Organization in technical writing

Material Information

Organization in technical writing views from three disciplines
Ward, Barbara Jean
Publication Date:
Physical Description:
120 leaves : ; 29 cm


Subjects / Keywords:
Technical writing ( lcsh )
Technical writing ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 111-120).
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Science, Technical Communication.
General Note:
Department of English
Statement of Responsibility:
by Barbara Jean Ward.

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:
20983843 ( OCLC )
LD1190.L67 1989m .W37 ( lcc )

Full Text
This thesis for the Master of Science degree by
Barbara Jean Ward
has been approved for the
Department of
Technical Communication
by .
n. fit i

Ward, Barbara Jean (M.S., Technical Communication)
Organization in Technical Writing: Views from Three Disciplines
Thesis directed by Assistant Professor Colleen E. Donnelly
This thesis examines the need for organization in technical
writing and methods for achieving that organization. The
discussion includes concepts and approaches from three disciplines:
cognitive psychology, linguistics, and rhetoric. These disciplines
focus on the ways in which people learn material, how they use
language to communicate, and how they shape communication to reach
specific ends.
Concepts from psychology utilized here include structures
of memory such as short-term and long-term memory, network models
of memory, hierarchical structures, and schema theory. These
concepts inform us that readers need a hierarchical structure in a
technical document and that the schemata that readers have for
various types of documents and content must be taken into account.
The section on linguistics examines the role of
propositions and macropropositions in structuring meaning, the
nature of gists, the given-new contract, speech act theory, and the
importance of context. Linguistic theory indicates that the
writer's purpose in communicating information should shape the
structure of the text and that new material must be related to
material the reader already knows.
The discussion of rhetoric explores principles of classical
rhetoric and its modern applications in invention, classification
of discourse types, the role of topic-specific knowledge, and the
role of audience and social context. This discussion also examines


specific rhetorical practices, such as organizational models and
structures for paragraphing. The principles derived from rhetoric
emphasize the need to structure a communication to meet an
audience's needs.
Several general principles can be identified that are
common across the disciplines. A text must have a controlling
purpose, which should determine sequencing and level of detail.
Information presented in a document must be linked to information
the reader already has available. Readers' expectations must be
anticipated and dealt with in the design of material. Conventions
for particular types of material must be followed or it must be
made clear why they are not followed. The technical writer must
balance the need to follow expected conventions with the unique
demands of each communication situation. Structure, the writer's
intent, and the needs of particular audiences all must be attended
to in organizing technical prose.
The form and content of this abstract are approved,
its publication.
I recommend
Colleen E. Dopj^elly\^~

X. INTRODUCTION .............................................. 1
Why Is Organization Important? ........................ 3
Is Examination of Models Useful in Fostering
Organizational Skills? ............................. 4
Will Organization Still Be Important
in the Future?........................................ 6
Can Research Findings Contribute to Development of
Organization in Technical Writing? ................... 7
How We Process Information ........................... 12
Closure............................................... 16
Expectancies......................................... 17
Schema Theory ........................................ 20
Hierarchies and Storage of Information ............... 23
Methods of Creating Structures ....................... 25
Implications for Technical Writing ................... 29
III. CONTRIBUTIONS OF LINGUISTICS ............................ 39
Propositions ......................................... 41
Gists................................................. 44
Given-New Contract ................................... 47
Speech Acts........................................... 51
Context........................................... . 54
Text Continuity....................................... 57

Implications for Technical Writing .................... 59
IV. CONTRIBUTIONS FROM RHETORIC ............................. 69
Classical Rhetoric .................................... 71
Modern Rhetoric ....................................... 74
Invention............................................ 74
Classification Systems .............................. 76
Knowledge of Topic and Conventions .................. 80
Role of the Audience and the Social Context .... 81
Strategies for Developing Organization ................ 83
Organization Models ................................. 84
Paragraphing....................................... 8 6
Topic Sentences...................................... 89
Implications for Technical Writing .................... 90
V. CONCLUSIONS.............................................. 98

1. Two Hierarchies for the Same Information............... 36
2. Proposition Structure for Repair Manual ............... 61

For the technical writer, faced with the task of
communicating information to an audience, the question often is not
"what do I say?" but "when do I say it?" In many situations, the
information to be written about is available but in daunting
amounts and complexity. Technical writers must make sense of this
information for their readers, solving multiple problems of
organization. Basic questions arise: what should come first; what
should come second; what should be in an appendix; how much detail
should be introduced at any given point?
When beginning to document a computer language, for
example, the writer may find that the reader needs to understand
considerable terminology to learn the language. But explaining the
terms before explaining the general concepts may overwhelm the
reader, who has no framework for understanding the terms, and who
may be bored or frustrated by being forced to go through terms
rather than getting to the "meat" of the information immediately.
As one technical writer in our organization said, "They have to
know it all at once. What do I tell them first?"
Stories of poorly organized and incomprehensible computer
documentation are numerous. Readers speak of the difficulty of
finding information and of the difficulty in figuring out what the

document is trying to tell them. We have all seen the writing of
engineers and programmers in which the writer simply dumps
knowledge on the pagewith no markers for the reader as to why the
information is important, and no guidance as to whether details are
included because the writer knew them or because they are critical
for the reader to understand. Although we instinctively look for
organization when reading documents and try to invent it when it is
not present (Flower 1981, 137), little is known about how we
achieve it or even how to define it. It is one of those things,
like pornography, that we seem to know when we see it.
Wider evidence of the importance of organization exists.
One study finds that mid- and top-level managers rank inability to
organize as one of the four main problems in their own writing
(Aldrich 1982, 298) Studies of teacher ratings "found the ,
influence of sentence structure considerably less important than
the influences of content and organization upon raters' judgments
of quality" (Faigley et al. 1985, 50). In a survey of nonacademic
writing, the authors find that
problems in word choice and in sentence structure are certainly
common in nonacademic writing; in many documents, however, the
most critical problem is a larger onethe document is not
organized to help the reader.... If readers can't find the
information they need, the well-written sentences may go
unfound and unread. (Redish et al. 1985, 129)
Flower and Hayes (1981, 382) tell us that one of the major problems
with inexperienced writers is that they simply string ideas
together in lists and do not provide an organizing structure.
However, the research literature offers little guidance on
organization to the technical writer. Much research exists on

cohesion, the connectives and repetitions that help to make the
transitions within and between sentences. Similarly, investigators
have focused on construction of paragraphs. However, research on
structuring larger pieces of prose is rare. As a recent compendium
of writing research says,
So far, we lack a theory of text structure that goes beyond
the immediate features of the language under a microscope;
there is no comprehensive theory, for example, to explain the
controlling glue that holds an entire book together (Bridwell
and Beach 1984, 11) .
Whv Is Organization Important?
Organization provides a logical progression through a
sequence of ideas, arranging those ideas so that their relationship
is clear. It implies putting ideas in order and establishing their
relative importance so that the reader knows which elements are
most important to attend to. Thus, he or she does not have to
spend time deducing the hierarchy of the information.
Organization takes place at many levels, from ordering
words within a sentence; to ordering sentences; to ordering
paragraphs, sections, and major topics. While some of the same
mechanisms used in organizing paragraphs and linking sentences
together also control larger segments of text, a document is more
than the sum of its words, sentences, and paragraphs.
This thesis assumes a working technical writer will have a
basic level of competence in the mechanics and syntax of writing
and facility with the tools of cohesion. It seeks to investigate
the mechanisms we use to create sense in larger structures. For

instance, is there a "grammar" of documents, like that of
sentences? Is there a set of rules to follow regarding topic
order, elements that must be included, and structures that must be
parallel, for instance?
To the technical writer, organization must be a key
consideration. This is not to deny the importance of organization
in other types of writing. However, structure is critical to
technical writing because of the amount of information and the
complexity of the information being conveyed. A popular article on
heart disease may cover three pages and make two or three major
points. An environmental impact statement may contain thousands of
pages, each section making points that must contribute to the
You cannot build a house without lumber and nails; however,
you also cannot build a house without a plan, a vision of what the
house should be. It is the construction and execution of such a
vision that will concern us here.
Is Examination of Models Useful in Fostering
Organizational Skills?
One of the difficulties in technical as well as other
writing instruction is that often writing has been taught by
preceptproviding students with a set of rules to follow.
Textbooks supplied models that students were expected to imitate;
teachers then examined the imitation for errors. Until recently
texts for technical writers presented a model proposal, report,
research paper, and business letter and told students to imitate

the models. (See, for example, Houp and Pearsall, 1977.) Little
assurance was available that the models held up were necessarily
good ones. Models often tended to reinforce the practices used by
convention, for example, passive, third-person constructions, even
though current investigations show that active constructions and
identification of human agents for the actions make texts easier
for readers to comprehend (Flower, Hayes, and Swarts 1983, 56). In
addition, models and specifications by themselves do not ensure
that texts will be usable. For example, a group of writers
preparing documents to meet military specifications failed to
produce comprehensible documents even though another group produced
usable documents following the same specifications (Duffy, Post,
and Smith 1987, 386-387).
The eighties, have seen a reaction to the use of models;
recent educational and cognitive process literature emphasizes the
process the writer uses in developing material (Bridwell and Beach
1984, 4-6; Cooper and Matsuhashi 1983, 3-6). This approach
stresses the writer's role in using heuristics or rules of thumb to
develop ideas rather than slavishly imitating models.
But the process approach also has its drawbacks. Writers
may be able to develop structures, but they may not intuitively
come up with those structures that readers expect. Although novel
structures may have a place in technical writing, they must be used
carefully because readers are typically reading for information and
do not want the additional burden of trying to understand
unexpected structures.

The tension between concentrating on the written product
and concentrating on the writer's process is evident in much of the
literature oh teaching writing (Cooper and Matsuhashi 1983, 5-6).
Although we must take into account the process the writer goes
through as we attempt to improve our own technical writing and that
of writers we work with, our obligation is to create a product that
communicates. In an attempt to find principles for working
technical writers, this thesis looks at written products to explore
principles that affect the reader's comprehension of that product.
To create such products, writers may want to consider guidelines
from the extensive literature on process in composing.
Will Organization Still Be Important in the Future?
It may be that the current interest in writing research is
an ironic one. We may be investigating skills that our children
will never need. Advances in computer technology may allow a
reader to enter a text at any level. This technology, exemplified
in the HyperCard program from Macintosh, has the potential to let
the reader construct a path through a text (Jong 1988, ATA30-31).
Writers may be able to simply create blocks of textfor example,
one computer screen's worthand store them. The writer then does
not have to organizethe reader chooses words or phrases in the
text about which to obtain more information and presses a key to
move to a screen with more details. The reader in essence creates
his or her own "webs of thought" (Rockley and Graham 1988, ATA12).

Completely reader-organized prose is probably far in the
future. Additionally, someone will have to program so that readers
can move from one screen to another; at present, readers cannot get
additional information on every word. However, writers may not
necessarily have to work in as linear a fashion as before. We
still need to predict the questions, but perhaps the order of the
questions will not be as critical as it is in a linear piece of
text. We will be even less able to assume that the reader has read
necessary preceding material. The problem of designing text for
readers who may enter at different locations may multiply with new
Technical communication is evolving with the technology it
describes and the technology it uses. As technical writers, we may
move increasingly to instructional design or to new media; however,
many of the same needs will still exist. Someone has to determine
the questions that will be askedand how to convey the answers
Can Research Findings Contribute to Development of
Organization in Technical Writing?
Research in writing is not the province of any one
discipline. In fact, writing was neglected during the first part
of the twentieth century, as English departments turned to the
critical examination of literature (Connors, Ede, and Lunsford
1984, 1-8). Interest revived with national awareness in the 1970s
that students were not writing well (Shells 1975), although
interest among professionals flourished during the 1960s (Bridwell

and Beach 1984, 2-3). With renewed interest in writing has come a
recognition that it is a complex process, involving more than
simply teaching grammar and punctuation. Interest has focused on
how people read and learn, how they use language to make meaning
and communicate with each other, and how they plan and structure
their communications.
How different is technical writing from other forms? Can
the body of research informing the process and product of writing
be applied to technical communication? Technical writing requires
the ability to interpret material, to impose a plan and an
interpretation upon disparate facts, and to take account of
audience needs just as other writing does (Rutter 1985, 703-704).
Research affecting our knowledge about writing is drawn
from a number of disciplines, among them cognitive psychology,
linguistics, rhetoric, education, computer science, and
sociolinguistics. This thesis examines the contributions of
cognitive psychology, linguistics, and rhetoric in informing
guidelines for the organization of technical writing. These
disciplines appear to have the most direct relationship to
questions involving the organization of technical writing, since
they explore how people use meaning and language to obtain
knowledge and communicate.
Cognitive psychology deals with how people comprehend,
assimilate, and recall information. It is also concerned with how
people use knowledge to solve problems. The focus here will be on
how to design materials to facilitate the reader's comprehension

and recall. From an examination of how people store information in
memory, we may be able to extrapolate ways to present material so
that it can be stored and therefore retrieved efficiently.
Linguistics is the study of language. Much of the work in
linguistics has been at the micro-level, the level of the phoneme,
word, and sentence. Although spoken language has been the
predominant focus of linguistic research, linguists have begun to
investigate written language as well. In addition, traditionally
linguistic studies have been interested in the study of rules for
an ideal language rather than the study of language as it is
actually used (van Dijk and Kintsch 1983, 2) However, research
has recently begun to take into account the content, or meaning, of
utterances and the circumstances in which utterances are produced.
The branch of linguistics known as "pragmatics" deals with language
as it is used. In investigating this use, some linguists have
attempted to develop a grammar of texts, similar to the grammar
that exists for sentences. Attempts to diagram texts, however,
prove difficult because diagramming sentences takes no account of
meaning. Linguists recognize the need for people to be able to
extract a text's meaning efficiently, which may require more than a
Rhetoric has traditionally been more concerned than
linguistics with the content of utterances and their effect on an
audience, although as noted previously, this divergence is
lessening. Another important feature of rhetoric is its habit of
classifying texts into types. This classification may be useful to

technical writers in terms of informing them what essential
elements must be included in various types of texts.
Other disciplines could be considered. However, they do
not contribute as directly to a concrete understanding of what the
written product should be likeeven though they may contribute
indirectly. For example, education research may inform us of the
problems people face in learning to write, problems that may be
evidenced in the texts produced by technical writers.
Sociolinguistic research can alert us to the importance of context
and the differences in how different audiences may interpret
documents. However, the three disciplines considered in this
thesis provide the most direct perspectives on how a text must be
shaped to facilitate a reader's comprehension and recall.
This thesis selects and explores concepts from three
disciplinescognitive psychology, linguistics, and rhetoricthat
can contribute to our understanding of requirements for
organization in technical writing. It then discusses how these
concepts can be applied in practice. As we shall see, organization
is situation and content specific, meaning that recommendations for
practice must be guided by the requirements of a particular writing
task. This thesis examines principles that can be applied and
presents suggestions for situations commonly encountered, framed in
light of the concepts from the disciplines considered. It also
includes suggestions for striking a balance between adherence to
existing models for technical writing and the need to develop new

Cognitive psychology is the study of mental processes
(Schank and Abelson 1977, 5) As such, it addresses the processes
used by the reader in comprehending and storing information as well
as the processes used by the writer in producing information.
Lately, much interest in the area of cognitive psychology as
related to writing has focused on the writer's process (Flower and
Hayes 1981; Flower 1981; Steinberg 1980, 159-163). However, Cooper
and Matsuhashi (1983, 5) argue that as well as considering process,
we must return to a consideration of text when discussing writing.
To make this transition, we need to look at writing from the
reader's perspective as well as the writer's, asking ourselves what
the reader needs to comprehend and remember text.
Cognitive psychology provides insights about how humans
learn and remember material. Current research in this area
indicates that humans do not learn material in isolated bits; they
must be able to associate new information with information they
have already available in order for the new information to be
meaningful. The ways that we structure our learning have important
implications for how textual material can be organized most

How We Process Information
Cognitive psychologists have postulated that humans have
several types of memorysensory input, short-term memory, and
long-term memory (Smith 1982, 45-52). Sensory input is the
mechanism that transfers the information from the printed page to
the brain. Short-term memory contains information that is
remembered verbatim, is transient, and must be constantly rehearsed
to be maintained. Long-term memory, on the other hand, contains
information stored as summaries or condensations of meaning, not as
a verbatim transcription. The information is stored so that it can
be retrieved after a period of time; it does not require constant
rehearsal (that is, constant repetition as one would repeat a
telephone number) to be maintained.
As Miller's classic work informed us, short-term memory is
limited in capacity, storing only five to nine items (van Dijk and
Kintsch 1983, 352). But, the size of an item is not fixed. We
"chunk" or group information into larger entities that can be
stored. However, information can only be chunked if it is
presented in a meaningful pattern. Studies of chess masters and
electronic technicians indicate that we chunk information according
to content that is meaningful in relation to the subject (Chase and
Simon 1973, 78-81; Egan and Schwartz 1979, 155-158) For example,
chess masters recalled the location of chess pieces grouped in what
was to them a meaningful pattern but did not remember pieces
scattered randomly on the chessboard. Because we can maintain only
five to nine chunks in short-term memory, we must find a way to put

chunks into long-term memory or they will be lost from short-term
memory as we encounter new chunks.
Chunks can also be subsumed into larger chunks. For
example, an instruction might tell us to follow three steps in
starting a car: put the key in the ignition, turn the key, and
step on the gas pedal. We will probably need to remember only the
chunk "starting a car" to retrieve the included information and can
begin to process other chunks, for instance, instructions about how
to drive the car.
Researchers studying prose comprehension are moving away
from the emphasis on short-term memory to a concept of working
memory (Miller 1984, 327-329). Although similar to short-term
memory, working memory is not static. It allows rapid access to
and manipulation of its contents. Miller postulated that since
working memory does not have room for all the chunks of information
in a paragraph, a subset is kept in working memory so that the
chunk being considered can be linked to the rest of the text.
The types of memory are differentiated by the amount of
information they can contain. As far as is known, there is no
limit on the amount that can be stored in long-term memory, while
short-term and working memory appear to be limited to five to nine
information chunks. The types of memory are also differentiated
according to how much we elaborate information before we store it
(Wessells 1982, 140-141). Elaboration refers to the extent to
which we relate an item to other items already in memory. For
example, we may link the word "spinster" to related concepts such

as "female," "unmarried," "elderly." In short-term memory,
information is not elaborated. A telephone number or items in a
grocery list typically do not have a set of associations and must
constantly be repeated to keep them in memory. So that we can
store information in long-term memory, the information must be
elaborated so that we can relate it to things we already know or
construct a new set of associations.
It appears that we place information in long-term memory in
networks of information. The spreading activation model (Anderson
1976, 252-270) indicates that we put information in networks
according to context. Links between concepts are stronger or
weaker depending upon the immediacy of the association. Linked
concepts are activated depending on the strength of the link.
Thus, we may immediately recognize that a canary is yellow but may
be slower to realize that it has skin (Howard 1983, 217).
The strongest links are not necessarily strictly
hierarchical, but are also based on familiarity. We recognize that
a dog is an animal faster than we recognize that it is a mammal
even though mammal supposedly is a more general term that logically
would be higher in a hierarchy. However, the association links are
stronger for the first classification (Howard 1983, 217) .
Additionally, concepts are linked not to only one other but to many
other concepts. Canary may have numerous associations, including
yellow, skin, sings, and so forth. Any one or a combination of
these associations may be used to access the concept of canary.

Requirements for storage in long-term memory have a bearing
on organization in writing. If memory is organized as networks of
information, we must be able to determine where a piece of
information "fits" before we can move it from short- to long-term
memory. Information presented in terms of weaker rather than
stronger links takes longer to process, as discussed previously.
As writers, we must examine the links we are using, since a
strictly hierarchical arrangement based on a predefined
classification scheme may not be the one that occurs naturally to
readers. Because readers may have different links to information,
we may need to provide more than one access link in our texts.
When organizing material, we need to consider when we
require people to access long-term memory because accessing long-
term memory affects processing speed. Subjects read a sentence
fastest when the sentence agreed with the semantic and syntactic
focus of the previous sentencefor example, when the syntax was
either all active or all passive for the series of sentences
because the needed knowledge structure did not have to be retrieved
from long-term memory (Miller 1984, 338-340).
The implications of the different types of memory for
technical writing lie in the necessity for making it easy for
people to remember technical writing. Unlike much expository
prose, in which people need to remember only main ideas, in
technical writing people often need to be able to remember both
main ideas and details. We can help them do this by presenting
material in expected patterns, so that it can be chunked for better

remembrance, and by making explicit associations for our readers,
so that they can place information in the appropriate place in
their long-term memory networks.
Since in short-term memory we can store information only
for a short time and can retain only a limited amount, there is a
need for closure, that is, "the tendency for people to complete an
incomplete message or figure" (Nystrand 1986, 9). We need to reach
closure so that a thought can be completed, a set of associations
developed, and the thought stored away.
Hirsch (1977, 108-137) discusses the problems of
uncertainty at the clause level as well as at larger levels. If
information is interjected between the beginning of a clause and
the point at which closure is reached, the reader's task is
complicated by having to keep the beginning of the clause in short-
term memory and in effect repeating, or rehearsing, it until the
portion that creates the closure is reached. The intervening
material must also be retained in short-term memory, potentially
straining capacity. However, closure cannot always be immediate.
As Hirsch says "readability demands rapid closure to avoid taxing
the reader's short-term memory. But readability also demands
explicit constraints on meaning in order to guide the reader's
understanding." (1977, 134)
Similar constraints exist in organization units larger than
the sentence. If the reader has to hold ideas in mind for a long

period of time before seeing their application, he or she is likely
to lose the train of thought and have to go back and reread the
beginning. Thus, for example, the classical technical organization
of presenting a theory of operations of a piece of equipment and a
history of its development before describing its use and
installation may be in error. The reader does not know the
application for the information and has difficulty holding it in
mind until reaching the sections where pieces of it undoubtedly do
Human beings expect patterns. This expectation is a factor
of our experience:
the world we live in is highly structured and predictable, for
particular events are far more likely to occur in some contexts
than others....we form expectations about what things we are
likely to see in various contexts, and these expectations may
influence how we recognize patterns (Wessells 1982, 59) .
Hirsch (1977, 103) also notes that we tend to assume the pattern of
a new situation will continue in a uniform way.
We develop strong expectancies when we read. As we come to
the conclusion of a sentence, the possible choices become more and
more constrained (Hirsch 1977, 102). At a micro-level, Hirsch
finds that the expectancies set up in one phrase should be
fulfilled by the next. He notes that a reader processes a word
faster if that word is one that the reader expects to see. In
fact, "cloze" tests, which measure how easy it is to read a text,
are based upon this principle. In these tests, every fifth or so

word is left out and the reader must supply them. Readers can
predict missing words with varying degrees of accuracy, depending
on the "readability" of the passage. This readability is affected
both by familiarity with the subject matter and by the number of
repeated words, new content words, and function words. Fewer new
content words and subject familiarity tend to promote ease of
In a study of reading comprehension, subjects' attempts to
predict the contents of a given paragraph became increasingly
accurate as they progressed through the paragraph, suggesting that
a paragraph becomes increasingly constrained as it develops.
Subjects appear to use these constraints to refine their selection
of world knowledge needed (Miller 1984, 342). Only information
relevant to the topics likely to be discussed is retained in
working memory. Thus, if expectancies are violated, the reader
must retrieve information from long-term memory and begin a new
processing cycle.
Elements other than the text itself can also set up
expectations. Readers' expectations will be quite different for
these two titles: "Byting Bits in First Grade" and "The Effects of
Text Editing on the Cognitive Processes of First Graders" (Nystrand
1986, 112) A reader might expect to find the first in a popular
magazine and might anticipate that it would discuss children's
computer learning. The second would be likely to appear in an
academic journal. The reader uses clues such as title and format

to determine what the text is about and will be confused if those
expectations conflict with the text.
A study of prediction skills in reading showed that people
had greater difficulty predicting what would happen in an essay
than in a story (Olson, Duffy, and Mack 1984, 268) In each case,
people formed tentative hypotheses as they read each sentence.
Their expectations were much clearer for the events in a story.
Comments on the essay's structure revealed people's expectations
about how an essay should be written, but predictions about subject
were not as accurate. Although people knew generally how an
argument in an essay should progress, they could not foresee the
content or the logic that the author would use.
As an aid in prediction, the reader develops a set of
questions while reading. As each sentence is understood,
the reader revises and elaborates the set of information still
needed to have the developing story make sense. These
informational needs interact with what is presented in the next
sentence to generate a new set of information needsor, if you
will, a new set of questionsthat guide the reader's
comprehension through the succeeding parts of the text. (Olson,
Duffy, and Mack 1984, 278-279)
The tendency to expect patterns means that similar material
should be treated in the same way throughout a text. Chapters
dealing with similar topics should contain the same organization of
headings, for example. Since the reader is searching for
information, we should not introduce variety for variety's sake,
particularly in organization, although there should be sufficient
variation in wording to prevent boredom and inattention.

Hirsch (1977, 104) indicates that large-scale constraints
or expectations are based on linguistic rules and semantic
conventions. Because readers have expectations about how
paragraphs and larger sequences of text are constructed, writers
should not violate those expectations unless the violation is
clearly signaled. Readers typically form their expectations on
the basis of their previous experience and their knowledge of the
world. Although we cannot know exactly what each reader's
individual expectations are, we can develop general notions of
readers' expectancies through the examination of common schemata.
Schema Theory
As indicated in the previous section, expectancies are
heavily influenced by readers' knowledge of the world. Schema
theory is a way of explaining how this knowledge of the world is
In its simplest form, schema theory means that we use our
background knowledge to fill in the gaps in a speaker's or a
writer's message. A schema is a set of organized knowledge about a
particular topic (Bartlett 1932, 197-201). Although this concept
has been given a number of names, including scripts (Schank and
Abelson 1977, 36-43) and frames (Minsky 1975, 212-213), all refer
to an existing mental framework about a topic into which new facts
are placed.
In brief, a schema allows us to understand many components
of a topic without being specifically told about them. A

frequently used example in the literature is the restaurant schema.
Americans familiar with restaurants have a common set of
expectations for what will happen in one. When someone talks about
a restaurant, the general schema is activated and we can use
typical information for information that is not explicitly stated
(Wessells 1982, 314).
A schema can represent complex, global events in a
relatively simple manner and can also guide the processing of new
information. The power of schemata lies in the ability to induce
inferences from the reader (Huckin 1983, 92) However, schemata
must be activated before they can be used. A frequently cited
example concerns a passage on washing clothes (Bransford and
Johnson 1973, 400-401). If the reader is aware that the passage is
about washing clothes, the passage makes sense, but without this
awareness, the passage appears as a series of disjoint sentences.
We have schemata for more than everyday events. We have
schemata for reading narratives, for example. If a story is
scrambled, comprehension is not as accurate as for a story in an
expected order (Thorndyke 1977, 95-98). Similarly, in a study
comparing reading times for short fables with sentences in the
conventional order and those in which the sentence order was
scrambled, sentences in the disorganized texts proved more
difficult to integrate (Mitchell 1984, 85). Subjects paused longer
at the ends of sentences when reading disorganized texts.
A schema is an important part of the reciprocity between
writer and reader. If a common set of assumptions is not shared

with a reader, the writer is obligated to explain considerably
more. Since schemata cannot be used by the reader until they are
activated in some way by the writer, the writer must also provide
contextual knowledge that activates the relevant schema. Nystrand
(1986, 26) argues that texts, exactly like speech, are not
autonomous and that each text can only be understood in its
Because the text must result in shared knowledge, what is
said must make a bridge to what is already known. The extent of
elaboration in technical writing typically is controlled by the
needs of the audiencefor example, a technical writer provides
more detail for nontechnical readers than for technical ones.
Since the reader attends to both the structure and the details,
text is coherent to the extent that the schemata create
expectations about the details to follow and the details fulfill
these expectations (Huckin 1983, 93).
Hirsch (1977, 135) indicates that ease of comprehending a
text depends on the writer's correct assessment of the relevant
knowledge already possessed by the audience. There are numerous
invisible constraints imposed by theme, genre, and the reader's
prior knowledge, all of which are part of the schema activated by
the writer.
Technical writing has a number of schemata of its own. The
form of a computer tutorial, a computer reference manual, an
operations manual, or a scientific journal article has become
fairly standardized. People develop a set of expectations based on

the conventions of these forms. Thus, the technical writer must be
familiar with these forms and conventions.
In addition, readers have schemata for different types of
content. For particular topics, certain knowledge can be assumed
on the part of the reader and certain methods of presentation may
be expected. As an example, readers knowledgeable about a topic
can infer far more than those with little background information.
Following a schema does not solve all questions of
organization for the technical writer. Narrative writers are also
constrained by highly predictable forms. The technical writer does
not just fill in the blanks, any more than the story writer does,
although technical material may impose more constraints.
The techical writer should generally try to use familiar
schema and to analyze his or her readers' familiarity with the
schema being used. For example, the writer must acquaint first-
time computer users with the conventions of using a computer manual
as well as with the subject matter. The writer must provide
sufficient content information to enable all readers to understand
but not enough to trivialize the topic for the knowledgeable user.
Hierarchies and Storage of Information
Huckin (1983, 95) found research in cognitive psychology
shows that readers tend to process a text hierarchically, assigning
more importance to material that comes high in the hierarchy than
to that which is lower. Tenebaum (1977, 534) found that a

hierarchical structure enhanced recognition of the main idea while
lists did more for instructions.
People focus on where information is located in a
hierarchy. This is called the "levels effect" and has been studied
by a number of researchers (Huckin 1983, 95; Wessells 1982, 318).
Access to lower level concepts depends on prior access to higher
level ones. If important details are listed instead of
subordinated, the hierarchy is flattened.
Van Dijk and Kintsch postulated a "leading-edge" strategy,
in which information is stored and remembered as a function of
height in the text hierarchy and recency of presentation (van Dijk
and Kintsch 1983, 44). As they note,
...the selection of propositions to be retained at each cycle
was made on the basis of the leading-edge strategy, which
favored the selection of semantic units high in the textbase
hierarchy. In case of ties, recency was made the basis for the
selection (van Dijk and Kintsch 1983, 44).
Important superordinate propositions are frequently retained for
more than one. processing cycle. The "leading-edge" is the point at
which information is either removed from working memory or retained
to be used in further processing.
Indeed, the most powerful variable related to recall
appears to be the content structure, that is, the organization of
the information in a passage (Meyer 1977) Information high in the
content structure is more likely to be remembered. However, if the
reader uses a strategy and schema different from that of the
writer, recall may not relate to content structure. Readers who
opposed the concepts presented in a passage tended to remember the

material that supported their point of view rather than material
high in the hierarchy.
Although the reader initially encounters and processes
individual pieces of a text, at the same time he or she is
constantly engaged in constructing a larger picture (Matsuhashi and
Quinn 1984, 319). Like other researchers, these authors posit that
individual ideas are connected hierarchically so that they can be
stored. Since all information cannot be remembered, the most
important concepts, those high in the hierarchy, tend to be
retained and information at lower levels tends to be lost.
Although we store material in memory in a spacial
arrangement that may or may not be hierarchical, it appears that
encountering material structured in a hierarchy helps us know what
to remember. Thus, the technical writer cannot simply present
details all at one level and expect the reader to remember them.
The hierarchy supplies a way for people to chunk information so
that it can be remembered. In fact, if we do not provide
hierarchies for readers, they will create them for themselves
(Flower 1981, 137), which may impede or at best slow comprehension.
Methods of Creating Structures
Given that cognitive psychology indicates that people
remember information hierarchically with information at the top
remembered more readily than that at lower levels, how can we
structure prose so that it is in fact organized hierarchically?
Flower, Hayes and Swarts (1983, 42) present what they call the

scenario principle. Under this principle, material is organized by
the tasks people actually do, rather than by categories of
descriptive information. Headings are often presented as questions
to be answered rather than as noun phrases. The authors devised
this principle after watching readers attempt to find meaning in a
set of government instructions for small-business owners. The
readers attempted to restructure the information to make it
functional for themselves by asking questions and then answering
them. They also tried to put the information in a personal context
by creating examples. Their strategies indicated that a different
kind of organization and elaboration was needed.
Cognitive science indicates that the best representation of
knowledge on a given topic depends on how that knowledge will be
used (Flower, Hayes, and Swarts 1983, 48) Readers probably
approach functional documents with a top-down strategy, that is, a
strategy that looks for overall meaning before grappling with
details. The writing must be structured so that the reader can
act. One strategy is to organize around the tasks the reader must
perform; another is around the questions he or she will ask.
Another strategy for organization is the use of headings to
activate schemata. Headings can set up the framework for a
document and can clearly indicate subordination, superordination,
and coordination. Investigations have shown that versions of
documents with headings are more easily comprehended than those
without (Huckin 1983, 97). Unfortunately, other research
contradicts this view, indicating that overviews of material to

come and logical connectives in the text are more important than
headings in conveying hierarchical relationships (Spyridakis and
Standal 1986, 251-253).
Herrstrom (1984, 228) postulates that technical writing
attempts to lay bare the structure of a text. He feels that while
a novelist strives to make structure implicit, the technical writer
tries to make it explicit. Thus, the organization plays a very
important role in technical writing because a major part of its
function is to move the reader through the material efficiently,
providing ample signposts along the way.
According to Herrstrom, many people understand material by
making a diagram of it. He finds that we map facts onto a design
so that we can see the fundamental relationships between the parts
of our explanation. This map can be schematic (locational), an
explanation of a process or flow (flowchart), or an explanation of
logical hierarchies (tree). Herrstrom's three maps are intimately
related to the type of content being dealt with. They can be used
to examine the reader's expectations for various types of technical
Further insights into cognitive processing of structure can
be gained from computer science. Harrington and Walton (1984)
recommend the use of Warnier-Orr diagrams to structure technical
prose. This diagram is somewhat like an outline in that it lists
major and subordinate headings. However, its structure groups
headings at the same level together, forcing the writer to examine

whether headings are consistent and whether headings at equal
subordination levels carry information of equal weight.
Why should we be concerned if information related to
headings is of equal weight? As noted earlier, we can process from
five to nine pieces of information, tending toward the lower
number. More headings at a particular level force the reader to
absorb them in two rather than one installment. Human expectations
of patterns lead readers to expect that sections at the same level
will contain roughly equivalent types and amounts of information
(Hirsh 1977, 152); otherwise, the pattern is broken and the reader
does not know which pattern to expect.
As Harrington and Walton (1984) note, the Warnier-Orr
diagram forces the writer to think both vertically and
horizontally. By giving equal weight to the vertical and the
horizontal, the writer is required to think of organization
(vertical) and development (horizontal). Both are essential,
although a text cannot be developed until it is organized.
Instruction manuals are often well organized by task but lack
developmentthat is, they do not explain the purpose of the task
to be doneso that the appropriate schema cannot be activated.
The authors note that "the converse fault, lack of organization,
will yield a piece of writing resembling a jumble of unrelated
detail entwined with some general claims" (Harrington and Walton
1984, 198). In this case, the reader is forced into a bottom-up
method of processing, acquiring details without an overall
understanding of how those details fit together.

Implications for Technical Writing
Cognitive psychology provides a number of useful insights
for those interested in creating organized technical prose. It
emphasizes the need to facilitate storage of material in readers'
long-term memory, to meet readers' expectations, and to reduce the
processing required to store and retrieve material. Based upon
these insights, we can examine features that should be present in
technical writing and strategies for including these features.
Most importantly, cognitive psychology emphasizes the need
for structure in learning and remembering material. Since a
primary aim of technical writing is to facilitate learning and use
of information, the writer is responsible for providing this
structure for the reader. We cannot ask or expect the reader to
take on this task.
As seen earlier, the nature of memory influences how
material should be presented for ease of storage and retrieval.
Since short-term memory can contain a limited number of items for a
limited time, that capacity must somehow be enhanced and the
transfer into long-term memory facilitated. Short-term memory
capacity can be enhanced by helping readers to chunk information.
Information must be grouped in meaningful patterns so that readers
can remember the major concept and use it to work back to details,
if necessary. To assist the reader in chunking, we must place
information into groups. For example, if we have a list of
computer commands, we may want to structure them by function, that
is, what they are used to do. If we are discussing criteria for

locating a dam, we might group data by impacts on various entities:
wildlife, population growth, agriculture, recreation, and so forth.
When recalling information, readers can go to the chunk and then
"unpack" the information contained in the chunk.
To aid the movement of information into long-term memory,
we must establish links to information already existing in memory
and then activate those links so that readers have a framework to
use when encountering new information. Otherwise, they do not how
to associate the information and must go back and reread when the
appropriate link is finally activated. Thus, a technical writer
should link material to what is already known and make that link
explicit. When presenting new material to novices, the writer can
make analogies to that which is likely to be known. For example,
in describing a computer operating system, a writer might say, "An
operating system is like an office manager. It manages tasks like
getting files copied and reports printed."
Once a link is activated, it should be maintained. Shifts
seemingly as small as changing from active to passive voice require
readers to return to long-term memory and retrieve new structures.
In the preceding example, shifting from "operating systems" to
"programs" might require activation of new links. This is not to
say that new links should not be activated, but the change should
appear planned and the number of activations in a particular
section of text should be limited. For example, a writer could
talk about both programs and operating systems under the topic
"Understanding Your Computer System" but should not switch from

operating systems to programs to operating systems if the topic is
operating systems.
Even when readers have a framework for storing new
information, they cannot hold many partially completed ideas in
mind. Readers must achieve closure of an idea so that they can
store the concept in memory. This need for closure implies that we
must be careful about interjecting ideas before a first idea is
completed. Otherwise, readers will not know which idea they should
consider and may forget one or the other. For example, the
following sentence violates the principle of closure: "The store,
which stocks milk, meat, and vegetables as well as hardware, is
closed." The sentence does eventually complete a thought, but the
reader must maintain the subject of the sentence, "store," in mind
while attempting to process the items in the store. By the time
the reader reaches the sentence predicate, "is closed," the subject
may well be confused with the modifiers. Although sentences of
this type can be effective, they are typically more difficult to
process and should be used sparingly. A more significant violation
of the principle occurs if, for example, a writer interjects a
detailed discussion of how a particular part works into a general
overview of the functions of a machine.
We need to reach closure on ideas that have been presented.
We also have strong expectancies about material that will be
presented. Expectancies are a powerful tool for the technical
writer. These expectancies are set up both by the patterns the
writer establishes and by the readers' expectations created by

their schemata for form and content. Expectancies lead us to
believe that patterns will be regular and that text will follow the
same logic and contain the same elements as similar texts
encountered in the past. DeBeaugrande makes the point that "the
actualization process of writers and readers are geared toward
continuity and regulation" (1980, 18) .
Although the concept of expectancies implies that we should
set up and maintain patterns in our writing, the extent to which
patterns should be maintained is an open question. For example,
expectancies would indicate that headings should be parallel
grammatically, that chapters treating similar subjects should- have
the same organization, and that phrases introducing the same types
of instructions should be identical. But a rigid adherence to
patterns may be self-defeating. An organization by tasks often
leads to headings beginning with* gerunds (for example, Editing
Text). Although typically all headings should be parallel, gerund
headings for sections that do not describe tasks can be misleading
if forced to fit a parallel structure. For a section listing
commands of various types, "Command Classifications" might be a
reasonable title. "Classifying Commands" could be misleading,
implying that the reader will learn how to classify commands rather
than finding out what classifications exist. Patterns should be
maintained in situations that are truly identical; for example, we
might always introduce sets of instructions with the phrase "Use
the following steps...."

While we .do not have to adhere slavishly to patterns, we
must be aware of those we are setting up. We should use the
patterns we have established unless there is some good reason
breaking them, a reason that we can identify. We should make it
apparent to the reader when a pattern is being broken and indicate
that there is a reason. For example, in a set of procedures each
containing the sections "Objectives" and "Actions," one set might
not have actions. We would have to let the reader know that the
omission was deliberate.
In addition to the patterns that writers provide, readers
use their existing schemata to make guesses about texts, applying
their background knowledge to infer information that the writer has
not explicitly stated. These schemata may include information
about the form or the content of a document. Writers cannot know
exactly what schemata their readers will have available; they must
make educated guesses based on their analysis of who their readers
will be and their likely level of knowledge. We should attempt to
ascertain our readers' level of subject matter knowledge so that we
can activate relevant schemata about the subject matter. Too much
background material will trivialize the topic and annoy the reader
who is knowledgeable about the content; too little background
material and an inexperienced reader will be lost.
In addition, we must activate schemata about the text forms
being used. A number of schemata are available for various types
of technical content. For example, computer manuals for products
geared to a general audience typically have a tutorial (which

provides step-by-step instructions for the user to follow as an
introduction), a task section that describes how to do various
tasks, and a reference section that lists concepts or commands
alphabetically. Proposals typically have an introductory section
describing how the work will be done and why the organization
making the proposal is the best for the job, followed by supporting
documentation about specifics of schedules, costs, and personnel.
Scientific articles usually contain the following sections:
statement of problem, review of literature, methods, results, and
discussion of results. Descriptions of physical entities often use
a spacial schema, moving from top to bottom or left to right about
an object. Readers have learned to use these and other schemata
for technical writing based upon their past experiences.
Although readers have schemata for existing document types,
they can learn to use new ones. For example, early computer
manuals often consisted of alphabetical lists of commands, which
proved difficult for novices to use. If you did not know what
commands existed, you had to work very hard to discover how the
commands could be used and how they worked together. As the
community of computer users expanded, usability became more
critical. Lists of commands proved insufficient for people without
a basic understanding of how commands are used. The new schema
typically includes a tutorial and task-oriented section with the
alphabetical list relegated to the back of the book or a separate

Typically we can refer to existing documents and our
knowledge of content to determine whether we should use existing
schema, modify them, or create new ones. New ones are most likely
to be needed when either the subject being discussed or the
audience for the documents has been in the process of change.
Readers automatically structure material in hierarchies to
remember and store it. Since they tend to remember information
highest in a hierarchy, we must be careful that the material in the
higher positions is actually the most important to be remembered.
And we must be sure to provide a hierarchy, since otherwise people
will construct them on their own and may not construct the ones
intended by the writer.
When creating hierarchies, we must carefully analyze the
material we are discussing. As writers, we must constantly ask
ourselves questions like the following: "What is the most
important point to make? What are five or so points that support
it or explain it? What are points that support each of these
points?" The nature of the points is determined by the purpose of
a document. For example, a description of two types of programming
counters might emphasize the types of counters available or
highlight their characteristics, as shown in the two hierarchies
that are displayed in Figure 1.

Types of Counters Available (Hierarchy 1)
Number available
How incremented
Number available
How incremented
Description of Counters (Hierarchy 2)
Number available
How incremented
Figure 1. Two Hierarchies for the Same Information
In the hierarchies in Figure 1, notice that the same order
for subpoints is used in each section and the same subpoints are
included in each section. In the second hierarchy, for example, we
would not suddenly decide to discuss how a third type of counter
was incremented. To discuss a third counter, we would have to
include it in the hierarchy and either fill in all the subpoints
for that counter or explain that they did not exist.
When we begin to structure a piece of technical writing, we
must first ask ourselves what schemata the readers are likely to
have available that we can build upon and what hierarchies we wish
to emphasize in the material we are presenting. In the process of

determining an appropriate schema and hierarchies for material, we
can also begin to determine an appropriate structure for the
document. These decisions are not necessarily made sequentially,
but typically occur interactively with decisions in one area
affecting those in another.
As we have seen, readers pose questions and hypotheses to
themselves as they read, predicting the general shape of what will
come next based on their knowledge and previous patterns. One
method of structuring a document is by building on the questions
readers naturally ask. To do this, the writer must first generate
a series of questions and then place himself or herself in the role
of the reader, asking which questions are most important, which
would be asked first, second, which are really subsets of other
questions, and so forth. Another potential framework is to
organize material according to the tasks the reader will perform.
Once tasks are identified, they can be organized in a variety of
ways. Tasks may be classified by the person likely to perform
them. Major tasks may be presented first followed by minor ones,
or tasks may be ordered in the order that they would typically be
performed. If tasks are always performed in the same order or are
dependent upon being performed in a particular order, a
chronological structure is advisable. The guiding principle must
be a combination of the nature of the information and the way that
the readers are expected to use that information.
In the process of creating a structure, the writer must
also ensure that expectations will be met and closure reached.

Specific details should not violate the hierarchical structure and
the organization should make clear to the reader why details are
subordinated to main ideas as they are, why topics and subtopics
are grouped as they are, and why topics are ordered as they are.
Insights from cognitive psychology tell us how people
process information and how they solve problems. Psychologists and
linguists have worked closely together to investigate how our use
of language influences the way we think and learn. The next
chapter examines organization from a linguistic point of view.
Some of the topics discussed under linguistics are also examined by
cognitive psychologists, even though they may have slightly
different names and emphases. The separation of the fields is not
a hard and fast one at their boundaries and some authors could be
discussed in both categories. However, using the perspective of
the different disciplines enables us to look at some of the same
topics from slightly different points of view.

Linguistics in America in the first part of the twentieth
century was primarily concerned with discerning the structure of
language. Seeking to determine the rules by which language is
produced, the study of linguistics attempted to isolate language
from context (Booth 1986, 453). In addition, linguistics prior to
the 1970s was concerned primarily with the construction of
individual sentences, not with larger units (van Dijk and Kintsch
1983, 2).
Recent developments in linguistics have included an
increasing interest in discourse and an attempt to determine
whether there are rules for discourse as there are rules for
sentences (van Dijk 1977, 5). Although the notion of a grammar of
texts is a seductive one, it fails to take account of the many
variables involved in producing text larger than a sentence,
including how the communication will be used and what the
communication means. And, in fact, even traditional sentence
grammars could not account for sentences found acceptable in
certain real-world situations even though they did not conform to
the rules for acceptable grammatical structures. A prescriptive
grammar covering all sentences that could be generated proved
impossible. Grammars seem most effective for describing sentences

already uttered, although they have been used as a method of
generating sentences. The same situation may exist for textswe
can generate principles for describing elements of successful
existing texts but we cannot provide universal rules for producing
them. In describing successful texts, we may be able to identify
elements that they have in common and principles that cannot be
violated, even though, as with sentences, there may be more than
one method of successful presentation. Similarly, as with
sentences there may be methods best suited to particular
situations, but grammars do not provide any way of gauging
appropriateness to a situation.
Linguists have come to the view that discourse processing
is a strategic process rather than a rule-governed one (van Dijk
and Kintsch 1983, 10-11). The strategies are like working
hypotheses, not rigid rules. While grammar can be rule governed,
understanding of text is principle governed (Enkvist 1985, 13-14) .
We tend toward regularities and strategies rather than rules and
laws (deBeaugrande and Dressier 1981, xv) .
This chapter examines general principles derived from
linguistic studies that can be applied to organization of writing
in general and technical writing in particular. These principles
include how we perceive and organize information, how we integrate
new information with what we already know, and how our expectations
and our knowledge of external factors influence our comprehension
of text. It also discusses the implications for technical writing
that can be drawn from these principles.

Some concepts, such as propositions, memory networks,
gists, and schemata, have been examined by scholars in both
cognitive psychology and linguistics. Accordingly, there is some
overlap between this and the previous chapter.
Kintsch's early research (1974, 12-15) indicated that we
process and store text as propositions. Propositions are the
minimum information units we remember from a sentence. This
minimum information must consist of a relationship between two
concepts, a predicate and an argument (deBeaugrande 1980, 29). For
example, we store the sentence "The dog barks" as the proposition
(BARK, DOG). Longer sentences contain more than one proposition;
propositions can contain more than one argument.
A memory representation of text is made in cycles that
begin when a proposition enters short-term memory and end when
enough information has been stored to make sense (Kintsch and van
Dijk 1978) This need for closure indicates that propositions
cannot be simply strung together. The reader must be able to make
a meaning to store information and must be able to discern the
place where the information would be appropriately stored. Because
this is a cyclic process, readers can relate propositions to higher
or lower level propositions with part of one proposition related to
the meaning for a subsequent one.
Although a number of sentences may have the same
propositional structure, the strategies for turning a proposition

into a sentence are complex and depend upon what the speaker wants
to emphasize (Kintsch 1974, 14; van Dijk and Kintsch 1983, 280-
285). The proposition (CHASED, CAT, MOUSE) might be represented as
either "the cat chased the mouse" or "the mouse was chased by the
cat." Whether these sentences are stored differently is dependent
on other factors, such as the importance of the difference to the
meaning of the utterance. Thus, authors have choices as to how to
represent meaning.
Just as individual propositions can be written in different
ways, groups of propositions can be linked in different ways.
Typically, however, propositions are part of a hierarchical
structure. Kintsch (1974, 16) sees the organization of
propositions as a hierarchy in which one set, or list, of
propositions, is subordinate to a higher level proposition, or
macroproposition. As Kintsch says, "The macrostructure of a text
is given by the hierarchical structure. The macrostructure of a
text is implicit in the proposition lists that constitute the text
base" (1974, 17) These macropropositions likewise form a set, or
list, subordinate to yet a higher level macroproposition. These
nested sets of propositions and macropropositions form the
macrostructures of the text.
Only the proposition used as a macroproposition and those
that directly refer to it are retained in working memory for the
next cycle of processing (Miller 1984, 336). The macroproposition
offers guidance for processing the rest of the text by focusing
attention on topics that are likely to be discussed. Since

knowledge structures are brought into working memory and compared
to text propositions, knowledge-based constraints are also
established. If expectancies or constraints are violated, the
reader must retrieve information from long-term memory and begin a
totally new processing cycle. Thus, it is more difficult to store
material that stops the processing cycle.
The notion of propositions stresses the importance of
meaning in our storage of information. We do not simply store
propositions linearly. We perform operations with propositions,
deleting, generalizing, and integrating information before we store
them (van Dijk 1977, 143-146). Some propositions not directly
presented are inferred during reading (Kintsch 1974, 105).
Inferences are part of comprehension; they are part of our
operations on propositions.
The fact that we store information as propositions stresses
the importance of meaning in our comprehension and recall of texts.
Propositions also imply that we tend to reduce information to its
essentials. However, propositions must be linked or we must
activate new information from long-term memory to use them. Simply
linking propositions has the potential to leave us with
unintegrated lists of facts or concepts unless we have some way to
structure and relate them. We appear to structure propositions by
developing macrostructures that enable us to determine the "gist"
of a set of propositions.

We remember the "gist," that is, the substance of
utterances, not the verbatim form of the communication (Wessells
1982, 90) In fact, our memory for text is "anything but
verbatim." (Kintsch 1974, 252). Instead, it is reconstructive and
multi-level. It is easier to reproduce the gist of a story than
its details (van Dijk 1977, 157).
In fact, information we take in is reduced to
macrostructures simply because we do not remember detail (van Dijk
1977, 156). A macrostructure is much the same as the gist, theme,
or topic of a section of text (van Dijk and Kintsch 1983, 15). We
require macrostructures in understanding because we have to be able
to develop higher level concepts to organize and reduce details so
that we can deal with them.
Even when remembering lists of words, it is easier to
recall the separate words when they can be related to some type of
organizing principle. A person might recall that a list contained
five words about cooking, five about furniture, and so forth, and
can then proceed to reconstruct each set of five words separately.
It is easier to recall five lists of five words, each with its own
topic, than one list of twenty-five words not organized into topics
(Wessells 1982, 141-143).
When we develop gists, each proposition is processed with a
corresponding macroproposition. Macropropositions are processed
with corresponding macropropositions, eventually making it possible
to condense meaning into a gist. If there are gaps between

propositions or missing links, the reader must link the proposition
with one already in memory (Cooper 1988, 357) Searching of long-
term memory or constructing inferences demands more processing,
making comprehension more difficult. If the reader must link
propositions in this way, comprehension becomes a bottom-up
process, in which the reader merges propositions into
macropropositions to arrive at a gist. Although comprehension can
proceed in this manner, it proves a burden for the reader, forcing
him or her to identify the macropropositions. In fact, just
because this method is so difficult, the writer should make the
macropropositions evident to the reader.
Metadiscourse also aids readers in finding the gist of a
passage (Vande Kopple 1985, 85-87). Metadiscourse does not add new
propositions to a text, but instead gives the reader clues as to
how to organize and interpret the material presented.
Metadiscourse can be as simple as reminders about material
presented earlier in the text or sequencing words such as "first"
or "next." It can also include pointers to the writer's attitude
or can indicate the author's opinion of the validity of the
material discussed. Although it guides the reader's interaction
with a text, metadiscourse does not expand the informational
content of a text. Metadiscourse is communication about
communication that shows us how we relate individual propositions
to make a coherent text.
Although the signals provided by metadiscourse may help
readers to discern gists, they by themselves are not enough to

create text coherence. As readers, we also have to have an
organizing principle to which we can relate information. In a
similar fashion, sequences of text that lack a specific topic of
conversation, that is, a gist or organizing principle, can be
linearly coherent but fail to make sense (van Dijk 1977, 50-52).
For example, the following sequence of sentences include references
to the sentences preceding them but do not make sense.
The quarterback threw the ball to the tight end.
Balls are used in many sports.
Most balls are spheres, but a football is an ellipsoid.
The tight end leaped to catch the ball. (Witte and Faigley
1981, 201)
The notion of a gist or organizing principle is related to
the writer's plan for the text. A discussion of examples of
student work indicates that
macrostructure, the paragraph or the composition as a whole,
also has a great impact on coherence. Many of our samples show
that the writer has formed no definite idea of what purpose a
paragraph is to serve or what means will be used to develop it.
(Goodin and Perkins 1982, 63)
How do we as readers determine gists? Partly through
inferencing, combining our world knowledge with the information
provided by the author. But inferencing may not be sufficient.
The writer must provide a gist for the reader to be able to discern
one. Van Dijk and Kintsch (1983, 272-273) make the assumption that
the creation of macrostructures requires a plan. The writer should
be able to state the gist of a text and of sections or paragraphs
of that text. If the writer cannot make such a statement, the
reader will undoubtedly have difficulty in discerning it. The
writer should be able to answer the question "What do I want my

reader to remember?" for each section of the text and also for the
text as a whole.
Although the writer must be able to state a gist, this does
not mean that a writer must signal each gist explicitly, as for
example, beginning each chapter with "The purpose of this chapter
is...." It does mean that the writer has to be able to construct
such a statement, even though this statement may never be
explicitly written in the text.
Given-New Contract
The linking of propositions with information already in
memory indicates that we try to relate new information to
information already available. So that readers can place the new
information in their networks, the writer must connect the new
information to information that is already known by the reader.
Clark and Haviland (1977) indicate that readers' understanding of
the antecedent proposition (the proposition that comes before)
depends on how clearly the old information is referred to, the
distance between the old and new information, and whether an
antecedent is in fact specified. The listener should have one and
only one antecedent for any given piece of information. Although
this antecedent can consist of more than one entity (such as a
compound subject or predicate), it must be identifiable as a single
thought unit. If the reader cannot find a direct antecedent, he or
she will try to build a bridge to something already known. If such

a bridge cannot be constructed, the reader may have to add a new
structure to memory, a laborious and error-prone process.
Clark and Haviland (1977) postulate three requirements in
the "given-new contract": (1) the given should be known and the
new should be unknown, (2) the listener must be able to determine
an antecedent that is unique, and (3) the listener must have
sufficient knowledge and skill to be able to determine the
antecedent. Before readers can compute an antecedent, they must
know the topic of the communication and they must have the level of
knowledge assumed for them; otherwise, the given may not be given
for them at all.
Goodin and Perkins (1982, 59-63) find the concept of given-
new very useful for analyzing incoherence (that is, a lack of
organization). A sentence may have little or no new information,
simply rephrasing or repeating previous information. This
situation raises a question in readers' minds as to whether the
information is really different. Readers may question the author's
knowledge and skill or their own understanding, in either case
slowing the transfer of information. Similarly, a sentence may
lack given information. Sentences of this type are incoherent with
their context, either leaving the reader unclear about the topic or
providing a set of topic sentences with no amplification or
The theory of given-new information allows the writer to
connect the proposition being developed with those that come before
and after. The writer decides the direction of a proposition (what

is to be said) and also determines which is "given" and which is
"new" information so that the sentence can be linked to the
sentences around it (Cooper and Matsuhashi 1983, 26-30) Sentence
recognition is easier when sentence pairs move from given to new
(Glatt 1982, 97-102; Vande Kopple 1982, 52-55).
Given-new research at the sentence level informs us about
how new information is integrated in memory but does not tell us
how much this model applies to production of larger discourse
structures (Matsuhashi and Quinn 1984, 313). It seems logical to
assume that in a longer organization a writer must still precede
from the known to the unknown. At the sentence level we can tell
that antecedents are necessary. But do we have to have a single
antecedent for a paragraph? A section? How specific must that
antecedent be?
A strict application of the given-new principle can easily
result in sentences that contain references to the sentences that
come before them but do not make sense, as seen in the preceding
example about the football. Rigidly applied, it can also result in
very boring prose. However, the principle is an important one in
its larger implications, indicating that writers cannot suddenly
present new topics or ideas without relating them to what has gone
before in some manner.
The technical writer is under more of a constraint than the
novel or article writer to observe the "given-new contract" because
expectations for technical writing are different. Readers may
expect to have difficulty with a novel or may anticipate that their

questions will be answered in time. Since the primary purpose of
technical writing is to convey information, readers expect to have
new material explicitly linked to old material so that learning of
the information is facilitated. Details must be carefully
structured and their relationship to each other and to the purpose
of the communication made clear.
The distinctions between what is "new" and what is "given"
are based on an agreement between the reader and writer as to what
is known and what is not. But, at a level above that of the
sentence, how does the writer know what is given and what is new
for the reader? All given material cannot be explicitly spelled
out in texts, or we would never reach the new material. And, as
noted previously, readers can infer much given information.
The writer must make assumptions about the readers' levels
of knowledge, based upon an understanding of the knowledge likely
to be possessed by those interested in reading a particular text.
The assumptions that we make are an important part of
communication. They are based upon our knowledge about
communication situations, our estimate of the reader's
understanding, and the reader's estimate of the writer's competence
and authority for conveying particular information. These
assumptions are an important component of speech act theory, which
enables us to look at communication as an action.

Speech Acts
Pragmatics is the branch of linguistics that deals with how
language is used in actual human communication. Pragmatics tells
us what makes utterances acceptable for speakers of a language (van
Dijk 1977, 2). Syntax and semantics, other branches of
linguistics, are concerned with how language is structured and how
words convey meaning. Pragmatics is also the domain of plans and
goals (deBeaugrande and Dressier 1981, 31). Utterances are
designed to communicate; however, without a plan, a speaker may
simply ramble. A speaker may ramble to fill up time, or a writer
may ramble to fill up a given number of pages. They may hope that.
someone else will be able to make sense out of their thoughts or
that sense will serendipitously come to them.
Speech act theory examines the plans that must be produced
to enable us to use language to make events happen. Speech act
theory comprises the context, intentions, attitudes, and
expectations of the speakers taking part in a communication. The
focus of speech act theory is upon the role of an utterance as an
act. It defines a speech act as an utterance act performed by a
speaker in a context with respect to an addressee (Traugott and
Pratt 1980, 229).
Searle defined three levels of speech actions: locutionary
acts produce recognizable utterances, illocutionary acts describe
acts that can be performed only if appropriateness or felicity
conditions exist, and perlocutionary acts achieve the desired
effects (Searle, 1969). In other words, locutionary acts encode a

message; illocutionary acts are defined by the purpose the speaker
has for speaking; and perlocutionary acts achieve the effect the
speaker wishes to have on an audience (Cooper 1984, 113). An
illocutionary act accomplishes a communicative purpose; the author
can represent, direct, commit, express, or declare a meaning.
Since illocutionary speech acts enable one to perform an action
for example, making an arrest, pronouncing a prison sentence, or
declaring a marriagethose participating in the act must know how,
when, and who can perform the act. A marriage performed by someone
with no authority to perform it is not a valid speech act. The
conditions for successful performance of speech acts are agreed
upon in a language community (Cooper 1984, 114-124).
Traugott and Pratt note, "In the area of pragmatics, the
concept of appropriateness plays a role somewhat analogous to the
role played in syntax by the concept of grammaticality" (1980,
232). An utterance must be made in a context that is
linguistically appropriate, and the purpose of an utterance must be
appropriate to one's ability to fulfill it (Kasher and Kasher 1986,
77-78) .
When people write or speak, each utterance is designed to
perform a function. Speakers expect listeners to understand the
function and to act accordingly. However, a description of types
of speech acts proved insufficient, for the same reasons that most
schemes of classification fail. There are always utterances that
fail to fit the categories or utterances that could fit more than
one equally well.

An amplification of speech act theory appears in Grice's
cooperative principle, which indicates that people work together to
create meaning by cooperating with each other. The cooperative
principle is the following: "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 1975, 45) .
Grice enlarged upon this principle with four maxims: (1)
Quantity: Do not give more or less information than required in a
given situation; (2) Quality: Say only that which you believe to
be true and for which you have adequate evidence; (3) Relation:
Make your contribution relevant with respect to the previous
discourse and to the context; (4) Manner: Make your discourse
nonambiguous, nonobscure, well-ordered, and brief (Grice 1975, 45-
56) .
These maxims are equally applicable to writing, which also
can be regarded as a speech act. The cooperative principle assumes
people want to accomplish purposeful communication (Traugott and
Pratt 1980, 237). For example, a person asking the time is
expected to be interested in the answer. A person reading a book
is expected to make an honest effort to understand the information
presented and to understand when a book is not appropriate to his
or her knowledge and skills.
The speaker can vary use of these maxims related to the
purpose of what is being communicated. But the speaker must

remember that too much complexity, redundancy, or subjectiveness
will destroy the communication (van Dijk 197 6, 46) .
The cooperative principle does not demand that all
information be made explicit. In fact, the quantity maxim requires
just the oppositethat we say only what our hearer or reader needs
to know. We can obtain much information by inference, but, our
expectations for conventions lead us to assume that events or
objects mentioned will be discussed at some point and that those
not mentioned can be presumed to occur in ordinary ways (Pratt
1977, 158) .
So that we can know what those ordinary ways are, speech
acts must be anchored in a frame of reference, because
"establishing shared worlds, maintaining coreference, and
distinguishing between what information is shared and what is
unshared in the context are important factors contributing to the
coherence of discourse" (Traugott and Pratt 1980, 283). To
understand a speech act, the hearer has to work to assess the
probable intent of the speaker given the context. Thus, speech
acts cannot take'place without a context, because we use context to
determine meaning.
Context is critical to our understanding of speech acts.
It enables us to make connections with the speaker, determines what
for us is given and new, and sets the expectations with which we
approach a communication. Meaning depends among other things upon

our knowledge of social conventions. For instance, we cannot
evaluate the meaning of certain actions, even actions as common as
shaking hands or a soccer goal, unless we are familiar with what
they represent (Potter and Wetherell 1987, 56-60). The same events
may have a different meaning in different contexts. For example,
biochemists typically adhere to the traditional view of science
hypotheses are made and tested, advances come through objective
studieswhen writing, but advance a different view in
conversation. In conversation, they indicate that advances are
also the result of personal connections and successful use of
politics to garner support for a point of view. The scientists
present events differently depending upon the context of the
communication situation.
The importance of context can be seen at the level of words
as well as utterances. Typically words do not have a one-to-one
relationship to meaning; thus, the context in which words are used
becomes all important (Booth 1986, 462-463). For example, the word
"patient" might mean a person who is in a hospital or a person
willing to wait, depending upon the context in which it is used.
Just as we make assumptions about the meaning of a word in its
context, we also make assumptions about the meaning of a discourse.
A text requires knowledge, intent, and a progression between
states. The text is confined by social conventions (deBeaugrande
1980, 12-13).
The identity of an audience is best understood in terms of
a social context. The social nature of discourse becomes evident:

speech-act theoryand sociolinguistics in generalhas taught
us to see all discourse as representing action performed within
and conditioned by a social situation.... All discourse...can
be described as part of a social transaction that has defined
roles for both writers and readers. (Park 1986, 482)
Language cannot exist without a context, or ambiguity would
overwhelm us. Because of one's knowledge and culture, one can make
informed guesses about the meaning to come (Halliday 1978, 61).
Three variables affect our choice of how to communicate in a
particular context: field (setting), tenor (relationship of
participants), and mode (channel of communication adopted and other
choices related to the use of language in a situation). The
wording and structures that we use will be different when we are at
home or at school, when we are talking with friends or a boss, when
we are arguing or when we are being friendly, when we are writing
or when we are speaking (Halliday 1978, 142-145).
Since many features of language are related to the context
of a situation, they are therefore predictable in many aspects.
Thus, we are able to listen and read with expectations and in fact
we always do so. However, the text can modify our expectations by
explicitly indicating a change in direction or a new emphasis. The
relationship of text and environment shifts constantly; an
essential feature of text is interaction. An understanding of text
consists of a constant compromise among knowledge presented in the
text, knowledge that the reader has available, and the reader's own
cognitive disposition and biases (deBeaugrande and Dressier 1981,

Text Continuity
Text is more than a supersentence. It has its own
structure, which is more than sentences in combination. Although
discourse can be considered as a sequence of individual sentences,
it must also be viewed as "wholes" structured at higher levels than
that of the sentence (Traugott and Pratt 1980, 241).
These "wholes" are created on a number of different levels.
A communication occurrence must meet seven standards: cohesion;
coherence; intentionality; acceptability; informativity;
situationality; and intertextuality (deBeaugrande and Dressier
1981, 3-11) The text must also be efficient, effective, and
appropriate. Many of these requirements are similar to those
implied by the cooperative principle, only at a slightly different
level of detail.
Although all seven requirements are important to
organization, cohesion, coherence, and intentionality are essential
elements. Cohesion includes the overt structures that link
utterances, including recurrence, partial recurrence, parallelism,
paraphrase, ellipsis, and use of referents such as pronouns.
However, cohesion rests on an assumption of an underlying coherent
world view. In fact, it is possible to have coherence without the
use of overt cohesive structures (Witte and Faigley 1981, 197-202).
Coherence depends upon a continuity of sense in which the concepts
activated match the receiver's prior knowledge of the world.
Coherence is a network of knowledge spaces centered around main

This coherence or continuity of sense in part occurs
because text users create a text world (deBeaugrande and Dressier
1981, 94). An acceptable text may not depend upon its correctness
or relevance to the real world but its believability and relevance
to participants' outlook, regarding the situation. For example, a
fairy tale may bear little resemblance to the real world but may be
believable and coherent within the confines of the text world it
Intentionality concerns the plans made by the speaker or
writer. Since the production and reception of texts are
probabalistic operations, not rule-governed ones, the main
consideration for decisions is the context of the discourse. The
speaker makes a plan and predicts the contributions of others. To
carry on a conversation, topic concepts must be activated, and the
participants must have control centers to display main ideas. Far
more is involved than comparing sentences with an all-purpose
grammar. Participants have to know how much knowledge is already
shared or being conveyed and how participants are monitoring or
managing the situation.
Humans apply consistent strategies of structuring the
world. Some of the strategies or orderings we use are spatial,
temporal, linear, sequential, and hierarchical. For example, when
asked to describe their apartment, all subjects used very similar
spatial strategies, staring at the front door and going from front
to back and left to right to describe the rooms (Clark and Clark
1977, 233-236). In a like manner, when asked to tell a story about

a frightening event, all subjects followed narrative strategies
that had many elements and a basic order in common (Pratt 1977, 38-
51) .
Implications for Technical Writing
Linguistics informs us how people use language to
communicate meaning. Although the principles of linguistics have
most often been studied with respect to spoken discourse, they can
be applied to written discourse as well. This application must be
made with the understanding that the negotiation of meaning cannot
be as immediate as for spoken communication. In a conversation, we
can judge our partner's knowledge level and reactions reasonably
accurately by his or her responses. When writing, the audience
cannot be monitored or observed to the same degree; therefore, we
must take into account more variance in the recipients' knowledge
and social context when constructing the message.
As technical writers, we must remember that people store
information in terms of meaning, looking to structure to help them
determine that meaning. We do not say "this is a noun, this is a
verb" when reading, but we notice if the function supplied by such
words is missing. Likewise, we may not consciously identify an
introduction or an amplification of a topic, but we expect such
structures to be present.
Meaning and language construct each other. The construct
of reality is part of our construct of language (Halliday 1978, 1).
The manner in which we express a fact or concept becomes a part of

the meaning that we impart to our audience. Thus, the technical
writer does not merely perform as a scribe, translating content
from technical specifications or engineering diagrams into
grammatical English. Organization, and with organization the
interpretation of material, must come from the writer's intention
and purpose in presenting the material, not strictly from content
(Cooper 1984, 112).
Although content constrains how the writer can present
material, the writer's intention is a major element in determining
what the text's organization will emphasize or downplay. For
example, we would not expect a hazardous waste disposal report to
focus solely on expense of disposal methods without any
consideration of this risks of various methods. However, the
writer can choose to make the expenses or the risks the major focus
of the report.
Macrostructures enable a writer to structure material to
convey his or her intentions. Like the hierarchies postulated in
cognitive psychology, macrostructures are used to relate
subordinate concepts, to signal concepts that should be considered
together, and to indicate the relative importance of various
concepts. The technical writer must ensure that the hierarchy or
structure implied by the macrostructures in the text matches the
intent of the communication. For example, a maintenance manual
might contain the proposition structure shown in Figure 2.

diagnose check brakes check tires check ignition
repair car correct remove defective part repair part replace part
test run test correct if needed
Figure 2. Proposition Structure for Repair Manual
"Diagnose," "correct," and "test" form a list of macropropositions
that are themselves subordinate to the macroproposition "repair
car." To emphasize how to repair particular types of parts, a
different structure would be needed.
Writers also must ensure that the propositions relate to
the macroproposition; a discussion of repair methods should not be
included in the diagnostic section, for example. Proposition lists
should be roughly equivalent. Although there are three
propositions listed for "diagnosing" and "correcting" and only two
for "testing," the overall sense of structure is maintained. A
writer must also be careful not to invent or segment material to
force it to fit a predefined structure.
Logically, the gist of a text could be derived from the
proposition structure making it up. However, a gist is more than a
summary or a list of propositions making up a text. It is also a
representation of meaning in a form that is useful to the reader.
It may appear that gists are not critical in technical material
because technical writers are supposedly presenting only facts.

However, readers must have a framework in which to understand
facts. As seen in the description of the scenario principle
(Flower, Hayes, and Swarts, 1983, 44-49), readers constantly try to
make sense of the facts they are given, attempting to recast
material so that they can determine meaning. They also create real
world examples in an attempt to make the abstract language of the
documents being used relevant to their situation. The nature of
the framework or gist is, of course, dependent on specific content.
But, the gist must be related to the readers' own experience both
in terms of overall meaning and specific details or they will not
be able to use it.
We need to guide the formation of gists, but we can only
restrict them to a certain extent. For example, a novice reading a
linguistics text may develop a very different gist than an expert
would and may have considerably more difficulty in developing that
gist. A structure and a message that seem apparent to a writer may
not be as clear to a reader, because the reader may not have the
benefit of knowledge that the writer assumes the reader has, or the
gist may be lost in the presentation of detail.
What are some typical gists that technical writers might
use? For a maintenance manual, the gist might be how to repair
something, when to repair something, or both. For a research
article on biology, the gist might be that a particular treatment
worked or did not work or that further research is needed. For an
accounting program, the gist might be to present step-by-step
instructions that are simple to use for calculating taxes. Any

approach is feasible, as long as the gist is apparent and the
details of the material are presented in such a way as to support
the gist.
We do not discern gists in a vacuum. We bring our own
world knowledge to the text. As readers, we expect that material
that is new will be related to information we know because
otherwise we have no good way to place it in memory. The notion of
a "given-new contract" holds for larger sections of text as well as
for sentence-to-sentence connections. For example, in a
programming manual, a writer devoted several chapters to various
tasks: "Generating Text," "Sending Text," and so forth. The types
of statements available for these tasks were introduced in a
preceding chapter, "Understanding the Elements of a Program."
These elements were further introduced by describing them as
similar to elements found in other programming languages. The
document's audience of programmers was assumed to have some
familiarity with basic programming concepts; thus, no further
amplification was needed.
The given-new contract, if applied without regard to
meaning, can result in the sequential incoherence seen in the
example about the football in Chapter 2. But, the principle that
we cannot introduce new material without linking it in some way to
something known is a sound one.
A text, however, requires more than skillful creation of
gists and adherence to the given-new principle. A text must be
regarded as a communication, and as such it is subject to the same

rules that exist for spoken communication. Both the writer and the
reader must agree to cooperate; the reader cannot willfully
misunderstand, and the writer must provide information this is not
over- or under-elaborated and that is accurate, relevant to the
reader's needs, and appropriate to the situation.
How do we know when we have met these principles? We can
never know exactly, but we can make approximations. As rules of
thumb, it is typically better to use more rather than less
elaboration and to lean toward being redundant, because as writers
we do not know which sections our writers will choose to read.
However, too much redundancy or elaboration will annoy readers, who
will begin to question why they are seeing the same material
repeated or amplified.
Accuracy is not an simple requirement, since a writer can
easily slant material by its position in a hierarchy or by its
links to other items. Accuracy can only be judged by the writer
and by those in a position to judge technical content (such as
technical specialists) in the context of the situation.
The criterion of relevance instructs the writer to add
enough detail so that the readers understand but so much as to not
to overwhelm them. We can only assess this balance through a
knowledge of the audience and an analysis of our intent in
presenting material. Do we want people to remember many details?
Are there main concepts that should be judiciously illustrated with
examples? Technical writers must continually ask themselves this
type of question. Common sense would dictate that a manual for

novices should contain less detail than one for experts. A
procedural manual for obtaining zoning permits would require
considerably more detail than an announcement warning people that
permits were necessary, for example; a computer manual intended for
word processing novices would contain less detail than one intended
for those who routinely prepare large documents.
Although appropriateness must also depend upon the
situation, a markedly informal tone is rarely appropriate in
technical material. A computer manual that strikes an
inappropriate tone describes a command as "busy" (because it can be
used for many things) and choices as "plain vanilla" or "mocha."
Although these constructions may have sounded "cute" to the writer,
they are tedious when encountered more than once. This type of
writing also impedes communication because the reader, if not
familiar with ice cream flavors or lacking the writer's sense of
humor, is left trying to figure out what "mocha" is and why it is
relevant to the situation. Speech act theory would indicate that
such text violates the principles of relevance and appropriateness.
The context in which particular communication strategies
are used affects whether a text is judged relevant and appropriate.
We often assign meaning based upon elements external to a text,
such as our own knowledge or the conventions of a situation. This
context sets up expectancies for the reader that the writer must be
aware of. As mentioned previously, we do not expect a computer
manual to be "chatty." A writer preparing an article for a science
journal can assume that certain conventions will be expected.

In making our communication convey our intent, we must also
ensure that it maintains continuity. If the links between
sentences, thoughts, paragraphs, and sections are missing, either
in terms of meaning or explicit connectives, the reader will have
difficulty in understanding why elements were included. He or she
may conclude that perhaps the writer did not know either,
undermining the writer's authority and impairing the communication.
We may find that standard patterns of speech acts, such as
directing, committing, or representing, can be related to
rhetorical structures such as cause/effect or analogy (Cooper 1984,
123). Linguistics has not yet dealt with the reasons for the
effectiveness of certain patterns. At this point, linguists are
just becoming aware that they exist. However, since people can
communicate successfully, there must be a limited set of powerful,
regular strategies at work (deBeaugrande and Dressier 1981, 220).
This chapter makes several major points for the technical
writer. We must structure gists for our readers, making it clear
how micropropositions relate to macropropositions and making sure
that propositions relate to propositions already activated. We
must be aware of the demands of the given-new contract and ensure
that new material is related to material already known. We must
remember that writing is a speech act and as such must follow the
cooperative principles of speech, mindful not to either bore or
overwhelm the audience while making the communication relevant to
the audience1.s interests. We must also take into account the fact
that communication is only meaningful in a context. We cannot

assume the others automatically understand the same context we do,
but must be careful to establish common ground and assumptions from
the outset. Once established, context must be maintained.
Linguistics informs us of the importance of intent when
structuring text. We do not simply provide structure for
structure's sake, but use that structure to both to convey and to
support our intent. Intent is of critical importance when
structuring communications because a person tries to go beyond the
surface to figure out what the speaker intended him to understand
(Clark and Haviland, 1977).
The concept of intent adds an additional dimension to the
requirements for structure and schemata described by cognitive
psychology. Structure, although essential, by itself is not
enough; the structure must be used to convey a meaning. That
meaning is derived from the author's purpose iri writing and what he
or she expects the audience to gain from the communication. We
must take this knowledge into account in developing gists, in
planning structure, and in matching our communication to the
context and knowledge level of our audience.
Although linguistics informs us of the need to shape our
structure and provides general principles regarding requirements
for communication, it does not provide specifics on which
structures are appropriate for which purposes or how to create
these structures. These specific approaches have been the province
of rhetoric. In our examination of rhetoric, we will review some

approaches that have been proposed and investigate whether they
compliment or contradict the insights gained from linguistics.

Rhetoric is the art of speaking or writing effectively.
Unlike psychology and linguistics, which illuminate general
principles of communication, rhetoric traditionally has been more
concerned with providing rules for practice in particular
situations. Since rhetoric is oriented toward production of
effective communication, it leans toward recommendation of specific
This emphasis on the practical and upon specific strategies
brought rhetoric into disrepute at the close of the nineteenth
century. The strategies came to be regarded as a set of techniques
used to persuade people to accept positions often of dubious truth
or benefit. Best signified by the phrase "empty rhetoric," the
discipline came to stand for an art that emphasized technique over
substance, rules over creativity. The emphasis on technique and
the doubtful benefits of those techniques made rhetoric an unfit
subject for departments of English, which became more interested in
the study of literature and left the mechanics of communication to
other disciplines, a trend that continued through the first half of
the twentieth century (Connors, Ede and Lunsford 1984, 2-11). Only
in the last twenty-five years has interest in the application of

rhetorical principles to writing instruction revived in departments
of English.
Technical writing has been affected by the trends in
rhetorical instruction. Often technical writing has been reduced
to a set of rules that mandate style, emphasize mechanical
correctness, and ensure that writing meets standards established by
readability formulas (Lutz et al. 1987, 300). Specific techniques
have been elevated and forms have been developed to try to limit
human error (Whitburn 1984, 235). However, rule-based systems have
not met the needs either of communication in general or of
technical communication specifically. The emergence of a wider
community and a changing community with access to technical
materials is forcing us to rethink traditional approaches and to
provide technical writers with more broadly based guidelines
(Redish and Schell, in press).
And, in fact, modern interpretations and amplifications of
classical theoretical theory are enabling us to do just that. This
chapter provides a synopsis of the major principles of classical
rhetoric and then discusses the modern interpretation and use of
several of these principles, among them invention, the
classification of texts, the role of subject knowledge, and the
role of audience. Because rhetoric does focus upon practice, it
also provides numerous specific recommendations for the technical
writer, some of which are examined here. Finally, this chapter
discusses the implications of these principles and practices for
the technical writer.

Classical Rhetoric
Aristotle defined the principles of classical rhetoric in
his Rhetoric, which systematized and summarized classical thinking
about the use of speech to persuade. According to Aristotle,
meaningful communication is an attempt to persuade an audience
(Baum 1988, WE109). Aristotle found that a successful speech was
the result of the interaction of three factorsthe speaker, the
audience, and the subject (Cooper 1932, 16). The nature and
knowledge of the audience become critical elements to consider in
designing a communication, since different strategies are best
suited to different types of audiences and subjects.
To classical rhetoricians, form and content were
interdependent. Gorgias noted that knowledge is incomplete,
relative, and uncertain, contingent on circumstance and human
nature. Since knowledge is not an absolute, the manner in which
material is presented shapes perception. Cicero also felt that
form and content could not be separated (Katz 1987, 362-365).
Thus, the appropriate form of development had to be selected for
the topic at hand.
Classical rhetoricians postulated that certain strategies
were best suited to particular types of speaking: the example for
deliberative speaking (encouragement or dissuasion) and the
enthymeme for forensic speaking (accusation or defense). The
enthymeme is a partial syllogism that presumes on the audience's
acceptance of unstated assumptions. These strategies provide
reasons for the amount of detail or example to be used and take

into account the knowledge the audience possesses before hearing
the speech.
Aristotle's Rhetoric discussed three essential elements of
a text: invention, style, and arrangement. Of particular interest
in studying organization are invention and arrangement. Style,
while of concern to writers, is not the primary focus of this
In his Rhetoric. Aristotle gave most attention to
invention. Appeals can be made to reason, the emotions, or
morality. To structure these appeals, Aristotle defined some 28
topoi, or topics. These topics are lines of argument that can be
used when developing a persuasive argument. Examples of these
topics are cause and effect, comparison, definition, and part to
whole. The topics "constituted a method of probing one's subject
in order to discover possible ways of developing that subject"
(Cooper 1932, 35). Aristotle's discussion of invention involves
structure, since some of the topoi imply structurefor example,
developing a discourse around a definition, cause and effect,
opposites, or part to whole.
In his discussion of arrangement, Aristotle touched upon
the need for structures that aid the listener's memory. In his
view, narration requires a prologue and an epilogue, both of which
aid memory. The prologue organizes the information for the
the proem gives a hint at the plot so that we may promptly know
what the story is about, and our minds not be left hanging
since the indefinite is bewildering. So when the teller puts

the gist of the action into your hand, as it were, he enables
you with this hold to follow the story. (Cooper 1932, 222)
A summary or conclusion also assists the audience in remembering
what has been said. Thus, Aristotle recognized the need for a gist
to organize a discourse and the need to reiterate this gist so that
the audience would recall it.
Much of Aristotle's advice on the construction of narration
sounds strikingly similar to Grice's maxims. Like Grice, Aristotle
indicates that one must make a selection from facts to show a
point, since too many details become hard to recall. Facts that
are generally known need not be repeated. Narration must not be
too long or short; the author must say just so much as will make
matters plain. This duplication suggests that there are enduring
principles of communication that can be agreed upon across
different disciplines, even though specifics must vary with
different topics, different times, and different audiences.
Many of the concepts currently being discussed by
psychologists and linguists were known and advocated by Aristotle,
although he developed them from a practical perspective, observing
their effect rather than analyzing the structure of the mind and of
language. Much of what Aristotle says is directly applicable to
technical writing today (Baum 1988, WE108-110). Unfortunately,
later rhetoricians often regarded his work as a set of rules to be
applied in specific situations. Modern rhetoricians are now
examining principles developed by Aristotle to determine how
rhetoric can be used to advise modern communication.

Modern Rhetoric
Modern rhetoric has taken the principles of Aristotle and
other classical rhetoricians and amplified and expanded upon them,
examining their applicability to today's discourse. Among the
issues that have been examined are the role of invention, the
usefulness of classification schemes such as the topoi and others,
the role of the writer's knowledge in shaping a communication, and
the role played by the audience in communication.
Modern rhetoric stresses a return to invention (Freedman
and Pringle 1980, 108-182; Young 1978, 34-35; Young 1980, 54-55).
In the 19th century, emphasis on invention declined and the topoi
were "displaced as inventional strategies and later appear as
organizational principles" (D'Angelo, 1984, p. 63) Invention
became less important because people came to believe that ideas
existed before the act of writing and the duty of the writer was
simply to order them and put them down. However, it later became
apparent that in many cases problems arose, not because people
could not order their ideas, but because they had nothing to say
(Cooper and Odell 1978, xi-xii).
Modern rhetoricians have developed numerous strategies for
inventionfrom Larson's classical questions, to tagmemics, to
Burke's Pentad, to Flower and Hayes' heuristics (Lauer 1984, 136).
These strategies all encourage the writer to develop ideas. Larson
develops a set of questions based upon the classical toppi.

Tagmemics asks the writer to look at a subject from different
aspectsas a static particle, as a part of a dynamic process, and
in relationship with a larger context. The Pentad requires the
writer to define an action, agent, agency, scene, and purpose. The
writer in effect must be able to fill in the blanks. Flower and
Hayes also develop a set of questions for the writer to answer.
These strategies have different purposes and must be used
appropriately; for example, some guide the writer to produce
information from memory, while others can be used to manage
research data (Lauer 1984, 135-136) .
However, modern rhetoricians do not tell us how to go from
the ideas generated to the way in which they are best structured.
Both inventional and structural models may be necessary (Winkler
1983, 120). Inventional models suggest ways to create information,
guiding the writer in generating the substance of discourse.
Structural models, like the schemata for prose forms discussed
earlier, provide suggestions for structure, aiding the writer to
give form to the substance. Models cannot be prescriptive, for
"both structural and inventional models should serve as heuristics,
rather than algorithms. These models do not guarantee success;
they simply increase the writer's chances of consistently producing
more effective discourse" (Winkler 1983, 120).
Strategies for invention such as the classical topoi and
the modern strategies discussed previously can help writers to
develop and structure ideas. However, writers must remember that
these strategies are only guidelines. Invention should be more

than filling in the elements of a form or answering a list of
questions, although when students are being taught to write
acquiring skill in mastering a form often takes precedence over
using invention to examine one's ideas (Gage 1984, 167). For the
experienced writer, both the invention strategies and the resulting
forms should arise from a desire to solve a communication problem.
The classical topoi, instead of a prescription for specific
situations, should be a way to discover answers to questions (Gage
1984, 158).
Classification Systems
There is a danger is using classification schemes such as
Aristotle's topoi because the rules for a particular topic may be
simply applied without a careful analysis of their appropriateness
for a particular situation. Nevertheless, it is useful to attempt
to classify types of discourse, since typically different
communication strategies are used for different discourse types.
Numerous classification schemes exist, designed to help the writer
examine a piece of prose and determine its type in order to ensure
that the requirements for that type of prose are met. These
classification schemes are of two kindsthose based on the purpose
of the discourse and those based on its organizational features.
And, in fact, the key elements of analyzing longer documents are
purpose and structure (Harris and Witte 1980, 94) Purpose should
define the other elements, for ideally, "the aim of a discourse
determines everything else in the process of discourse" (Kinneavy
1971, 48).

Kinneavy (1971, 18-24) suggested that the interaction of
the writer, the reader, and the subject of a text determines the
nature of the discourse. The importance of each of these three
factors has a different weight depending upon the aim or purpose of
the discourse.
According to Kinneavy, there are four aims for discourse:
expressive, reference, literary, and persuasive (1971, 63) .
Reference writing, of most interest to technical writers, includes
scientific and informative writing. In reference writing, the main
concern of the discourse is the reality under consideration, that
is, the subject or the text. Reference writing demands a logical
organization, which can be one of the following: deductive (set up
a theorem of logic and draw inferences from it); inductive (draw a
conclusion from particulars); or inverted induction (give the
generalization first followed by the particulars that are its '
evidence). Although reference writing demands factuality and
comprehensiveness, facts cannot organize themselves; they must be
given some type of an organizing principle (Kinneavy 1971, 160) .
After a writer choses an organizing principle suited to the aim of
the discourse, he or she can amplify the organization by using one
or more of the four modes of discoursenarration, description,
classification, or evaluationeach of which has its own organizing
Another classification scheme divides writing into three
basic categoriesthe transactional, the expressive, and the poetic
(Britton 1978, 18-19). Transactional writing is writing to get

things done. Expressive writing presents the writer's reactions,
feelings, preferences, speculations, and so forth. Poetic writing
makes the language the focus of attention. The range from
expressive to transactional involves writing in the role of a
participant, that is, as one involved in the action, while the
range from expressive to poetic involves viewing writing in the
role of a spectator, where one contemplates rather than acts on the
events presented. Organization is more evident in transactional
writing than the other types because the writer must make it
possible for the reader to comprehend and if desired act upon the
message being conveyed by setting up a coherent movement for the
Although our classification of text types is fuzzy, people
do make classifications and use these classifications in processing
information (deBeaugrande 1980, 196-199). People base their
classifications on the situation, topic, and knowledge being
addressed. Among the text types we recognize are descriptive,
narrative, and argumentative text. Since readers activate
different processing controls for different types of text, these
expectations may limit the material that can reasonably occur in
different text types.
Given that a number of discourse theoristsMoffett and
D'Angelo in addition to Britton and Kinneavysuccessfully identify
categories of writing, it seems evident that basic categories of
writing do exist and that the task of writing changes when the
function changes (Kinneavy 1980, 47-50). Classifications are not

necessarily ironclad and may overlap; however, a classification can
assist a writer in identifying the modes or syntax of discourse
most appropriate for particular discourse aims.
For example, in writing a document that could be classified
as reference writing, we would want to ensure that a logical
organization existed (deductive, indictive, or inverted inductive).
We would also strive to ensure that facts were organized in light
of an organizing principle, that is, they supported a topic of
discussion rather than being presented as isolated pieces of
information. In writing a scholarly paper, the writer might want
to examine the balance between an expressive and a transactional
presentation, making sure that the expression of personal thoughts
and opinions was supported by transactional sections providing a
basis or a rationale for that expression.
The classification schemes discussed previously focused on
the purpose of the discourse and derived organization from the
purpose. Classification schemes also exist that attempt to
describe the linkages existing in an entire document without regard
for the content of the document.
For example, Haswell (1986) designed a classification of 14
patterns of organization, basing his scheme on an examination of
structure. In his view, discourse can be either "chained or
"unchained" and either symmetrical or asymmetrical. An unchained
discourse consists of one overall logical unit, while a chained
discourse consists of two or more unchained patterns with each new
pattern grafted on to the end. In symmetric patterns, elements are

all members of the same class; in an asymmetrical pattern they are
not. This classification system ideally should force the writer to
examine the logical patterns actually followed in a piece of
writing and determine if they are used consistently. Like the
classification schemes based upon purpose, this system might be
useful in enabling a writer to look at a completed document and
determine if it met the system's requirements for connectives and
In addition to understanding that there may be general
requirements for different types of documents, the writer must also
be aware of constraints created by the topic being written about
and by the conventions of particular disciplines and subject areas.'
Knowledge of Topic and Conventions
Organization is shaped by the writer's knowledge of the
subject matter and knowledge of appropriate discourse structures.
When you write about something you know, you are more likely to
write fluent, well-organized prose (Newell and MacAdam 1987, 157) .
As Aristotle tells us, "whatever the subject on which we have to
speak or reason...we must have some knowledge, if not a complete
one, of the facts. Without it, you would have no materials from
which to construct an argument" (Cooper 1932, 156) .
However, it is not just the quantity of knowledge but the
quality or organization of prior knowledge that has an impact on
the structure of writing (Flower and Hayes 1984, 142) If we do
not have top-level structures and hierarchies of knowledge for a
particular topic, we cannot present them to our readers. Too

often, we assume that writing skills will simply transfer from one
subject area to another. Although grammar and syntax skills appear
to do so, the interaction of context and content of a rhetorical
situation may not (Teich 1987, 196-1978;). Unfortunately, "no
specific subject matter
ey et al. 1985, 85). There
knowledge and the ability to
theorist has defined operationally how
knowledge helps shape composing" (Faigl
are few studies on the relationship of
generate text based upon that knowledgej.
We also need to be able to write about different subjects
according to the conventions of particular disciplines or discourse
communities (Faigley et al. 1985, xi).
know the conventions of various discipl
different disciplines, for good reason,
Technical writers need to
ines and to understand that
may have different
conventions. For example, disciplines 'such as biology, philosophy,
and art history have very different conventions. An examination of
such conventions can help writers learn
essential relations of
form, content, and audience (Moore and Peterson 1986, 467-470) .
Role of the Audience and the Social Context
Essential to Aristotle's concept of rhetoric is adapting
the discourse to the audience. He identified different types of
audiencesthe old, the young, the poor
appeal that affects each. Rhetoricians
present have emphasized that prose must
needs of the audience. However, many novice writers create
"writer-based" prose. Such prose is defined as follows:
, the richand the kinds of
from Aristotle to the
be fashioned to meet the

the organization of a piece of writer-based prose faithfully
reflects the writer's own discovery process and the structure
of the remembered information itself, but it often fails to
transform or reorganize that knowledge to meet the different
needs of the writer. (Flower and Hayes 1981, 372)
Structure appears to be one of the major differences
between writer-based and reader-based prose (Flower 1979, 27-29) .
Creation of structure is in large part determined by the goals that
writers set for themselves. Experts generate goals in response to
the rhetorical situation. Novice writers generate the most ideas
in response to the writing topic.
Although some expert writers may not appear to think much
about their audience, it may be because they know the conventions
and the audience they are writing for so well. For example,
Selzer's (1983, 183-184) study of the composing process of an
engineer found that the engineer revised very little, but spent a
great deal of time in planning and knew his subject area and
audience very well. He was virtually filling in the slots in a
formula. Similarly, an experienced article writer did not revise
much in creating an article, spending more time in planning
(Faigley et al. 1985) It may be that the expert simply knows
moreabout strategies, about topics, and about a prospective
audience. Expert writers surveyed by Flower and Hayes may simply
have had more strategies to fall back upon, both more knowledge
about how one writes in specific situations as well as topic
Strategies must adapted to the audience. For example, the
importance of structure appears to vary with the audience. Readers

that are familiar with a topic are more likely to notice the
presence or absence of structure that those less familiar with it
(Pace 1982, 16). In addition to the idea structure of a text,
aspects such as the purpose for reading, the audience for the text,
the interest the reader has in a passage, and the situation in
which a text is read also affect a reader's comprehension and
In constructing a text, we cannot lose track of the
importance of the situation. Prose does not exist by itself. In
fact, all language acts can be regarded as social events that occur
in a specific context (Halliday 1978, 9-16). The context and a
person's past history with similar context influence the reader's
interpretation. Kinneavy (1971, 22-23) notes that discourse is
characterized by individuals acting in a specific place and time
and that these constraints also affect the nature of the text. The
writer should develop a clear picture of the audience and its
expectations for the type of text being encountered.
Strategies for Developing Organization
The preceding sections discussed principles for technical
communication derived from classical and modern rhetoric. This
section amplifies upon those principles by discussing strategies
for developing organization that can be applied to technical
writing. Among the strategies included are organization models,
paragraph construction, and use of topic sentences.

Organization Models
A number of specific methods for organizing technical prose
have been advocated by professionals active in the technical
writing field. A model outline can serve as a guide for organizing
routine communications (Hays 1982, 4-5). Among the models Hays
provides are outlines for a progress report, trip report,
experiment, letters, and meeting minutesnone of which typically
pose especially complex organizational problems. A model may be
less effective when the demands of the situation are not typical.
Another strategy, the sentence outline, enables the writer
to carefully analyze the structure of a piece of text while
composing it (Plung 1982, 8-9) Since each point and subpoint is
expressed as a complete sentence, the writer is forced to make
logical connections explicit and to think in terms of specifics
rather than placeholders.
Although the content outline (whatever its form) is
advocated by many in the field of technical writing, it may serve
to emphasize the pattern of the thing rather than the thing itself.
Outlines make organization predominant, but may leave out other
important features (Whitburn 1984, 242-243). Writers may focus on
the demands of completing the outline or use a standard outline
rather than experimenting to see if a new structure would be more
suitable. Too often, the outline becomes the entire solution
rather than part of the solution. Other factors, such as style,
use of illustration and analogy, and visual aspects, to name a few,
must also be considered.

While an outline may help in creating a structure, a writer
must also have an organizing principle. Several models have been
suggested for such an organizing principle.
When presenting technical descriptions, masses of detail
are often included that overwhelm the reader (Britton 1982, 14) .
Although engineers often want to tell everything they know on a
subject at once, real skill in technical description "requires the
fine art of omission and delay" (15). A "control statement" for an
entire document should be developed. This statement, which can be
explicit or implicit, should present the most important aspect of
the subject to which everything else will be related. For example,
in a product description, the most important aspect might be a
unique feature of a product that differentiates it from other
products of the same type.
A problem-solving model can also be used to provide an
organizing principle. This model duplicates the steps actually
taken to solve the problem (Annett 1982, 12-13). The model starts
by identifying the problem and moves through steps required to
select, implement, and test the best alternative. Since this
organization results in an inductive strategy, a careful abstract
and summary are also recommended.
A contrasting model for an organizing principle for
technical material indicates that a report should not be a summary
of the scientists' process; it should be a creation of the process
the reader must go through to understand (Samuels 1982, 309). A
report should open with a conclusion, using the what and why to

provide a context for the how. This model is appealing because
facts in isolation make no sense without a purpose, conclusion, or
unifying factor. In fact, many authors find that a deductive
structure is typically most effective, that is, a structure that
presents the conclusion first, followed by the supporting facts
(Hays 1982, 5; Plung 1982, 10; Britton 1982, 15; Huston and
Southard 1988, 180).
A generalization can also be a powerful organizing feature
(Britton 1978, 25). The generalization can indicate why particular
facts are included or not included and can also indicate why a
reader should be interested in a particular collection of facts.
Recognizing that a model is only a typical approach that
must be adapted to the situation, Hays (1982, 4) feels that model
outlines can serve an important function for unskilled writers,
suggesting approaches and aiding them to group ideas and check for
inclusion of vital parts. Similarly, Annett (1982, 13) states that
the intent of suggesting a model is not to produce a "fill-in-the-
blank" format but to provide a starting point for writers. Thus,
we must be ready to assist novice writers in moving beyond the
models as soon as they can handle the models effectively.
A number of rhetoricians have examined the structure of the
paragraph. This examination is useful to the understanding of the
organization of longer texts because some of the techniques may
also be useful for longer structures.

One structuring method familiar to both students and
professionals is to set up questions and then answer them: "The
prevalence of questions both in professional writing and student
prose suggests that question asking is as natural in written
discourse as it is in conversation" (Smith, 1984, p. 21). However,
in writing, the writer has to provide both the question and the
answer, unlike participants in a conversation. Smith sees the
paragraph as a unit of implied discoursetypically a question and
its responseand identifies several question structures that can
be used in a paragraph. For example, the writer may answer a
question the reader can be expected to have, answer a question the
writer poses, or provide several parallel answers to a question.
A paragraph can also be unified by centering it about a
particular term (Markels 1983, 454-464). This method implies a
concept of form relatively independent of meaning. Like Smith,
Markels identifies different structures that can be developed for a
paragraph, but uses one term or several terms as a unifying device.
Typically major terms create one or two chains and other terms must
be related to those. Although a paragraph can move linearly, there
must be a vertical structure as well as horizontal movement.
Otherwise, this situation can occur: "John likes cheese. Cheese
is made in Wisconsin. Wisconsin is the home of my best friend"
(Markels 1983, 461).
Other methods also attempt to make explicit the structure
of a paragraph. Warner (1979) uses a system in which the elements
of a paragraph he definestopic sentence, limiting sentences, and

development sentencesare represented mathematically, enabling
students to see how the basic elements of paragraphs are related
and how they can be transformed to produce different effects. Nold
and Davis (1980) identify the sentences in a paragraph as
superordinate, subordinate, or coordinate to one another and define
a system of movement for a paragraph, noting that one cannot return
to a subject after the movement of the paragraph has passed it by.
These structural systems do not take into account larger
organizational schemes such as the modes of organization or the
topoi. They may be useful in that they give us a common vocabulary
for talking about and improving writing (Warner 1979, 155). They
may more be useful as analysis and editing tools, but do not seem
as valuable as generative aids (Eden and Mitchell 1986, 318). It
is probably most useful for the technical writer to use these
structures in examining created prose to make sure that terms are
related, for example, or that a subordinate idea has meaningful
links to a superordinate idea.
Some authors have argued that a paragraph really is not a
unit of organization, but instead is a visual device, and that
actual organization spans more than a paragraph (Eden and Mitchell
1986, 417-419). Although paragraphs are not necessarily units of
meaning, they are needed to meet the reader's expectations.
Paragraph expectations are governed by the type of text one is
reading (Eden and Mitchell 1986, 317). Newspaper stories have more
and shorter paragraphs than philosophy texts, for example. Readers
also expect paragraphs to be unified and they expect the initial

sentences of a paragraph to orient them. In fact, "even a unified
paragraph can be ineffective if its initial sentences do not orient
the reader so that the unity is readily apparent" (Eden and
Mitchell 1986, 425) Illustrating the effect of sentence and
paragraph placement, Eden and Mitchell organize the same paragraphs
to provide a factual and then a more inflammatory stance, providing
support for their position that organization, both of paragraphs
and larger structures, flows from the purpose of the discourse.
Topic Sentences
Although the point has been made that often professional
writers do not use topic sentences (Braddock 1974, 299-301),
research summarized by D'Angelo (1986, 438-439) indicates that
readers tend to remember topic sentences and that they remember
more of what they read when the topic sentence is first in a
paragraph. If a writer does not present topic sentences or a
similar form of structure, readers "must engage in the time-
consuming process of making inferences and constructing topical
propositions of their own." (p. 436).
Because Braddock's (1974, 299-301) study, which found that
professional writers do not use topic sentences, surveyed articles
in popular magazines, his findings may not be valid for scientific
and technical writing. A study of academic research articles in
various subject areas found that topic sentences were used
extensively (Popken 1987, 223-225). Articles in the areas of
humanities and the social sciences had more explicit topic
sentences than engineering and scientific articles. However,

engineering and scientific articles typically used very short
paragraphs and more headings than the humanities and social science
articles. It appeared that in technical and natural science
writing the role of the topic sentence was in part replaced by
headings. The necessity for topic sentences may be greater in
academic than popular material because the information is dense and
the relationships and logic of the reasoning must be made clear to
the readers.
The importance of topic sentences is supported by other
studies (Selzer 1982, 294-298; Huckin 1983, 97). However, the use
of topic sentences must be planned with the audience in mind.
While the nonspecialist may rely on structural features like
headings and topic sentences, explicit aids may not be needed by
specialists and in fact, may get in the way (Huckin 1983, 98) .
Implications for Technical Writing
Do the lessons from rhetoric apply to technical writing?
Are invention, awareness of various types of texts, and attention
to audience important to technical writers or are these concerns
valid only for other types of writing? Is technical writing simply
a matter of learning various conventions and filling in the slots?
Invention is a major aspect of classical rhetoric and of
its modern interpretation. Invention should be a set of strategies
for exploring a topic, not hard and fast rules for particular
situations. The question arises, does technical writing require
invention? Often the content is provided for the writer, in the

product specifications, the physical product, or the data from an
investigation. Invention in this case must be regarded not so much
as the generation of ideas, but the examination of relationships
and appropriate levels of detail in the material available and the
generation of a proper form and organization. Invention strategies
can help the writer ensure that there are no holes in the
information, that questions that will occur in the reader's mind
have been answered, and that details do not overwhelm the main
point of the discussion.
The technical writer must be wary of simply using a set of
invention strategies such as Aristotle's topics or Flower and
Hayes' questions. Invention, uncontrolled, can create a set of
unconnected ideas unless the ideas, once they are generated, are
then carefully arranged and related to one another. Arrangement
and selection of information must be based upon invention guided by
a controlling purpose that dictates the subordination of parts, the
amount of detail to include, and the order of presentation.
Do classification schemes based on the purpose of discourse
serve to identify this controlling purpose? As seen earlier,
readers have schemata for different types of discourse and their
understanding will be hindered if the conventions of these schemata
are not met. Classification schemes based on the purpose of the
text can be useful in suggesting characteristics that must be
present for particular types of text. However, classification
schemes as broad as Kinneavy's or Britton's describe types of text
that exist. They do not provide direction, except of a most

general nature, about how to produce effective communication when
using the various types of text. Kinneavy (1971, 152-159)
describes organization strategies such as deductive and inductive
organization that can be used for reference writing and cites the
need for a controlling purpose to organize facts, but does not
provide specific examples of situations calling for various
organizations. He presents examples of existing organizations, but
does not tell us how to use these examples to generate
organizational schemes for new circumstances. And indeed, this may
not be a desirable or even reasonable goal. Since by their nature
rules cannot be defined to cover every contingency, rhetoricians
prefer to suggest guidelines that can be carried over to other
similar situations and to stress the need for an evaluation of each
unique communication situation.
Although aids for invention do appear to help writing, it
is an open question whether writers really make use of explicit
structural knowledge while composing (Faigley et al. 1985, 30). It
is not certain whether classification schemes actually help writers
in creating text and whether they in fact reliably identify
features of real writing (Odell, Cooper, and Courts 1978, 8-9) .
And, it is not clear whether writers primarily acquire these
schemes as tacit knowledge or whether they can be taught.
Classification schemes based on structure are probably more
useful for analyzing and discussing existing prose than for
generating material. Usually when we write, we put down our ideas
first, and then we determine whether their presentation is

sufficiently consistent with the forms and structures we have been
taught. Technical editors may find it helpful to use these schemes
when discussing text with writers; writers may be able to use them
to analyze their own prose in order to revise it, but a fair degree
of sophistication would be required. We tend to see the
subordinations or chains in our work that we intend to be there,
and it usually takes a fresh and objective eye to apply the
structural classification schemes and determine whether their
requirements have been met.
In designing documents to meet readers' expectations, the
technical writer has the benefit of conventions already established
for particular scientific and technical discourse communities.
These conventions often limit the choices a technical writer has,
unlike the fiction or "poetic" writer, who is allowed greater
flexibility. However, these conventions may prove inadequate for
particular communication situations. An experienced computer
audience may be bored or irritated by a tutorial, for example. In
using these conventions, the technical writer must constantly
evaluate their appropriateness, ensuring that they are modified, if
necessary, to meet changing needs and circumstances.
As evidenced by Huston and Southard (1988, 179-180), old
rhetorical structures may not be sufficient for new content. For
example, organizing the information in software manuals often
involves much more than using traditional rhetorical patterns of
organization. Many manuals must be designed so readers can first
use them linearly to learn and then access them as a reference

work. The needs of the users and the purpose of the manual often
dictate a task-oriented organization, mirroring the way the users
will use the software. Although a writer may use traditional
rhetorical patterns in sections and paragraphs, typically these do
not serve as an organization for the entire piece. We may have to
generate new rhetorical structures following the principles upon
which the old were developedexamining the audience's needs and
fashioning structures accordingly.
Invention strategies and organization development methods
such as model outlines and content outlines can lead to a overly
narrow view that does not take into account all aspects of a
communication problem. Models and outlines developed from
solutions to previously encountered communication problems are
useful as background information but cannot be completely adequate
to solve the unique problems presented by each new context for
writing (Lutz et al. 1987, 308) Rote structures may lead us to
use our knowledge of past communication solutions as a sufficient
guide for new situations (Whitburn 1984, 244). Although this
danger must be recognized, a structured method of approach can be
very useful, particularly for the novice technical writer. For the
novice, it is probably better to use a tested solution that fits
the situation approximately than to attempt a totally new solution
that may or may not communicate with readers. As a writer gains
experience, adapting to meet a situation becomes more possible
because he or she is familiar with the solutions that have already
been tried and the shortcomings of these solutions for particular