of such differences would be an important enterprise, for it would enhance the effectiveness of educators in their attempts to impart the knowledge and improve the abilities which underlie these vital processes. The success of such an enterprise is still far off in the future. For the present time, a lesser goal must be chosen, one which ahs a more realistic probability of success but whose attainment is a prerequisite for the success of the larger and more crucial enterprise. This is the description of the dimensions of variability of discourse and the identification of evaluation criteria for the quality of discourse.

1. Levels of analyses. The analysis of discourse can proceed at various levels. At the lower, "micro" level, attention is focused on individual units of discourse such as lexemes and phrases. At some higher "macro" level, one is concerned with larger units such as the sentence, the paragraph, the essay, the play and the novel. The choice of unit determines the kind of observation and generalization that can be induced about discourse. Thus inferences about ratiocination (the analytic quality of thought), organization, inventiveness, require the examination of larger units such as the paragraph and multiples thereof. Inferences about creativity and originality are possible at all levels, although the nature of these processes may not be of the same form at the various levels. For instance, the originality of a writer in the figurative use of words and the construction of metaphoric expressions may not be matched by a parallel creativity in the organization and inventiveness of an essay. Command over the manipulation of complex syntactic constructions may not be related to the ability to choose words whose indexical features facilitate the reception of the intended thought.

The analyses of discourse must take into account the three-way relation between the sender, the message, and the receiver. In order to evaluate a piece of composition, information about all three aspects of this relationship is required: What is the goal of the writer? Is the content original, interesting, worthwhile? Has the author chosen words and syntactic constructions that allow for the derivation of his intent? Is the reader's inferential reasoning accurate, perceptive, sufficiently complex?

In traditional discussion about "style" the nature of the role of the author and that of the reader is often not explicitly recognized. Disagreements amoun literary critics about the quality of a piece of writing and the low correlation among judges of scores assigned to student essays (Braddock. Lloyd-Jones and Schoer, 1963) are clear indications of the reader's role in the evaluation of discourse. The appreciation of universally recognized literary contributions ("the great literature of the world") is a skill that has to be learned and individuals can be expected to attain different levels of achievement in this skill. The same sentence might mean one thing to one person and something quite different to another person. A rhetorical scheme effective in one context might be quite inappropriate and tedious in a different context. it has been stated that if there is one absolute rule in rhetoric it is that no rule is absolutely true.

A relativistic viewpoint to the analysis of discourse and the evaluation of style should not be allowed to obscure the fact that there are certain inherent characteristics in the indexical feature of words and utterances, in the inferential reasoning of people, and its relation to the syntactic organization of linguistic expressions. Consider each of these in turn.

An extreme instance of the inherent variability of words in terms of the possibilities for their indexical breadth is represented by the difference between function words and content words. Novel uses of functors are extemely limited, and ths restriction is determined by the nature of their definition. A less extreme example is provided by the use of conventionalized symbols and expressions: fad words, code words, technical words, proper names, acronyms, greetings, trite phrases, idioms, proverbs etc. The "difficulty" of words is not simply a function of familiarity. Thorndike and Garrettson (1968) report that high school pupils find abstract words more difficult to understand than concrete words, and root words are more difficult than compound words. These differences in difficulty were independent of word-count frequency. Words differ also in the intensity of their connotation. Some words like "snake," crime," "death" evoke intense negative affective responses; others like "stick," "February," "five" are almost neutral in connotation (Center for comparative Psycholinguistics, 1969). Some words have definitions which allow for diffuse and unspecific reference (e.g. "wonderful," "feathery," "thing," "anywhere") while others are restrictive and specific (e.g. "screwdriver," "melting point," "naturalized").

At the syntactic level, different codes exist which vary in flexibility and complexity. Bernstein (1964) distinguishes between "elaborated" and "restricted" codes. The latter is characterized by the following: short, grammatically simple sentences with little subordination of clauses; repetitive and predictable with frequent use of trite phrases, expressions, and "in" words; rigid and limited use of adjectives and adverbs; active declarative sentences predominate. The "elaborated" code is characterized by grammatical complexity and by the use of subordinate clauses with a range of conjunctions; prepositions indicating logical, temporal, and spatial relations are frequent; passive constructions with impersonal pronouns are common; there is discriminative use of a range of adjectives and adverbs. In short, the elaborated code makes use of the theoretical possibilities of expressing individuated thought while the restricted code conveys intent by relying on non-linguistic information provided by the social structure which governs the relationship between the two individuals.

The quality and complexity of inferential reasoning is the the third factor which determines the indexical possibilities of linguistic expressions. Two encyclopedias may have a very similar index, yet one may contain much richer information than the other. Individuals differ not only in their encyclopedic knowledge of the world but also in their habitual mode of realizing the indexical possibilities of expressions, viz., the extent to which they simply scan the information in their encyclopedias or review it in depth. A particular poem can be either rich or poor in meaning depending on the extent to which the reader explores the indexical possibilities it contains. On the other hand, discourse in a restricted code contains little indexical potential, in which case the reader's inferential reasoning capacities will be limited value.

In summary, then, the levels of analysis of discourse can vary from the lexemes the micro-unit to the larger compositional whole and the choice of level will determine the nature of the critical inferences one can make about the discourse. Since meaning derives from the indexical feature of linguistic expressions rather than from their absolute denotation, discussions about style and the quality of discourse must take into account the three-way relation between writer, composition, and reader (or speaker, utterance, and listener). The reader's inferential reasoning abilities set a limit on the extent to which he recognizes the writer's intent and explores the indexical possibilities of the compositional matter. But equally important is the quality of the discourse in terms of its potential for indexical possibilities at all levels of analysis.

2. Types of analyses. De Vito (1967a) identifies six areas of focus in the analysis of style: psychological: drawing inferences about the psychological state of the speaker; rhetorical: evaluating the effectiveness of the discourse upon the listener; semantic: examining the effectiveness with which words convey reference and cognitive intentions; literary: describing the poetic and literary form of a piece of writing; sociological: indentifying the social relation between speaker and listener; and linguistic: describing the lexical and syntactic variability of a passage (e.g. type frequencies).

Heinberg, Harms, and Yamada (1969), whose chief interest lies in oral speech-communication, define satisfactory style as the ability to transform sentences for the purpose of getting the listener to respond in the intended funtional way. They specify a number of such functions: (a) Question-Answere: To transform a simple declarative sentence into a question which succeeds in getting a response that clarifies the declarative sentence; (b) Confirmation-Negation: to transform a declarative sentence which receives a "Yes" or "No" answer; (c) Generalization-Specification: to transform a declarative sentence into a question which requires the listener to expand or restrict the applicability of the original declarative sentence; (d) Elaboration-Summarization: to ask a question which requires elaboration (more wordiness) or summarization (less wordiness) of the original declarative sentence; (e) Predicting: formulating a declarative sentence which anticipates questions and provides answers for them before they occur.

Chatman (1967) reviews the literature on style and categorizes the various definitions that have been proposed into four groups: (a) The noramtive sense: deals with the question of what is "good writing"; (b) Individual manner of expression; (c) Form: features of expression that belong to form rather than content; and (d) Tone of speaking: sometimes referred to as levels of style" (e.g. formal vs. colloquial). Chatman finds shortcomings with each approach taken singly and proposes a synthesis of his own in terms of a hierarchy of distinctions at five different levels: (1) a three-pronged division of "purport" into cognitive, emotive, and general; (2) a subdivision of general purport into "casual" and "non-casual" discourse; (3) variability in casual discourse is identified with "register" and the focus can be either on the message ("situational register") or on the speaker ("personal register"). Variability in non-casual discourse is identified with "style" which has various forms: literary, legal, etc.; (4) the focus of style forms can be either on the author or on the work; (5) finally, at the lowest level, and for all preceding branches, there is a bifurcation into an analysis of "semantic texture" (micro level) and "semantic structure" (macro level).

Further approaches to the analysis of style is exemplified by Hughes and Duhamel (1966). They distinguish between motive, diction, sentence movement, and schemes. The motive of style is carried through in thre modes: description, narration, and exposition. The diction of style is cast in three forms: formal, informal, and colloquial. Sentence movement can be described as balanced (syntactically symmetrical), curt (short and direct), loose (rambling), and familiar. Finally, 15 rhetorical schemes are identified (e.g. alliteration, antithesis, anaphora, inversion and so on).

The availability of large computers in recent years has sparked renewed interest in an approach to the analysis of style that makes use of frequency counts of lexical and syntactic elements and various derivative ratios based on these counts (see the review by DeVito, 1967b). This approach is sometimes referred to as "stylometrics." It is generally recognized by users of this approach that type frequencies are merely indicators of style rather than inherent features of it. By developing style indicators that correlate with inherent features that have more face validity, but are also more subjective and less reliable, the proponents of this approach hope to increase the objectivity and reliability of essay grading. For example, Slotnick and knapp (1967) developed the MOMSR computer program (Machine Oriented Measurement of Student Writing) which utilizes some 40 mechanical indicators based on Page (1967) and Bernstein (1964). In a recent application of the MOMSR program, Slotnick (personal communication) obtained a multiple correlation of .67 between teacher evaluations of style and four of the mechanical measures. The data was based on 40 essays written by college freshmen on an assigned topic (literary criticism of a character in a novel). The four measures were: number of rare words (an indicator of "vocabulary richness"), number of temporal prepositions (negative correlation), total number of sentences, and total number of quotations.

In a series of articles, De Vito (1967c, d/ 1966a, b; 1965) reports differences in type frequency characteristics between written and oral discourse produced by the same persons. Reliable differences were found for the following types (in each case, the difference is stated in favor of the written material): greater number of Friesian Class I (nouns) and Class 3 (adjectives) type words and a correspondingly smaller number of Class 2 (verbs) and Class 4 (adverbs) type words: greater number of abstract words relative to the number of finite verbs and definite articles and their nouns; a smaller number of terms of self-reference, of "pseudo-quantifiers" (much, many, a lot), of qualification (however, but, except), of allness (none, never, always, all), and of "consciousness of projection" (apparently, it seems to me); greater number of "difficult" words (low frequency and polysyllable); larger diversity (type/token ratio); and , finally a greater number of content words. None of these differences seem surprising and as De Vito points out, they are understandable in terms of the conventionalized requirements of written expository scholarly writing.

A more complex approach within the stylometric tradition was recently developed by Hunt (1965) who replaced the traditional measure of sentence length by the "T-unit" ("minimal terminable units"), which is defined as a main or independent clause together with its subordinate elements. Hunt found that the length of the T-unit in the compositions of school children gradually increases as their writing skill improves. He argues, however, that the maturing of style does not consist merely of lengthening the T-units but in the way in which the T-unit is made longer. Hunt concludes that the significant factor is the development of "sentence embedding transformations" whereby adjectival clauses are lengthened by the use of noun clauses and phrases (in place of nouns and pronouns) and modifiers embedded before and after nouns. Christensen (1968) questions this conclusion and argues instead that the more relevant indicator of a "mature style" is the extensive use (frequency and length) of free modifiers, especially in the final position, while the base clause is kept relatively short.

It is clear even from this brief and selective review that the analysis of discourse can be approached from a large number of vantage points. In fact, it would appear that discourse analysis is an open-ended enterprise, in which relevant features for study can be invented without limit in number. This is hardly surprising given that the variability of language is infinite and that new stylistic devices can be created which have never existed before. These remarks apply to the approaches developed and used thus far, all of which are descriptive. It is obvious that no taxonomic approach can give a complete description of an open-ended system. Until more powerful theoretical devices are invented, the analysis of discourse must perforce remain incomplete, descriptive, merely enumerative.

3. Evaluation Criteria. The question of how many criteria are needed to evaluate the style of discourse is not particularly significant. The answer surely depends on some arbitrary classification of elements one may wish to dream up. The more serious question is the corollary. "Needed for what purpose?" If the purpose is to distinguish between the writing of high school seniors and regular contributors to Harper's, Christensen's (1968) measure of the number and length of free modifiers in final position is quite adequate. But this same measure will also distinguish between the writings of Hemingway and James; the latter's passages will typically show a much higher count than those authored by the former. Yet it is plain that the difference in style between a Hemingway and a James is not of the same sort as the difference between a high school senior and David Halberstam.

Hughes and Duhamel pose the problem as follows: "There is no such thing as a 'good' stylistic device or a 'bad' stylistic device; there are only functional and non-functional devices. If a style satisfies and completes the writer's intention, then it is successful. If it frustrates or is contrary to his intention, then it is a failure" (1966, p. 229). This is the "autotelic" approach to the evaluation of style; it asks, first, what is the writer's intention and, second, do the components of his writing fulfill that intention.

There are several difficulties with this type of evaluation criterion. In the first place, how is one to determine the intent of the author? When confronted with the works of great literary figures, the skilled reader has no difficulty in recognizing the intent of the author and he can appreciate the masterful way in which the linguistic composition effectively promotes his purpose. But surely it is the "effectiveness" of their writing that makes their intent so clear to grasp --which means that the determination of intent is not independent of the quality of the discourse, hence the autotelic definition suffers from circularity.

Furthermore, the contribution of the reader in the indentification of intent is a difficulty to be recognized. Granted that training for reading appreciation, assuming it can be standardized (another difficulty!), would improve the reader's ability to infer intent, it still is to be expected that individual differences would remain in inferential reasoning. Thus, one critic might "see" things which another might not.

Finally, what if the discourse suggests to the reader new perceptions and new cognitions which the author himself never intended (at least consciously) or never even realized during the composing process. After all, discourse is indexical, as discussed earlier, and its cognitive possibilities may be very stimulating indeed. It is not unreasonable to suppose that new indexical possibilities are realized upon second, third, and even subsequent readings of a passage of prose or poem. At which point, then, is one to decide whether the discourse promotes or frustrates the author's intent?

It would seem that the problem of developing evaluation criteria is the problem of determining what questions the evaluator wishes to ask about the discourse. Heinberg, Harms, and Yamada (1969), concerned as they are with speech communication (see above), asks the question of whether the speaker can get the listener to respond in a predetermined way. If he can accomplish this within the restriction imposed upon the communication situation (time, number of questions asked, etc.), then his style is "satisfactory," "effective," etc. for this particular purpose. If the literary critic decides (by any means whatsoever) that the intent of a writer in a particular passage is "obviously intended to suggest an atmosphere of terror and foreboding" (Hughes and Duhamel, 1966, p. 182), then the first paragraph of Edgar Allan Poe's The fall of the House of Usher is "effective," for indeed the trained reader (and possibly the untrained as well) does "get the message." If the critic decides that the purpose of metaphors and similes is to suggest to the reader novel recombinations of perceptions, then Shakespeare's writings are replete with schemes of this sort that are supremely effective, certainly insofar as the trained reader is concerned, and possibly even for the untrained reader. A professor's lecture might be entertaining but uninstructive, instructive but boring, or both entertaining and instructive; then, the lecture is "effective," "satisfactory," "creative," "original" depending on what the critic decides the purpose of the lecture is or ought to be.

To summarize the argument thus far, the determination of evaluation criteria for the style of discourse is strictly a question of deciding, first, what its function is and, second, whether the function is fulfilled effectively under appropriate conditions. The preceding sentence italicizes "deciding" and "under appropriate conditions" because this is where the difficulties pertaining to the autotelic definition must be resolved. The "intent of the author" is but one of the many functions. Then, the determination of whether the chosen function is accomplished effectively must take into account the limiting conditions: the trained or untrained reader or listener, the social-psychological background of the receiver to whom the discourse is addressed, of the critic who uses himself as the "guinea pig" to test out his reactions, and so on.

The problem of developing evaluation criteria for assessing the style of discourse is after all not different from the problem of evaluating any human endeavor. The principles are the same whether the object is art, a teaching method, a piece of machinery, or a scientific theory. But, whereas the decision of the function of a piece of machinery easily commands common consensus among its users, such consensus is more difficult to atttain about the decision of the function of a work of art. This is because the indexical possibilities are so much richer for a work of art than for a machine.


Discourse is the product of the composing process. But what is composing? All acts of composition, in whatever mode they occur--linguistic (e.g. discourse), visual (e.g. painting), auditory (e.g. music), etc. --have this thing in common, that they are all sequencing operations, organized in time-space. There are at least two fundamentally different kinds of sequencing operation: the mechanical and the creative. Mechanical sequences are generated by an algorithm, which is a computational device that specifies the order and nature of the steps to be followed in the generation of a sequence. Thus, a computer program is an algorithm. Algebraic problems can be solved by the application of an algorithm that specifies the order of some specific manipulations to be performed. A music score is an algorithm for generating a tune. Electro-magnetic devices have been developed for generating random sequences. In this case, the product, the composition, is not predictable in advance, but the method of generation is mechanical, through the repeated application of the same series of steps, an algorithm. "Abstract" paintings can be produced by the mechanical repetition of a specified series of steps whereby a paint gun (or some equivalent device) is activated in front of a canvas. "Aleatoric" musical compositions ("computer music") can be generated by writing a program, an algorithm, which controls "sub-programs" in predetermined (restricted) or in random variations. 2

Creative sequencing is one which is produced without a known algorithm. The emphasis is on "known" because it is possible to claim that all sequences are produced by some algorithm. Certainly it is the case that a sequence produced by an unknown algorithm can, at some future time, be discovered. At that point, the composition process changes from a creative to a mechanical one. The famous mathematician and developer of computers, John von Neumann, has characterized two kinds of "games" in just terms discussed here for the distinction between mechanical and creative processes. He defines an "interesting" game as one whose solution presents some unsolved problems. An "uninteresting" game is one whose solution is completely known. Thus, checkers ceased to be an interesting game (in the mathematical, not psychological sense) when all the problems connected with playing it were solved. A computer can now play a perfect game of checkers. The simulation of a process via computer is an ipso facto demonstration of its interesting status. The program which controls the computer's game is the algorithm of the process. It has become mechanical.

Chess is today still an interesting game. No algorithms exist which would allow a computer to play an unbeatable game. In fact, chess playing computers consistently lose against human opponents that have the standing of "International Grandmaster." Von Neumann predicted that chess will cease to be an interesting game in the decade of the 1970's. (And with that he also foretold the demise of international chess tournaments.)

Discourse, being a sequence of linguistic elements, can be produced by either a mechanical or a creative process. Algorithms for the generation of sentences are still in an early stage of development but they already exist and further strides are certain to come. That discourse can be generated through a mechanical process is no longer the issue. The pertinent question today relates to the comparison of the two products--and, here the issues dealt with earlier in this paper become salient. There is no doubt that discourse produced by a mechanical process can be "interesting" (in the psychological sense), novel, effective, etc. at some point and for some listeners. But it can never be creative. To the programmer who developed the algorithm for the production of such discourse, there can be no surprise in the product, no novel rhetorical schemes, no creation of new stylistic devices not known to him. In short, to him and other knowledgeable individuals, mechanically produced discourse is a stale piece of composition.

These considerations have important implications for the teaching of composition. Examples of algorithms that are being used in the teaching of "good writing" consist of the "rules" one finds in "how to" type cookbooks of rhetoric and English composition: the "do's" and "don't's" of punctuation, of sentence length and subordination, of theme organization, and so on. the result of such teaching, when it is successful, can never in and of itself produce anything but stale writing. The importance of such teaching, when it is successful, can never in and of itself produce anything but stale writing. The importance of such teaching should not, however, be minimized. There are two reasons for this. In the first place, stale writing of this sort is often more effective for communication than the undisciplined composition of the untrained, insensitive individual. In the second place, teaching of this sort may not impart merely the knowledge of the algorithms involved.3 The human mind can be of a staunchly independent spirit: it can resist indoctrination and control by algorithms. The question whether the teaching of rhetorical algorithms facilitates or inhibits such independence or creativeness is one for which we do not now have the answer. On it hinges the value of our educational system.

Center for Comparative Psycholinguistics
University of Illinois, Urbana


Bernsein, B. Aspects of language and learning in the genesis of the social process. In Dell Hymes (Ed.), Language in Culture and Society. New York: Harper and Row, 1964, Pp. 251-263.

Braddock, R., Lloyd-Jones, R., and Schoer, L. Research in Written Composition. Chanmpaign, Ill.: National Council of Teachers of English, 1963.

Center for Comparative Psychollinguistics. A World Atlas of Affective Profiles, University of Ilinois, Urbana, 1969. (In preparation.)

Chatman, S. The semantics of style. Social Science Information, 1967, 6 (4), 77-99.

Christensen, F. The problem of defining a mature style. English Journal, 1968, 572-579.

Deese, J. The Structure of Associations in Language and Thought. Baltimore: The Johns Hopkins Press, 1965.

De Vito, J. A. Style and stylistics. An attempt at definition. The Quarterly Journal of Speech, 1967a, 53, 248-255.

De Vito, J. A. Oral and written style: Directions for research. Souther Speech Journal, 1967b, 33, 37-43.

De Vito, J. A. Levels of abstraction in spoken and written language. The Journal of Communication, 1967c, 17, 354-361.

De Vito, J. A. A linguistic analysis of spoken and written language. Central States Speech Journal, 1967d, (May), 81-85.

De Vito, J. A. The encoding of speech and writing. The Speech TEacher, 1966a, 15,55-60.

De Vito, J. A. Psychogrammatical factors in oral and written discourse by skilled communicators. Speech Monographs, 1966b, 33, 73-76.

De Vito, J. A. Comprehension factors in oral and written discourse of skilled communicators. Speech Monographs, 1965, 32, 124-128.

Goodenough, W. H. Componential analysis. Science, 1967, 156, 1203-1209.

Haagen, C. H. Synonymity, vividness, familiarity, and association value ratings of 400 pairs of common adjectives. Journal of Psychology, 1949, 27, 453-463.

Heinberg, P., Harms, L. S., and Yamada, June. Speech-Communication Learning System. Speech-Communication Center, Department of Speech, Univerity of Hawaii, 1969.

Holzmakn, Mathilda S. Ellipsis in discourse: A psycholinguistic problem. eliot-Pearson Department of Child Study, Tufts University, 1968. (Mimeo.; date uncertain)

Hughes, R. E. and Duhamel, P. A. Principles of Rhetoric. Englewood Cliffs, N. J.: Prentice-Hall, 1966.

Hunt, K. W. Grammatical Structures Written at Three Grade Levels. Champaign, Ill.: The National Council of Teachers of English, 1965. (Research Report No. 3.)

Jakobovits, L. A. The act of composition: Some elements in a performance model of language. Paper prepared for a Conference on the Composing Process sponsored by the National Council of Teachers of English, Colorado Springs, November 3-6, 1968. (Proceedings are in press.)

Katz, J. J. and Fodor, J. A. The structure of a semantic theory. In J. A. Fodor and J. J. Katz (Eds.), The Structure of Language. Englewood Cliffs, N. J.: Prentice-Hall, 1965.

Osgood, C. E. Interpersonal verbs and interpersonal behavior. Institute of Communications Research, University of Illinois, Urbana, 1968. (Mimeo.)

Osgood, C. E., Suci, G. J., and Tannenbaum, P. H. The Measurement of Meaning. Urgana: University of Illinois Press, 1957.

Page, E. Statistical and linguistic strategies in the computer grading of essays. Paper presented at the International Conference on Computational Linguistics, Grenoble, France, August 23-25, 1967.

Paivio, A. Abstractness, imagery, and meaningfulness in paired-associate learning. Journal of Verbal Learning and Verbal Behavior, 1965, f, 32-38.

Russell, W. A. and jenkins, J. J. The complete Minnesota norms for responses to 100 words from the Kent-Rosanoff Word Association Test. Studies on the Role of Language in Behavior, Technical Report No. 11 for Contract Nonr-66216, Department of Psychology, University of Minnesota, August 1954. (Mimeo.)

Sircello, G. Expressive qualities of ordinary language. Mind, 1967, 76, 548-555.

Slotnick, H. B. and Knapp, J. V. Machine oriented measurement of student writing. Office of Instructional Resources, University of Illinois, December 6, 1967. (Mimeo.)

Steinberg, D. Analyticity and features of semantic interaction. Instistute of Communications Research, University of Illinois, Urbana, 1969. (Mimeo.)

Thorndike, E. L. and Garrettson, Judith. What makes a word difficult? Teachers College, Columbia University, 1968. (Mimeo.)

Thorndike, E. L. and Lroge. I. L. The teacher's word book of 30,000 words, New York: Teachers College, Columbia University, 1944.

von Neumann, John. The Computer and the Brain. New Haven: Yale University Press, 1958.


1 These examples and the arguments are based on Sircello's (1967) paper.

2 Whereas the application of the algorithm, once developed, is a mechanical process, the discovery or invention of algorithms is a creative process since they are not controlled themselves by algorithms. Algorithms for the discovery of algorithms is not a certain achievement that man can count on. It seems to me that "stale art" is so because it is produced by known algorithms (the artist--in this case, "technician" would be a better descriptor--merely repeats what has already been done before).

3 Some types of "rules" might be useful in this respect. I am thinking of Christensen's (1968) rule for developing "mature style": use long free modifiers in final position in conjunction with short base clauses. I might call this rule a semi-algorithm since it does not specify the manner of constructing free modifiers--and for just cause since they can be infinitely variable.

Applied Psycholinguistics in Social Psychology
Effects of Mere Exposure: A Comment
Rhetoric and Stylistics: Some Basic Issues in the Analysis of Discourse
Semantic Satiation and Cognitive Dynamics
Semantic Satiation as a Function of Initial Polarity and Scale Relevance
Some Potential Uses of the Cross-cultural Atlas of Affective Meanings


An Evaluation of People's Attitudes Toward Technostress and Techniques on How to Overcome it.
To Fear or not to Fear the Computer


Leon James Home Page