You are on page 1of 31

Text Cues and Strategies Successful Readers Use to Construct the Gist of Lengthy

Written Arguments
Author(s): Marilyn J. Chambliss
Source: Reading Research Quarterly , Oct. - Nov. - Dec., 1995, Vol. 30, No. 4 (Oct. - Nov.
- Dec., 1995), pp. 778-807
Published by: International Literacy Association and Wiley

Stable URL: https://www.jstor.org/stable/748198

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/748198?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

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 support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Wiley and International Literacy Association are collaborating with JSTOR to digitize, preserve
and extend access to Reading Research Quarterly

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Reading Research Quarterly
Vol. 30, No. 4
October/November/December 1995

Marilyn J. Chambliss ?1995 International Reading Association


(pp. 778-807)
University of California, Berkeley, California, USA

Text cues and strategies successful


readers use to construct the gist of
lengthy written arguments

Lengthy written argument is an important form of had few encounters with lengthy written argument may
communication. In newspaper editorials, maga- not know how authors cue claims and evidence in
zine articles, and books, authors offer evidence for lengthy text nor have the necessary strategies to make
claims, their arguments spanning several para- use of the cues to reconstruct a mental representation of
graphs, a number of sections, or even many chapters. the argument.
Work by Gilbert (1991) suggests that readers who suc- Although it is risky to infer possible instructional
cessfully comprehend a lengthy argument must mentally approaches based on what successful readers do, identi-
represent the author's claim and evidence as it is re- fying the text cues and strategies they use is a plausible
vealed across at least several paragraphs of text. Only first step. This work relied on Toulmin's three-part mod-
those readers who have successfully comprehended the el of argument (Toulmin, 1958), Meyer's text compre-
argument will be prepared to weigh the evidence and hension model (Meyer, 1985; Meyer, Brandt, & Bluth,
accept or reject the claim (Kuhn, 1992). 1980), and a framework of text structures (Calfee &
Unfortunately, a sizeable proportion of readers Chambliss, 1987; Chambliss, 1994a; Chambliss & Calfee,
may reach adulthood never learning how to represent 1989, in press) to suggest the text cues and strategies
the claim and evidence in a lengthy written argument. used by competent readers to construct the gist of a
Test results from the National Assessment of Educational lengthy argument. Comprehension of written argument
Progress (NAEP) suggest that the majority of adults in has not been as thoroughly examined as comprehension
the U.S. do not competently comprehend written argu- of other text types (e.g., Meyer & Freedle, 1984). Though
ments longer than a few paragraphs (Kirsch, Jungeblut, it is true that researchers have developed models for
Jenkins, & Kolstad, 1993). Part of the problem may be how people of various ages weigh evidence and accept
lack of practice. Textbook writing, the genre with which or reject claims (e.g., Kuhn, Schauble, & Garcia-Mila,
school-aged readers have the most contact, rarely incor- 1992), they have typically not used reading tasks nor fo-
porates well-written argument of any length. In a survey cused on how people comprehend an argument in the
of nine social studies textbooks (Calfee & Chambliss, first place. Furthermore, comprehension work, customar-
1988) and 12 science textbooks (Chambliss, Calfee, & ily employing texts of one or two short paragraphs (e.g.,
Wong, 1990) for Grades 4-12, Calfee and his colleagues Hare, Rabinowitz, & Schieble, 1989), has neglected to fo-
found only a handful of arguments, each of which was cus on how readers comprehend lengthy text. Neither
no more than a few paragraphs long. Readers who have comprehension research nor work on argument reason-

778

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
I ABTRATS

Text cues and strategies successful readers use to construct the gist of lengthy written arguments
COMPREHENSION OF written argument has been less thoroughly grade advanced placement English students read arguments varying
examined than comprehension of other text types, even though, ac- in argument structure, content familiarity, and argument signaling
cording to National Assessment of Educational Progress results, and completed either written survey or think-aloud protocol com-
many adults cannot competently comprehend lengthy written argu- prehension tasks. Both text structure and signaling in introductions
ments (Kirsch, Jungeblut, Jenkins, & Kolstad, 1993). Three experi- and conclusions consistently influenced their responses. They used
ments based on Toulmin's (1958) argument model, Meyer's (1985) these cues to (a) recognize the argument structure in a lengthy text,
comprehension model, and Calfee and Chambliss' (1987) text design (b) identify the claim and evidence, and (c) construct a gist repre-
model identified text cues and comprehension strategies used by sentation. Outcomes have both general theoretical value and practi-
competent readers comprehending lengthy arguments. Eighty 12th cal implications for designing instructional materials.

Pistas textuales y estrategias usadas por los buenos lectores para construir el tema de argumenta-
ciones escritas extensas

LA COMPRENSION de la argumentaci6n escrita no ha sido exami-


zados de ingles de 12Q grado leyeron argumentaciones que varia-
ban en estructura argumentativa, familiaridad de los contenidos y
nada tan profundamente como la comprensi6n de otros tipos de tex-
to, si bien seguin los resultados de NAEP (Evaluaci6n Nacionalsefialamiento
de de argumentaci6n y completaron tareas de compren-
Programas Educativos), muchos adultos no pueden comprender si6n por medio de informes escritos o protocolos de 'pensar en voz
competentemente argumentaciones escritas extensas (Kirsch,
alta'. Tanto la estructura del texto como los sefialamientos en las in-
Jungeblut, Jenkins & Kolstad, 1993). En tres experimentos basados
troducciones y conclusiones influenciaron consistentemente sus res-
en el modelo de argumentaci6n de Toulmin (1958), el modelo puestas.
de Usaron estas pistas para a) reconocer la estructura argu-
comprensi6n de Meyer (1985) y el modelo de disefio textualmentativa
de en un texto extenso, b) identificar la afirmaci6n y la
Chambliss y Calfee (Chambliss, 1987), se identificaron pistas tex-
evidencia y c) construir la representaci6n del tema. Los resultados
tuales y estrategias de comprensi6n usadas por lectores competentes
tienen valor te6rico general e implicancias pricticas para el diseio
de materiales didacticos.
al comprender argumentaciones extensas. Ochenta estudiantes avan-

Texthinweise und Strategien von Lesern, die erfolgreich den Kernpunkt langatmiger Argumentationen
erfassen wollen
DAS TEXTVERSTEHEN geschriebener Erirterungen ist vergleichs-
die sich hinsichtlich ihrer Argumentationsstruktur, inhaltlicher
weise weniger untersucht als das Verstehen anderer Textsorten, Verwandtheit
ob- und Argumentmarkierung unterschieden und legten
wohl laut NAEP-Ergebnissen viele Erwachsene das kompetente entweder einen schriftlichen Bericht vor oder bekamen mondliche
Aufgaben zum Textverstehen. Sowohl die Textstruktur und Hinweise
Lesen lingerer Er6rterungen nicht beherrschen (Kirsch, Jungblut,
Jenkins, Kolstad 1993). Drei Untersuchungen, die auf Toulmins
in den Einleitungen als auch Schlugfolgerungen beeinfluSten ihre
Argumentenmodell (1958) und auf dem Textmodell von Caffee und
Antworten. Sie benutzten die Schlisselstellen, um a) die
Chambliss (1987) beruhen, sollten nun Textschlissel und
Argumentationsstruktur im Langtext zu erkennen, b) Absicht und
Selbstverstdndnis des Textes zu erfassen, und c) das Wesentliche
Verstehensstrategien aufkldren, die von kompetenten Lesem zum
wiederzugeben. Die Ergebnisse haben generellen theoretischen Wert
Verstehen lingerer Abhandlungen eingesetzt werden. 80 fort-
und praktische Implikationen, um Lehrmaterial zu konzipieren.
geschrittene Englischschiiler der 12. Klasse lasen die Er6rterungen,

779

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
ABSTRAC,

_tS- 5-Ef AL~~t\d c~E a~~~~~;~`:Qae-5 ; ~ \a

NAEP(National Assessment
?( ?/ of
7"-? Edu-
s T-A ? N 4.%?1f -W -Dk r
cational Progress) Oa*S2 If48 0 AfiAMOA AViON )
t LG4:4

StvC 74!- t7 (Kirsch, rt 0)A W kT 60


Jungeblut,
Jenkins, & Kolstad, 1993). Toulmin
7) A3 M)L jmo)) ? T i , -j U -Ct
(1958) a6Di ( -F- a --E ?f:)L,
Meyer(1985) 0a Y)?f:
"Jz) a>-,
tz &) / 0)c~
I~ tJ~tf~5 ~5f
70) 'U *fA t-a-:
AM I
"=E )t5 , LZUT Chambliss YN tr) 01 L m *a
Calfee
(1987)6D "' x K f+ EMC.
0) -C & 7cD 0

t tz 5_- o YI,,, OD U L OD -A, !,

Indices textuels et stratggies des bons lecteurs pour saisir l'essentiel d'arguments dcrits d'un texte long
ON S'EST moins interesse ' la comprehension de l'argumentation structure argumentative, la familiarite du contenu, le signalement
&crite qu'a celle d'autres types de textes, meme si, suivant la argumentatif,
NAEP et ont effectue des tiches de comprehension utilisant
(Programme National d'Evaluation de l'Education), de nombreux
un questionnaire &crit ou un protocole de verbalisation de la pensee.
adultes ne comprennent pas bien les arguments ecrits d'un texte
Leurs reponses ont et? fortement influences par la structure du texte
long (Kirsch, Jungblut, Jenkins, & Kolstad, 1993). Trois experiences
et par les signalements de l'introduction ou de la conclusion. Ils ont
basees sur le modele argumentatif de Toulin (1958), le modele de
utiliseces indices pour a) reconnaitre la structure argumentative dans
la comprehension de Meyer (1985), et le module de planification un texte
de &tendu, b) identifier le propos et les preuves, et c) construi-
texte de Calfee & Chambliss (1987), ont identifie les indices textuels
re une representation de ce qui est essentiel. Les retombtes ont une
et les strategies de comprehension qu'utilisent des lecteurs experts
valeur theorique genmrale et des implications pratiques pour la fab-
pour comprendre des arguments 6tendus. Quatre vingts 6tudiants rication du material didactique.
d'anglais de 120 annie ont lu des arguments qui variaient suivant la

780

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 781

Figure 1 An argument text diagra

Evidence Warrant Claimhn

Tiny mouse bites snake)


on back repeatedly
until it drops the mouse
that it is holding in its
mouth.

Animals that attack a


much larger predator
show amazing courage.

Birds that attack a


Two geese fight off a Animals can show
predator show amazing
fox. amazing courage.
courage.

Mother animals that risk


their lives to save their
babies show amazing
courage.

A mother skunk
battles a flooded
river to rescue her
trapped babies.

ing has demonstrated the text cues and strategies that ventional structure in written arguments as specified in
successful readers use to comprehend lengthy argument. the rhetoric and a rhetorical schema used by readers to
comprehend argument text. Indeed, at least some current
The Toulmin argument model textbooks in rhetoric and composition explain argument
The study of argument has a long, prestigious histo-structure according to the Toulmin model (e.g., Hairston,
ry. Beginning in antiquity, rhetoricians have analyzed the1992; Ramage & Bean, 1995).
features of effective spoken and written argument (e.g., Figure 1 depicts the structure in a written argument
Aristotle in Cooper, 1932; Bain, 1867; Brooks & Warren, according to the model. Based on the discipline of ju-
1972; Ramage & Bean, 1995). The rhetoric is typically en- risprudence as more representative of real-world argu-
countered today in composition books (e.g., Barnet & ments than syllogistic structures, Toulmin's model has
Stubbs, 1990; Crews, 1992; Hairston, 1992), which uni- three parts: a claim (an assertion analogous to a plea of
formly include chapters on how to craft an argument ac- guilt or innocence), evidence (facts and examples), and a
cording to these rhetorical features (Calfee & Chambliss, warrant (a law-like generalization that links the two). In
1987). Focusing on reasoning instead of writing, logicianssome respects the warrant is the most innovative feature
have developed the logical calculus to represent formalof the model, but also the most elusive and inexplicit.
relations among argument parts (Alexander, 1969; Mates,Toulmin conceived it as analogous to the major premise
1972). The philosopher, Stephen Toulmin (1958), ex- in a classical syllogism and proposed that the warrant is
plained that as logic has become more mathematical, itthe underlying reason offered for accepting or rejecting
has moved away from real-world arguments. Van Dijk the evidence as support for the claim. As such, it often
and Kintsch (1983) suggested that Toulmin's (1958) argu- makes or breaks the argument. Consider the warrant in
ment model could effectively characterize both the con- Figure 1, "Animals that attack a much larger predator

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
782 READING RESEARCH QUARTERLY October/November/December 1995 30/4

general
show amazing courage." The reader who is reluctant to background knowledge, text knowledge, a
attribute human traits, such as courage, to animals repertoire
may of comprehension strategies assuming a sta
reject this warrant and the claim that it underlies.toward a text with particular features within a social
Toulmin (1958) proposed that the claim-evidence-
text (Ruddell, 1994). This study highlighted certain fe
tures of these models and downplayed others. I hav
warrant structure characterizes an argument independent
of content. Brooks and Warren (1972), echoed by
already hypothesized that competent readers have
ground
Nickerson (1991), point to an interesting ambiguity in knowledge about argument specified by th
Toulmin
the word "argument" itself. As used by the logician or (1958) model as qualified by Pollard (1982
Becauseofthe purpose of argument is to communicat
rhetorician, argument refers to a reasoned justification
some claim. It can be used for any content, as the exam-
author's thinking, I expected readers to take an effer
stance as proposed by Rosenblatt (1978, 1994), whe
ple in Figure 1 would suggest. In common parlance,
they would search for text cues so that they could ab
however, argument is a verbal fight between disputants
(Brooks & Warren, 1972; Nickerson, 1991) where the structure, and remember the claim and its evi
stract,
dence as a mental representation of the text's gist
content is necessarily contentious. The goal for both
(Meyer,
combatants, whether face to face or linked as author and 1985; Spivey, 1987). How well and in what
the text
reader by a text, is to win the argument rather than to cued the argument would also be important.
Readers can be strongly affected by text characterist
accept or reject a claim through reasoning (Nickerson,
(Dole,on
1991). Nickerson's notions suggest that readers intent Duffy, Roehler, & Pearson, 1991), an effect tha
winning an argument have no need to represent the au-
should be particularly potent for readers who have as
thor's argument accurately-to comprehend it-in order
sumed an efferent stance. In order to highlight reade
to weigh evidence and accept or reject a claim. gument
Their knowledge, comprehension processes, and
influence of text features, the study did not focus on
goal, instead, is to build a case for their own "side,"
picking and choosing from the argument the evidence reader general knowledge or the social context f
ther
and claims that match whatever they hold to be true reading.
and
ignoring the rest. According to this second conceptual- Although it does not address written argument per
ization of argument, the text depicted in Figure 1 se,would
a comprehension model proposed and tested by
Meyer and her colleagues (Meyer et al., 1980; Meyer &
not be considered contentious enough to be character-
ized as an argument. Freedle, 1984; Meyer & Rice, 1982) suggests specific
Even reasoned arguments may have important con-
strategies that successful readers may apply when they
tent characteristics, although not of a contentiousassume an efferent stance. According to the model, com-
nature.
In a review of research on reasoning, Pollard (1982) con- know a set of rhetorical structures, or
petent readers
cluded that people typically recognize as claimsschemata
asser-(van Dijk & Kintsch, 1983; Meyer, 1985), that
tions for which they cannot readily remember eitherauthors use to organize text (Meyer & Freedle, 1984). If a
confirming or disconfirming examples or scenarios reader can identify one of these structures in a text, he or
(Tversky & Kahneman, 1973). They can comprehend an for its parts and link them together into a
she can search
text representation. Readers who have used this structure
argument when the warrants and evidence are familiar.
strategy recall the text as a gist, a summary of the impor-
Pollard explained that if the claim is familiar, people
may fail to recognize the presence of an argument. tantIf
content organized comparably to the organization in
warrants and evidence are unfamiliar, they may relythe texton (Kintsch & van Dijk, 1978; Meyer et al., 1980).
what they already know rather than construct a Teaching
repre- less competent readers to find the author's
sentation of the argument presented to them. structure improves their comprehension (Armbruster,
According to Toulmin's model and Pollard'sAnderson,
qualifi- & Meyer, 1991), lending support to the model.
cation, I hypothesized that an unfamiliar claim-familiar Figure 2 displays a three-stage model based on a
synthesis
evidence-familiar warrant pattern appears in written text of Toulmin's (1958) and Meyer's (1985) no-
as a set of argument cues. I also proposed that thetions. pat-
The model portrays the strategies used by compe-
tern characterizes an argument schema known bytent readers to construct the gist of a lengthy written
suc-
cessful readers. I avoided contentious content thatargument.
mightIn this study I hypothesized that competent
cue readers to case build rather than comprehend the
readers use strategies specified by the model.
author's claim and support. In the first stage, as soon as the reader notices an
unfamiliar claim-familiar evidence-familiar warrant in a
Reader strategies: A comprehension model text, he or she uses an argument schema to identify the
Current models of reading comprehension text (see as an argument. Meyer and Rice (1982) discovered
Ruddell, Ruddell, & Singer, 1994) depict a reader with
that readers rely heavily on structure cues, such as the

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 783

Figure 2 A comprehension model


according to three stages

Recognizing the Read text


argument stage
no?

yes?

Argument recognized.

Identifying the claim Identify as a H Ignore


and evidence stage

not match ?
Evaluate facts and examples.

match ?

Identify as E

( Claim and evidence identified.

Constructing an argument Use warrants to relate and


representation stage to relate

Evidence supports entire claim? Evidence supports different


Sparts of claim?

Argument representation constructed.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
784 READING RESEARCH QUARTERLY October/November/December 1995 30/4

passage of time, and signal words, such as "in compari-


structure in the text (e.g., Meyer et al., 1980; Meyer &
son," to recognize the structure in a short text. They
Freedle,at-1984). The structure of the recall is affected by
tend particularly closely to text openings, such asthe
titles
structure of the text (Meyer & Freedle, 1984). Lorch
(Meyer & Rice, 1982). Current research in argument and Lorch (1985) also demonstrated that accurately sig-
suggests that successful readers may be able tonaling
use a text's structure in an introductory paragraph im-
Toulmin's argument structure in the same way. Kuhn proves the accuracy of the recall. Signaling in a conclud-
(1992) focused on whether people could provide ingevi-
paragraph should have an even stronger effect since
dence for their beliefs (or claims). The majority of her could use it to repair any "mistakes" in their rep-
readers
resentation. I hypothesized that competent readers use
160 participants offered plausible scenarios to support
their claims rather than actual examples or facts,warrants
appar- to link the claim and evidence they have identi-
ently unable to distinguish between claims andfied evidence.
into a coherent representation of the argument; that
However, participants in her study who had a college
their recall matches the argument's structure; and that
education typically did offer genuine evidence for their
their accuracy is affected by signaling in the concluding
claims. Readers who can distinguish these two argument
text paragraph.
parts may be able to recognize them as argument cues
in text, particularly if the claim and evidence areExtending
sig- research results to lengthy text
naled by phrases such as "I claim that..." and "EvidenceResearch thus far has used short texts, focusing on
that...." what readers do with sentences within paragraphs for
During the second stage, the reader searches thefor
most part (e.g., Hare et al., 1989). The reader who
and identifies the claim and evidence in the text. Two would comprehend a lengthy argument must be able to
types of text cues may influence this search: whether the look for text cues and perform strategies on amounts of
claim is present in the text and whether certain facts and text larger than sentences. I hypothesized that competent
examples support the claim. The claim in an argument readers notice the same cues and use the same strategies
may be analogous to the topic sentence in a paragraph. for lengthy texts of at least seven paragraphs as they
Consistently, readers are more accurate at identifying the have been found to use for paragraphs of several sen-
main idea in a short text when the topic sentence is tences.

present than when it has been omitted (Bridge, Belmore, Calfee and Chambliss (1987; Chambliss, 1994a;
Moskow, Cohen, & Mathews, 1984; Hare et al., 1989). Chambliss & Calfee, 1989, in press) have proposed a
Apparently, when no topic sentence is present, readers taxonomy to be used with texts of any length. The tax
often try to "add up" the subjects and predicates in the onomy has three wellsprings. The first comes from th
paragraph, and construct a superordinate that subsumesrhetoric. By surveying college composition books, C
everything (Kieras, 1980; Williams, 1984). Not surprising- and Chambliss (1987) identified several common pat
ly, their accuracy suffers. In this study I hypothesized terns, which they distinguished by the author's presu
that competent readers identify claim statements present purpose: to inform, to argue, or to explain (Chambliss
in written arguments. They infer absent claims. 1994a, 1994b; Chambliss & Calfee, 1989, in press). T
The search for evidence should be affected by second wellspring comes from Simon's (1981) sugge
whether particular facts and examples match the claim- that all human artifacts (including written text) can b
evidence-warrant relationship in the overall argument. characterized according to separate elements, linkag
Successful readers should be able to distinguish betweenamong the elements, and a theme to hold the design
evidence and nonevidence. Again, research on para- gether. Extending these notions to written text, Calfe
graphs is suggestive. Readers as young as sixth grade and Chambliss (1987; Chambliss, 1994a) characterize
can accurately distinguish between detail sentences that texts of any length to have elements (sentences, para-
either match or mismatch the rhetorical structure in a graphs, sections, chapters) linked (the text's structure
paragraph (Englert & Hiebert, 1984). I hypothesized that according to a theme (the author's purpose). The thi
competent readers identify evidence and distinguish it exploits the usefulness of graphics to represent compl
from nonevidence, justifying their answer by referring to relationships (Tufte, 1990). The taxonomy specifies
the structure of the argument. graphics such as topical nets, matrices, hierarchies, an
In the third stage, the reader constructs a mental Toulmin's model to depict the elements, linkages, a
representation of the author's argument whereby the theme in a text's design (Chambliss, 1994a, 1994b;
warrant links the claim to the evidence. One of the most Chambliss & Calfee, 1989).
consistent research findings to support the Meyer model Figure 1 demonstrates the approach. Each insta
is that successful readers asked to recall a text produce a of evidence in the figure represents an entire paragra
summary of important content organized to match the of details, which concludes with a sentence that exp

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 785

es the warrant. An ed theintroductory


hypothesis that successful readers construct the parag
claim. If the modelgist ofin Figure
an argument 2 familiar
by using an explicitly stated is corre
warrant to link the claim and evidence
successfully comprehends the together accord-
argumen
ognizes it as an argument,
ing to the particular configurationfinds
in the text. It varied the cl
duction, realizes that each
whether the text adhered paragraph
to a simple argument structure
instance of evidence,
or a more complex and uses
variant, whether explicit warrantthe wa
each paragraph to statements
link connected
thethe claim toparts
the evidence, whetherinto a
sentation. warrants were familiar or unfamiliar, and whether the text
conclusion summarized the argument. Readers produced
a summary recall of the argument as a measure of the
Overview of study methods mental representation they had constructed.
This study included three experiments, one for
Participants
each stage in the argument comprehension model: rec-
ognizing the argument pattern, identifying the claim and The 80 participants in the study were members of
evidence, and constructing a gist representation. Two three advanced placement English classes, one class
general types of text cues suggested by research re- from School 1 and two classes taught by the same
viewed above varied across these experiments: patternteacher from School 2. Both schools were in the same
characteristics and signaling. Measures varied by experi- middle-class suburban community of 175,000 in the San
ment depending on the particular strategies being stud- Francisco Bay area, California, USA. The classes tended
ied. The same students from three 12th-grade advanced to reflect the ethnic makeup of the schools, with approx-
placement English classes completed the comprehension imately 50% of the students belonging to minority
tasks for all three experiments. groups with Latino, Asian, or Indian surnames. School 2
Eperiment 1: Recognizing the Argument Pattern drew from a slightly higher socioeconomic area than
tested the hypothesis that competent readers would rec- School 1 and typically had higher means on the
ognize as an argument a text that followed the unfamiliar
California Assessment Program test. In the year of this
claim-familiar evidence-familiar warrant pattern and notstudy, the mean of School 1 in reading was 278, with a
otherwise. Texts were structured either as arguments or percentile ranking of 72. The mean of School 2 was 322
informational topical nets (Chambliss, 1994a; Chambliss with
& a rank of 93. Scores were on a scale of 100-400.
Calfee, in press), claims were unfamiliar or familiar, The 12 females and 14 males from School 1 had
claims came early in the text or in the conclusion, and applied for the class. Their selection was based on their
the pattern was signaled with individual signal words or standardized vocabulary and comprehension test scores,
not. Readers indicated whether they could recognize an prior grades in English, and teacher recommendations.
argument by choosing the author's purpose in writing the These separate indices were combined into a single
text (to argue or to inform), predicting the pattern they score, and the top 26 scorers in the school were chosen.
expected throughout the text (to provide evidence for the The 26 females and 28 males from School 2 had been
claim or information about the topic), and completing selected by a screening process that began in the fresh-
several measures that demonstrated whether their mental man year according to standardized vocabulary and
representation matched the argument or the information- comprehension test scores, which determined their eligi-
al pattern. bility for the freshman honors English class. Each subse-
Experiment 2: Identifying the Argument Parts tested quent year, their English teacher had decided that they
the hypothesis that successful readers would identify a should continue in the honors program, culminating in
claim explicitly stated in the text, infer a missing claim, their placement in advanced placement English for their
and identify as evidence all examples and facts that senior year. I chose these students as representative of
matched the argument pattern. This experiment em- the most successful readers produced by the public
ployed arguments exclusively and varied whether the school system.
claim was explicitly expressed in a text, whether individ- The high school English program was similar for
ual paragraphs within the text matched the argument both schools. During the 4 years, composition and litera-
pattern, and whether introductions and conclusions sig- ture instruction were typically integrated. However, in
naled the evidence accurately. Students listed the claim the 10th grade, students studied composition as an iso-
and evidence to show whether they had identified them lated discipline. During a presession, the majority of the
and explained why the author had included particular students from both schools reported to believe that they
facts and examples in the text (as evidence or not). had learned about text patterns, including argument, in
Experiment 3: Constructing the Argument's Gist test- this composition class.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
786 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Across the three experiments, readers were ran- EXPERIMENT 1:


domly assigned to one of two groups. Most readers
Recognizing the Argument Patter
were in a survey group where they completed paper/
Experiment 1 focused on the first stage of the
pencil tasks for the entire study. I expected survey re-
sults to provide an estimate, albeit somewhat prehension model. I designed passages and compr
superficial, of text cues and strategies used by the sion measures to identify the text cues and strategi
group
as a whole. The two teachers administered theused by competent readers recognizing the argum
survey
tasks to these students in their classrooms. A smaller pattern in a lengthy written text.
group of students met individually with either me or
one of my colleagues in another room to complete Method
think-aloud protocols, 8 students for the first two exper-
iments and 6 students for the third experiment. The
think-aloud technique has been successfully used to Participants
identify specific text cues and strategies unknowable Seventy-one students participated in this experi-
from survey data (Afflerbach & Johnston, 1986; Wade, ment, 23 from School 1 and 48 from School 2. Sixty-
Trathen, & Schraw, 1990). I expected protocol results three
to completed paper/pencil comprehension tasks.
provide an in-depth view of the cues and strategies Eight students completed think-aloud protocols.
used by students.
Design
Materials The design was a 2 (Claim Familiarity) X 2 (Clai
The passages for all three experiments came from Position) X 2 (Text Signaling) x 2 (Text Replicate) x
(Order)
an initial pool of 20 written arguments about different X 2 (Text Structure) mixed design. Text
Structure was a within subjects factor, all other factor
types of animals from natural history books and maga-
zines. As suggested by the Toulmin (1958) model, claims were between subjects, and Replicates was nested un
were superordinate statements subsuming a substantial Claim Familiarity. For the protocol group, Order was
amount of the content in the passage (e.g., Geese are counterbalanced but not a factor, and the design wa
fractionalized in half with one reader per cell.
surprisingly useful.) Evidence was at a basic level that
could be readily verified (e.g., Farmers use geese to
weed large fields.) Passages were several paragraphsMaterials
long, typically exceeding 1,200 words. Texts. The Claim Familiarity and Text Replicate fac-
In a presession, participants provided examplestors required four base texts with familiar evidence and
for claims and rated the truth of claims, evidence, andwarrants, two with unfamiliar and two with familiar
warrants from the text pool. Since warrants are the un-claims. The Text Structure factor required that each base
derlying premise, they often are not explicitly statedtext
in be rewritten to have two versions differing accord-
written arguments. Therefore, I inferred them by con- ing to structural pattern: an argument with a claim, evi-
structing generalizations that subsumed both the claim dence, and warrants (Figure 3) and a topical net (Figure
and evidence analogous to major premises in syllogisms 4). The four argument versions included only content
(e.g., Birds that farmers can use to weed large fields consistent
are with the argument diagram. The four topical
net versions included all other information from the orig-
surprisingly useful.) If the majority of participants pre-
dicted that they could give no examples for a claiminal selection. Since passages differed in the amount of
and
had no idea whether it was true, I judged the claim to this ancillary information, I added content from other
be unfamiliar (Tversky & Kahneman, 1973). I judged ev- sources to control for length.
idence and warrants to be familiar if the majority of par- Claim Placement and Signaling were the last two
ticipants rated them as true or probably true. Results text factors. The claim appeared in either the introduc-
tion
from the pretest allowed me to manipulate the familiari- or conclusion of the text. Versions either did or did
ty of claims, evidence, and warrants across the three not signal the overall text structure and its parts. Phrases
experiments. such as, "This evidence points to an important truth
The study required 14 base texts. I rewrote the about topic," and "I will share information about topic",
original passages to match the experimental designs.signaled
In the text pattern. Other phrases, "For example,
we have evidence that ," "I propose that claim," and
addition, rewrites shared two features: introductory and
concluding paragraphs that signaled the structure, and"Biologists inform us that " signaled pattern parts
paragraph topic sentences. (Shopen & Williams, 1981; Vande Kopple, 1985).

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 787

Figure 3 The introductory paragraph

The introductory paragraph tha


In many obvious respects elephants are not one
whether we compare body shape and size, nose
care for both very young and very old membe
elephants frequently behave like people do.

The argument diagram that or

Evidence Warrant Claim

Elephants figure out how to


avoid being shocked by an
electric fence.

Figuring out solutions to new


situations is a human behavior.

Two elephants court and "go


steady."

Courting and going steady are


human behaviors.

A mother elephant disciplines her


baby for going in deep water.
Disciplining the young for
misbehaving is a human behavior.
Elephants frequently
behave like people.

A mother elephant mourns her


Mourning the dead is a human
dead baby and buries it. behavior.

Taking care of the old is a human


Young elephants take care of an behavior.
old elephant until he dies.

Remembering kindness is a
human behavior.
A young elephant remembers the
kindness of a human helper for a
long time.

Rewrites based on these two factors resulted in 32 text ments relating the instance of evidence to the claim (see
versions. Figure 3). For topical net versions, one of the instances
All versions were approximately three pages of sin-of evidence was included as a subtopic to provide the
gle-spaced typing. Each instance of evidence or subtopic
same amount of elaboration for the claim as paragraph
in the diagram was a paragraph long. Paragraphs in ar-
topic sentences had. Both argument and topical net ver-
gument versions concluded with explicit warrant state- sions included two superordinate statements: the claim -

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
788 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Figure 4 The introductory paragraph and topical net diagram for the informational top
version about elephants

The introductory paragraph that summarizes the entire text


In many obvious respects elephants are not one whit like people. The elephant trunk is their most distinctive fe
course. There are actually two types of elephants, Asian and African, which, while different in important respec
acteristics, such as eating and sleeping habits, skin care needs, and swimming ability. Both types are also surpris
their feet. Furthermore, despite the obvious physical differences, elephants of both types frequently behave lik

The topical net diagram that organizes the entire text

Trunk Two types


characteristics of elephants

Eating
Swimming habits
Elephants
ability

Reasoning Skin
ability care

Dexterity

and an "innocuous" generalization that waspurpose not elaborat-


(to present information about the topic or to
ed in the body of the text. For the two elephant support a point). The second and third items assessed
texts,
this generalization was, "In many obvious respects, ele- of the representations that readers had
the configuration
phants are not one whit like people." The generalization constructed for the two types of patterns. Item 2 instruct-
was included as a distractor for the measurement instru- ed participants to summarize the text in a single sen-
ment described below. tence. Item 3 presented participants with 10 sentences
To prompt protocol readers to think aloud, I from the text: the innocuous generalization, the claim, 4
placed small red dots at the end of each paragraph of paragraph topic sentences, and 4 detail sentences. In-
their text copies. Otherwise, their versions were identicalstructions told readers to distribute 100 points among the
to those used for the survey task. 10 sentences according to how important each sentence
The pattern recognition (PR) measure. The PR mea- would be in summarizing the text.
sure assessed the effect of text pattern type and signaling Because of the claim-evidence-warrant structure in
on the pattern that readers recognized. The paper/pencil an argument, readers should be able to use the claim as
version contained four items. The first item measured a single sentence to summarize the text if they have rep-
whether readers could distinguish the argument and top- resented it as an argument. Likewise, they should assign
ical net patterns, asking them to identify the author's a much larger share of the 100 points to the claim than

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 789

Table 1 A sample of one-sentence sum


how closely summaries matched the

Rating
One-sentence summary (2-0)

Claim: Elephants frequently behav


"* Elephants behave in a manner similar to humans. 2
"* Elephants are similar to people in many aspects because of the way
yet they're very different in their physical needs and features.
"* Elephants have many interesting and unusual characteristics. 0

Claim: Geese are surprisingly useful animals.


"* Geese are useful animals in a very broad variety of respects. 2
"* Geese can be useful to humans in many ways. 2
"* (No summary of the argument or topical net versions of "Ge
"* Geese are not ducks: they are intelligent and clever, have
by domestication, and are excellent guard animals.

Claim: Birds build an astonishing variety of nests.


"* Despite the fact that bird nests are all used for the same
"* This tells the purpose, construction, and variety of bird nests. 1
"* This passage gives us facts about birds and their nests. 0

Claim: Leopards are very dangerous predators.


"* Leopards, because of their many impressive characteristics, are truly dangerous predators. 2
"* The leopard is a beautiful, but dangerous animal which rarely attacks man. 1
"* The author gives information about leopards. 0

to the other sentences from the


measures. text,
Because including
of the Order factor, half of thethe
pack- in-
nocuous generalization. A etstopical net
presented the text has
with an a structure
argument differentfirst.
tern, represented by a topic that
Instructions subsumes
at the top of each passage a series
instructed readersof
subtopics. If readers have accurately
to read for the author'srepresented
message without worryingthe about in
details. The
formational text, they should useend of each passage directedother
something readers to than
the claim to summarize it and should distribute the 100 complete the PR measure without looking back at the
points approximately equally across the topic sentences text. Packets for the protocols omitted the PR measures
in the text, one of which is the claim in the argument since the index cards served the same purpose.
version. Item 4 asked participants to explain their ratio-
nale for assigning points. If their representation matched Procedure
the text pattern, I expected their rationale to refer to the Survey packets were stacked in random order for
claim and its evidence for the argument and subtopics each class. The classroom teachers followed the same
for the informational text. written instructions, distributing the packets face down
To measure pattern recognition for the protocol and giving readers 15 minutes to read and complete the
group, large 5" x 8" index cards directed readers to PR measure for each text. Both teachers were unaware
choose the author's purpose, predict what they expected of the purpose of the experiment and the experimental
would come next in the passage, and justify their an- conditions.
swers. Readers who had recognized an argument should Simultaneously, the protocol participants met indi-
choose the "argue" purpose, predict that they would vidually with me or one of my colleagues in an adjoin-
read about evidence, and point to the claim-evidence- ing classroom. We instructed each participant to read for
warrant structure in the text as justification. Those who the author's message rather than details and to pause at
had recognized an informational pattern should choose red dots to answer the questions on the index card. After
an "inform" purpose, predict that they would be reading completing each text, readers wrote a free recall to
information about the topic, and point to the subtopics demonstrate whether their representation of the text
to justify their responses. matched either an argument or an informational pattern.
Text packets. Packets for survey tasks combined Participants were given unlimited time to complete all
two passages differing only in text structure with two PR tasks within the constraints of a 50-minute period.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
790 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Table 2 Means for Claim Position, Text Structure, and Text Replicate for three dependent mea
recognizing the argument pattern

Author's purpose Summary Claim rating


(1-2) (0-2) (0-100)

Between-subjects facto
Claim Position
First (N= 32) 1.56 (.50) .89 (.89) 22.17 (13.39)
Last (N= 31) 1.48 (.50) 1.26 (.92) 23.34 (17.25)

Text Replicate"
Unfamiliar Claim
Elephants/geese (N= 17) 1.56 (.50) 1.00 (.89) 23.94 (14.94)
Geese/elephants (N= 18) 1.53 (.51) 1.08 (.97) 20.05 (8.92)

Familiar Claim
Nests/leopards (N= 14) 1.41 (.50) 1.04 (.94) 23.48 (18.45)
Leopards/nests (N= 14) 1.59 (.50) 1.11 (.93) 24.15 (19.14)

Within-subjects factor
Text Structure
Argument (N= 63) 1.92 (.27) 1.76 (.53) 28.29 (17.33)
Informational (N= 63) 1.11 (.32) .33 (.62) 17.05 (10.31)

Note. Italics denotes p<.05; both italic and boldfaced p<.001.


a The first text in each pairing was structured as an argument; the second w

Assigning survey scores cause its subject and predicate subordinated the prepon-
Higher scores for three derance
of of thecontentmeasures
in the recall. If the recall had such a
collected
from the surveys indicated claim,
that
it was diagrammed
participants as having an argument
hadstruc- recog
nized the argument pattern. ture.For
Because the
text structure
first was item,
expected to "Suppor
have a
powerful influence
ing a point" received a 2; "Giving on pattern recognition and
information" a because
1. A
one-sentence summary received
the analysis procedure
a 2 had
for
been developed
a close for this
para-
study,
phrase of the claim, a 1 for antwo analysts
imperfectcompleted diagrams for all recalls.
paraphrase th
included information from Judgments
the matched
claim's for 93% subject
of the analyses (15
and recalls),
pred
with consensus
cate, but (a) used terms that were reachedeitheron the remaining
more recall. general
specific than the terms in the claim or (b) embedded th
claim as either a dependent clause Results or as one of a ser
of equally important clauses. All other answers recei
a 0. Scoring for this second item Regardlessrequired
of task or measure,
a high
Text Structure
degre
of inference. Two raters independently strongly and consistently affected
scored readers'
allresponses.
re- N
sponses. Their judgments matched other factor had for
either strong
92% or of
consistent
the effects.
score Res
with consensus easily reached for the two on
typesthe remaining
of tasks scor
are reported separately.
Table 1 presents a sample of 12 responses and their r
ings. The third item was Written survey tasks
quantitative and needed no
scoring. Responses to the fourth Table 2 displays
item large main effect
were means for the
categorical,
and I analyzed them separately. three measures.
TheRow headings in consistent
results, Table 2 show the fa
with the other three measures, tors; column are not
headings detailed
show the measures.in this
The table
report. also indicates significance levels resulting from ANOVA
tests conducted for each measure. Consistently low to
modest correlations among measures indicated the ap
Analyzing protocols propriateness of using univariate rather than multivari
Analyses for protocol responses focused on an- analyses. The cells in the experimental design had un
swers consistent with recognizing an argument pattern. equal sizes because of a mixup in distributing the pac
The analysis of the free recalls began with a search for a ets and failure of some students to return completed
superordinate statement that could serve as a claim be packets.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 791

Means for texts with general information


an about birds' nests. Next I expect to
argument st
consistently higher read about how birds
than make adaptations.
means for tex
tional structure regardless of measure.
Throughout both types of texts, participants tended
tured as arguments rather than topical
to reaffirm their initial decision. While reading an argu-
tended to choose an argument purpo
ment text, another reader explained,
379.17, p<.001, MSE = .06, use the cla
the text, F(1,42) = The purpose is to provide support
235.66, for a point, and I canMS
p<.001,
the claim as more tell because the paragraph supports
important the author's point
than topiof
view by giving an example. I think this is good support;
tails, F(1,42) = 28.47, p<.001, MSE =
Other differences were either small or inconsistent because the reasoning example shows elephants acting
similarly to people [the argument's claim].
across measures. Claim Position affected the summary
score. This result may reflect a recency rather than a Participants responded differently when they could
comprehension effect. Students who had just read the not locate the claim/evidence relationship. As they read
claim in the concluding paragraph were more likely to texts structured as topical nets, readers would explain,
use it to summarize the text than those who had read it
The purpose is to provide information. The text doesn't
in the introduction, F(1,42) = 15.76, p<.001, MSE = .31. state a point that it's supporting.... This text is not arguing
Both Text Replicates, F(2,42) = 3.22, p<.05, MSE = .09 a point. It's just giving you stuff.
and an interaction between Text Structure and Text
Signaling, F(1,42) = 8.18, p<.01, MSE = .06, affected the Readers' responses differed depending on the pat-
author's purpose score. In the first case, readers tended tern in the text. For texts structured as arguments they
to choose an informational purpose for a text about bird identified the "support" purpose, pointed to the claim/
nests regardless of text structure. Consequently, the pair evidence relationship as justification, and predicted addi-
of replicates where nests had an argument structure re- tional evidence, the precise pattern I expected. For the
ceived a lower score than the other pairings. The inter- topical net they chose an "inform" purpose, pointed to
action resulted from signaling increasing the difference signaling, facts, or the absence of argument parts as justi-
in scores for the two text structures. All students chose fication, and predicted they would read more informa-
tion
an argument purpose when a text was both structured
and signaled as an argument (M = 2.00) while they tend- As for the written comprehension tasks, signaling
ed to choose an informational purpose for a text both and claim placement had no consistent effects. Although
structured and signaled accordingly (M = 1.06). These readers responded no differently when signaling was ab-
differences were not as great for texts without signaling sent, they sometimes referred to its presence. Placing the
(M = 1.82 and M = 1.13). Neither of these effects oc- claim at the end of the text also had no effect. Readers
curred for the other two measures, however. had inferred it by the close of the second paragraph:
Protocols. The protocol data corroborated and ex- The author is starting to make a point about the fact that
tended the data from the surveys. Text Structure was the the leopard is a dangerous predator. Next I will be read-
one text characteristic consistently influencing how read- ing more support about the thesis, the point.
ers responded. Immediately upon recognizing the
This reader had correctly induced the claim from the
claim/evidence structure in a text, they would identify it
summary of all the evidence she read in the introduc-
as an argument. Failing to find that structure, readers
tion, the elaboration of the first instance of evidence,
would identify the text as informational. The following
and the warrant, which linked this evidence to the miss-
examples illustrate the types of responses made by par-
ing claim and occurred as the concluding sentence of
ticipants. Following the first paragraph of a text with an
the paragraph ("Surely any animal that can so cleverly
argument structure, one reader explained,
trick its prey is a dangerous predator.").
The author's purpose is to present support for a point. I
Free recalls supported the protocol results. Figure 5
can see evidence for the point in this introductory para-
depicts recalls of the argument and topical net versions
graph. I expect that next the author will give examples of
about elephants. The top panel shows the argument pat-
how geese help [the text's claim]. The thesis makes you
tern and the recall on which it was based. The reader
expect you'll get more reasons.
had accurately recalled both claim and evidence. He also
After reading the introductory paragraph for the included appropriate metadiscourse (talk about talk) in
topical net, the same reader answered, his introduction.

The author's purpose is to give information about birds' Hugh O'Brien [sic], the author of the essay, proposes that
nests. This paragraph was an introduction with lots of elephants are similar to humans in the way in which they

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
792 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Figure 5 Verbatim recalls and text diagrams of two readers after reading either the argument
about elephants

Argument text
Evidence Warrant Claim

Hugh O'Brien (sic), the author of the essay, proposes


that elephants are similar to humans in the way
Twoin whichelephants showed
affection.
they behave. In the introductory paragraph, he stated
that elephants do not appear to be like humans; Elephants
however, they do act like humans. showed definite
human behavior.
His examples to support his thesis included specific
instances in which elephants showed definite Elephant
human be- mother
havior. One of the examples included two elephants
punished naughty cub.
showing affection to one another.

Also, elephants care for each other the way humans Elephants care Elephants are
do. The author stated several examples in which ele- similar to hu-
for each other
phants demonstrated human traits. A mother elephant, like humans do. mans in the way
Young elephants in which they
after saving her cub from drowning, waited to see if it Elephants
protect old. behave.
was all right before she punished him. Young elephants demonstrated
have been seen protecting the old and weak elephants human traits.
of the herd. Another example was when a mother ele-
phant was mourning after her dead newborn cub. She
walked around without eating and very temperamental
and eventually buried it. Mother elephant
mourned dead cub.

Elephants not only act like humans but they can rea- Elephants
son like humans as well. The author stated an occurance reason like
humans.
in the jungle where an expedition put up electric fences
to prevent the elephants from entering. The elephants "Elephants figured out
found out that when the lights were on, the fence was that their tusks did not
not electrified and that their tusks did not conduct elec- conduct electricity.
tricity.

Informational text

Body hair

Skin care Trunk Tusks Uses


What I remember from this article was that there are
two types of elephants, the African and the Asian. The
elephants are different, but have a few things in com-
mon, such as skin care, and of course, a trunk. The ele-
phant's trunk is used for eating, drinking, breathing,
Similarities Differences
smelling, and self defense. Elephants also take very
good care of their skin, since it is tough and leathery.

The differences between the Asian and African ele-


phants are the tusk size, amount of hair growth, and
also the Asian elephant is usually the one seen in circus
Two types
shows, etc. The elephants have a big appetite and eat a
lot also. Elephants are also very good swimmers. They
are also very smart, and can reason well. Elephants are
also light on their feet, and cannot (will not) really
crush anything.
Elephants

Light on
Appetite feet

Good swimmers

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 793

Figure 6 A diagram of the argument st

Evidence Warrant Claim Ancillary content

Ant species tend aphids to be


able to collect and drink the
Different ant-castes are partly
liquid they produce.
determined by the amount of
kohlrabies fed to the larvae and
partly determined by whether the
larvae develop from fertilized
Fierce army ants move about in eggs.
a compact herd, attacking all
animals in their path.

Any animal that maintains


herds, invades and conquers
Farmer ants cultivate crops. new lands, cultivates crops,
weaves "fabric," and captures Ant tribes pursue occupations
slaves is pursuing occupations startlingly like humankind's.
startlingly like humankind's.

Weaver ants construct cradles


for their larvae.

Slave-making ants capture oth-


er ants to do their work for
them.

behave...His examples to support his thesis included spe- EXPERIMENT 2:


cific instances in which elephants showed definite human
behavior. Identifying the argument parts
The bottom panel of Figure 5 displays the recall The second experiment focused on the secon
and diagram from another participant who read the topi-
stage of the comprehension model. Both texts an
cal net version about elephants. She recounted what she
sures were designed to identify the text cues and s
remembered from the text she had read as topics and
gies used by competent readers to find the claim
subtopics rather than as a claim and evidence. These dif-
ferences in recalls were typical. evidence in a lengthy written argument.

Conclusion
Method
Results suggest that good readers use the presence
Participants
or absence of a claim-evidence-warrant relationship to
recognize a written argument and distinguish it from Sixty
an participants completed written tasks; eight
informational topical net. Whether a text had signaling,
completed protocols.
claims were unfamiliar, or claims were placed in intro-
ductions had virtually no effect on reader performance.
Design
For texts with an argument structure, readers chose an
The design was a 2 (Order) X 2 (Signaling) x 2
"argue" purpose, expected to find argument parts in the
(Text
text, summarized it with the claim, and structured their Replicate) x 2 (Claim Presence) X 6 (Paragrap
recalls as arguments. Their answers followed quitemixed design with Order, Signaling, and Text Replic
a dif-
the between-subjects factors, Claim Presence a within-
ferent pattern for informational texts where the claim-
evidence-warrant relationship was missing. subjects factor, and Paragraphs a within-text factor.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
794 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Table 3 A sample of claim statements that readers constructed after reading arguments with explicit a
claims and the ratings assigned to the statements according to how closely they matched the a
claim

Rating
Claim statement (3-0)

Claim: The coyote is par


"* Coyotes take the best advantage of their environment. 3
"* Coyotes are very adept and can adapt to almost any environment well. 2
"* Coyotes, contrary to the saying that they are inferior members of the wolf family, ar
they have a better chance to survive.
"* Many people tend to overlook the remarkable qualitites and habits that coyotes possess. O

Claim: Ant tribes pursue occupations startlingly like humankind's.


"* Ants have occupations amazingly similar to mans [sic], (both past and present). 3
"* Ants, in their duties and roles, are much like humans in their respective roles. 2
"* Ants act very much like "civilized" humans do. 1
"* Ants are industrious and society oriented. 0

Materials Procedure
Texts. Two passages from the original pool of 20 Participants completed the two types of tasks as in
texts were used for this experiment, both with unfamiliarthe first experiment with one exception. They were told
claims, familiar evidence, and familiar warrants. Repli- that both passages had been written to support a point.
cates were prepared comparably to the first experiment,
with three exceptions. All versions had an argument
Assigning survey scores
structure. According to the Paragraphs factor, one para-
The author's claim scores distinguished among dif-
graph in each text was ancillary to the relationship in the
claim, related to it topically only. A summary warrant inferent degrees of accuracy. For each claim, I identified
the concluding paragraph related the evidence to the two concepts that each instance of evidence in the argu
claim instead of individual warrants at the end of each ment exemplified ("taking advantage" and "environment"
paragraph. Figure 6 depicts the resulting structure for for the argument about coyotes; "occupations" and "lik
one of the texts. humankind's" for the argument about ants). A claim
These structures formed the basis for the four repli- statement that related both concepts received a 3. If both
cates. Because of the Claim Presence factor, introduc- concepts were present, but the meaning of one or both
tions either did or did not contain the argument's claim. was slightly different, the statement received a 2 (e.g.,
The Signaling factor required that introductions and con- "situation" for "environment"; "roles" for "occupations").
clusions either summarized the argument accurately by A statement with only one of the concepts received a 1.
including only content from the evidence paragraphs or Anything else received a 0. Because scoring of claim
summarized the entire passage by including content statements involved a high degree of inference, three
from the ancillary paragraph as well. raters scored the responses independently. On 98% of
School 1 participants, completing the tasks first, the responses, two of the three raters assigned identical
seemed consistently to misinterpret the meaning of the scores. Raters reached consensus on the remaining rat-
claim slightly for one of the passages. Consequently, I ings. Table 3 lists a sample of claim statements and their
reworded the title and rewrote several paragraph topic accompanying scores. Evidence list responses received
sentences for students from School 2.
one point for each of the six text paragraphs from which
The Identification Measure (IM). The IM asked students included at least one piece of information.
readers completing written surveys to (a) write the au-
thor's point and (b) summarize in a single sentence each
of the five pieces of support in the argument. Prompts Analyzing protocols
asked students completing protocols to explain why they Protocol analyses focused on how the presence or
thought the author had included the paragraph they had absence of the claim affected readers' identification of

just read, share any thoughts they had while reading, the author's claim throughout the text, particular text
state the author's point for the entire passage, and justify characteristics that they gave as rationale for their an-
their answer. swers, and how they handled the ancillary paragraph.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 795

Table 4 Means for Signaling, Claim P


Experiment 2: identifying the argu

Author's claim Support list


(0-3) (0-1)

Between-subje
Signaling
Exact (N = 29) 1.93 (.87) (N = 348) .77 (.42)
Ancillary (N= 31) 1.22 (.89) (N = 372) .70 (.46)

Within-subjects factor
Claim Presence
Present (N = 60) 1.80 (.92) (N= 720) .75 (.43)
Absent (N= 60) 1.33 (.93) (N= 720) .72 (.45)

Within-text factor
Paragraphs
Evidenceb (N= 600) .79 (.41)
Ancillary (N= 120) .47 (.50)

Note. Italics denotes p<.05; boldfaced, p<.01; bo


a Within-text factors were measured only by the
b The mean for evidence has been computed a

Results paragraphs was significant, F(1,31) = 11.67, p<.001, MSE


= .24. As hypothesized, participants were more likely to
Whether introductions and conclusions signaled
list evidence as support and leave ancillary content off
the argument accurately, claim statements were present
the list when the argument was summarized accurately in
or absent, and content in paragraphs was relevant or an-
the introduction and conclusion than otherwise.
cillary influenced the claim and evidence that readers Claim Presence affected the author's claim score, as
identified in the argument. expected, F(1,31) = 7.71, p<.01, MSE = .54. Readers re-
sponded more accurately when the claim was present
Written survey tasks than absent. An interaction with school indicated that the
Table 4 displays the means for large main effects
effect was more evident for School 1 than School 2 par-
for each of the two measures and the significance levelsF(1,31) = 7.71, p<.01, MSE = .54. School 2
ticipants,
from analyses of variance. Consistently modest toreaders
low were equally accurate at stating the claim regard-
correlations among the two measures indicated lesstheof ap-
claim presence.
propriateness of univariate rather than multivariate analy-
The contrast between evidence and ancillary con-
ses. An ANOVA for each measure was conducted on a
tent had an effect on the evidence list, [(1,31) = 34.16,
subset of the data. To test the effect of slightly different
p<.001, MSE = .24. Readers were almost twice as likely
text versions for each of the schools, I added School to as list
a evidence as they were to list ancillary content. A
factor. Since almost twice as many students came from quadratic contrast conducted on the five evidence para-
School 2, I eliminated School 2 data until the design graphs
was suggested that readers were somewhat more like-
complete with three students per cell and ran the ly to list evidence from the beginning and end of the
ANOVA tests on a subset of the data.
text than the middle, perhaps reflecting primacy and re-
Signaling in introductions and conclusions affectedcency memory effects, F(1,31) = 6.13, P<.05, MSE = .11.
both measures. Participants were more likely to produceSince ancillary content appeared as the fifth paragraph
an accurate statement of the author's claim when the in-
out of eight in both texts, some readers may have left
troductory and concluding paragraphs summarized only ancillary content off the list because they forgot it rather
the argument, F(1,31) = 19.08, p<.001, MSE = .83, an out-than rejected it as support.
come I had not anticipated. They also listed more items
overall on the evidence list when introductions and con- Protocol tasks
clusions summarized the argument accurately, F(1,31) = Protocols matched results from the written surveys.
4.40, p<.05, MSE = .13. The interaction between Signaling
Most participants recognized the claim in the argument
and an orthogonal contrast of evidence with ancillary and identified as evidence paragraphs that supported the

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
796 READING RESEARCH QUARTERLY October/November/December 1995 30/4

claim while rejecting as evidence paragraphs only topi- realized how all those examples fit in. [The
cally related. When the claim was missing, some readers author] compared humans and ants."
quickly inferred it, apparently from patterns they recog-
Other participants were willing to accept that texts
nized in the evidence. Others frequently changed their
with missing claims were arguments. They readily in-
minds about the claim as they successively read the
duced claims, which changed after each paragraph. In
paragraphs in the text. They seemed particularly influ-
addition, they tended to accept ancillary content as evi-
enced by the ancillary paragraph, broadening the claim
dence. The accuracy of their answers was affected by
to incorporate it rather than realizing that it did not fit
both the absence of the claim and the presence of
the rest of the argument. Two readers seemed unable to
nonevidentiary information.
infer a missing claim.
In the first example, the reader found the claim in Introduction: "I can't tell the point yet."
Paragraph 2: "No matter what the environment, the coy-
the introduction, stated it as the author's point through-
out the text, and accurately distinguished evidence from ote tries to solve problems. It will find a
ancillary content. way to live."
Paragraph 3: "The coyote will eat no matter what the
Introduction: "The point is that ants' occupations are go- cost; it will take advantage of every oppor-
ing to be like humans. It says it right here." tunity."
Paragraph 2: "It's still the same. Ants are like humans." Paragraph 4: "The coyote is smart. He will make the best
(Note that "occupations" dropped out, a
of any situation where he can get a free
change that occurred frequently among
meal. He always thinks before he acts."
readers.)
Ancillary: "The same point. Coyotes make do with
Paragraph 3: "The same point. Ants have human-like
what they have. Also, that they are smart....
qualities."
The author could be comparing them to hu-
Paragraph 4: "Same [point]. Ants are like humans."
mans. I'm not really sure. But their children
Ancillary: "Same [point]. This paragraph is not good
are important to them-a lot like humans.
support. Humans don't do this."
They make do with what they have, just
Paragraph 6: "Same [point]. Ants are human-like. Now
like humans. Anyone will take a good deal.
he's back on track to support the point."
Humans do that too."
Paragraph 7: "Ants are human-like."
This reader's statement of the author's claim changed
Readers intent on looking for a claim in the text
throughout the passage. Note particularly her last re-
were thrown by versions that did not explicitly state it.
sponse. Upon reading the ancillary paragraph, she re-
After reading the introduction, one reader explained,
viewed what she had read about the coyote thus far and
Introduction: "This passage will just give information.
seemed to be searching for one generalization to subsume
There is no opinion in the first paragraph."
even the ancillary information about how coyotes parent.
He was unable to identify a claim until after reading the Other readers were more accurate at inducing the
warrant at the end of the passage, which specifically re- claim from the evidence, and consequently less strongly
lated the evidence to the claim. influenced by the ancillary paragraph.

Paragraph 2: "No opinion yet." Introduction: "The coyote can survive the environment. It
Paragraph 3: "Not that I can tell." is a fighter. It does a lot of adaptive stuff. I
Paragraph 4: "Just an essay on ants." can tell [that this is the point] because this
Ancillary: "It's an informative research paper on ants." introductory paragraph kept saying how
Paragraph 6: "No opinion. Just about how ants make they got food, and examples of how it
nests for larvae." adapted." (She had induced the claim from
Paragraph 7: "It's just making statements about ants." a summary of the evidence in the introduc-
tion.)
Once he had read the warrant in the concluding para- Paragraph 2: "The coyote is a survivor. [I can tell be-
graph, he explained, cause] the paragraph explains different
Conclusion: "The [concluding] paragraph stated an opin- changes and adaptations."
ion. I was thinking as I read this paragraph Paragraph 3: "The coyote is able to survive in the envi-
that the author finally stated an opinion. ronment. The paragraph continues to show
The thesis came at the end of the para- different ways."
graph. It is, 'the similarity between the soci- Paragraph 4: "The coyote is a smart animal. The para-
ety of ants and the society of humans.' I graph continues to give examples."

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 797

Figure 7 A diagram of the complex argument


familiar warrants

Evidence Warrant Claim

The shrew is one of the


smallest mammals. An animal rated among
the smallest mammals is The tiny mammal, the
tiny. common shrew,

The shrew attacks, kills, "Any animal that will at-


and devours animals twice tack, kill, and devour ani-
its size. mals twice its size is a
fierce fighter.

The shrew overpowers its


Any animal that overpowers
prey after heavy fighting, is a fierce fighter
its prey after heavy fighting,
bites it in the face and
bites it in the face and
throat, paralyzes it, and
throat, paralyzes it, and be-
begins to eat while the
gins to eat while it is still
,prey is still alive. alive is a fierce fighter.

Any animal that eats large


The shrew eats constantly.
amounts of prey constantly
is a voracious predator.

and a voracious predator.


The shrew eats insects, Any predator that will eat
eggs, fish, small reptiles, almost anything is likely to
mammals, and birds. be voracious.

By the end of the introduction, the reader had induced a to identify the claim and evidence in a written argume
superordinate from the summary of the evidence. She They readily identified the claim statement in a text a
maintained this superordinate as the claim for two of themore accurately produced the claim when it was stat
next three paragraphs. Her next response follows the an-
They pointed to patterns in the text to justify both the
cillary paragraph. statements of the claim and their rejection of ancillary
Ancillary: "[The point is] still the same, but changing. content. Apparently they know the defining features o
At this point it seems to be different. I'm notclaims and evidence well. All participants produced
sure what it is yet. The paragraph is no claim statements that were generalizations at a superor
longer about how [the coyote] survived. It dinate rather than a basic level. In contrast, they all pr
just kind of gives facts." duced as evidence basic level instances or facts. The
Paragraph 6: "This paragraph is more like the whole pas-
claims and evidence tended to be related. Readers justi-
sage-how the coyote survives the environ- fied their statement of the claim by pointing to the evi-
ment. I can tell because the paragraph
dence and distinguished evidence from nonevidence by
supports the point through the breeding
referring to the claim. Their accuracy was influenced by
example."
the caliber of introductory and concluding summaries.
Finally, results suggest that good readers tend to
Conclusion use three strategies to identify an argument's claim and
evidence. They may locate the claim in the text, identify
Results suggest that good readers use claim state-
it, and use it to distinguish evidence from ancillary con-
ments, patterns in the content presented in text para-
graphs, and summaries in introductions and conclusions
tent. Failing to find the claim in the text, readers may use

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
798 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Figure 8 Verbatim summary recalls, argument diagrams, and scores for two readers recalling a
argument with a complex structure and familiar warrants

A summary with a complex argument pattern


Summary Pattern Scores
Evidence Warrant Claim

The author argues that although


Two shrews most of the
ferocious and deadly mammals are the big cats
on a scale Although
like lions and tigers, the tiny mouse-like
register an shrew most
is of the Claim = 1
just as deadly in its own right.
ounce. ferocious and (2-0)
deadly mam-
Although it takes two shrews on a scale to mals are the
Shrews will
register an ounce, they will attack animals twice
big cats like Support = 5
its size to temporarily satisfy attack
itsani- voracious lions and (5-0)
appetite. Because the shrew mals has twicesuch a high
tigers, the
their size.
metabolism, it must spend a majority of its life in
tiny mouse-
search of food. Match = 1
like shrew
Because of (1-0)
The shrew attacks by biting at an animal's
neck and face and paralyzing them with powerful high metabo-
lism, a shrew
venom. They even eat their prey alive! They Pattern = 2
consume their body weight in food every three must spend
most of life (2-0)
hours.
in search of
food.
They attack anything unfortunate enough to
have crossed their paths, including rats, snakes,
small mammals, insects, and even other shrews. Shrews attack is just as
in face and deadly in its
neck and own right.
paralyze with
venom. They
even eat prey
alive.

They will
attack
anything.

A summary with a simple argument pattern


Summary Pattern Scores
Evidence Warrant Claim

Shrews are one of the most ferocious crea- They attack in


tures despite their small size. They attack their neck and bite
prey in the neck and bite into it with their razor- with razor-sharp
sharp teeth, depositing poisonous venom which teeth, depositing
sort of paralyzes their victim. They eat their prey poisonous venom Claim = 1
sometimes while it is still alive. They are con- which sort of par- (2-0)
stantly searching for food and will even eat mam- alyzes victim.
mals bigger than themselves. They devour every Shrews are
They sometimes
part of the animal or insect or mammal they one of the Support = 3
eat prey while still
catch. Once the author of the passage gave a alive. most (5-0)
shrew to his snake. When he came back, every ferocious
bit of the snake was gone. The shrew can starve creatures

in one day if he isn't constantly eating food. The Shrews will eat Match = 0
despite their
shrew is a very fierce creature. mammals bigger small size. (1-0)
than themselves.
Devour every
part. Shrew ate Pattern = 1
every bit of snake. (2-0)

Shrew can starve


in one day if he
isn't constantly
eating food.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 799

one of of a strategies.
alternate complex claim. Two of these
two passages had may
They familiar
claim on from warrants
early
the and two had unfamiliar warrants. Figure and
introduction 7 de-
picts the and
instances of evidence structure for one of the complex arguments
subsequently
with familiarmay
guish evidence. Or they warrants. The other four passages hadall
identify sim-
ples as evidence andple structures
search similar to all arguments used in Experi-
repeatedly f
subsuming claim. ments 1 and 2, two with familiar and two with unfamiliar
warrants.

Warrant Structure and Argument Summary wer


EXPERIMENT 3:
the last two text factors. Each evidence paragraph in
Constructing the gist text ended with its own warrant, the concluding text
paragraph included one summary warrant, or the tex
The third experiment focused on the third stage of
had no warrant statements. Texts concluded with a sum-
the comprehension model. I designed texts and mea-
mary of the argument or with an innocuous paragraph
sures to identify the text cues and strategies competent
that neither summarized the argument nor added new
readers employ to construct a gist representation of a
meaning. Rewrites to incorporate these last two factors
lengthy written argument.
resulted in 48 text versions.

Method Procedure
Procedures were similar to the first two experi-
ments with three exceptions. Participants were told that
Participants
both texts were written to support a point with the claim
Fifty-one participants completed surveys. Six com-
present. After reading, survey participants wrote a sum-
pleted protocols.
mary recall of each passage. Protocol participants sum-
marized the evolving argument after each paragraph.
Design
The design was a 2 (Argument Summary) X 2
Assigning survey scores
(Warrant Familiarity) X 3 (Warrant Structure) x 2
To score the summary recalls, I diagramed them
(Order) X 2 (Argument Complexity) x 2 (Text
using a procedure identical to the analysis of written re-
Replicates I) X 2 (Text Replicates II) mixed design with
calls in Experiment 1. Because this procedure involved
Argument Summary, Warrant Familiarity, Warrant
high degree of inference, a colleague familiar with the
Structure, Order, and Text Replicates I and II as be-
structures diagramed a randomly selected subset of 12
tween-subjects factors and Argument Complexity as a
recalls. Her diagrams matched mine on 11 recalls for a
within-subjects factor. Text Replicates I and II were nest-
interrater agreement of 91%. We easily reached consen-
ed within both Warrant Familiarity and Argument sus on the last recall.
Complexity. Text Replicates I had simple argument struc- Figure 8 displays two recalls for the complex argu-
tures and either familiar or unfamiliar warrants. Text
ment about shrews (see Figure 7) and their accompany-
Replicates II had complex structures with either familiar ing diagrams. The recall in the first panel of Figure 8 has
or unfamiliar warrants.
a complex structure that matches the pattern in the text,
The 96-cell design exceeded the number of readers mentions all five instances of evidence in the text, but
who would be completing surveys. I fractionalized in half does not accurately restate the complete claim or include
the part of the design containing all two-level between- any warrant statements. The recall in the second panel
subjects factors and crossed it with the between-subjects has a simple structure that does not match the text pat-
three-level factor. The six combinations for the readers
tern, mentions only three of the five instances of text ev-
completing protocols came from the larger design. idence, and also does not accurately restate the complete
claim or include warrant statements.
Materials
Each resulting diagram received four scores: Claim
Except for the variations required by the text fac-
score and support score measured comprehensiveness;
tors for this experiment, replicates were prepared com-
match score and pattern score measured organization.
parably to the two earlier experiments. Warrant The claim score ranged from 2 to 0. If all major concepts
Familiarity, Argument Complexity, and Text Replicates in Ithe claim appeared in the summary, the claim score
and II required eight passages from the original pool wasof2. If only some of the concepts were present, the
20 texts. Four of these passages had complex structures
score was 1. If none of the concepts was present, the
whereby instances of evidence supported different parts
score was 0. To guard against rater bias on this high in-

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
800 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Table 5 Means for Argument Summary and Argument Complexity for four dependent me
constructing the gist

Claim Support Match Pattern


(0-2) (0-5) (0-1) (0-2)

Between-subjects facto
Argument signaling
No (N= 25) 1.22 (.82) 320 (1.77) .66 (.48) 1.08 (.73)
Yes (N= 26) 1.50 (.64) 4.04 (1.20) .71 (.46) 1.12 (.62)

Within-subjects factor
Argument complexity
Simple (N= 51) 1.43 (.81) 3.82 (1.63) .82 (.39) .82 (.39)
Complex (N= 51) 1.29 (.67) 3.43 (1.47) .55 (.50) 1.37 (.77)

Note. Italics denotes p<.05; boldfaced, p<.01; both italics and boldfaced, p<.001.

ference measure, a colleague independently


per cell, I used the pooled three-way rescored
and higher interac- all
claims, unaware of condition. Our
tions for theratings agreed
error term in the analyses. for 83%
of the summaries, with consensus Whether the easily reached
argument was summarized inon the
the con-
clusion affected
remaining summaries. The support how accurately
score, rangingreaders reproduced
fromthe 5
claim, F(1,12)
to 0, was determined comparably to = the
5.26, p<.05, MSE = .65, and
support listsupport,
scores in Experiment 2 and F(1,12) = 10.17, p<.01,
assessed how MSE = many
1.80) in theirof
summary.
the
five instances of support in each
Means passage
for both the claim and occurred
support scores were in
high- a
recall. Match and pattern scores er than were
otherwise when
closelythe argument was summarized.
related.
The pattern score, ranging from Whether the
2 argument
to 0, had measured
a simple or complex
which
struc-
argument pattern organized ture the affected
recall.
how readers
A 2 organized
represented
their summary as a
complex structure; a 1 a simple measured by the pattern score,
structure; and F(1,12)
a 0= 71.06,
any- p<.001,
thing else. The match score, MSE from = .10. The
1 mean
to 0, for complex
assessedarguments was higher
whether the structure of the recall matched the structure than for simple arguments, indicating that more readers
of the text, with a match receiving a 1. Figure 8 reports organized their recalls of a complex argument with a
these four scores for each of the two summaries. complex structure than for the simple case. Readers
tended to use the same structure in their summary as
Analyzing protocols had appeared in the argument they had read. However,
To analyze readers' summaries of the evolving ar- means were higher for the simple argument than the
gument, I noted their statement of the author's claim, complex argument on the match score, F(1,12) = 21.28,
any support they included, and any expressions of war- p<.001, MSE = .09, indicating recalls were more likely to
rants linking the two. match the simple structure than the complex. These two
results suggest both intra- and interindividual differences.
Some participants used different structures in recalling
Results
the two types; other participants recalled both complex
and simple texts using the simple structure.
The complexity of a text's argument and whether
final paragraphs summarized the argument affected the The interaction between Order and Warrant
summaries that readers constructed. Familiarity was statistically significant for the support
score, F(1,12) = 15.69, p<.001, MSE = 1.80. Readers who
Written survey tasks read simple arguments with unfamiliar warrants first re-
Table 5 displays the means of large main effects for called less support overall than readers who read simple
each of the four measures and the significance levels arguments with familiar warrants first, while the reverse
from analyses of variance. Measures were highly corre- was true for readers who read simple arguments second.
lated, indicating the appropriateness of a multivariate
This interaction is difficult to interpret. Since it does not
analysis of variance. MANOVA and ANOVA results were appear with the other measures, it probably should not
identical. Since the fractionalized design had one reader be taken too seriously.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 801

Protocol results The author gives background to when African egrets first
The protocols illustrated the effects of complexity got here and how they began to spread. He is giving
and signaling on readers' summaries. All participants background to support the statement that egrets have
thrived because of the environmental changes that hu-
summarized simple arguments accurately. They restated
mans have made on the environment.
the claim after each paragraph and added the new evi-
dence onto their summary. Readers varied in how accu- He rejected the second instance even more forcefully.
rately they summarized the complex structure. Although "Cattle egrets have become a nuisance in Texas. This
two summarized the complex structure accurately, theparagraph is not supporting the point."
other four readers either summarized complex argu- Upon reading the first evidence about the influ-
ments as if they had simple structures, or were unable ence of human changes, he dropped the first two evi-
to summarize the text. dence paragraphs from his summary of the evidence.
The first example chronicles a reader accurately People have made changes in the environment that give
summarizing a complex argument. The claim was, "The egrets exactly what they need to thrive. Improved irriga-
common crow is a feathered genius and full of mis- tion has increased farm production. As farm production
chief." The first three instances of evidence supported has increased, so has the insect population.
the crow's genius. The last two supported the crow's
By the end of the passage, this reader had dropped
mischievousness. Following the second paragraph, the
reader summarized the claim and first instance of evi-
"thriving" out of the claim statement as well.
dence: Humans have made changes that increased the egret pop-
ulation. Increased farm production increased the insect
The crow is intelligent and full of mischief. Crows mimic population. Increased numbers of cattle kicked up the in-
people and other animals perfectly and appropriately. sects. Increased gasoline powered machinery kicked up
more insects for the egrets. The egret food supply has in-
The fifth paragraph introduced the first evidence
creased, which allows an increase in the egret population.
for crow mischievousness. The reader added this evi-
dence to her summary of the claim and the evidence Reading a complex argument with an innocuous
for crow intelligence. conclusion first confused two of the participants so com-
Crows are intelligent and full of mischief. They mimic pletely that their summaries were incoherent. Both read-
people and other animals. They have been impossible to ers summarized the simple argument text, which they
eradicate and have thrived instead. Crows use special read second, in an orderly, coherent fashion. After read-
ing the innocuous conclusion, one reader summarized
strategies for getting the food they like. They steal nesting
materials from each other. the egret text as:

The egret is unique. It has been able to migrate success-


After reading the last paragraph, she summarized the en-
fully and build its own niche. The egret has been able by
tire argument.
itself to adapt to the environment and build its own
Crows are intelligent and full of mischief. They mimic niche. It gave support for how it was able to survive and
people and other animals. They have been impossible to find its own niche. Egrets will follow the cattle so they
eradicate and have thrived instead. They use special can pick up insects which is their main form of food.
strategies for getting food they like. They steal nesting Agricultural changes in the landscape have increased in-
materials from each other and play games and practical sects, and egrets will follow that as well, and man-made
jokes. environmental changes such as machinery for building
roads or farm machinery, and these expose insects for the
A second example demonstrates how other readers egrets as well and the egrets will follow these. And the
forced the complex argument into a simple structure by conclusion is that the egret is a unique bird that has been
shortening both the claim and evidence in the text. The able to migrate from Africa to North America and adapt to
argument's claim was, "African egrets have thrived in the environment and be able to thrive on the availability
North America because people have made changes in of the food supply which is the insect basically. And de-
pending on the amount of the food supply, they will fol-
the environment that give egrets exactly what they need
low... I had trouble with that one.
to thrive." The text first presented evidence that egrets
have thrived. The remainder of the argument gave evi- The concluding innocuous paragraph discussed
dence of the beneficial changes that people have made. how unique the egret is to have made a niche for itself
After reading the first instance of evidence, this reader on a new continent. The influence of this paragraph on
reproduced the complete claim, but dismissed the evi- the reader's final argument summary was strong and
dence as "background." confused his shaky grasp of the argument.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
802 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Conclusion design, the following sections focus on a discussion of


Results suggest that good readers use two typesthe
oftheoretical and practical implications of these results.
text cues to construct a gist representation of a written
Text cues used by successful readers
argument: argument structure and concluding paragraph
summaries. First, participants tended to structure theirThe relationship between a claim and its evidence
summaries comparably to the structure in the text.was crucial for these readers. They understood claims as
superordinate to basic level facts and examples offered
Second, readers who had read a final summary para-
as evidence or support. Only a superordinate statement
graph typically recalled an argument more accurately
that indeed subsumed the facts and examples in an
than those who had read a final innocuous paragraph.
argument could serve as its claim. Only facts and exam-
This effect was so potent for two protocol readers that
they were unable to produce a coherent summaryples of subsumable
a by the claim could count as an argu-
complex argument that ended with an innocuousment's para- evidence. Although all readers in the study could
identify
graph. In contrast, the characteristics of warrants had no these two parts independently when asked and
effect. Both the accuracy and structure of readers' had
sum-technical vocabulary that they possessed to discuss
them ("claim," "point," "thesis," "evidence," and "sup-
maries were unaffected by the familiarity and structure
of the argument's warrants. Furthermore, only 6 of port"),
the what they identified as a claim tended to be de-
pendent
57 participants included warrants in their summaries. on the evidence they identified and vice versa.
Results also suggest, however, that good readers These readers also seemed to make good use of
use an argument schema in conjunction with text cuesintroductions
to and conclusions that summarized the argu-
ment. Their answers suggested that they could recognize
link the claim and evidence together into a gist represen-
tation. Apparently, they represent the claim and addthe claim-evidence-warrant relationship in the text based
on the skimpiest of summaries in introductions and con-
each new piece of evidence to their evolving representa-
tion of the text according to what they know aboutclusions
argu- (a sentence expressing the claim and a few
compound
ment structures. Virtually all participants seemed to have and complex sentences that summarized all
a schema for simple arguments, recalling them with ofathe evidence). If introductions and conclusions did
not accurately
simple argument structure. They differed in how they re- summarize the argument, the comprehen-
called complex arguments, however. Some readers sion of some readers suffered. In fact, otherwise success-
sim-
plified complex structures by omitting evidence as ful readers could fail completely to find the claim/
if it
evidence relationship in a text with a complex argument
were ancillary and shortening the claim accordingly.
They appeared to know the simple structure only. structure and a nonsummarizing concluding paragraph.
Content familiarity and placement of argument
parts were text cues that did not affect reader compre-
Discussion
hension as expected. Whether the claim was familiar or
The successful readers in this study recognized unfamiliar or appeared in the introduction or conclusion,
lengthy written arguments structured according readers
totended
the to recognize the claim-evidence-warrant
three-part Toulmin (1958) model, identifiedrelationship. claims and Whether a warrant was familiar or unfamil-
evidence even when the claim had to be inferred and iar or came at the end of each paragraph, in the con-
the evidence distinguished from ancillary content, and cluding paragraph, or was missing entirely, they linked
constructed a gist representation that tended to match a claim to its evidence to construct a gist representation.
the claim-evidence relationship in the author's argument. Explanations for why readers seemed not to use
For the most part, the Toulmin (1958) model character- these two types of text cues differ for claims and war-
ized both an argument schema they knew and text cuesrants. Competent readers appeared to know the
they used. Furthermore, participants searched for argu- claim/evidence relationship so well that they ignored
ment cues and carried out strategies on entire para- both the content (at least in these noncontentious argu-
graphs within lengthy written texts. The performance of ments) and placement of claims. If they recognized the
students from three different classes in two schools tend- presence of the relationship, they recognized the text as
ed to be consistent across 14 separate argument topics, an argument regardless of where the claim came and
two types of tasks and analyses, and three data collec- how familiar it was.
tion sessions. Results supported the initial hypotheses. Warrants differ from claims. As Toulmin (1958)
Results also suggested specific text cues and strate- conceived them, warrants are the reasons that link evi-
gies employed by these competent readers. Because dence to claims. In some ways, the warrant is the most
specific outcomes may well be the most useful for sug- important, and also the most elusive, argument part
gesting instructional strategies and improving textbook since it typically is the least explicit. Indeed, obvious

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 803

warrants are patternexpressed


rarely will proceed through the otherin
two stages almost
infor
identically regardless
versation. The arguer of either the features of the
presents the particu- ev
and assumes that lar
the relationship
pattern or the caliber of cues provided by the text. Abet
warrant) is certainAny
obvious. automaticity is argument
implied by this model. wit
In contrast, readers giving by
parts has been categorized think-aloudclassica
protocols in
enthymeme (Cooper, 1932;
this study demonstrated Corbett,
flexibility and conscious plan- 1
ning to comprehend
The Rhetoric (Cooper, the author's argument.
1932) recomme They evalu-
present enthymemesated author choices, e.g., "This paragraph
rather than is not support-
argu
parts because he thought an
ing the point," or "I think this audienc
is good support, because
entire argument inthe reasoning
mind example shows
easily.elephants acting similarly
Autho
to people." They
his advice and omitted predicted: "I expect that next parti
warrants, the au-
arguments. thor will give examples of how geese help. The thesis
makes you expect you'll
Readers in this study had get more no reasons." They
vocabu
changed their
identify the warrant, minds. "I was thinking as I read
referring tothis para-
it as
graphit
ion if they noticed that the
in authorthe
finally stated an opinion.... Iat
text real- all
spontaneously and ized how all those examples fit in."
correctly usedThey reflected on
term
and evidence. They their ownseemed to
comprehension: "I had trouble withhave
that one," h
with warrants expressed or "The author could explicitly
be comparing [coyotes] to humans.in t
made little use of them. I'm not really sure." Since protocol and survey results
Likewise, warrant familiarity had no systematic ef- supported one another so consistently, it seems possible
fect on reader comprehension, even though I had ex- that readers sitting quietly in their classrooms were em-
pected unfamiliar warrants to leave readers scratching ploying some of the same strategies.
their heads and muttering, "I don't see the connection." Successful readers chose their strategies to match
Readers rated as unfamiliar the warrant, "Racing across text cues. Readers typically noted and made use of
the plain, circling back, and leaping forward each pro- explicit text cues. For example, their accuracy in identi-
vide exercise that would develop muscle mass, increase fying claims increased when claim statements were pres-
bone size and shape, and maximize the heart pumping ent. Protocol participants often pointed directly to the
ability needed for optimal development of a baby ani- cues. "The point is that ants' occupations are going to be
mal's neuromuscular system." Subsequently they were like humans. It says it right here." Most readers switched
able to summarize the entire argument (which had a to a different strategy if cues deviated from what they
simple structure) with accuracy even when this unfamil- expected. Following through on the same example,
iar warrant had to be inferred. It is likely that warrants readers could not point to a claim statement that was
rated relatively unfamiliar in the pretest actually were not missing from the text. Instead, most of them seemed to
unobvious, but simply previously unknown. Only a search for a superordinate statement that could subordi-
more powerful manipulation than the one in this study nate the facts and examples in the text and to identify it
could suggest whether explicit warrants have a place in as the author's claim. Readers typically used one of two
the argument relationship used by competent readers. strategies: (a) identify a possible claim early on and use
(Complex arguments, such as the one I am making in it to distinguish evidence from nonevidence or (b) post-
this article, require authors to deal explicitly with war- pone identifying the claim for certain until reading the
rants, I suspect.) entire text. For the task in this study, the first strategy led
to greater accuracy than the second since it allowed the
Strategies used by successful readers reader to distinguish ancillary content from evidence and
Although results from this study supported the gen- to resist broadening the claim.
eral comprehension model (see Figure 2), they also sug- Earlier work by Williams, Taylor, and de Cani
gested that the participants frequently were being more (1984) portrays the same two strategies. Another series
strategic than the model implies. Dole et al. (1991) dis- of studies on the effect of elaboration on comprehension
tinguished between skills, which are "highly routinized, by Mohr, Glover, and Ronning (1984) suggests that as
almost automatic behaviors" (p. 242), and strategies, progressively more ancillary content is added to an argu-
which are "conscious, instantiated, and flexible plans" ment, readers become less able to use the first strategy.
(p. 242). To be sure, the general comprehension model They become less accurate in identifying the author's
is not really a skill-based model. It pictures a reader rec- claim. Possibly the text is not providing sufficient cues
ognizing, identifying, and constructing. Nonetheless, the about the claim/evidence relationship that the author has
model predicts that a reader who has recognized a text in mind.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
804 READING RESEARCH QUARTERLY October/November/December 1995 30/4

Some otherwise competent readers in this study


signals they could glean from text openings and clos-
ings,obvi-
demonstrated a certain lack of flexibility. The most a skill they had learned well.
ous and widespread example concerned readers en-Whether readers in this study employed strategies
countering a complex argument that apparently didor skills,
not they seemed to rely heavily on an argument
match their argument schema. Toulmin (1958) schema
described
analogous to other rhetorical schemata hypothe-
sized by Meyer (1985). Knowing and being able to use a
backing, qualification, and rebuttal as argument features
present in complex arguments. Typical persuasive text schema may be particularly important for readers
pieces, e.g., newspaper editorials, exhibit these comprehending
features lengthy texts. They can employ the
(Chambliss, 1994b). Some readers apparently had an ar-
schema to recognize a pattern at the top level of the text
and search for its parts among the scores of words, sen-
gument schema that did not allow for additional argu-
tences, and paragraphs. The schema helps readers pre-
ment features beyond a simple claim and its evidence.
They failed to recognize the complex relationship dict inwhata is to come, separate the wheat from the chaff
so to speak, link the separate parts, and later recall the
complex argument, typically leaving out the complexity
and recalling the text as if it had a simple structure.
text's gist.
A second rather widespread example concerns
reader reliance on text introductions and conclusions. In Case building versus comprehending
two of the three experiments, introductions and/or con- the author's argument
clusions did not summarize the argument accurately for In many ways the context for the study, although
some text versions. The accuracy with which some read- not particularly unusual for many school situations, was
ers identified the argument parts and constructed a gist artificial. Arguments had noncontentious content.
representation suffered accordingly. These readers Participants had no authentic purpose for reading other
seemed to rely on introductory and concluding signals than having volunteered to participate in the study.
unquestioningly. Readers could have recognized a mis-Many of the comprehension measures, certainly includ-
match between a text's conclusion and overall argument ing the think-aloud protocols, were unusual tasks de-
structure, recycled back through the text to focus on thesigned to address hypotheses for the study rather than to
text as a whole, and subsequently ignored the conclud-be meaningful to the participants. It is certainly possible
ing signals. However, no protocol participants seemed that
to only excellent students accustomed to completing
notice discrepancies between the text's argument and its all school tasks would have demonstrated the strategies
introduction and conclusion or reread claim statements used by participants in this study.
and instances of evidence in the body of the text. Readers probably are far more likely to read argu-
Instead, a few protocol readers did reread or point to ments with contentious content than otherwise and to
features in introductions and conclusions when they build a case than represent the author's argument accu-
were experiencing difficulty. rately. Indeed, a preliminary study of readers compre-
Both of these examples may be instances of a con-hending texts with persuasive content, deliberately cho-
flict between skill-based prior knowledge and strategy sen to be contentious, showed participants to engage in
use. One of the two teachers admitted teaching her stu- "biased" processing. Reading texts about either logging
dents to use the simple structure to the exclusion of in the Pacific Northwest or Operation Desert Storm, col-
more complex argument patterns. "If a student wrote a lege students from a university in the northwest U.S. typ-
paper with this more complex structure," she explained, ically failed to remember either claims or evidence that
"I would grade it lower as poorly organized. I never contradicted what they believed to be true prior to read-
thought about teaching students to use more complex ing. Even more striking, so intent were readers in build-
structures." Some of her students had learned the lesson ing a case for their beliefs that different readers holding
about the one "acceptable" relationship between a claimopposing claims often chose the same evidence as sup-
and evidence so well that they failed to comprehend a port for their conflicting beliefs (Chambliss, 1994b;
complex argument accurately. Gamer & Alexander, 1991), findings consistent with
Work by Lorch and Lorch (1985) demonstrated the work in social psychology (e.g., Lord, Ross, & Lepper,
effect that introductions can have on readers. This work 1979).
highlighted the power of text conclusions. Apparently, The results of this study as well as the work on
readers expected authors to signal important information case building have interesting implications for the role of
about a text in its opening and closing sections, some- education. Kuhn (1992) concluded that education makes
what analogously to how people begin and end conver- a difference; apparently people can learn in school to
sations (Calfee & Chambliss, 1987). This expectation mayrecognize claims and evidence, distinguishing between
have predisposed readers to accept uncritically whateverthem so that evidence can be weighed and claims either

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 805

maintained or changed. As I have already noted, compo- construct a text representation that accurately reflects the
sition textbooks consistently present the defining fea- structure in the text itself. Since argument may be a text
tures of arguments (e.g., Hairston, 1992). Logicians (e.g., type where accurate comprehension is particularly im-
Mates, 1972) and rhetoricians (e.g., Ramage & Bean, portant, however, writers should be aware that requiring
1995)-the academicians who study argument-specify readers to infer the point or including content extrane-
the same features. Argument, as it is supposed to be ous to the structure can influence the accuracy with
taught in school, seems to have a formal, recognizable which readers will comprehend the text.
structure that is independent from content and that can In the introduction I suggested that adults who
lead to a reasoned response rather than case building. cannot comprehend lengthy arguments may have had
The participants in this study believed that they had little experience with the argument structure and there-
learned the argument pattern in composition class. One fore never learned either the cues or the strategies to
of their teachers believed that she had taught them to employ them because instructional materials seldom
use the simple argument structure in organizing their exhibit a claim-evidence-warrant structure (Calfee &
writing to the exclusion of complex structures. If these Chambliss, 1988; Chambliss, Calfee, & Wong, 1990).
observations are valid, participants apparently have been Most students are expected to read thousands of pages
able to transfer what they learned about their own writ- of written instructional materials during their 12 years of
ing to the comprehension of arguments written by other public schooling. Redesigning textbook writing to exhib-
authors-at least when reading noncontentious content. it this structure would offer practice comprehending ar-
However, even well-educated readers may be prone to guments that many students may need.
case building when the content is contentious. The col- In some important respects, the results from this
lege students reading texts about logging in the Pacific study are reassuring. Competent readers were able to
Northwest and Operation Desert Storm (Chambliss, recognize lengthy written arguments, to identify the
1994b; Garner & Alexander, 1991) presumably were all claim and evidence, and to put the parts together into a
competent readers. Nonetheless, they consistently re- summary that tended to match the argument. It seems
called the text as if they were building a case for their reasonable that many of these readers had learned in
own beliefs. Tversky and Kahneman (1982) have con- school what the argument cues are and what to do with
cluded that even well-educated, experienced researchers them; that instruction can make a difference. Identifying
are prone to case build when the content is contentious
the specific instructional approaches that would support
and when they are thinking "intuitively." students who have not learned either argument cues or
appropriate comprehension strategies for lengthy texts is
In conclusion an important next step.
Because of the risk in drawing practical implica-
tions from theoretical work, I have resisted suggesting
REFERENCES
AFFLERBACH, P.P., & JOHNSTON, P.H. (1986). What do expert r
instructional strategies implied by the results of this
ers do when the main idea is not explicit? In J.F. Baumann (Ed
study. It does seem possible that what these competent
Teaching main idea comprehension (pp. 49-72). Newark, DE:
readers knew about argument and what they did while International Reading Association.
reading were influenced strongly by their instruction in
ALEXANDER, P. (1969). An introduction to logic: The criticism of
composition class. Furthermore, questions asked of arguments.
pro- New York: Shocken Books.
ARMBRUSTER, B.B., ANDERSON, T.H., & MEYER, J.L. (1991).
tocol participants consistently elicited strategic behavior
from them. Both of these outcomes could suggest specif- Improving content area reading using instructional graphics.
Reading Research Quarterly, 26, 393-416.
ic instructional approaches to use with readers who have
BAIN, A. (1867). English composition and rhetoric: A manual. N
not learned the text cues and strategies employed by theYork: D. Appleton.
competent readers in this study. BARNET, S., & STUBBS, M. (1990). Practical guide to writing (6t
Drawing implications for improved textbook mate- Glenview, IL: Scott, Foresman/Little, Brown Higher Education.
rials seems to be more warranted. Readers were consis- BRIDGE, C.A., BELMORE, S.M., MOSKOW, S.P., COHEN, S.S., &
MATHEWS, P.D. (1984). Topicalization and memory for main
tently affected by the characteristics of the passages they
in prose. Journal of Reading Behavior, 16, 61-80.
read. Otherwise competent readers comprehended less BROOKS, C., & WARREN, R.P. (1972). Modern rhetoric (3rd ed.).
accurately when claims were missing and when content York: Harcourt Brace Jovanovich.
extraneous to the structure was introduced. Their accura- CALFEE, R.C., & CHAMBLISS, M.J. (1987). The structural design fe
cy seemed to be particularly affected by misleading in- tures of large texts. Educational Psychologist, 22, 357-378.
CALFEE, R.C., & CHAMBLISS, M.J. (1988, April). Structure in socia
troductions and conclusions. In a few cases, reader
studies textbooks: Where is the design? Paper presented at the a
comprehension was seriously disrupted. To be sure, it is al meeting of the American Educational Research Association,
not always either necessary or desirable for readers to Orleans, LA.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
806 READING RESEARCH QUARTERLY October/November/December 1995 30/4

CHAMBLISS, M.J. (1994a). Evaluating textbooks for diverse MEYER,


learners.B.J.F., BRANDT, D.H., & BLUTH, G.J. (1980). Use of top-level
Remedial and Special Education, 15, 348-362. structure in text: Key for reading comprehension of ninth-grade
CHAMBLISS, M.J. (1994b). Why do readers fail to change theirstudents.
beliefs Reading Research Quarterly, 16, 72-103.
after reading persuasive text? In R. Garner & P.A. Alexander
MEYER,(Eds.),
B.J.F., & FREEDLE, R.O. (1984). Effects of discourse type on
recall. American Educational Research Journal, 21, 121-143.
Beliefs about text and instruction with text (pp. 75-89). Hillsdale,
NJ: Erlbaum. MEYER, B.J.F., & RICE, G.E. (1982). The interaction of reader strate-
CHAMBLISS, M.J., & CALFEE, R.C. (1989). Designing science textbooks gies and the organization of text. Text: An Interdisciplinary Joumrnal
to enhance student understanding. Educational Psychologist, for24,
the Study of Discourse, 2, 155-192.
307-322. MOHR, P., GLOVER, J.A., & RONNING, R.R. (1984). The effect of re-
CHAMBLISS, M.J., & CALFEE, R.C. (in press). Textbooks for learning: lated and unrelated details on the recall of major ideas in prose.
Nurturing children's minds. Cambridge, MA: Blackwell. Journal of Reading Behavior, 16, 97-108.
CHAMBLISS, M.J., CALFEE, R.C., & WONG, I. (1990, April). Content NICKERSON, R.S. (1991). Modes and models of informal reasoning:
and structure in science textbooks: Where is the design? Paper pre- A commentary. In J.F. Voss, D.N. Perkins, & J.W. Segal (Eds.),
sented at the annual meeting of the American Educational Informal reasoning and education (pp. 291-309). Hillsdale, NJ:
Research Association, Boston, MA. Erlbaum.

COOPER, L. (1932). The rhetoric ofAristotle. New York: Appleton- POLLARD, P. (1982). Human reasoning: Some possible effects of avail-
Century. ability. Cognition, 12, 65-96.
CORBETT, E.P.J. (1965). Classical rhetoric for the modern student. RAMAGE, J.D., & BEAN, J.C. (1995). Writing arguments: A rhetoric
New York: Oxford University Press. with readings (3rd ed.). Boston: Allyn & Bacon.
CREWS, F. (1992). The Random House handbook (6th ed.). New York: ROSENBLATT, L.M. (1978). The reader, the text, the poem: The transac-
McGraw-Hill, Inc. tional theory of the literary work. Carbondale, IL: Southern Illinois
DOLE, J.A., DUFFY, G.G., ROEHLER, L.R., & PEARSON, P.D. (1991). University Press.
Moving from the old to the new: Research on reading comprehen- ROSENBLATT, L.M. (1994). The transactional theory of reading and
sion instruction. Review of Educational Research, 61, 239-264. writing. In R. Ruddell, M.R. Ruddell, & H. Singer (Eds.), Theoretical
ENGLERT, C.S., & HIEBERT, E.H. (1984). Children's developing aware- models and processes of reading (4th ed., pp. 1057-1092). Newark,
ness of text structures in expository materials. Journal of DE: International Reading Association.
Educational Psychology, 76, 65-74. RUDDELL, M.R. (1994). Vocabulary knowledge and comprehension: A
GARNER, R., & ALEXANDER, P.A. (1991, April). Skill, will, and thrill: comprehension-process view of complex literacy relationships. In
Factors in adults' text comprehension. Paper presented at the an- R.B. Ruddell, M.R. Ruddell, & H. Singer (Eds.), Theoretical models
nual meeting of the American Educational Research Association, and processes of reading (4th ed., pp. 414-447). Newark, DE:
Chicago, IL. International Reading Association.
GILBERT, D.T. (1991). How mental systems believe. American RUDDELL, R., RUDDELL, M.R., & SINGER, H. (1994). Theoretical mod-
Psychologist, 46, 107-119. els and processes of reading (4th ed.). Newark, DE: International
HAIRSTON, M.C. (1992). Successful writing (3rd ed.). New York: W.W. Reading Association.
Norton. SHOPEN, T., & WILLIAMS, J.M. (1981). Style and variables in English.
HARE, V.C., RABINOWITZ, M., & SCHIEBLE, K.M. (1989). Text effects Cambridge, MA: Winthrop.
on main idea comprehension. Reading Research Quarterly, 24, SIMON, H.A. (1981). Sciences of the artificial(2nd ed.). Cambridge,
72-88. MA: MIT Press.

KIERAS, D.E. (1980). Initial mention as a signal to thematic content SPIVEY,


in N.N. (1987). Construing constructivism: Reading research in
technical passages. Memory and Cognition, 8, 345-353. the United States. Poetics, 10, 109-192.
KINTSCH, W., & VAN DIJK, T.A. (1978). Toward a model of text com- TOULMIN, S.E. (1958). The uses of argument. Cambridge, UK:
prehension and production. Psychological Review, 85, 363-394. Cambridge University Press.
KIRSCH, I.S., JUNGEBLUT, A., JENKINS, L., & KOLSTAD, A. (1993, TUFTE, E.R. (1990). Envisioning information. Cheshire, CT: Graphics
September). Adult literacy in America: A first look at the results of Press.
the National Adult Literacy Survey. Washington, DC: Office of TVERSKY, A., & KAHNEMAN, D. (1973). Availability: A heuristic for
Educational Research and Improvement, U.S. Department of judging frequency and probability. Cognitive Psychology, 5,
Education. 207-232.
KUHN, D. (1992). Thinking as argument. Harvard Educational TVERSKY, A., & KAHNEMAN, D. (1982). Judgment under uncertainty:
Review, 62, 155-178. Heuristics and biases. In D. Kahneman, P. Slovic, & A. Tversky
KUHN, D., SCHAUBLE, L., & GARCIA-MILA, M. (1992). Cross-domain (Eds.), Judgment under uncertainty: Heuristics and biases (pp.
development of scientific reasoning. Cognition and Instruction, 9, 3-20). Cambridge, UK: Cambridge University Press.
285-327. VANDE KOPPLE, W. (1985). Some exploratory discourse on metadis-
LORCH, R.F., & LORCH, E.P. (1985). Topic structure representation course. College Composition and Communication, 36, 82-93.
and text recall. Journal of Educational Psychology, 77, 137-148. VAN DIJK, T.E., & KINTSCH, W. (1983). Strategies of discourse com-
LORD, C.G., ROSS, L., & LEPPER, M.R. (1979). Biased assimilation and prehension. New York: Academic Press.
attitude polarization: The effects of prior theories on subsequently WADE, S.E., TRATHEN, W., & SCHRAW, G. (1990). An analysis of
considered evidence. Journal of Personality and Social Psychology, spontaneous study strategies. Reading Research Quarterly, 25
37, 2098-2109. 147-166.
MATES, B. (1972). Elementary logic (2nd ed.). New York: Oxford WILLIAMS, J.P. (1984). Categorization, macrostructure, and finding the
University Press. main idea. Journal of Educational Psychology, 76, 874-879.
MEYER, B.J.F. (1985). Prose analysis: Purposes, procedures, and prob- WILLIAMS, J.P., TAYLOR, M.B., & DE CANI, J.S. (1984). Constructing
lems, Parts 1 and 2. In B.K. Britton & J. Black (Eds.), Understand- macrostructure for expository text. Journal of Educational
ing expository text (pp.11-64, 269-304). Hillsdale, NJ: Erlbaum. Psychology, 76, 1065-1075.

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms
Constructing argument gist 807

Received May 30, 1991 per based on the dissertation was presented at the 1991 IRA annual
convention in Las Vegas.
Final revision received November 9, 1994
Accepted December 14, 1994 I gratefully acknowledge the support and encouragement of
the members of my dissertation committee: Robert Calfee, principal
advisor; Richard Snow; and Stephen Witte. I also thank Janice Dole,
AUTHOR NOTES
Ruth Garner, Roger Bruning, and Robert Calfee for their helpful com-
This article is based on a dissertation submitted to the School
ments on earlier drafts of this article. Their input, as well as the feed-
of Education, Stanford University, in partial fulfillment for the degree back of four anonymous reviewers and the three editors of Reading
of Doctor of Philosophy. The dissertation was among the 10 finalistsResearch Quarterly, helped to fine-tune my thinking and clarify my
in IRA's Outstanding Dissertation Award 1990-1991 competition. A writing.
pa-

Call for volunteer consultants

The International Reading Association has recently created a Division of International Develo
will serve to coordinate Association support for literacy activities in the developing world. A
jective of this Division is to establish and manage a volunteer consultancy service in order to of
perienced professional educators on short-term assignments to those developing countries pur
for all. This will be done in conjunction with the IRA ad hoc committee-Volunteers for
Professional Programs.
Interested applicants with strong experience in teacher education and educational assessm
send a brief Curriculum Vitae to the attention of the Director, International Development Div
particular interest will be applicants who have experience in multicultural and internati
Curricula Vitae will be used to build an active database and submission in no way represents a
on the part of the applicants.

International Reading Association


800 Barksdale Rd., PO Box 8139
Newark, DE 19714-8139, USA

This content downloaded from


200.75.19.153 on Sun, 22 Aug 2021 21:55:27 UTC
All use subject to https://about.jstor.org/terms

You might also like