What Makes a Text Coherent?

Author(s): Betty Bamberg

Source: College Composition and Communication, Vol. 34, No. 4, Coherence and Cohesion: What
Are They and How Are They Achieved? (Dec., 1983), pp. 417-429
Published by: National Council of Teachers of English
Stable URL: http://www.jstor.org/stable/357898
Accessed: 25-05-2015 05:38 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

National Council of Teachers of English is collaborating with JSTOR to digitize, preserve and extend access to College
Composition and Communication.

http://www.jstor.org

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
What Makes a Text Coherent?

Betty Bamberg

Pedagogical interest in coherence has its roots in the nineteenth century, its
probable beginnings in Alexander Bain's first rule of the paragraph: "The
bearing of each sentence upon what precedes shall be explicit and unmistake-
able."' By the end of the nineteenth century coherence, along with unity and
emphasis, was an established canon of paragraph structure. The view of
coherence in some of today's popular composition texts still closely resem-
bles Bain's original formulation. For example, McCrimmon's Writing With a
Purpose, one of the most widely used freshman composition texts, defines
coherence as follows:
A paragraphis coherent when the reader can move easily from one
sentence to the next and read the paragraphas an integrated whole,
ratherthan a series of separatesentences.
McCrimmon then advises writers to make paragraphs coherent by weaving
sentences together with "such connective devices as pronouns, repetitive
structures, contrast, and transitional markers.2
Coherence is generally accepted as a sine qua non in written discourse;
writing that lacks coherence will almost certainly fail to communicate its in-
tended message to a reader. Even though most composition texts and
rhetorics have routinely included a section on coherence, interest in this
topic has intensified during the last five years. In part, this increased interest
grows out of linguistic research, which has evolved from a focus on the sen-
tence to a consideration of "texts," or extended sections of discourse. But the
interest also arises from the renewed emphasis on writing instruction and the
recognition, in writing instruction, that many problems in writing require
attention at the level of the whole discourse rather than at the level of the
word of sentence.
In 1975, as we began to see other evidence of a decline in students' writing
skills, the National Assessment of Educational Progress (NAEP) reported

Betty Bamberg is Director of the USC/California Writing Project and Director of Special
Programs for Freshman Writing. She has published essays in Researchin the Teaching of English,
CollegeEnglish, and California English. The work reported in this essay is part of a larger study
on coherence in the writing of thirteen- and seventeen-year-olds, funded by the NCTE Research
Foundation.

College Composition and Communication, Vol. 34, No. 4, December 1983 417

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
418 and Communication
CollegeComposition
that the average level of performance on one of their writing tasks had de-
creased significantly between 1969 and 1974. Students had been asked to
write an impromptu essay describing something they knew about "so that it
could be recognized by someone who has read your description." NAEP's
preliminary analysis of these essays concluded that lack of coherence con-
tributed substantially to the lower scores received by the 1974 essays.3 Al-
though NAEP's analysis of these essays identified coherence as a major prob-
lem, it could offer no guidelines for writing instruction because the essays
had not been analyzed to determine what difficulties in the texts might be
described as "lack of coherence." To identify these difficulties, I recently
read over 800 essays written by thirteen- and seventeen-year-olds for the
1974 Assessment. I was particularly interested in the essays written by
seventeen-year-olds because their writing was likely to be close to the level
of writing of freshman composition students.
My first reading of the essays made clear that the traditional view of coher-
ence derived from Bain was too limited to account for the coherence prob-
lems I found. This traditional view treats coherence as a phenomenon some-
what similar to what many linguists and rhetoricians now call cohesion. In
Cohesion in English, Halliday and Hasan define "cohesion" as a relationship
between two textual elements in which one is interpreted by the other. For
example, in the sentence "He said so," "He" and "so" presuppose something
that has gone before. We cannot interpret this sentence (as opposed to de-
coding it) unless we can relate "He" and "so" to words in a preceding sen-
tence. Such relationships between words create cohesive "ties" and allow us
to differentiate sentences that constitute a "text" from sequences of unrelated
sentences. Halliday and Hasan define a text as "a semantic unit: a unit not of
form but of meaning ..." that "may be anything from a single proverb to a
whole play."4
However, this view of "cohesion" differs from the traditional concept of
coherence as presented by Bain. His formulation of "coherence" stressed
between-sentence connections that created tightly-structured, autonomous
paragraphs which were then linked together into a larger text by transitions.
As discussed by Halliday and Hasan, however, "cohesive" ties often connect
adjacent sentences (or more precisely, adjacent T-units), but they may also
connect "remote" rather than "immediate" sections of text. In a "remote" tie,
the cohesive elements are separated by at least one t-unit and may extend
across paragraph boundaries. "Cohesion," therefore, describes a linguistic
system that extends through the text and binds together larger chunks of
discourse, in addition to forming smaller discourse units.
Research and theory in discourse analysis now view cohesive ties as part of
what makes a text coherent; however, these ties are not, by themselves, suffi-
cient to create coherent text. After analyzing cohesive ties in student essays,
Steven Witte and Lester Faigley differentiated between cohesion and coher-
ence, concluding that "coherence defines those underlying semantic relations

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
WhatMakesa Text Coherent 419

that allow a text to be understood and used" and "coherence conditions are
governed by the writer's purpose, the audience's knowledge and expecta-
tions, and the information to be conveyed."5 The text linguist T. A. van Dijk
points out that cohesive ties create only "local" coherence (his term for cohe-
sion) and are unable by themselves to create discourse-level or "global"
coherence.6 He illustrates the difference between local and global coherence
in the following example:
I bought this typewriterin New York. New York is a largecity in the
USA. Largecities often have serious financialproblems....

Although this passage contains lexical cohesive ties-repetitions of "New

York" and "large city"-readers will not consider the text as a whole to be
coherent unless they can discover a broader topic that incorporates buying
typewriters, large cities, and financial problems. Van Dijk further notes that
essays, in addition to being unified around a theme or topic, must have an
overall form or structure if readers are to find them coherent over the whole
discourse.7
Research on artificial intelligence has found that we rely heavily on con-
ventional structures of knowledge known as scripts, frames, or schema to
organize experience and knowledge so that we can understand it.8 A script
may prescribe a sequence of actions to be followed in a particular situation.
For example, in a restaurant we expect to order food from a menu, to be
served by a waitress or waiter, and to pay for our meal after eating it. How-
ever, the restaurant script will vary, depending upon the type of restaurant
we select. Buying a hamburger at MacDonald's follows a script different from
ordering a meal at Howard Johnson's, and both differ from the script pre-
scribed at Chasen's or Le Pavillon. Textual information also follows conven-
tional patterns. Labov found that oral narratives could be divided into five
parts: abstract, orientation, complication, resolution, and coda. More re-
cently, a number of discourse theorists have constructed "story grammars"
that describe narrative structure.9 Scientific reports also have a predictable
structure: they begin with an abstract and are then divided into introduction,
method, results, and discussion. Form is more fluid in expository essays, but
certain methods of development-classification, comparison/contrast-tend
to provide familiar structures. Despite their different forms, all schema serve
the same function: they help readers anticipate upcoming textual informa-
tion, thereby enabling them to reduce and organize the text into an un-
derstandable and coherent whole.
The complexity of any inquiry into the sources of coherence is further
increased by recent research on reading and on discourse comprehension,
which shows that the reading process "constructs" a text and that the reader's
prior knowledge-both conscious and tacit-affects the understanding of a
text.'0 Meaning and coherence are not inscribed in the text, this research
shows, but arise from readers' efforts to construct meaning and to integrate

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
420 and Communication
CollegeComposition
the details in the text into a coherent whole. Although readers are guided by
textual cues, they also draw on their own knowledge and expectations to
bridge gaps and to fill in assumed information. Louise Phelps argues that
failures in coherence occur either because writers undercue-provide too
few cues for readers to let them perceive the relationships between parts of a
text--or because they miscue-give conflicting or misleading cues."1
The interaction between text and reader and its effect on judgments about
coherence can best be understood in the context of psycholinguistic reading
theory. To comprehend a text, fluent readers do not read word for word,
but, in what Kenneth Goodman has called a "psycholinguistic guessing
game," predict meanings from graphic, semantic, and syntactic cues, sampling
only enough of the text to confirm their predictions. They predict meaning
from nonvisual information-their prior knowledge and expectations-as
well as from the letters on the page.12 For example, tacit knowledge of the
English language (acceptable letter combinations and their relative frequency
of occurrence, along with syntactic rules) helps readers predict the meanings
of individual words by limiting the possible choices of meaning'3 Similarly, a
knowledge of discourse conventions helps readers predict meaning and struc-
ture. Preschool children have learned that "once upon a time" signals the
beginning of a story, and experienced readers of research reports are able to
read an abstract and then turn to sections of interest, perhaps skipping the
review of research or description of methodology and going directly to the
results and discussion.
While reading, readers draw on their tacit knowledge at the level of the
sentence and of the whole discourse by using a "top-down," "bottom-up"
strategy. That is, as they process individual words and sentences at the begin-
ning of a text, they attempt to form an overall conception of the structure
and meaning of the whole text into which they can fit the information that
follows.14 This anticipation of structure and meaning, in turn, directs their
guesses about individual words and phrases. Clearly stated topic sentences,
an obvious organizational pattern, statements of topic and purpose, and head-
ings which indicate divisions of the text-these are all cues that facilitate a
reader's integration of details in a text into a coherent whole. When such
cues are missing, readers may be unable to make this integration.
Theoretical discussions of coherence-whether of sentences or of a whole
discourse-usually analyze hypothetical texts constructed to highlight the
principles being considered. Although useful for theoretical discussions,
these passages do not resemble the writing of our students, nor do they help
us discover which coherence conditions students have the most difficulty
meeting. Examining the presence or absence of coherence in student texts
such as those written for the National Assessment can not only give us in-
sights into students' difficulties in producing coherent texts, but can suggest
ways for us to intervene productively in students' writing processes to help
them write more coherently.

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
WhatMakesa Text Coherent 421

I had initially intended to compare a coherent and incoherent essays using

ratings from the National Assessment analysis. The original NAEP analysis
consisted of two parts; a holistic general impression score and a detailed de-
scription of specific features (particularly grammatical and mechanical errors)
known as the "mechanics" scoring. Coherence ratings were given during the
mechanics scoring; raters evaluated the coherence of each paragraph accord-
ing to a set of criteria provided. In a preliminary analysis, I discovered that
NAEP's rubric and scoring procedures for coherence rendered many of the
coherence ratings invalid. To begin with, the analysts rated the coherence of
individual paragraphs, so essay- or discourse-level coherence was not rated
unless an essay consisted of a single paragraph. In addition, the coherence
scoring guide did not include all coherence features, omitting particularly any
consideration of overall structure or form, a feature often critical for coher-
ence through the whole essay, and the guide also confounded coherence with
paragraph development.'5 To overcome the limitations of NAEP's mechanics
analysis. I developed a new rubric that assessed coherence at both the local
(or sentence) and global (or discourse) levels. The rubric asked raters to
focus on all textual features that affect coherence and to assign a holistic
score for the entire essay, ignoring paragraph indentations and development.
Results from these holistic ratings were then used to separate coherent from
incoherent essays.'"
The coherence scale I used to rescore the NAEP essays has the strengths
as well as the weaknesses of other holistic scales. Although it reliably rank
orders writing according to a set of criteria, it does not identify which fea-
tures are present or missing in particular essays.'" To identify specific fea-
tures that result in perceived incoherence, I analyzed the coherence features
in a subset of essays, looking for the presence or absence of linguistic fea-
tures or rhetorical structures specified in the rubric. Not surprisingly, essays
with coherence scores in the lower half lacked coherence between sentences
and through the whole discourse:
He had huge body framethat obtain shock, bumps, cuts, but it is also
in some ways powerful, smart,gracefullike a dancer;quick on his feet,
speed like lightning,a eye for directionand a mind that keep sayingpush
on and fear of anybody. Hand that slap feet that are step, leg that are
broken, arm that bend, and a head that get push in the ground, but into
all of those thing a personthat is determineto push on becauseglory is at
the end.18
The perceived absence of local coherence in the essay above is due primar-
ily to mechanical and grammatical errors, a type of miscue not usually con-
sidered in theoretical discussions of coherence. However, such errors do
interfere with a reader's attempt to construct a text and integrate details into
a coherent whole. The writer of the essay above left out thirteen inflectional
endings, including five "ed" past tense verb endings, two third-person singu-
lar "s" verb endings, and eight noun plural "s" inflections. The writer also

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
422 and Communication
CollegeComposition

omitted words ("had huge body frame," "fear of anybody") and used other
unexpected words and phrases--"obtain shocks," "eye for direction," "into
all of those." To process the text, readers must backtrack and reread to fill in
missing words and grammatical inflections. Not only do the errors interfere
with the processing of sentences, but readers' focus on individual words and
letters distracts their attention from global cues.
However, editing the essay above to eliminate grammatical and mechanical
errors would not make it coherent. The essay above illustrates a problem
common even among 17-year-old writers: the writer failed to identify the
topic-to tell the reader who or what the essay is about. Readers can infer
that the essay is describing a large man ("he," "huge body frame") and find
enough details to form some impressions of the man's physical and mental
qualities, but they cannot identify him. The description suggests a fighter, but
is it a particular fighter-Muhammed Ali or the hero of Rocky-or is it some-
one known only to the writer? Writers of incoherent essays often failed to
identify their topic even though the instructions for the writing task re-
minded them to "name what you are describing." These writers apparently
assumed that readers shared the information they possessed and that the
topic was, therefore, "given" information."9 Their descriptive details enabled
readers to infer a superordinate term-man, building, monument, etc.-but
not to identify the topic unless it was a very familiar object or place (i.e., the
Statue of Liberty).
Writers of essays with clear topic statements, on the other hand, regarded
the topic as new information that should be stated at the beginning of the
essay. The most skillful writers not only identified their topic but provided
an introduction that oriented the reader to the situation and placed the sub-
ject in context by identifying time, place, and circumstance.20
When you cross the Golden Gate Bridge from San Francisco to
Sausalito,on your right you see an ominous object in the middle of the
bay. It's an islandcalled Alcatraz.Watersmashesup againstthe rocksand
gulls fly over this godforsaken,lonely, deserted prison. There are many
buildingswhich now lay crumblingto ruins. A lone water tower juts up
from the rest of the building. Once the buildingswere white but now
only gray buildingssun crackedremain. It sits out in the middle of the
San FranciscoBay by itself cryingfor it has been the ruin of manymen. It
was almostunescapableuntil it was shut down. The Indianstook it over
and much of it was burnedand devastatedin the earlyseventies.The only
way to reach it was by small boat. No one was permitted on it unless
given specialgovernmentalpermission.In August 1974 it was opened to
the public,

The description of Alcatraz begins successfully because the writer establishes

the setting (San Franciso Bay) and the tone ("ominous object") in addition to
identifying the topic. More often, writers announced the topic, but expected
the reader to fill in the context:

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
WhatMakesa Text Coherent 423

I am describingan ancientEgyptianstatue of the goddess Bastet.

The statue is made of bronze, and is extremely old, dating back to
perhaps 1087 b.c. The figure of Bastet is slender, with fine contours of
the body. The goddess is cat-headed,with large erict ears and slanting
eyes. She is wearinga tight-fittingsheath-likegarment,which emphasizes
her gracefulfemale body and noble carriage.At her feet are severaltiny
bronze kittens.
The writer of the essay above clearly identifies her topic, even emphasizes it
by setting it off as a paragraph, but fails to create a context for the descrip-
tion. A reader must draw on his world knowledge-ancient statues are usu-
ally viewed in museums-to construct the context. Whether readers can con-
struct a context will, of course, depend upon their background and their
prior knowledge as well as on the information provided by the writer.
The "describe" topic tended to elicit from writers a list of attributes, a
more difficult organizational structure for readers to process and recall than
structures such as comparison, problem-solution, or cause-effect.21 When
writers failed to arrange these details according to an overall plan, readers
often had difficulty integrating the details into a coherent whole. Writers of
coherent essays used several different ordering plans to create a global struc-
ture. One effective arrangement orders descriptive details according to a plan
of movement such as general to particular, whole to part, and container to
contained.22 The essay on Alcatraz begins with such a strategy: the writer
first gives general statements describing the whole complex ("This godforsa-
ken, lonely, deserted prison"), then moves to particular details describing
parts of the prison ("buildings crumbling to ruins," "a lone water tower"). He
does not sustain this order through the essay, however, and the last section
simply lists events that occurred after the prison was closed. The writer who
described Bastet arranges her description more effectively by sustaining the
whole-to-part, general-to-particular plan throughout the essay and by arrang-
ing particular details describing the statue in a spatial order from top to bot-
tom. The reader thus sees Bastet much as he might if standing before a glass
case in a museum-first the whole statue, then her "cat's head," her "graceful
female body," and finally the "tiny bronze kittens" at her feet.
The "riddle" structure was another plan used-with surprising frequency
and varying success.23 In "riddle" essays, the writer deliberately withheld the
identity of the subject, listed attributes that would let the reader "guess" the
subject, and optionally revealed the "answer" in the final sentence. Gener-
ally, riddle essays were judged coherent when the writer finally identified the
topic or included enough description details to enable the reader to infer the
subject. The most successful riddle essays did both, the final identification
confirming the reader's guess.
Other writers organized descriptive details in the form of a "tour." In these
essays the writer seemed to be taking the reader on a guided tour of the
place being described, a strategy that transformed spatial information into a

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
424 CollegeCompositionand Communication

chronologically organized narrative.24 In the essay below, the writer first

created a context for the reader by narrating the events leading up to a family
visit to Pala Dura Canyon and the initial stage of the trip. She began her tour
as her family's car entered the canyon:
As we entered the entrancegates of PalaDurawe could gaze down and
look at the tall trees and cliffswhichwere much lower thanwe were. The
narrowpaved road was built on the side of a mountainand the far side
was protectedby guardrailings.Manysigns were up warningtouristsnot
to park on this road because of the possibilityof falling rocks. As we
continuedon down the roadour ears soon becamestopped up becauseof
the rapiddecreasein altitude.When we finallyreachedthe bottom of the
canyon,we could look up and see what we had seen coming down.
By carefully signalling each stage of the tour through the canyon-"As we
entered the entrance gates"; "As we continued on down the road"; "When
we finally reached the bottom of the canyon"-the writer recreated the visit
for readers, thereby providing a narrative sequence into which descriptive
details could be woven.
The whole-to-part, riddle, and tour plans used to arrange descriptive de-
tails share a common feature: they are structures of knowledge familiar in our
experience and are thus readily accessible to both writers and readers. The
whole-to-part, general-to-particular and spatial orders reflect the way we
normally perceive places or objects-first grasping the gestalt, then observing
particulars, usually in a systematic spatial order. Riddles and tours, both part
of the typical elementary school child's world, are well practiced and familiar
patterns by age seventeen. Writers who failed to draw on these or other
ordering strategies produced essays that were essentially lists of unordered
details:
It is a excitingplace with characterssuch as Mickey Mouse or Donald
Duck runningaround.
The rides are fun, but the lines can become quite long. The rides all
seem to be real. And in the president'sHall of Fame,the presidentsactu-
ally breathe. When Lincoln is giving a speeches, two other president's
turn towardeach other and startwhispering.
The groundsare alwaysclean. There are boys and girls who go around
with a trayon a handleand brooms and pick up everythingeven cigerate
butts.
When you first go in, you buy a book of tickets. And when you ride
something,you give them a ticket. Also, when you leave your motel, you
ride on a monorail.
The motels are nice. They both have heated pools and you can rent
boats (peddle) and ride in the lagoon.
The experence is will worth your time and your money. But it is better
to go duringschool, so it won't be as crowded.
Although many readers can probably infer that the essay above is describ-
ing Disneyland, they will find no structure to help them organize the details.
Even though each paragraph indentation signals a different set of descriptive

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
WhatMakesa Text Coherent 425

details (the rides, the grounds, the motels, etc.), the details follow no logical,
spatial, or temporal order. Placing them in separate paragraphsgives readers
little help in integrating them into a coherent whole.
The failure of 17-year-olds to achieve coherence because of incomplete
announcement of the topic, failure to establish a context, and/or selection of
an inadequate organizational plan may seem surprising, particularly since the
topic makes relatively low-level cognitive demands.25 Essays on more com-
plex topics often lack global or overall coherence because writers have not
yet discovered the main point they wish to make about their subject or have
not yet fully understood the relationships among their ideas. In most of the
essays in this study, however, writers simply failed to give readers necessary
information about the topic or to organize the details adequately.
Writers (students and others) may have difficulties in focusing on topic and
selecting a plan of organization, or in creating a context for their readers,
because they continue to struggle with the production of words and sen-
tences. Fear of error can halt the flow of discourse and cripple a writer's
attempts to project and sustain plans.26 However, many essays lacking
discourse-level coherence were not particularly error-ridden. (See, for exam-
ple, the essay above that describes Disneyland.) These essays fit Linda
Flower's description of "writer-based" prose: retaining an egocentric focus,
ordering ideas with a narrative framework or merely listing them randomly in
survey form, and relying on general "code" words or phrases that draw their
meaning from the writer's individual experience and are, therefore, not ac-
cessible to readers. Flower views the composing of "writer-based" prose as a
functional strategy because it allows writers to avoid overloading the short
term memory, lets them concentrate on retrieving information from mem-
ory, and postpones the task of forming complex concepts or considering the
reader's needs. Flower argues that "writer-based" prose should be seen as a
"half-way place," writing that can later be transformed into reader-based
prose that takes into account the reader's needs and purpose in reading.27
Accordingly, impromptu essays such as these should be regarded as first
drafts, not as the best writing that students can produce.
Nevertheless, the writing elicited by the topic illustrates a full range of
writing skills among 17-year-olds. The better writers were able to take the
reader's perspective from the beginning and produced well-structured, co-
herent essays. This ability seems to be a consequence of both the writer's
skill and experience and the relative lack of difficulty in the writing task.
Based on the holistic ratings of coherence, the percentage of students able to
write coherent or partially coherent first drafts on this topic increased from
39% to 51% between ages 13 and 17, a reflection of the writers' developing
control over written discourse. However, some writers of coherent first
drafts on this topic would almost certainly be unable to duplicate their suc-
cess if given a conceptually more demanding topic. The National Assess-
ment's first analysis of mechanics found that essays written in 1974 contained

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
426 College andCommunication
Composition
fewer coherent paragraphsthan those written in 1969,28and some "back-to-
the-basics"critics interpretedthese results as further justificationfor more
instructionin grammarand in productionof correct sentences. But another
look at the failuresof coherence in these essays indicatesthat the most seri-
ous coherence problemsoccurredover the whole discourse.
The difficulties students experienced in writing coherent essays for the
National Assessmentgive us a startingpoint for identifyingways in which we
can intervene in students'composing processes to help them produce more
coherent texts. Essaysthat have only local coherence problemscan often be
improved by careful editing or proofreading-adding words and phrases to
make relationshipsbetween sentences more explicit or correctingmechanical
errorsthat interferewith the readingprocess. However, essays that are inco-
herent overall almost always require major restructuringand rewriting to
provide the unity, organization,and identificationof context needed by a
reader.
Writers cannot always take the reader's perspective during their initial
planning and drafting, but they can revise their first efforts into coherent
texts if they know what makes a text coherent and know how to revise so
that a text meets those conditionsfor coherence. However, researchon revi-
sion indicatesthat most studentwritersdo not revise effectively.They revise
primarilysentences or words, making minor adjustmentsthat are largely
cosmetic or have a minimal effect on the text. Others, especially weaker
writers,edit their work prematurelyand are unableto sustaintheir focus on a
topic or maintaina plan of organization.Only a few seem able to make the
substantivechangesthatwould convertwriter-basedto reader-basedprose.29
If producingwriter-basedprose is to be for students a functionalgoal of
composing,we must be able to teach them to revise effectively. Findingef-
fective teachingstrategies,however, is not easy. Peer group readingof essays
in progress, one of the most frequently recommended techniques derived
from researchon the writingprocess, promotes effective revision by helping
studentsinternalizethe needs of readers.It not only gives them responsesto
their writingfrom a real audience,but also makes them awareof the needs of
that audienceas readerswhen they respond to essays written by their fellow
students. Peer readers are particularlygood at pointing out places where
more detail or informationis needed in order to orient readersor to fill in
gaps in the text, where terms are unclear,where connections between sen-
tences need to be more explicit, and where digressions or shifts of topic
occur. But although peer readers can often identify poor organizationas a
problem,they are less often able to propose a plan or overalldesign that will
eliminatethe difficulty.
Both compositionteachersand textbook writersrecognize the importance
of organization. But, like peer readers, both groups are often unable to help
students find satisfactory designs for their essays. Perhaps the best known
attempt to propose an essay structure which students can start with has re-

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
WhatMakesa Text Coherent 427

suited in the Procrustean formula for the five-paragraph essay, a prescription

rightly rejected as too rigid and limiting. Nevertheless, the most successful
NAEP essays were those in which writers projected and sustained an overall
design-such as the whole-to-part and tour plans-that controlled the entire
essay. But even though these plans successfully organize descriptive details,
they are likely to have limited applicability to other types of writing. They
could not, for example, serve as adequate designs for the more analytic ex-
pository and persuasive discourse commonly demanded by universities and
businesses. Moreover, the NAEP essays were quite short, and longer essays
attempting more complex tasks may need to draw at various times on several
different plans to organize their information. Writers need to learn practical
and flexible strategies that will help them organize the information in their
essays-even when these are fairly short-and to learn ways to help readers
remain aware of these plans.30
When we look at coherence in its broadest sense, we become aware that
almost any feature-whether seen locally or over the whole discourse-has
the potential to affect a reader's ability to integrate details of a text into a
coherent whole. But although effective writing will be coherent at both
levels, attention to overall coherence must precede most concerns about local
coherence. First drafts are likely to have global or overall coherence only
when the writing task is routine or when it makes relatively low-level cogni-
tive demands. Our goal as writing teachers must be to create a classroom
setting that enables students to understand what makes a text coherent and to
teach them ways of revising their writing to meet those conditions.

Notes

1. Bain's six paragraph rules first appeared in the 1866 edition of English Compositionand
Rhetoric, a manual designed for his classes at the University of Aberdeen. All six rules, adapted
from the 1866 and 1887 editions, are restated in Ned A. Shearer, "Alexander Bain and the
Genesis of Paragraph Theory," QuarterlyJournal of Speech, 58 (December, 1972), 413. For a
discussion of these rules and the development of paragraph theory in the second half of the
nineteenth century, see Paul C. Rodgers, Jr., "Alexander Bain and the Rise of the Organic
Paragraph,"QuarterlyJournal of Speech, 51 (December, 1965), 399-408.
2. Writing With a Purpose, Short Edition (Boston: Houghton Mifflin Company, 1980), pp.
104-110.
3. National Assessment of Educational Progress, Writing Mechanics 1969-1974, Writing Re-
port No. 05-W-01. (Washington, D.C.; U.S. Government Printing Office, 1975).
4. Cohesionin English (London: Longman, 1976), pp. 1-30.
5. "Coherence, Cohesion, and Writing Quality," CollegeCompositionand Communication, 32
(May, 1981), 189-204.
6. See van Dijk's discussion in Macrostructures(Hillsdale, NJ: Lawrence Erlbaum Associates,
1980), pp. 32-46. Local coherence differs from cohesion in that, for local coherence, logical
relationships between sentences need not be explicitly stated. In this essay, I am using Van
Dijk's more inclusive term "local coherence" rather than Halliday and Hasan's term "cohesion."
7. Teun A. van Dijk, Text and Context: Explorations in the Semanticsand Pragmaticsof Dis-
course(London: Longmans, 1977), p. 149.

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
428 CollegeCompositionand Communication

8. Roger Schank and Robert Abelson, Scripts, Plans, Goals and Understanding, (Hillsdale,
NJ: Lawrence Erlbaum Associates, 1977).
9. "The Transformation of Experience in Narrative Syntax," Language in the Inner City:
Studies in the Black English Vernacular(Philadelphia: University of Pennsylvania Press, 1972), pp.
354-396. One of the earliest and best known story grammars is David Rumelhart, "Notes on a
Schema for Stories," in D. Bobrow and A. Collins, ed., Representationand Understanding (New
York: Academic Press, 1975), pp. 237-272. For a recent summary of this research see Teun A.
van Dijk, "Story Comprehension: An Introduction," Poetics 9 (June, 1980), 1-21.
10. Marcel Just and Patricia Carpenter, eds., Cognitive Processesin Comprehension(Hillsdale,
N.J.: Lawrence Erlbaum Assoc., 1977).
11. "Rethinking Coherence: A Conceptual Analysis and Its Implications for Teaching Prac-
tice," in Marilyn Sternglass and Douglas Buttruff, ed., Building the Bridge Between Reading and
Writing, (Conway, AR: L&S Books, in press).
12. "Reading: A Psycholinguistic Guessing Game," in Doris V. Gunderson, ed., Languageand
Reading (Washington, D.C.: Center for Applied Linguistics, 1970), pp. 107-119.
13. Frank Smith, UnderstandingReading (New York: Holt, Rinehart & Winston, 1971).
14. Marilyn Adams and Allan Collins, "A Schema-Theoretic View of Reading," in Roy O.
Freedle, ed., New Directions in DiscourseProcessing,Vol. 2 (Norwood, NJ: Ablex Publishing Co.,
1979), pp. 1-21.
15. In the "1974 Mechanics Scoring Guide" (mimeo, 12 pp., n.d.), paragraph coherence was
defined as "the interconnectedness among sentences and among the ideas of those sentences."
The scoring guide listed four major ways of achieving coherence: (1) consistency in time, voice,
person, number, subject; (2) clear pronoun reference; (3) use of transitional markers; and (4) use
of rhetorical strategies such as lexical repetition, parallel structure, antithesis. Paragraphs were
both coherent and developed if, in addition to meeting the above criteria, they had an expressed
or implied topic sentence that "identifies and limits the central area of concern in the paragraph"
with each subsequent sentence adding to or explaining something about the main idea in an
orderly manner. Paragraphs which were "neither coherent nor developed" according to the
criteria above, along with most one-sentence paragraphs, were grouped into a category labeled
"Paragraph Used." This amorphous term allowed NAEP to group together paragraphs that
lacked coherence and paragraphs in which the indentation was essentially a graphic device. The
number of paragraphs in this last category was used by the National Assessment to estimate the
level of coherence in National Assessment essays. A short form of the scoring guide may be
found in Writing Mechanics 1969-1974, 1975.
16. Essays meeting the following criteria were rated "fully coherent" (4) on the four point
holistic coherence rubric: writer identifies the topic; writer does not shift topics or digress;
writer orients the reader by describing the context or situation; writer organizes details according
to a discernible plan that is sustained throughout the essay; writer skillfully uses cohesive ties
such as lexical cohesion, conjunction, reference, etc. to link sentences and/or paragraphs to-
gether; writer often concludes with a statement that gives the reader a definite sense of closure;
writer makes few or no grammatical and/or mechanical errors that interrupt the discourse flow or
the reading process. Essays that were partially coherent (3) met enough of the criteria above so
that a reader could make at least a partial integration of the text. Essays were rated incoherent
(2) when some of the following prevented a reader from integrating the text into a coherent
whole: writer does not identify the topic and the reader would be unlikely to infer or guess the
topic from the details provided; writer shifts topics or digresses frequently from the topic; writer
assumes the reader shares his/her context and provides little or no orientation; writer has no
organizational plan in most of the text and frequently relies on listing; writer uses few cohesive
ties such as lexical cohesion, conjunction, reference, etc. to link sentences and/or paragraphs
together; writer creates no sense of closure; writer makes numerous mechanical and/or grammat-
ical errors, resulting in interruption of the reading process and a rough or irregular discourse
flow. Essays receiving the lowest score (1) were literally incomprehensible because missing or
misleading cues prevented readers from making sense of the text. Inter-rater reliability on the
holistic rescoring was .85. Although the holistic scores and original NAEP coherence ratings
overlapped, almost 40% of the 17-year-olds' essays containing one or more paragraphs not given
a "coherent" rating by NAEP were reclassified as coherent by the holistic scoring. Most of these
essays contain brief one- or two-sentence paragraphs that were on the topic and fit into the

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions
What Makes a Text Coherent 429

overall organization of the essay but were excluded from NAEP's "coherent" paragraphcategory
because of their brevity.
17. For a discussion of holistic and other types of evaluation scales, see Lee Odell and Charles
Cooper, "Procedures for Evaluating Writing: Assumptions and Needed Research," CollegeEn-
glish, 42 (September, 1980), 35-43.
18. This and all subsequent essays used as examples were written by 17-year-olds for the
1974 National Assessment. Essays are transcribed exactly, including all grammatical and mechan-
ical errors. Essays from the Writing MechanicsAnalysis are available on computer tape from the
Education Commission of the States, Suite 700, 1860 Lincoln Street; Denver, Colorado 80203.
19. Herbert Clark and Susan Haviland, "Comprehension and the Given-New Contract," in
Roy O. Freedle, ed., Discourse Productionand Comprehension,Vol. 1 (Norwood, NJ: Ablex Pub-
lishing Co., 1977), pp. 1-39.
20. Teun A. van Dijk discusses state descriptions (descriptions of places or static objects) and
the importance of creating a context in Macrostructures,p. 35.
21. Bonnie J. Meyer, "What is Remembered From Prose: A Function of Passage Structure,"
in Roy O. Freedle, Ed., Discourse Production and Comprehension,Vol. 1 (Norwood, NJ: Ablex
Publishing Co., 1977), pp. 307-336.
22. Van Dijk, Macrostructures,p. 35.
23. Colleen Aycock and Kevin O'Connor, two graduate students at the University of South-
ern California, first identified the riddle plan in these essays.
24. My use of this term is taken from a study by Charlotte Linde and William Labov, "Spatial
Networks as a Site for the Study of Language and Thought," Language, 51 (December, 1975),
924-939. Almost all (97%) of the New Yorkers asked to describe the layout of their apartment
used a tour plan. Readers of transcribed apartment tours often found them confusing and hard to
follow, perhaps because of the complex spatial relationships. In the NAEP essays the tour plan
was usually an effective organizational structure.
25. In The Developmentof Writing Abilities (11-18), (London: MacMillan Education Ltd.,
1975), James Britton divides transactional writing into seven categories according to level of
abstraction. Within his divisions, most of the "describe" essays would be classified at the second
level, report, and a few at the next level up, generalized narrative.
26. Mina Shaughnessy, Errorsand Expectations (New York: Oxford University Press, 1977)
and Sondra Perl, "The Composing Processes of Unskilled College Writers," Research in the
Teaching of English, 13 (December, 1979), 317-336, document this phenomenon from different
perspectives.
27. "Writer-Based Prose, A Cognitive Basis for Problems in Writing," College English, 41
(September, 1979), 19-37.
28. Writing MechanicsReport, 1975.
29. For research on students' revision strategies see Richard Beach, "Self-Evaluation
Strategies of Extensive Revisers and Nonrevisers," College Compositionand Communication, 27
(May, 1976), 160-164; Nancy Sommers, "Revision Strategies of Student Writers and Experi-
enced Adult Writers," CollegeComposition,and Communication31 (December, 1980), 378-388;
and Lillian Bridwell, "Revising Strategies in Twelfth Grade Students' Transactional Writing,"
Researchin the Teaching of English, 14 (October, 1980), 197-222.
30. In "Toward a Linear Rhetoric of the Essay," CollegeCompositionand Communication(May,
1971), 140-146, Richard Larson illustrates a method of essay analysis that focuses on writers'
purposes as a way of perceiving their organizational plans. He suggests that this procedure may
help students construct plans for their own writing. In a recent paper "Pragmatics of Form: In
This Paper I Will Assert, Dispute, Confirm, Recount, and Recommend," presented at the Con-
ference on College Composition and Communication, Dallas, March 1981, Marilyn Cooper links
a writer's purposes to specific speech acts and discusses the conditions for their successful per-
formance. She argues that a writer's illocutionary intentions can be used to create a pragmatic
form-that is, form conceived as a series of purposes to be achieved rather than simply a series
of topics to be discussed.

This content downloaded from 138.251.14.35 on Mon, 25 May 2015 05:38:34 UTC
All use subject to JSTOR Terms and Conditions