Exploring computer conversation through communication theory

Material Information

Exploring computer conversation through communication theory
Rothberg, Lynn April
Publication Date:
Physical Description:
69 leaves : illustrations ; 28 cm


Subjects / Keywords:
Computer literacy ( lcsh )
Rhetoric -- Philosophy ( lcsh )
Information theory ( lcsh )
Computer literacy ( fast )
Information theory ( fast )
Rhetoric -- Philosophy ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 67-69).
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Science in Technical Communication, Department of English.
Statement of Responsibility:
by Lynn April Rothberg.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
19817931 ( OCLC )
LD1190.L67 1987m .R67 ( lcc )

Full Text
Lynn April Rothberg
B.H., Pennsylvania State University, 1974
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Science in
Technical Communication
Department of English

This thesis for the Master of Science degree by
Lynn April Rothberg
has been approved for the
Department of
Martin Tessmer

Rothberg, Lynn April (M.S., Technical Communication)
Exploring Computer Conversation through Communication Theory
Thesis directed by Associate Professor Richard P. VanDeWeghe
The computer-conversation model explored in this thesis selects and
combines elements from classical and modern rhetoric and from conversation
theory to create a computer-conversation model for online help. A
computer-conversation model accommodates the unique communication that
occurs when humans and computers interact.
Chapter I discusses how rhetorical theory developed from classical to
modern times and how this theory guides all communication models.
Chapter II examines a communication model based on conversation
theory and how it operates in people's everyday use of language.
Chapter III analyzes how classical and modern rhetoric and conversation
theory contribute to a communication model for human/computer inter-
Chapter IV presents an actual user/computer dialog that is not conversa-
tional and then explains how the system designed for computer-
conversation would have altered the exchange. This chapter points to the
importance of this research to professional communicators who develop
documentation for computer products.

I. A HISTORY OF RHETORIC .................................. 1
Classical Rhetoric ................................... 1
Aristotle's Rhetoric ............................. 3
Cicero's Rhetoric ................................ 5
Quintilian's Rhetoric .............................7
Rhetoric after the Romans .............................7
Saint AugustineA Transition to Medieval Rhetoric ... 8
Rhetoric in the Renaissance .......................9
Rhetoric Redefined at Harvard .....................9
The Continuing Tradition of Classical Rhetoric 10
The Movement Away from Classical Rhetoric 10
The Renewal of Scholarship in Discourse: Modern Rhetoric 11
Comparing Modern Rhetoric to Classical Rhetoric 13
Expanding the Domain of Rhetoric ................ 13
Redefining the Roles of the Communication Model 14
The Conversation Model 19
Conversational Maxims ............................... 23
Characteristics of Rules

Rules of Conversation ............................... 26
An Utterance Gains Meaning Hierarchically ...... 26
Rules Guide the Interpretation of Meaning ...... 29
Conversation Proceeds through Moves 31
III. THE COMPUTER-CONVERSATION MODEL ...................... 36
The Communication Triangle for Computer Conversation . 38
Ethos of the User .............................. 38
User's Pathos for the Computer ................. 40
Logos for the User ............................. 43
Ethos of the Computer ........................... 44
Computer's Pathos for the User ................. 44
Logos for the Computer .......................... 46
Conversational Features of User/Computer Dialog ..... 47
Environment Restricts Computer Conversation ......... 49
The Physical Environment ........................ 49
The Cognitive Environment 50
Topics 51
Common Language ..................................... 52
System Help Functions ............................... 53
The Computer as Advisor 57
Computer Conversation for the Familiar User 64
Computer Conversation for the Naive User 64

Computer Conversation for the Expert User 65
The Role of Professional Writers in Computer Conversation 66
BIBLIOGRAPHY ..................................................... 67

1. Hierarchical Meanings in Coherent Conversation ............. 29
2. Conversational Moves and Clue Words ........................ 33
3. Analysis of a Conversation through Moves ................... 34
4. Typical Human/Computer Dialog (Part 1) 62
5. Typical Human/Computer Dialog (Part 2) 63
6. Computer Conversation for the Naive User 65
7. Computer Conversation for an Expert User .................... 66

1. The Classical Model .......................................... 3
2. The Modern Model ............................................. 16
3. The Conversation Model ....................................... 20
4. The Computer-Conversation Model 38
5. Architecture for a System to Support User/Computer Dialog 45

Rhetorical theory describes a systematic approach to establishing
knowledge in social situations. Rooted in 2500 years of speaking and lis-
tening and writing and reading, classical and modern rhetoric provide the
foundation for all communication theory. The classical theorists recognized
the power of formal speech to spread knowledge and to persuade society to
action. Ancients like Aristotle and, later, Cicero and Quintilian carefully
analyzed oratory and structured its activity through their writings. The
influence of Gutenberg's invention of the printing press prompted modern
theorists to recognize and explore the power of printed information to spread
knowledge and to persuade society to action. The following overview of the
development of rhetorical theory, with the explanation of conversation
theory in Chapter II, are the bases for the computer-conversation model
investigated in this thesis.
Classical Rhetoric
Many subjects for academic pursuit in a culture's school system are
first revealed in the art of that culture. Interest in rhetoric appeared long
before 700 B.C., in Greece, where early Greek drama and literature showed
a concern for variations in modes of speech among different types of

persons (Murphy 1983, 5) and for speeches that were arranged in ways
that were calculated to achieve a desired effect (Murphy 1983, 3).
Concern with rhetoric grew from widespread interest in the oral
presentation of ideas, especially in Athens. Logos, defined as thought-plus-
expression, meant for Athenians that thought is useless without a way to
convey it, and mere expressive ability is worthless if it has nothing to
convey (Murphy 1983, 4). Antithesis, the pairing of contradictions to
display the necessity of choice between them, led to a systematic debate of
opposite sides in political assemblies, to pairing of accusation and defense in
the law courts, and to systematic exploration of contradictory statements in
early Greek logic (Murphy 1983, 5).
According to ancient tradition, the formal study of rhetoric started
in 476 B.C. with Corax, who devised a systematic approach to argument,
which he applied in lawsuits about property. In 431 B.C., in Athens,
Gorgias opened a school of rhetoric, in which he promoted the use of
parallelism and antithesis. The concept of antithesis pervades Greek culture,
where oratory served to resolve inherent conflicts to one extreme or the
Classical rhetoric as interdisciplinary study is evident in the
teachings of Plato, who encouraged the rhetor to study not only the soul,
but also the details of arguments, types of language, and modes of delivery.
(Murphy 1983, 18). Early descriptions of rhetoric furnished little practical
advice to speakers; they mostly defined nature of rhetoric. It was Aristotle,
Plato's famous student, who first attempted to make a more comprehensive
statement about the subject of rhetoric.

Aristotle's Rhetoric
Aristotle's legacy and the informing principle to the study of all dis-
course is the communication triangle of ethos, pathos, and logos. Figure 1
shows the relationship between the ethos of the speaker, pathos of the audi-
ence, and the logos of the subject in classical rhetoric.
Figure 1. The Classical Model
To Aristotle, rhetoric is the faculty of seeing in any situation the
available means of persuasion (Aristotle, Cooper edition 1932, 7) and all
speeches are directed toward a decision. His approach to persuasion
through rhetoric is through three kinds of artistic proofs:
1. Ethical proof, which pledge the speaker's good character to establish his
credibility (ethos);
2. Psychological proofs, which bring the auditor into a state of feeling
favorable to the acceptance of the speaker's arguments {pathos);

3. Logical proofs, which either make a case or appear to make it (logos).
The division of artistic proofs shows that rhetoric is connected on the one
hand to the study of probable proofs, or dialectic, and on the other to the
study of man's character, or ethics. To Aristotle, ethos and pathos
operate through logos...within the enthymematic chain (Murphy 1983, 47).
This means that, in Aristotle's communication triangle, a speaker controls
how the audience feels about him and his subject by carefully constructing
the enthymemes in his argument.
Aristotle recognized that rhetoric is a neutral tool that sees
impartially the arguments on both sides of a question (Murphy 1983, 25).
Enthymeme is Aristotle's technical term for argument from premises
(Murphy 1983, 27). It is a form of syllogism that contains two premises and
a conclusion. For example, the syllogism All crustaceans are scavengers, all
oysters are crustaceans, therefore all oysters are scavengers can be
diagrammed as:
All C are S
All O are C
All O are S
The syllogistic process takes place through the term C. Because C is the
subject of the first premise and the predicate of the second premise and
therefore equal, we infer the equality of O and S in the conclusion.
Throughout the Rhetoric, Aristotle defines aspects of persuasive
speech into a system of rhetoric. Aristotle assigned three classes of subjects
to rhetoric: the epideictic, the deliberative, and the judicial. The epideictic
is concerned with praise and blame of a particular person; the deliberative
subjects occur in political debates; the judicial branch concerns accusation

and defense (Murphy 1983, 98). In his system of rhetoric for these sub-
jects, Aristotle presents topics that include premises for different kinds of
discourse, forms of arguments, style for presenting proofs in argument, and a
taxonomy for the arrangement of proofs.
Aristotle's primary contributions to a rhetorical theory of communi-
cation (for this thesis) were that he defined a role for the audience, he
defined a value system behind persuasive arguments, and he presented a
system for organizing proofs in persuasive speeches. Through deliberate
analysis of interaction between the ethos of the speaker, the pathos of the
audience, and the logos of the subject, Aristotle provided an organized, prac-
tical system of rhetoric for later theoreticians, like Cicero, to refine.
Cicero's Rhetoric
Cicero assimilated the rules and precepts of the Greek model of rhetoric into
a Roman model, which he wanted to be a system of general culture, a mode
of life itself. He viewed the orator as a cultured mixture of philosopher,
lawyer, and politician (Murphy 1983, 107).
Cicero's writings elaborate the divisions and subdivisions of rhetoric
and oratory. In De Inventione, Cicero redefines the parts of rhetoric and the
parts of oration and he describes the different situations to which an orator
must respond with skill (Murphy 1983, 93).
Like Aristotle, Cicero assigns the three subjects of rhetoric as the
epideictic, the deliberative, and the judicial. The five parts of rhetoric are
invention, arrangement, style, memory, and delivery (Murphy 1983, 99-101).

1. Invention is the discovery of true or apparently true arguments that
make an argumentative case probable. As a primary step of invention,
Roman rhetoricians used a series of questions called a stasis document to
determine relevant issues and a topical method of questioning to investi-
gate a case completely.
2. Arrangement is the distribution of invented arguments in their proper
order. For arrangement, Cicero defined six parts of an oration:
the exordium, an introduction to win favorable attention; the
narratio, a statement of the case; the partitio, an announcement of
the headings under which the case was to be discussed; the
confirmatio, the constructive arguments; the refutatio, arguments
refuting the opponent's claims; and the peroratio, a summary, con-
clusion, and final appeal (Murphy 1983, 94).
3. Style is the fitting of the proper language to the invented materials.
The art of style involved the student of rhetoric in selecting words and
constructing sentences that possess four virtues: clarity, correctness,
appropriateness, and embellishment.
4. Memory is the firm mental grasp of arguments and language.
5. Delivery is the control of the voice and body adapted to the importance
of the material and language.
Like Aristotle, Cicero also emphasized the importance of ethos and pathos in
rhetoric. The primary contribution of Cicero and other Roman rhetoricians
was to refine and systematize the rules and precepts of rhetoric.

Quintilian's Rhetoric
The fall of the Republic and the rise of the Roman Empire in the
first century A.D. created social and political conditions that changed the
status of oratory in the Roman world: the length of speeches, the number
of advocates, and duration of court trials were reduced; orators ran the risk
of crossing the Emperor in every speech they gave (Murphy 1983, 152).
From 14-138 A.D., the theme of rhetoric was a new composition based on
themes that were invented and then argued by students. Training in rhet-
oric focused on expertise in traditional and novel approaches to style.
In this time, Quintilian produced a comprehensive Latin statement
of rhetoric that stressed morality and guided citizen-orators (Murphy 1983,
153). His work summarized the four stages of analysis for literary works:
(1) establishing the texta textual criticism needed because of poor manu-
scripts, (2) an expressive readingparts memorized, (3) expositiona word
by word content analysis, and (4) judgementdrawing moral lessons from
the reading (Kinneavy 1971, 6). Quintilian's primary contribution was to
maintain the influence of rhetoric in teaching at a time when the influence of
oratory waned.
Rhetoric after the Romans
By Quintilian's death, Rome had developed a large population that
demanded standardization of institutions like schools, armies, and the legal

system. Foreshadowing the next many centuries of public activity, this
standardization created an explosive growth in written documents and a
need for legal technicians who were lawyers first and orators only
secondly (Murphy 1983, 178). Oratory was activated primarily in situ-
ations about the conflict of contradictory laws.
Saint AugustineA Transition to Medieval Rhetoric
In the fifth century A.D., Saint Augustine urged the church to use
Cicero's model of rhetoric as a guide to preaching. Augustine's rhetoric is
Cicero's but his communication is based on Christian theology. Theology as
the primary communication of rhetoric places the works of Augustine firmly
in the Medieval Age and marks the end of the development of Classical
rhetoric (Murphy 1983, 183-4).
Educators in the Middle Ages combined literary, rhetorical, and
dialectical studies into language arts, which emphasized a more systematic
study of language through grammar. The main business of colleges in the
Middle Ages shifted from rhetoric, where academic success was determined
by delivery of a set speech, to dialectic debate, where defense of ideas in
debate were specifically tied to a student's progress in the school system
(Kinneavy 1971, 8-9).

Rhetoric in the Renaissance
The split of rhetoric into the two arts of preaching and letter writing
that started in the Middle Ages continued through the Renaissance. Since
the content of religious speeches was biblical, invention became less signif-
icant. For the clergy, style and delivery dominated the other parts of clas-
sical rhetoric. At the same time, Renaissance scholars increasingly applied
rhetorical principles to written discourse (Lindemann 1982, 42).
Rhetoricians of this era surfaced into three groups:
Traditionalists tended to appreciate the importance of all five departments
of rhetoric (Lindemann 1982, 43). They advocated constant practice of
the arts rather than rote drill as a teaching technique.
Figurists subordinated logic to rhetoric, emphasizing above all the impor-
tance of style (Lindemann 1982, 44).
Ramists tended to subordinate rhetoric to logic, assigning invention,
arrangement, and memory to logic, and grouping style and delivery under
rhetoric (Lindemann 1982, 44).
The Ramists dictated the activity and influence of rhetoric in the
Renaissance. Their separation of logic from rhetoric established the
dichotomy between reason and imagination (Corbett 1971, 611).
Rhetoric Redefined at Harvard
In seventeenth century America, the earliest rhetorical instruction
and theory occurred at Harvard. There, curricular reformers echoed the
Ramists and proposed that the primary purpose of rhetoric was to perpet-

uate the study of Latin, and its primary parts were elocutio and pronunciatio
(Connors 1984, 1). Harvard's definition of rhetoric to the narrow scope of
Latin studies foreshadowed the decline of rhetoric in the nineteenth century,
which is discussed in The Movement Away from Classical Rhetoric.
The Continuing Tradition of Classical Rhetoric
Colleges of the late eighteenth century applied the doctrine of clas-
sical rhetoric aimed at communication as art that synthesizes material from
a wide variety of fields, communication as a system, and communication as
a method to produce good citizens skilled in speaking (Connors 1984, 2). By
this time also, Gutenberg's invention of the printing press (1440) had made
printed information available to many. In 1759, John Ward presented
oratory as a system and reunited the offices of rhetoric by discussing
invention and arrangement as well as style and delivery (Connors 1984,
221). Ward represented a faction of rhetoricians that reestablished the con-
nection between rhetoric and writing.
The Movement Away from Classical Rhetoric
The nineteenth century saw the province of rhetoric...both trun-
cated (to a focus on style) and diffused (to emphasize the aesthetic appreci-
ation of literature rather than the active production of public discourse)
(Connors 1984, 2). Increased specialization of disciplines, steadily growing
enrollments, and the influence of the belles lettres (a movement to the appre-
ciation of literary works away from active production of discourse) led to

what we now view as a permanent institution: the department of English
(Connors 1984, 3) and also to the decline in the influence of classical rhetoric
in America's colleges.
Pursuit of literary scholarship completely shifted the commitment
of English to reading and writing rather than to speaking (Connors 1984,
5); and in 1914, scholars of rhetoric campaigned for separate departments
of rhetoric and public speaking (Connors 1984, 6). From 1900 to 1920,
especially in America, classical rhetoric was viewed less as a living tradition
and more and more as a cultural heritage, locked away in footnotes and
museums (Connors 1984, 6).
The Renewal of Scholarship in Discourse: Modem Rhetoric
In 1924, Baldwin published Ancient Rhetoric and began a tradition
of renewed critical scholarship in classical rhetoric that has lasted to the
present day (Connors 1984, 7). Aided by Lane Cooper's translation of
Aristotle's Rhetoric in 1932, classical rhetoric enjoyed a revival in speech.
English departments, which controlled the study of composition, did
not join the revival of classical rhetoric until the late 1930s. Then the com-
munications movement, which emphasized all four of the communicative
skillsspeaking, writing, listening, and readingand the growing discipline of
literary criticism (Connors 1984, 8) caught their interest. Authors like
Kenneth Burke and Richard Weaver built on principles of traditional rhet-
oric to establish the importance of human communication in the here and
now (Lindemann 1982, 49) of a complex society.

In his writings, Kenneth Burke asserted that rhetoric is interdiscipli-
nary, covering language and the humanities including all of the symbolic
means of inducing cooperation in beings that by nature respond to symbols
(Burke, Rhetoric 1969, 43). Burke's rhetoric centers on the theme that
people are motivated in a rhetorical situation to identify with one group or
another. He analyzes rhetorical situations using the pentad for examining
human motivation:
Any complete statement about motives will offer some kind of
answers to these five questions: what was done (act), when or where
it was done (scene), who did it (agent), how he did it (agency), and
why (purpose) (Burke, Grammar 1969, xv).
At a time when people felt the dehumanizing effects of technology, Burke
presented rhetoric as a cohesive device for social interaction (Lindemann
1982, 52). Relating rhetoric directly to people's social concerns revived the
purpose of classical rhetoric and broadened the scope and influence of rhet-
oric in the twentieth century.
By 1944, courses in communication were developed and taught by
faculty from both speech and English departments. Combining elements of
of both disciplines marked a return to applying the principles of classical
rhetoric in everyday academia. Rhetorical principles were reintroduced to
composition classrooms across America. At the University of Chicago, study
for applying classical rhetorical theory to written discourse centered with the
group called the Chicago Neo-Aristotelians. By 1957, Richard Weaver, a
member of the group, reintroduced to writing pedagogy such classical ele-
ments as the enthymeme and the topoi (Connors 1984, 8-9).

By 1968, the founders of Rhetoric Society of America, an interdisci-
plinary mix of scholars, created an intellectual force that spearheaded the
movement toward renewed cooperation among disciplines (Connors 1984,
11). This cooperation directs language instruction in colleges today. Sharing
a theoretical base between disciplines renewed our acquaintance with clas-
sical rhetoric and also accelerated the movement of modern rhetoric.
Modern rhetoric extends this cooperation to a holistic concept of man's
knowledge and interaction in the world and reevaluates the roles of ethos,
pathos, and logos that Aristotle presented in his communication model.
Comparing Modern Rhetoric to Classical Rhetoric
Many philosophers of rhetoric separate classical and modern rhet-
oric by acknowledging subtle differences between them. The informing prin-
ciple of classical and modern rhetoric is the same. It is the relationship of
ethos, pathos, and logos in discourse. The primary differences in rhetoric
from classical to modern times are (1) the expanded domain of situations in
which to apply principles of a theory of rhetoric and (2) new roles for the
participants in a communication model.
Expanding the Domain of Rhetoric
Classical rhetoric was the study of persuasion as the science of aca-
demic eloquence. Modern rhetoric now means a general study of communi-
cation (Kinneavy 1971, 7). Classical rhetoric addressed formal

communication in the limited rhetorical experiences of speeches. Modern
rhetoric adds to oratory new forms of persuasion that earlier ages didn't
addressfor example, rhetoric in a conference room.
Modern rhetoric includes informal situations that encompass all
verbal arts because we believe in mutual persuasion as a way of life
(Booth 1965, 8-12). This defines an expanded domain for rhetoric in
everyday communication. (Chapter 2 of this thesis discusses conversation
theory as part of the expanded domain of rhetoric and Chapter 3 shows
human/computer interaction as a new situation in which rhetoric functions.)
Redefining the Roles of the Communication Model
The classical communication triangle is a rigid model that closely
shows one-way communication. The roles of audience and subject were
limited to a speaker's conception of them. Classical rhetoric directed a
speaker to be credible, knowledgeable about the subject, skilled in argument,
and able to manipulate an audience to openly accept his ideas. The audi-
ence was stereotyped and then acted only as an empty receptacle into which
the speaker's ideas were poured.
In the classical communication triangle, the speaker controls
meaning; the speaker conveys meaning when he properly manipulates the
subject matter and the audience. The classical model therefore assumes
perfect communication whenever a speaker meets these qualifications.
Modern rhetoric realizes, however, that communication does not occur like
that. In reality, participants in a modern communication model interact

with each other and with the subject; they negotiate meaning in the area of
the subject based on their knowledge and experience with the subject.
In oral and written communication, an audience is not simply a
blank page on which ideas are imprinted, and the subject is not only what a
speaker knows about it. An audience certainly brings knowledge and experi-
ence to a communication that affects how much meaning about the subject
is actually understood. Modern rhetoric recognizes that, in any communi-
cation, an audience interprets words about a subject and actively constructs
meaning in those words based on what they believe about the subject.
Modern rhetoric views the knowledge and experience of perceivers as reality,;
it is two-way communication in which a speaker or writer encodes, that is,
chooses the verbal signals of the message and a listener or reader decodes
that message. Where members of the audience sense ambiguity, they inter-
pret meaning (Rosenblatt 1978, 19). (The message is the words about a
subject, which, in this thesis, include signal, speech, "text, utterance,
or context.)
Classical rhetoric established a relationship between the ethos of the
speaker (encoder), the pathos of the audience (decoder), and the logos of the
subject (signal) for communication. Modern rhetoric describes that relation-
ship as interactive within the reality of the situation and the knowledge and
beliefs of participants. Figure 2 shows communication as an inverted tri-
angle that pivots on reality. Communication is essentially the interaction of
the classical components of rhetoric (ethos, pathos, and logos) in the context
of reality.

Figure 2. The Modern Model. (Adapted from Kinneavy 1971, 47.)
In contrast to the classical model in which a speaker controls
meaning in a communication through his stereotype of the audience and his
knowledge of the subject, the modern model shows the role of audience as
independent of the speaker but operating within the realm of reality. The
speaker and audience cooperate in the area of the subject to establish
The history of rhetoric presented here shows how different beliefs
about the elements of rhetoric drew different followings. The shifting
emphasis between logic and style and attempts to separate and mix the ele-
ments of discourse drew attention to all of the elements. Study and contro-
versy over the domain of rhetoric led to todays belief that rhetoric
encompasses all communication, formal and informal. This belief reveals the
link from rhetorical to conversational theory.

Conversation theory is a logical extension of the communication
theory that developed from classical and modern rhetoric. It contributes to
realizing the primary goal of classical rhetoric to see in any situation the
available means of persuasion (Aristotle, Cooper edition 1932, 7) and a
goal of modern rhetoric to develop a rhetorical theory that embraces all
communication. Ancient and modern rhetoricians gave definition to rhetoric
and attempted to provide detailed theory and procedure for using rhetoric.
Conversation theorists are attempting to do the same with conversation.
The primary differences between conversation and the written and
oral discourse described for classical and modern rhetoric stem from the
formal realm of rhetoric versus the informal nature of conversation. The
rules of rhetoric that hold for formal speeches and writing may not hold for
their natural counterpart of conversation (Grice 1976, 43). Conversation has
participants and discourse moving through contexts. Conversation has levels
of meaning which are derived from behavior that follows rules. The activity
of conversation progresses hierarchically from one level of meaning to the
next. The elements of the conversation model are: (1) participants, (2) dis-
course, (3) maxims, (4) rules, (5) context, and (6) moves.

1. Participants embody the roles of ethos and pathos from classical and
modern rhetoric, except that their roles have been expanded and they are
2. Discourse contains the logos of the subject described by the classical and
modern communication models. It also directs the communication; but,
in conversation, discourse is spontaneous and it changes depending on
the feedback of participants.
3. Maxims provide the framework in which conversation occurs. They gen-
erally describe the conditions that presuppose all discourse: that partic-
ipants will say what is true, relevant, and needed to express a subject.
4. Rules guide how participants behave to establish and interpret meaning
in conversation. Classical and modern rhetoric explained rules in terms
of pathos for persuading audiences, ethos for establishing the credibility
of the speaker, and logos for arranging and supporting the rhetors dis-
course of the subject. Additional rules in conversation guide new goals
and control meaning for all participants.
5. Context is the situation or episode in which communication occurs. This
is the reality shown in the modern communication model, which oper-
ates within the formal context of oral and written discourse. The context
of conversation changes frequently and introduces the informal episodes.
6. Moves describe the purpose of the segment of discourse, called a
context. Some moves appear in classical and modern rhetoric, such as
making a claim and supporting the claim in formal and persuasive

communication. Conversation permits more sets of moves to define
context in informal communication, which has more purposes than just
The Conversation Model
Conversation is a relatively informal social interaction in which the
roles of speaker and hearer are exchanged in a nonautomatic fashion under
the collaborative management of all parties (McLaughlin 1984, 271). Con-
versation is communication; its model contains the same elements as any
other communication model: ethos of the speaker, pathos of the listener,
and logos of the subject. A conversation model, however, must show how
participants interact and exchange roles.
Conversation involves much more than just talking about this and
that. It involves us in doing things, in getting others to do things, in
eliciting information, in refusals, apologies, promises, and threats, and
in a whole list of other activities. Conversation is not just talking; it is
also doing (Wardhaugh 1985, 162).
The conversation model in Figure 3 on page 20 shows how the the
components of communication (ethos, pathos, and logos) interact in conver-

Figure 3. The Conversation Model
All factors of communication for Person A and Person B operate
within the model. The triangles overlap in the domain of logos, which
includes the context of the situation, mutual knowledge of the participants,
and the words that express the subject. Participants in conversation each
have goals and need to consider the goals of one another. The subject is
guided, but is, for the most part, spontaneous. The goals of participants and
the course of the conversation may change as parties provide feedback to
one another. There is always imprecision and vagueness in conversation;
however, meaning is made clear in conversation by the interactions of partic-
ipants and their spontaneous feedback. For example:

Ive got that meeting I told you about.
What meeting?
The one to go over the accounts.
Oh, that one. So when will you be home?
A little late, I suppose.
How late?
Six-thirty perhaps. Maybe a little later.
From As remarks, B concludes that A does not know exactly when she will
be home. He eventually tells her not to rush home (Wardhaugh, 31-32).
The meaning of a conversation, therefore, is something that is negoti-
ated during the course of the conversation rather than directly
expressed. What is going on, what is meant, depends on what has
gone before, what is currently happening, and what may or may not
happen. It is not fixed, but subject to constant review and reinterpre-
tation. However, at each point in a conversation the parties to it
must have some view of where they are in order to make sense of
what is going on, but it is not a view that they can or should hold
rigidly for, if they do, misunderstanding is likely to be the conse-
quence. (Wardhaugh, 33).
Conversation includes behavior other than talk (Goodwin 1981, 2),
like paralinguistic behavior that includes gestures (nods), facial expression
(raising eyebrows), eye movements, and stress and intonation (Goodwin
1981, 8-9). Gaze, for example, is itself a social act that has types and social
characteristics. The movement of gaze can be side-by-side or face-to-face.
And gaze can signal a display of listening, disengagement (Goodwin 1981,
30), or "choice points, places where the future action of the speaker is con-
tingent on the subsequent action of his hearer (Goodwin 1981, 32).
Intonation can be rising or falling. Rising intonation expresses uncertainty
or finality and is found in hesitations and ends of questions; falling
intonation expresses finality and is found at the ends of statements
(Goodwin 1981, 26). Conversation also includes more positions than

speaker and listener, for example, those who overhear or are not specifically
addressed in the conversation (Goodwin 1981, 4). Although non-verbal
aspects of conversation and additional positions for participants certainly
exist, these features of conversation are not relevant to the analysis in this
thesis and will not be pursued here.
Investigations into conversation theory currently center on the belief
that conversation is a highly organized activity that is governed by the
notion of a rule. The virtues of a theoretical perspective of rules for conver-
sation appear extensively in literature about communication.1 The belief that
conversation follows rules implies that conversation is a system in which ele-
ments follow a sequential structure. To demonstrate that these sequential
patterns exist, Clarke tested whether or not subjects could reassemble an
unfamiliar dialogue with greater than chance accuracy when presented with
its component utterances in random order (Clarke 1983, 45). Subjects were
able to reassemble the conversation with greater than chance accuracy.
This suggests that the original dialogues did embody some systematic
relationship between verbal context and the nature of each utterance;
that the pattern is one on which the native speaker can pass judge-
ment from his own knowledge of the speech community, and that a
consensus may be reached by the judges which is reasonably accurate
(Clarke 1983, 49).
1 Pearce 1979 and McLaughlin 1984, works which appear in the
list of references for this thesis, show several sources that advocate this

Describing the features of the sequential structure that conversation
follows is currently the substance of much research and theoretical dis-
cussion. Theorists assume that talk exchange develops from purposeful and
rational behavior, which is governed by a precise system of rules. However,
little consensus about the criteria for evaluating rules research or the formal
requirements of a rules-based theory has emerged from these treatises
(Pearce 1979, 75). Knowing that an accepted set of rules for conversation
has not yet developed, this thesis generalizes about what conversational theo-
rists knowthe maxims, rules, and other lore that conversation
followswhich makes conversation a system.
Conversational Maxims
Maxims are widely held assumptions that the expected behavior of
conversationalists is to provide accurate, economical, clear, polite, ethical,
and supportive contributions to spoken interaction (McLaughlin 1984,
274). This theory stems from the maxims proposed by H. P. Grice, which
include the general guiding maxim of conversation, which he calls the coop-
erative principle, and its four attendant conversational maxims of quantity,
quality, relation, and manner. The cooperative principle states: Make your
conversational contribution such as is required, at the stage at which it
occurs, by the accepted purpose or direction of the talk exchange in which
you are engaged (Grice 1976, 45).

The cooperative principle implies four attendant maxims that assure
cooperative contribution to conversation if followed:
1. The quantity maxim relates to the amount of information that you
provide. It advises you to make your contribution neither more nor less
informative than is required. Conversation that is over informative may
be confusing and raise side issues. Conversation that is underinformative
may lead to misinterpretation.
2. The quality maxim requires that you state only that which you believe is
true and that for which you have adequate evidence.
3. The relation maxim relates to the communicative intent or purpose of the
conversation. It states simply "Be relevant and requires your contrib-
ution to be immediate and pertinent at each stage of the conversation.
4. The manner maxim relates to how what is said is to be said. It advises
that you refrain from being obscure, ambiguous, verbose, or unorganized
in your speech.
Grices maxims present a global framework in which conversation
exists, stating that talkers will generally proceed in the manner that these
principles prescribe (Grice 1976, 48). To Grices maxims, Kraut and Higgins
(Zeidner 1986, 273) add the sensitivity maxim, which requires that your
message adapt to your listeners characteristics, including what you believe
your listeners know. This additional maxim clearly shows that conversation
is affected specifically by how the perceiver will interpret the speakers
message in its unique situational context.

Characteristics of Rules
In conversation, the goals and roles of participants and the content
of the subject change continually. The activity of satisfying goals by negoti-
ating for information in the course of conversation has a structure that is
best understood by recourse to the notion of rule, as opposed to that of a
law. If governed by law, the structure of conversation would be enforced
through consequences to those who do not obey. Rules are about activities
over which we have control (like conversation) . rules have less predictive
power than laws (McLaughlin 1984, 18). To explore the conversation
model as a rules-based system, we need first to understand the character-
istics of a rule and the properties of a rule system.
Rules are products of agreement in a society that encapsulate the
pervading wisdom about appropriate behavior. They serve the society to
predict, interpret, and evaluate behavior. Each rule is a followable
prescription that states what sort of activity is obligated, preferred, or
prohibited in particular contexts (McLaughlin 1984, 16). For example, the
prescription for rules that interpret behavior can be translated in the form: If
situation X occurs, do (do not do) Y (McLaughlin 1984, 21). Whatever
behavior a particular rule prescribes, it is clear that the behavior is situation-
Unfortunately, as mentioned earlier, no real theory of exact situ-
ations has yet appeared to which such efforts can be linked to develop a rule
hierarchy (McLaughlin 1984, 26). Rules are best viewed globally as prop-
ositions that guide action (McLaughlin 1984, 16) and locally as suggestions,

since they can be revised at this level. Without a theory of situations to
order rules in general, we cannot order the rules of conversation absolutely
in a rule hierarchy. Currently, many authors have different theories about
the rules of conversation. This thesis describes points and theory about con-
versation that seem commonly accepted among authors and are relevant to
the study presented here.
Rules of Conversation
As rule-based activity, conversation follows procedures that partic-
ipants negotiate as the activity progresses. Conversation is a special case of
what Erving Goffman calls focused interactionthe kind of interaction that
occurs when persons gather close together and openly cooperate to sustain a
single focus of attention (Goodwin 1981, 2). Rules are the controlling
devices with which we focus attention and guide the behavior of conversa-
An Utterance Gains Meaning Hierarchically
The lowest level of conversation occurs with the utterance. An
utterance is the stream of speech actually produced by a speaker in conversa-
tion (Goodwin 1981, 7-8). What follows the utterance is a collaborative
effort by participants to establish its meaning based on rules.

An utterance gains meaning hierarchically based on the roles and
goals of participants (ethos and pathos) and the context of the conversation
Note: The opposite view is that conversation develops linearly from
one utterance to the next. A linear model implies that each utterance
depends directly on the one before it. However, interpreting a current utter-
ance in conversation requires identifying how it fits with the entire context of
the conversation. The relevant utterance is only a portion of preceding dis-
course and is not necessarily that portion that linearly precedes the
utterance (Reichman 1985, 31-32). This thesis therefore rejects the view
that meaning in conversation builds linearly.
A hierarchy of conversation shows how an utterance develops
meaning. The five levels of meaning for information in conversation are: (1)
the stream of behavior, (2) propositions, (3) speech acts, (4) episodes, and (5)
1. The stream of behavior is the first form of information that any utterance
in conversation takes. It consists of raw sensory data including move-
ments, noises, and visual patterns that remain meaningless and undiffer-
entiated until punctuated and interpreted by an observer (Pearce 1979,
76). This level of information is the signal referred to in a modern
communication model (see page 15).
2. A proposition is the statement that results from how a signal is
decoded on the technical and semantic levels of communication. A pro-
position has a reference and a predication (Pearce 1979, 77), for

example, Would you like to step outside? The proposition does not
identify the significance of the statement to the listener in conversation.
3. The speech act identifies the significance or intent of the speaker's propo-
sition, for example, whether it is informing, asking, or advising. A speech
act tells what the speaker did to the listener by saying what he did
(Pearce 1979). This level establishes Would you like to step outside? as
a request. Speech acts acquire meaning from the conversational events,
called episodes, that occur before and after them.
4. Episodes are sequences of message which have a starting and a stopping
point and an internal structure (Pearce 1979, 78). Episodes place
speech acts within the larger meaning of the context of the conversation.
The request, Would you like to step outside? may result from the
episode of disagreement. Episodes are the particular instances in the
larger framework of archetypes.
5. Archetypes refer to the cross-cultural similarity in episodic structure
. . that humans share a common physiology and live in a world with
common physical properties (Pearce 1979, 78). They are the common
framing of particular streams of episodes that symbolize the regularities
of life in this world, for example, birth, death, ballgames, etc. Pictured as
a flare of temper in a bar, the episode of disagreement for Would you
like to step outside? would convey the archetype of a bar-room quarrel.

Rules Guide the Interpretation of Meaning
The activity of coherent conversation is coordinated by participants
according to rules that guide how information is processed through levels of
meaning. Pearce and Conklin suggest that the rules of coherent conversa-
tion, which guide utterances through the levels of meaning, are:
1. Rules of information processing
2. Rules of communication
3. Rules of sociation
4. Rules of symbolic identification
Table 1 shows how these rules guide the behavior to establish meaning in
Level 5: Archetypes: basic patterns of interpersonal relationships which are articulated as episodic structure i
1, Rules of symbolic identification I
Level 4: Episodes: 1 punctuated sequences of speech acts with an internal structure i
i Rules of sociation
Level 3: Speech acts: 1 interpretation of the significance of propositions in terms of what the speaker did to the listener by saying what he did 1
1 Rules of communication |
Level 2: Propositions: 1 references and predicates which identify perceptual units and the relations among them i
1 Rules of information processing i
Level 1: The stream of behavior: i raw sensory data including movements, noises, visual patterns
Table 1. Hierarchical Meanings in Coherent Conversation

1. Rules of information processing guide the behavior that elevates raw
sensory data . into identifiable perceptual units and establishes the
relationship between them (Pearce 1979, 79). When people translate the
raw data into propositions, they use personal knowledge to create
meaning. Information processing rules indicate that people share
meaning only at the level of sensory data; thereafter, meaning is individ-
2. Rules of communication are agreements about what elements at the level
of propositions count as particular elements in the level of speech acts
(Pearce, 1979, 80). People use rules of communication to label their per-
ceptions about propositions. As a speech act, you might recognize
Would you like to step outside? as a request. But is this statement
merely a friendly request to change location? Defining the speech act of
a particular utterance depends on what episode is being enacted.
3. Rules of sociation define the episode that is being enacted. They direct
subsequent activity in a conversation because they determine which
speech acts fit together in sequence and how those sequences should be
punctuated (Pearce 1979, 80). For example, little information in its pro-
positional content indicates that Would you like to step outside? will
end in a brawl. It may result in a brawl when uttered by a participant in
a bar-room quarrel. The episode, then, is a challenge. Participants
quickly understand the challenge because they share the situation and the
rules of sociation (Pearce 1979, 80-81).

4. Rules of symbolic identification guide conversation when participants
match episodes to archetypes. In some way, individuals recognize quite
dissimilar segments of the stream of behavior as alternative enactments
of the same episode (Pearce 1979, 81). Having processed proposition
"Would you like to step outside? through the hierarchy of rules to rules
of symbolic identification, a listener recognizes the event and episode as a
bar-room quarrel. These rules govern the symbols that people use to
conceptualize archetypes.
The levels of meaning and rules for behavior interact in conversa-
tion. They cannot be studied separately to accurately predict behavior, even
though theorists frequently telescope a particular level of meaning for
study. Propositions are linked with different levels of meaning depending on
the context.
Conversants can coordinate their meanings if both know which lower
level rules to use in interpreting utterances as propositions and speech
acts (which they do by inferring what episode is being enacted), and if
they agree that that episode is situationally appropriate (which they
do by following the rules of symbolic identification) (Pearce 1979, 82).
Conversation Proceeds through Moves
At a high level, conversation proceeds through a sequence of func-
tionally related frames or turns or moves that participants use to
comprehend levels of meaning. These terms are synonymous in conversation
theory and refer to the context of conversation:
Frame The conversational context includes a social frame within which a
text is constructed (while eating in a restaurant, . being interviewed,
etc.) and a relational frame that participants bring to a conversation
(Craig 1983, 136-7).

Turn The talk of one party bounded by the talk of others constitutes a
turn, with turn-taking being the process through which the party doing
the talk of the moment is changed (Goodwin 1981, 2).
Move Conversational moves represent the various kinds of semantic and
logical relations that can hold between utterances of a discourse
(Reichman 1985, 35).
This thesis uses the term conversational move to describe the activity
of beginning a new communicative act in conversation, which is one of two
types of utterances in conversation. All other utterances constitute
embellishment or continuation of a current topic. Racheal Reichman pre-
sents an analysis of conversation through conversational moves:
1. A conversation is a sequence of conversational moves, each of
which fulfills a particular function in the development of the dis-
2. Utterances in discourse are generated and interpreted in light of
their functional relation to preceding discourse units, called context
spaces, each of which contains a single conversational move.
3. The smooth flow of discourse rests on identification of these func-
tional relations between context spaces.
4. Context spaces may be independent discourse constituents or may
be dependent subconstituents of other context spaces.
5. Interpreting a current utterance requires identifying the currently
relevant discourse contextthe particular context spacein light of
which it is to be interpreted ... .in the course of conversation
various preceding context spaces are brought in and our of the
foreground, depending on their influence in the generation and
interpretation of subsequent utterances.
6. Understanding the point of a speaker's utterances partially
depends on distinguishing between the importance of the various
items mentioned by a speaker. 7
7. The processing of discourse is thus focused processing'. Each
utterance is interpreted in light of the currently relevant discourse

context to which it is relatedthe context space and its items that
are most influential at that point in a discourse.
8. Among the mechanisms that facilitate focused processing are dis-
course expectations, which allow conversants to predict expected
kinds of utterances (and the context to which they will relate) on
the basis of utterances that have already been generated. Dis-
course expectations allow mutual knowledge of what type of con-
versational move is likely and to which particular section of
preceding discourse it will be related (Reichman 1985, 31-32).
Conversational moves refer to a list of objectives which utterances
fulfill in relation to other discourse elements in a conversation. Some objec-
tives categorized as conversational moves involve: presenting, supporting,
and challenging a claim (including counter-challenging); resuming a previ-
ously interrupted topic; and asking for support. Participants in conversation
perform conversational moves to accomplish their objectives. These moves
are signaled in conversation by clue words. Clue words signal that a
context space boundary point has been reached; and simultaneously they
specify the kind of shift (the kind of conversational move) about to take
place. The following figure shows some conversational moves and the
common clue words that are associated with them.
Conversational Move Clue Words
Support Because . ; Like . ; Like when . .
Restatement and/or conclusion So . .
Interruption Incidentally . ; By the way . .
Return to previously interrupted context space Anyway . ; In any case . .
Indirect challenge Yes/Right but . ; Except, however, . .
Direct challenge (No) But . .
Table 2 (Part 1 of 2). Conversational Moves and Clue Words

Conversational Move Clue Words
Subargument concession All right/Okay but. .
Prior logical abstraction But look/listen/you see . .
Further development Now ...
Table 2 (Part 2 of 2). Conversational Moves and Clue Words
Some conversational moves are shown in the following example:
Utterance Conversational Move
And, of course, what's made it worse this time is the British army moving in. And, moving in, in the first place, as a police force. Its almost a Vietnam. Assertion of a claim
But, all within Northern Ireland? Challenge
All within Northern Ireland. Moving in as a police force, being seen by everybody as a police force that was going to favor the Protestants. Itd rather be like Syria being in Lebanon, right? Clarification Logical abstraction
I dont know enough about it to know, maybe. Concession
TheresWhere theres a foreign police force in one country. I mean, when you say its like Vietnam, I cant take Vietnam. Vietnam is North Vietnam and South Vietnam. Further development/indirect challenge
No, I meant war. You know, moving in and saying were a police action and actually fighting a war when you got there. Restatement
Oh, well, thats Syria, that obviously Syria, right? Who are implicitly supportingnot supportingcause actually its very similar in Lebanon, right? You have the Catholics and the Moslems. Thats right, thats Lebanon. Interruption/request for support
I suppose, yes. Concession
You have the Catholics and the Moslems, and then Syrias coming in and implicitly supporting the Moslems, because Syria itself is Moslem. Now, England is Protestant? Retum/direct challenge
Table 3. Analysis of a Conversation through Moves
Reichmans work on conversational moves continues to describe the
elements of moves, which includes type, derivation, goal, contextual-
function, speakers, status, and focus. Her analysis basically amounts to
fancy language parsing, which is a grammar that computers understand.

The theory of conversation and rules that guide conversation are
vague and imprecise currently. Unfortunately, the grammars that computers
use currently (context-free grammars) cannot accurately process human con-
versation, and context-sensitive grammars are beyond current technology.
Current research, however, addresses communication theory that
may someday lead to language parsing that enables conversation occur
between people and machines. Chapter III of this thesis examines research
in natural language processing and explores a computer-conversation model
for interactive online help documentation.

Centuries of advances in technology that increased the spread of
information and the mobility of people has broadened the scope of commu-
nication theory. Chapters I and II analyzed rhetorical and conversation
theory for human communication. Humans have always communicated,
formally and informally, through speeches, debates, books, drama, poems,
and everyday talk. Communication theory, however, gains new dimensions
when humans and computers interact through language.
This chapter compares communication between people and com-
puters with communication among people. It investigates computer conver-
sation in terms of the elements of the ethos, pathos, and logos of participants
and in terms of the structure and rules for processing from the conversation
model. These elements are uniquely restricted by the physical, cognitive,
and topic features of the environment in which computer conversation
occurs. The plans and goals of the user and the knowledge and capabilities
of the system define all interaction in the environment. Common language is
the necessary feature in directing this interaction. The computer-
conversation model explored in this chapter addresses the role of the com-
puter as advisor through online documentation known as system help.

The environment in which computer conversation occurs naturally
with users is an idealistic knowledge-based expert system shell. It is a
system comprised of a set of programs that manages user/computer inter-
actions. This system:
Builds and updates a profile of the user each time a user works with it
Has rules that circumscribe different levels of users
Automatically interprets typed text semantically and responds to that
text conversationally
This thesis does not define the computer-conversation environment;
it just presents how the environment will work. This thesis assumes that
communication between the computer and the user develops through an
application that runs under a knowledge-based manager.
The foundation for a computer-conversation model rests on a foun-
dation of rhetorical and conversational communication theory. Figure 4 on
page 38 shows a computer-conversation model that looks much like the con-
versation model presented in Chapter II (Figure 3 on page 20). Derived
from the communication models of classical and modern rhetoric and from
conversation theory, the computer-conversation model combines the same
elements and follows many of the same rules of those models.

Figure 4. The Computer-Conversation Model
The Communication Triangle for Computer Conversation
The relationship of pathos, ethos, and logos in traditional commu-
nication models is still embodied in the computer-conversation model
through interaction of the participants (computer and user) in the domain of
the subject. When a computer is involved, each part of the conversation
model changes some. Since human/computer interaction progresses like con-
versation, this thesis examines elements of the model for both participants.
Ethos of the User
In conversation, participants negotiate positions of equality by estab-
lishing and switching roles. In human/computer interaction, a users roles as
performer in completing a task functions and learner of the system equally
with the roles of the computer as both worker and advisor. This balance

maintains self-esteem for the user and fulfills the goal of the interaction.
Ethos deals with motivation and behavior in humans. Users of computer
systems are strongly motivated to accomplish work. To facilitate this goal,
users follow the set of conventions that they and the computer know. A
computer knows only that a user followed the rules or didnt. For example,
by using only legal commands and syntax, a users behavior informs the
system that he or she is correct and cooperative.
The computer interprets the ethos of users based on their behavior.
Levels of expertise reveal how much users know about the subject. In a
computer-conversation help system, the computer judges the character of
users through whether or not they pressed the right keys. Researchers have
studied and analyzed the salient traits for users, usually at the extremes of
naive and expert users.
1. New users have trouble learning to use computer systems. These naive
users generally learn menu-driven systems most easily. Precisely fol-
lowing menus helps them navigate through the system, since users that
are new to computing often create less appropriate mental models.
2. Experts, on the other hand, have discovery skills that allow them to find
answers to questions (Carroll 1987, 19). Menu systems are inappropriate
for experts, since working through menus inhibits discovery and slows
activity for expert users.

Between these extremes of naive and expert users, however, is the
largest user group. Todays users are familiar users who possess varying
levels of experience in working with computer systems. In learning a new
application, most users prefer to rely on exploration rather than on detailed
information. Annette Bradford describes these intermediate user groups:
Computer Limited: A user who is familiar with a computer system only in so
far as it fulfills a limited need. For example, some users buy computers
and use them as dedicated word processors. They deal with computing
only as much as is necessary to accomplish their task.
System Novice: A user familiar with one operating system who is learning
System Sophisticated: Two kinds of userssystem novices who have sophisti-
cated command of several operating systems or users who were system
novices on a second system and have learned that second system fairly
well (Bradford 1987, 3).
These different groups of familiar and experienced users have different infor-
mation needs and different cognitive styles, which must be considered in
developing a model for system help.
User's Pathos for the Computer
What the user understands about the computer, called a conceptual
model of the system, is an important component of user/computer inter-
action. The conceptual model of the computer that a user has, right or
wrong, is his or her picture of how the computer responds and what it
means in user/computer communication. This picture of the computer
affects every interaction in computer conversation.

Thomas and Carroll (1981) found that people almost always try to
learn new things by making use of past learning. New concepts are
typically explained in terms of old conceptsat least initially .... An
existent knowledge structure is loaded into memory and used as a
structural template for further learning (Bradford 1987, 7).
Research shows that new users have trouble learning to use com-
puter systems. The skills of users at most levels tend to stagnate at a medi-
ocre level because the motivation and time that it takes users to learn a
system conflicts with their motivation to use the system to do work (Carroll
1987, 14). Users construct a conceptual model (accurate or not) of the first
system they work with, which influences how they develop subsequent con-
ceptual models of other systems.
Users with a strong conceptual model of a system more accurately
predict interface actions (moves) that will achieve their task-level goals.
Also, they recover more easily from errors for that system than do users who
lack a strong conceptual model of a system. Users later apply their estab-
lished conceptual model of one system to their activities with the next system
they use. When the next product is conceptually different in structure and
language than the original, users have trouble working within and between
different products. (This observation highlights the need for a consistent
user interface. The goal of a consistent user interface through a computer-
conversation model for online computer documentation is the purpose of this

For example, in testing the influence of prior knowledge about other
systems to learning a new graphics system, Miller et al. (Miller 1987, 6)
selected two groups of subjects with different backgrounds in using com-
puters. Half of the subjects previously worked with only traditional
command language systems. The other half had at least six months experi-
ence using Apple Macintosh. Because of Macintoshs graphical orientation,
the researchers expected the Macintosh subjects to learn the new system,
VSTAT, more quickly. (The new product contained similar icons to
Macintosh, but its information was mapped differently.) The Macintosh
model of interaction was so overwhelming for those users that they could not
conceptualize the flow of information in the new system. In contrast, the
command language subjects had fewer problems learning VSTAT because
they did not have a strong and competing conceptual model of how it might
work (Miller 1987, 8).
In applying their conceptual models of the system, users reveal the
pathos, that is, their understanding of how the computer thinks, relates, and
operates to guide their interaction with the system. The ability of a system to
then help a user create and maintain appropriate conceptual models depends
on flexibility in the representation and reasoning processes to be used
(Miller 1987, 9) by the systems advisor. In this thesis, the advisor is the
computer products online documentation (Help).

Logos for the User
The types of knowledge that a user needs to participate in inter-
action with the computer through language include:
Knowledge of the computer version of the domain, including knowledge
of the domain and knowledge of the workbase version of the domain
Knowledge of the problem
Knowledge of system operations including its capacity for natural lan-
Knowledge of the physical interface
Knowledge of interface dialogue (Zeidner 1986, 272)
Computer conversation allows participants to explore a subject, which is
static but which evolves through interaction, in the same way that conversa-
tion does.
An important feature of user/computer interaction, however, is that
users probably accommodate some of the limitations inherent to computer
grammars. A good example of this extended cooperative effort by users to
accommodate a computer's limited vocabulary and syntax is evident in com-
puter adventure games like Infocom's Zork series. Typically, a user seeks to
map an unknown land and solve a mystery without getting killed. By
asking the right questions and moving through time and space, a user pro-
ceeds without dying. Dialog between the player and the computer is surpris-
ingly conversational. A player can include several sentences on one input
line by separating them with the word then or with a period, for example:
North.Read the book.Drop it then bum it with torch. Interacting through the
familiar structure of conversation encourages players to explore the system.

Ethos of the Computer
The ethical character of the computer in human/computer inter-
action is impeccable. The computer is honest, reliable, logical, and
hardworking. The formal logic that governs the development of computer
products often blinds the system designers to users needs. The lack of com-
munication between human factors engineers and system designers too fre-
quently produces inappropriate, simplistic, or incorrect user models (Zeidner
1986, 272).
Computer's Pathos for the User
Hammond and Barnard propose theoretical models of the user that
the computer builds while interacting with the user. The computer, then,
uses the model to converse with the user. Based on cognitive science, these
models include information models, goal models, dynamic models, and per-
formance models. Figure 5 on page 45 shows the computer software archi-
tecture, which Halping and Sheppard developed, for a system that supports
interactive dialog between user and computer.

Figure 5. Architecture for a System to Support User/Computer
Dialog. (Copied from Zeidner 1986, 273)
The components for this architecture:
interpret a command or request based on . understanding of the
user characteristics and goals and of the current state of the world.
The software then directs the sensors or effectors as required.
Data or knowledge gained by the system are used to update the
world model. The software draws on the constantly updated model
of the user to determine when, in the context of user characteristics,
goals, task context, and so on, there is meaningful change in the world
and calls on the output formatter to transmit the relevant informa-
tion to the user (Zeidner 1986, 274)
For computer conversation, the computer will build the user model.
It will track files on the users disk, which files the user accesses, which com-
mands the user issues (successfully and unsuccessfully), and the syntax of

conversations with the user. To create an accurate user model, the computer
must be aware of itself.
Logos for the Computer
The computer must also have knowledge to interact with a user. A
knowledge-based help system, for example, must have four domains of
1. Knowledge of the problem
2. Knowledge about communication processes
3. Knowledge about the user
4. Knowledge about the most common problems which users have in using
a system and about tutorial intervention (Zeidner 1986, 266)
Logos for the computer includes all of its knowledge about itself and
about the plans and goals of the user. Knowledge about itself is essentially
how it is programmed. A knowledge-based system runs basically on two
kinds of rules: (1) rules that apply to specific situations; and (2) meta-rules
that apply to the specific rules. Meta-rules apply when, for example, the
computer could invoke any of several rules for a particular situation. Meta-
rules are the front-end rules that decide which of a set of specifically pro-
grammed rules applies to the current situation.
Knowledge about the user is the model of the user that it creates to
map the plans and goals of the user. For example, when a user enters the
command copyfile, the computer anticipates that the user will also provide
information about the source and target files to copy from and to.

Conversational Features of User/Computer Dialog
Conversation shares its theoretical base with computer conversation,
which, like conversation, has participants, discourse, context, and moves.
Computer conversation also obeys maxims and rules. The following list
shows how these elements change in computer conversation:
1. Participants embody the roles of ethos and pathos described in classical
and modern rhetoric. A computer-conversation model has essentially two
participants, and one is a machine.
2. Discourse in this model is usually the text that appears on the display
screen. It is static and contains the logos of the subject described by the
classical and modern communication models. As in conversation, the
discourse subject evolves, which allows meaning to become clear to par-
ticipants. Language for a computer is formal and logical, while language
for people is less rigid in formality and logic. To present discourse in the
computer-conversation model, todays computers use text and graphic
displays with an occasional beep. People enter their discourse from the
computer keyboard (or with a light pen or mouse). Exchanges then
progress much like those among people.
3. Context is the situation that occurs between participants in the computer
environment. Instead of depending on the feedback of participants, the
context of conversation in a computer-conversation model depends on the
plans and goals of the user and on the capability of the computer system.
A user converses with the computer to accomplish goals that relate only
to the context of the computer environment. The computer has no world

knowledge, since the computers world is itself; so the context for the
computer exists only its own domain of knowledge and its knowledge of
the user.
4. A move, which describes the purpose of an utterance in conversation, is
essentially a unit of work in human/computer interaction. For example,
the types of utterances in conversation that moves describe are either ini-
tiation or continuation of a topic. The same is true for human/computer
interaction. In a study of interaction between users who asked questions
and experts who answered the questions, Pollack discovered that only
two turns were taken in most of the dialogs (Pollack 1985, 156).
The following exchange shows moves for logging on and copying
a file. Participants follow a convention for starting the interaction; they
next interact to copy a file. They make moves of request, response, chal-
lenge, and support.
Logon, Logoff, or Msg
logon 373
input source filename
input target filename
file already exists, enter rep to replace
5. Maxims in computer conversation are powerful assumptions about the
discourse that occurs in human/computer dialog. A computer environ-
ment assumes perfect compliance with the conversational maxims

described on page 24. By nature, computers tell the truth, and, in so far
as their programming allows, cooperate and generate pertinent informa-
6. Rules, which guide how participants behave to establish and interpret
meaning, are laws in computer conversation. If a computer is pro-
grammed to follow a rule, it will always follow the rule. The rules for
information processing that appear on page 30 also guide the behavior of
talk by participants in computer conversation.
Environment Restricts Computer Conversation
In computer conversation, as in all communication, meaning is
situation-bound depending on the context of the exchange. The key concept
in the computer-conversation model is that human/computer dialog is "like
conversation. Communication between animate and inanimate partners will
always have restrictions. Computer conversation, like normal conversation,
is restricted by 1) the physical environment of the interaction, 2) the cognitive
environment of the participants minds, and 3) the topic of the exchange.
The Physical Environment
The physical aspects of any communication involve participants, the
context, and the time and place of the exchange. In everyday conversation,
the elements of the physical environment describe the who (agent), when and
where (scene), and how (agency) of Burkes pentad (see page 12). For

example, in the movie Casablanca, two men in a smoky bar converse. One
participant sits at a piano. The physical environment for the conversation is
the time, location, circumstance, and physical characteristics, such as gaze
and intonation, that participants display.
In computer conversation, the physical environment has similar fea-
tures: the participants (who), the ergonomics of the work environment (when
and where), and the computer hardware and software screen design (how)
available for the exchange. For example, the user and computer sit face-to-
face and they interact through the keyboard and the screen. The physical
environment of computer conversation is part of the interface, the means by
which people and computers communicate with each other.
The Cognitive Environment
The cognitive environment of communication describes the purpose
of the exchange and involves the thought processes of participants. It
explains the why (purpose) in Burkes pentad for communication. In the
movie Casablanca, Humphrey Bogart is reminiscing; his purpose for con-
versing with the piano player is to relive a memory.
In computer conversation, the cognitive environment contains the
capabilities of the system software and the plans and goals of the user. For
example, computers are good at high-speed arithmetic; they distribute infor-
mation and process information tasks such as form-filling, tutoring, and
accounting. Users converse with computers to accomplish these information-
processing tasks.

The cognitive environment of the computer is also part of the inter-
face, the face of the computer that talks. Computers generally take on a role
of worker or advisor, depending on the context of the conversation. In the
example of human/computer dialog on page 48, the user wants to copy the
contents of one file to a target file. As worker, the computer tries to copy
the file. As advisor, the computer tells the user that the target file already
exists and explains how to copy over the existing file. (The computer-
conversation model for online help functions in this thesis describes the role
of computer as advisor.)
Conversation between people covers literally all topics. A topic in
everyday conversation is the subject in communication theory and the
what (act) in Burkes pentad. In the movie, Bogart says, Play it again,
Sam. Sam, the piano player, responds to Bogarts request, which lets
Bogart relive his memory.
Exchanges between people and computers generally comprise only
two acts, which are topic situations in computer conversation; they are 1)
performing tasks and 2) moving in and between tasks. As mentioned above,
users make requests and computers respond to complete the tasks. In the
copy example, the user essentially says, copyfile filel to file2 (Play it
again, Sam) and the computer attempts it. Performing a task in computer
conversation is usually a straight-forward procedure of issuing a command
or selecting from a menu.

The second topic of moving in and between tasks refers to the ability
of users to make appropriate moves to continue the conversation. In the
movie, Sam responds appropriately by playing the requested song on the
piano. In the copyfile example, the computer tries to copy the first file to
the target file but discovers that the target file already exists. The user must
now move appropriately to complete the task. This requires that the user
have specialized knowledge about the computers cognitive environment to
accommodate the twists and turns of computer conversation.
Common Language
Common language is the necessary feature in directing
human/computer interaction. This thesis suggests that the most flexible rep-
resentation and reasoning process for dialog between a user and the com-
puter is a computer-conversation model. The best way to learn, of course, is
in a familiar language that you can understand; and the best place to start
natural language processing for learning computer systems is with online
help. Help is where language and learning work together to keep a user
using the system.
Henry Ledgard tested the hypothesis that an interactive system
should be based on familiar, descriptive, everyday words and legitimate
English phrases (Ledgard 1981, 89). The experiment had three levels of
users (inexperienced, familiar, and experienced) using two text editors,
which were identical semantically in terms of editing power but were dif-
ferent syntactically (Ledgard 1981, 90). Experimenters redesigned the

surface syntax of a commercial editor so that the commands more closely
resemble English phrases, which resulted in far better performance. In this
experiment, users made no distinction between syntax and semantics
(Ledgard 1981, 102). This indicates that language designers for
user/computer interaction must be concerned with surface syntax as with
functional features, if they hope to design a product to optimize perform-
Humans can clearly adapt themselves to difficult systems including
ones that have been poorly human engineered. The question is, at what
cost? (Ledgard 1981, 104). A good software product seeks to maximize
human performance on an information processing task (Zeidner 1986, 249).
In the computer environment help system, language and learning of a pro-
duct's online help is key to the products success. The product and the
system that supports the product must support user interactions. The lan-
guage with which a computer talks to and responds to users must have the
same format and style as that in which the user converses normally. A
natural language is easy to remember.
System Help Functions
The specialized knowledge needed to navigate through a computer
environment is generally found in documentation about the computer soft-
ware. This information used to appear exclusively in books but is rapidly
being moved to online display. As computer documentation about system
software is moved from printed (hardcopy) to online (softcopy), computers

need to provide users with access to the information when they need help.
Most users need help to navigate through a system at some time. (The
computer-conversation model proposed here directly addresses this form of
computer documentation known as the system help function.)
Help functions are designed to educate the user on systems capabil-
ities, explain error messages, and assist in creating legal entries (Zeidner
1986, 266). They provide information to users trying to navigate through
complex levels of task activity and remind users of information they have
already learned but seldom use (Bradford 1987, 5). System helps replace
traditional users guides and reference manuals. The screens full of informa-
tion, called panels, created to replace the books, however, have historically
been little used and not very useable, but this is changing as walk up and
use becomes a goal in system design (Bradford 1987, 6). The level of
sophistication for system helps is developing this way:
1. No help is usually represented by a menu-driven system. The informa-
tion on its panels is designed to guide users through the task activities, so
no help is needed.
2. Online books, also known as dumb help, are scrollable, read-only infor-
mation. They are often created from data sets originally intended for
printed media, but are provided to customers in machine-readable form
for either the display screen or local printing (Bradford 1987, 6). An
example of an online book is the documentation for the Unix1 Man
1 Unix is a trademark of AT&T.

3. Static help divides reference information into help for commands, mes-
sages, panels, and fields. Typically, static help has little sense of the user
and it is usually destructive because it disrupts the users activity by
replacing the panel he or she is viewing with the help text. It can also
appear in a window, which does not cover the part of the screen that
motivates the user to ask for help (Bradford 1987, 6). Turbo Prolog2 has
static window help.
4. Static help with a partial user model presents static help to its typical
user. The Lotus1 program has this type of help. Lotus provides a help
window that asserts the help a user needs, based on its static, pro-
grammed model of a typical user.
5. Layered static help for different users groups information into layers of
complexity. Users decide what level of information they need, based on
how they view their skills (novice, experienced, expert) and based on their
goals. But people are notoriously bad at giving accurate descriptions of
their own information needs (Carroll 1987, 19) and advice seekers do
not always have well-formed plans about what they need to know
(Carroll 1987, 26). This dilemma is evidenced in the sample dialog on
page 61.
Still, help information should be available in a layered structure,
perhaps a tree structure that is accessible for top-down searches and spe-
2Turbo Prolog is a trademark of Borland Incorporated
3Lotus is a trademark of Lotus Development Corporation.

cific queries; and users should be given control of the level of detail that
they view in helps (Zeidner 1986, 266). Carroll cites research by Fikes
and by Stevens, Collins, and Goldin that describes these layers as mul-
tiple representations, which are necessary to provide enough basis and
flexibility to support effective advising (Carroll 1987, 17).
6. Context-sensitive help is able to sense the input field or the situation
that motivated the user to ask. for help. Expert system help facilities
might even track a series of user errors and then offer help based on a
diagnosis of need (Carroll 1987, 17).
An ideal help system contains layered information that users
obtain at different levels of expertise. It is a help that also allows users
to generate more information. Context-sensitive help may develop into
an ideal help, which builds a model of the user as he or she progresses
through the system. For example, if Lotus built a user model, it would
know not to ask if a user needs help correcting typing errors, when it has
seen the user backspace to correct errors himself. By contrast, if Lotus
knows that the user has never before asked for a range before, it can help
the user learn how to use ranges.
In addition to having levels of information, context-sensitive help
can be passive or active. Passive help waits for the user to ask for help
and then tries to figure out what the user needs. A knowledge-based
system may have its passive help facility built in so that, when it asks for
information, it has the ability to explain why it asked. For example, if a
the computer-diagnostic application for a nuclear reactor asked if there is

steam in the core, it could also respond (if asked), I already know that
the cooling system has low pressure; therefore, steam in the core means
that a valve is open. Active help interrupts user activity to provide advi-
sory comment on his actions (Carroll 1987, 16).
The Computer as Advisor
Help functions represent knowledge about a system or program
application that users need to do their work. To learn about a system, users
want the same things from help functions that they expect from traditional
learning situations. In the limited domain of the computer environment,
computers should therefore teachthat is, guide and trainusers who need
information to continue working with the system.
In academic settings, teachers use different methods to provide guid-
ance and training for a subject. A teacher can work collectively with many
students in a class or work individually with only one student. Spreading
information collectively through lectures, assigned readings, and exercises
gives structure to a subject and saves time and money in spreading informa-
tion. Teaching a student individually through focused narrative, answers to
questions, and exercises at the students pace gives depth and focus to that
subject for the individual student. However, one-on-one education is expen-
sive and time consuming.
Seen on a continuum, collective education appears on one side. It
consists of formal rhetoric that aims at a fictional audience, an audience that
is contrived based on what the teacher believes about the audience and what

he or she believes the audience should know. Collective teaching arranges its
subject carefully and reaches many students at a low cost.
Individual education appears on the other side of the continuum. It
consists of informal rhetoric that aims at a uniquely defined audience of one
person. Individual education proceeds much like a conversation with com-
munication that flows episodically and is negotiated by participants in the
current context. The difference in learning from these teaching methods is
basically breadth versus depth. Collective education usually offers concep-
tual information and a moderate level of detail about the subject to many
people (breadth). Individual education offers information that is tailored
and spontaneous to an individuals learning style and potentially offers a
deeper understanding of a subject for the individual.
Programming a computer to teach involves merging the benefits of
collective and individual instruction to design a human/computer interface
for computer documentation (system help) that is easily accessible, complete,
personal, inexpensive, and non-disruptive. To accomplish this, the informa-
tion about the subject must be layered so that users receive only information
that they need. A computer gives information to a user who knows how to
converse with it.
Today, computer users converse with a system based on prior know-
ledge about systems. Users work with many different software products
(systems and applications) to accomplish their information processing tasks.
Ideally, differences in these products are not apparent to users; but, usually
that isnt the case. Responding to the challenge of establishing the set of

concepts that guide behavior between people and computers is the goal of
the computer-conversation model.
The factors that influence the effectiveness of the modelhow
quickly and effectively a user learns a system and performs tasksare:
The users conceptual model of the computer system
The level of the users computer experience
The computers model of the user
The computers knowledge of the domain
The language of computer conversation
To explore a conversation model for developing computer documen-
tation, we need first to understand:
1. How the different needs of users and layers of contextual information
affect the way information for computers is presented
2. The linguistic devices that would adapt the conversation model to cre-
ating the human/computer interface
A help system based on the computer-conversation model integrates
time and effort spent on learning with the actual use of the system. In
order to store information and to act on the basis of that information, an
advice-giving system must exhibit three types of knowledge: general skills
about tutoring and natural language, domain knowledge that incorporates
task information, and a user model of the person who is interacting with the
system (Carroll 1987, 17). A computer-conversation model may provide
the link across systems to increase efficiency for todays multi-activity com-
puter users.

This chapter presents dialog between a typical user and system help
to reveal inadequacies in the design of online information today. Next the
chapter shows how these inadequacies could be corrected through a
computer-conversation model. The computer-conversation model is a the-
oretical model that has rhetorical and conversational features which enable
users to quickly comprehend system help information.
The following example shows a system devoid of the features
inherent in a computer-conversation model. It doesn't consider the user,
fulfill the user's expectations, or anticipate errors; it is generally frustrating.
The program described here, called Publish, is modeled from a real pro-
grammed application. Its name was changed to protect the guilty. Publish
is a menu- and command-driven interface to a word processing program. It
lets you format a file to print, print a file, or edit a file. It formats for a
variety of printers and for a 3270 display terminal. Its help system is fairly
sophisticated, featuring context-sensitive, field-level help. When you fill in a
field incorrectly, Publish displays a panel with help text that describes the
field and what goes in it. You can fill in the field on the help panel. When
you return to the original menu, the field is filled in correctly there. The
following real-life scenario shows how Publish and its online help interact
with a user. (Yes, it happened to me.)

Lynn is currently updating a 360-page reference manual that con-
tains some formatting errors. The online files for the reference book are kept
on a disk in the system that is not Lynns main disk. (The main disk is
called A; the subsidiary disk is called Q.) Users define and store files
using a naming convention of three words: filename, filetype, and filemode,
filename is a title of eight or fewer letters made up for the file,
filetype is the category of the file. Publish assumes that it is script for
printing, but the filetype can be letter or text, too.
fileinode is the physical place, called a disk, in the computer where you store
the file. Publish assumes the filemode is A, your main disk, but
you can have other disks.
Lynns control file contains several chapters, which are imbedded
when Publish formats it. Usually, Publish lists the formatting errors that it
encounters two ways: 1) on softcopy in the users mail and 2) on hardcopy
as the last pages of the printed document. Lynn wants to print only the
error messages, not the entire book and she does not care about softcopy of
the messages. Here is her dialog with the computer:

P: Enter filename filetype
P: Document not found
P: Enter filename filetype
P: Document not found
Menu appears.
Notice that Publish does not require the filemode.
Publish submits the job to the system.
Lynn has not seen this message before. The control file
has many imbedded parts. She thinks, perhaps one is
lost. No, the computer says it cannot find CONTROL. Sus-
pecting a typing error, Lynn tries again.
Menu appears.
Lynn notices that Publish does not ask for the filemode.
Lynn thinks that Publish wont print a file that resides on
her Q-disk and she consults a co-worker who suggests
using Publish in its command form.
L: PUBLISH CONTROL SCRIPT Q Publish accesses the file and prints the book. Unfortu-
(3812) nately, Lynn wanted only the error messages, which are
the last pages that print after the book; but Publish
printed the book and the list. Lynn keeps the last pages
and discards the 360-page book.
Table 4. Typical Human/Computer Dialog (Part 1)
Lynn works on the errors listed in the pages that she printed and
then wants to check her corrections before printing the book. Not wanting
to print the book again to get the new list of messages, Lynn enters the
command so she sees the menus.

Utterance Comment
L: PUBLISH CONTROL SCRIPT Q Menu appears that Lynn has never seen.
P: F = Format to the printer
T = Format to the terminal
E = Edit
L: F Menu from last failure appears. Lynn returns to the
previous menu and makes another selection.
L: T
P: Publish formats the book and displays the large file on
Lynn's terminal for her to scroll. The online pages
are different from actual hardcopy because of differ-
ences in the formatting devices.
L: Quit Lynn doesnt want to see this file but views it as a
necessary evil to obtain updated error messages. She
quits from the displayed file and accesses the softcopy
error message file on her A-disk.
Table 5. Typical Human/Computer Dialog (Part 2)
Through all of the dialog described above Lynn never did get exactly
what she wanted, a hardcopy list of errors for a file located on her Q-disk,
without printing the 360-page book. She settled for using the softcopy list.
Printing the entire book was an expensive lesson, since the existing
field-level help contained the information Lynn needed to achieve her goal.
Lynn had legitimate values in all fields, so none were flagged. She had to
know which field to change. The original Publish menu contained dozens of
fields, for which Lynn could have changed the default values. And, the help
contained descriptions of values for all of the fields on the menu. But Lynn
didn't know the particular value to change to get only hardcopy messages.
Ideally, Lynn should have entered the command: PUBLISH
filename filetype filemode (3812 MENU). This displays a menu on which
Lynn should have changed terminal to no in the message field. This tells

Publish not to print the document but still print the messages. Lynn may
never have discovered this solution.
One interesting point is that users are notoriously poor at judging
their own level of expertise. Since Lynn works with the computer at her job
every day, she does not consider herself a naive user, even when she uses an
application for the first time. This is why the ability for the system to create
an appropriate model of the user is essential to the computer-conversation
Computer Conversation for the Familiar User
The preceding scenario would not have occurred, if, originally, Lynn
could have entered: Format this book and print only the error messages.
Familiar users generally know how to navigate through the system and what
tasks they can perform.
Computer Conversation for the Naive User
Lynn, as a familiar user, knew what she wanteda printed copy of
the error messages. But suppose she printed the document unaware of the
formatting errors? What should the system do? Ideal help might initiate
this conversation:

Participant Computer: Utterance This document contains serious formatting errors. Do you want to print it anyway?
Naive Lynn: No. What is wrong?
Computer: The errors start on page 223. Do you want to print a range of pages to see the printed errors?
Naive Lynn: What is a range?
Computer: A range describes. .. You could print these pages to see the problem. Or you could print only the error messages and then correct your file. Do you want to print a range of pages?
Naive Lynn: No.
Computer: Do you want to print only the error messages?
Naive Lynn: Yes.
Table 6. Computer Conversation for the Naive User
In dialog with a naive user, computer-conversation help teaches the
user how to continue using the system. It defined a term that Lynn doesn't
know and explained what steps she can take.
Computer Conversation for the Expert User
A user with more experience than Familiar Lynn and Naive Lynn
may opt to print nothing but want to correct errors immediately. Computer
conversation between the computer and Expert Lynn may proceed like this:

Participant Utterance
Computer: This document contains serious formatting errors. Do you want to print it
Expert Lynn: No. Let's correct the errors now.
Computer: Error 1: List item found outside list on line 30 of CONTROL SCRIPT.
Expert Lynn: Edit CONTROL SCRIPT.
Table 7. Computer Conversation for an Expert User
The Role of Professional Writers in Computer Conversation
Applying the computer-conversation model as just described demon-
strates the ultimate help. We dont have that now. But researchers from
many fields are working to produce natural language processors that contain
the raw speed and sophisticated parsers needed to build appropriate user
models and thus develop cooperative user/computer interaction. Program-
ming interactive systems is the job for system programmers. Designing the
appropriate interface is the job for professional communicators trained in
many disciplines, such as linguistics, human factors, and computer science.
Even though the ultimate isnt available now, professional writers need to
keep the computer-conversation model in mind when designing and docu-
menting information for user/computer interaction.

Andriole, S. J. Interactive. Computer-Based Systems. Petrocelli Books,
Inc., New York, 1983.
Booth, W. C. The Revival of Rhetoric. PMLA, LXXX (May, 1965),
Booth, W. C. The Rhetorical Stance. College Composition and Commu-
nication,, 14 (October, 1963), 139-145.
Bradford, A. N. Levels of Sophistication in Computer-based Tutorials:
Writer and User Variables. IEEE, 1983.
Bradford, A. N. Writing and Using Online Tutorial Information. Paper
presented at The Conference on College Composition and Communi-
cation., March 30, 1984, New York, NY.
Bradford, A. N. Using Online Information with Your Product: A Deci-
sion Process for the Information Planner. Paper presented at Human
Factors Symposium, October, 1985, Dayton, OH.
Burke, K. A Grammar of Motives. University of California Press,
Berkeley, CA, 1969.
Burke, K. A Rhetoric of Motives. University of California Press,
Berkeley, CA, 1969.
Carroll, J. M., and McKendree, J. Interface Design Issues for Advice-
giving Expert Systems. Communications of the ACM, Volume 30
(January, 1987), 14-31.
Clarke, D. D. Language and Action. Pergamon Press, Elmsford, NY,
Connors, R. J., Ede, L. S., and Lunsford, A. A., editors. Essays on Clas-
sical Rhetoric and Modern Discourse. Southern Illinois University Press,
Carbondale, 1984.
Cooper, L. The Rhetoric of Aristotle. Prentice-Hall Inc., Englewood
Cliffs, NJ, 1932.
Corbett E. Classical Rhetoric. Oxford University Press, New York, NY,
Craig, R. T. and Tracy, K., editors. Conversational Coherence: Form,
Structure, and Strategy. Sage Publications, Beverly Hills, 1983.
Goodwin, C. Conversational Organization. Academic Press, New York,

Kinneavy, J. L. A Theory of Discourse. Prentice-Hall, Inc, Englewood
Cliffs, NJ, 1971.
Ledgard, H. F. Directions in Human Factors for Interactive Systems.
From the series Lecture Notes in Computer Science, Berlin, West
Germany: Springer Verlag, 1981.
Lindermann E. A Rhetoric for Writing Teachers. Oxford University
Press, New York, 1982.
Litman, D. J. Understanding Plan Ellipsis. Natural Language, 1986,
Martin, J. Design of Man-Computer Dialogues. Prentice-Hall, Inc.,
Englewood Cliffs, NJ, 1973.
McLaughlin, M. L. Conversation: How Talk is Organized. Sage Publica-
tions, Beverly Hills, 1984.
Miller, J. R., et al. The role of the system image in intelligent user assist-
ance. Microelectronics and Computer Technology Corporation, Austin,
TX, 1987.
Murphy, J. J., ed., A Synoptic History of Classical Rhetoric. Hermagoras
Press, Davis, CA, 1983.
Ochs, E., and Schieffelin, B. B. Acquiring Conversational Competence.
Routledge and Regan Paul, London, 1983.
Pearce, W. B., and Concklin, F. A. A model of hierarchical meaning in
coherent conversation and a study of indirect responses. Communication
Monographs, 1979, 46, 75-87.
Pollack. M. E. Information sought and information provided: An empir-
ical study of user/expert dialogues. In Proceedings of CHI 85 Human
Factors in Computing Systems (San Francisco, CA, April 14-17). ACM,
New York, 1985, 155-159.
Reichman, R. Communication Paradigms for a Window System. In
User Centered System Design, edited by D. Norman and S. Draper,
285-313. Hillsdale, NJ: Lawrence Erlbaum Associates, Publisher, 1986.
Reichman, R. Getting Computers to Talk Like You and Me. The MIT
Press, Cambridge, MA, 1985.
Rosenblatt, L. M. The Reader, the Text, the Poem. Southern Illinois
University Press, Carbondale, 1978.
Rubenstein, R. and Hersh, H. The Human Factor. Digital Press,
Burlington, MA, 1984.
Sanders, R. E. Cognitive Foundations of Calculated Speech. State Univer-
sity of New York Press, Albany, 1987.
Soderston, C. An Experimental Study of Structure for Online Informa-
tion. Paper presented at International Technical Communication Confer-
ence, Denver, CO, May, 1987.

Steinmann, M., Jr. Rhetorical Research. College English, XXVII
(January, 1966), 278-285.
Tannen, D. Conversational Style: Analyzing Talk among Friends. Ablex
Publishing Corporation, Norwood, NJ, 1984.
Wardhaugh, R. How Conversation Works. Basil Blackwell, Inc., New
York, NY, 1985.
Wheatley, J. Language and Rules. Mouton, Paris, 1970.
Zeidner, J. Human Productivity Enhancement, Volume l. Praeger Pub-
lishers, New York, NY, 1986.