You are on page 1of 16

Comparison of Comprehension Monitoring of Skilled and Less Skilled Readers

Author(s): Diane L. August, John H. Flavell and Rene Clift


Source: Reading Research Quarterly, Vol. 20, No. 1 (Autumn, 1984), pp. 39-53
Published by: Wiley on behalf of the International Reading Association
Stable URL: http://www.jstor.org/stable/747650
Accessed: 23-06-2015 22:12 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/747650?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

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

http://www.jstor.org

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

Comparison of comprehension monitoring of


skilled and less skilled readers
DIANE L. AUGUST
JOHN H. FLAVELL
RENEE CLIFT
StanfordUniversity

of this study was to gain a better understanding of the differences in


THEPURPOSE
comprehension and comprehension monitoring in reading between skilled and less
skilled readers. Sixteen fifth-grade students of each type, matched on intelligence, were
given five stories to read on a microcomputer screen. Three of the stories made little sense
because a page of the story was purposefully omitted to make the story inconsistent. The
skilled readers correctly reported that a page was missing on the inconsistent stories
significantly more often than did the less skilled readers. The same trend obtained for the
subset of subjects in each group who appeared to detect the problem at some lower level,
as evidenced by longer reading times, lookbacks, or the making of unwarranted
inferences at the point of the inconsistency. These group differences in comprehension
monitoring performance could not be explained by differences in intelligence, decoding
skill, or ability to recall crucial story information. In both groups, inferring that is not
marked by the subject as hypothetical seemed to account for more failure to report the
missing page than any other variable.

Une comparaison du contr6le de comprdhension de lecteurs


experimentes et moins experimentes
LE BUT DE CETTE etude 6tait d'acqu6rir un meilleur entendement des diff6rences en
compr6hension et contr6le de compr6hension en lecture parmi des lecteurs exp6riment6s
et moins exp6riment6s. On a donn6 &seize e61vesde sixieme de chaque type, d'intelligence
6gale, cinq histoires a lire sur un 6cran de micro-ordinateur. Trois de ces histoires
n'avaient pas beaucoup de sens parce qu'une page de l'histoire avait 6t6 omise
volontairement pour la rendre inconsistante. Les lecteurs exp6riment6s ont rapport6 de
faqon correcte qu'une page manquait dans l'histoire inconsistante beaucoup plus souvent
que ne l'ont fait les lecteurs moins exp6riment6s. La mime tendance s'est effectu6e pour le
sous-ensemble des sujets dans chaque groupe qui semblait d6tecter le problkme a un
certain niveau inf6rieur, comme cela a 6t6 d6montr6 par des temps de lecture plus longs,
des retours en arriere, ou la production de conclusions injustifi6es la oii l'inconsistance
avait lieu. On n'a pas pu expliquer ces diff6rences de groupe dans l'accomplissement du
contr6le de comprehension par des diff6rences d'intelligence, des comp6tences de
d6chiffrement, ou la capacit6 de se rappeler des informations cruciales dans l'histoire.
Dans les deux groupes, la conclusion qui n'est pas marquee par le sujet comme
hypoth6tique semblait justifier un plus grand 6chec dans le rapport de la page manquante
que dans toute autre variable.

Una comparacidn de comprobaciones de comprensidn de lectores de


mds y menos destreza
EL OBJETIVODE ESTEESTUDIOera avanzar los conocimientos

sobre las diferencias en la

comprensi6n y la comprobaci6n de comprensi6n de lectura de lectores hibiles y de los de


menos destreza. A 16 alumnos de quinto grado, de ambas destrezas lectoras y
equiparados en inteligencia, se proyectaron 5 cuentos para ser leidos en la pantalla de una

39

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

Tresde los cuentostenianpocosentidoporqueunade laspiginasdel


microcomputadora.
cuentohabiasido intencionadamente
omitidaparahacerel cuentoinconsistente.Losde
destrezalectora informaroncorrectamentela omisi6n de la pigina en los cuentos
con mis frecuenciaque los de menosdestrezalectora.
inconsistentessignificativamente
Se obtuvo la misma tendenciadel subgrupode alumnosen cada grupo que parecia
advertir.elproblemaa un nivel mis bajo, manifestadopor lecturade mis duraci6n,
relectura,o haciendo inferenciasinnecesariasen el punto de inconsistencia.Estas
diferenciasde grupo en la comprobaci6nde comprensi6n,no podianjustificarseser
debidasa diferenciasde inteligencia,habilidadde descifre,o a la habilidadde recordar
puntos importantesdel cuento. En ambos grupos,masque ningunaotra variable,la
inferenciaqueno eraindicadaporel alumnocomohipo-tica,pareciasermis responsable
por la falla de informarla omisi6nde la pigina.
A question that continues to perplex
educators is why some younger children and
less skilled readersof normal intelligence have
difficultyevaluatingtheir own comprehension.
Some researchers (Canney & Winograd,
1979; Garner, 1981; Markman, 1977, 1979)
suggest that children fail to understand that a
message is problematic because they do not
integrate material as actively as older or more
skilled readers, but instead process the
material word by word or sentence by
sentence. They do not construct a coherent
representation of the entire story that can
then be evaluated for consistency, i.e., engage
in constructive processing as they read.
Other investigators, mostly in the area of
referential communication, contend instead
that younger children may process the
information in the above constructive fashion
but that this processing does not result in
conscious awareness of noncomprehension.
Children may exhibit nonverbal evidence of
noncomprehension (Flavell, Speer, Green, &
August, 1981;Ironsmith & Whitehurst, 1978;
Patterson, Cosgrove, & O'Brien, 1980) but
not report having encountered a problem.
Current research on the detection and awareness of textual anomaly also supports the
hypothesis that younger children process
constructively but do not necessarily show
awareness of nonunderstanding (Capelli &
Markman, 1981; Harris, Kruithof, Terwogt,
& Visser, 1981).
There are several reasons why children
who have given evidence of on-line monitoring
might not report a problem. Capelli (1982)
suggests that these children: (a) may not
access the relevant information in making
40

judgments because of memory constraints,


limited processing capability, or insensitivity
to cues of nonunderstanding; or (b) may
access the necessary information but use it
differently than adults in making judgments,
i.e., they may set a higher threshold for
judging something incomprehensible, have
different criteria for what constitutes a
problematic message, or give less weight to
internal signals of noncomprehension.
Another possible explanation for the
lack of problem reporting is that children may
make unwarranted inferences to resolve the
problem. In a study conducted with college
students, Baker (1979) found that subjects
might have been better comprehension monitors than the data suggested had they not
spontaneously used "fix-up" procedures to
resolve the potential confusions. That is, once
the subjects established that a confusion
existed, they often drew upon priorknowledge
to render the text more comprehensible.
Garner (1981) suggested that even good
seventh- and eighth-grade readers might rate
inconsistent messages as "okay" rather than
"difficult to understand" because they made
assumptions to resolve the inconsistencies. In
a pilot study by August (1981), both skilled
and less skilled readers made inferences to fill
in missing information thus making the
message consistent or unambiguous. Differences in problem reporting occurred (skilled
readers knew the messages were inconsistent
or ambiguous) because unskilled readers did
not rememberthey had made inferenceswhile
skilled readers did.
The purpose of the present study was to
find out more about factors that contribute to
poor comprehensionmonitoringby comparing

* Fall 1984
READINGRESEARCH
QUARTERLY

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

XX/

skilled and less skilled fifth-grade readers.


Children were presented with five stories,
three of which had a missing page that made
the story inconsistent. It was hypothesized
that skilled readers would be likelier than less
skilled readers to recognize that a page was
missing in the case of these three stories.
Investigators in the area of comprehension monitoring (Garner & Anderson, 1982;
Winograd & Johnson, 1980) caution that the
results of studies that depend on the error
detection paradigm are difficult to explain
because there are so many factors that may
account for the lack of problem detection
besides poor comprehension monitoring.
This study attempted to reduce these difficulties in the following ways:
1. Children tend to assume that written
materials are adequate and are consequently
not prepared to look for or find problems in
textual materials. In addition, they may fail to
judge materials as incomprehensible if their
reasons for reading differ from the experimenter's; Baker (1979) found that college
students failed to detect problems in their
reading material because they were focusing
on the main idea, ignoring in the process a
sentence that did not fit in. In the present
study children were told that some of the
stories might have a missing page. This
alerted them to the possibility that the
materials might be defective and encouraged
them to criticize the stories. In addition, it
provided them with the same criteria for
judging comprehensibilityas the experimenter.
The stories were considered comprehensible if
they did not have a missing page.
2. Poor recall of what has been read or
making inferences to fix up the message may
account for failure to notice or to report a
problem. In this study children were asked to
retell the story after reading it. This informed
the experimenter whether children had the
relevant information to decide if a message
was problematic and whether they made
unwarranted inferences.
3. Verbal reports of cognitive activities
do not necessarily coincide with actual
behavior (Nisbett & Wilson, 1977). Having
children read the stories on a microcomputer
provided on-line measures of comprehension
monitoring--specifically reading time and
Comparison

of comprehension

monitoring

lookbacks at the inconsistent part of the


story. These on-line measures were used as
indications of whetherchildrenwere constructively processing the information.
4. Probes often used in previous studies
may have affected the children's subsequent
performance in ways that were difficult to
assess. In this study, by contrast, the children
were probed only after they had read the
stories once to make sure that the probes
would not affect subsequent performance.
5. The nature of the stories influences
performance. Garner and Anderson (1982)
found differentialerror detection performance
by two comparablegroups using a comparable
procedure but with different passages. In the
present study one half of the subjects in each
group read one series of stories and the other
half read another series in order to control for
story effect. For internal consistency, the
error type and its location were always the
same in each of the stories.
6. Nonverbal intelligence may account
for differences in error detection. Most
studies fail to control for this variable. In this
study each skilled readerwas matched with an
unskilled reader on nonverbal intelligence to
eliminate intelligence as a factor responsible
for between-group differences in comprehension monitoring.

Method
Subjects
Thirty-two fifth-grade students from two
school districts in Northern California participated in this study. The subjects were
classified as skilled and less skilled readers on

the basis of their scores on the comprehension


subtest of the Comprehensive Test of Basic
Skills, Level 2, Form S (1973), which was
administered in April, immediately prior to
the study. Skilled readers were defined as
those who scored at or above grade level (5.8
or above); less skilled readers were defined as
those who scored at least 1 year below grade
level (4.7 or below). Children in between
(i.e., those who were less than 1 year below
grade level) were excluded from this study.
All subjects were also given the Coloured
Progressive Matrices Test (Raven, 1962), a

AUGUST,FLAVELL,& CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

41

nonverbal measure of intelligence. Each order to determine readability level. Correskilled reader was paired with an unskilled sponding pages in each story had the same
reader with no more than 1point difference in number of syllables. Only simple sentence
their raw IQ scores. Sixteen pairs of subjects structures were used, and each story followed
were formed. The skilled readers had a mean the same story grammarformat developed by
reading comprehension score of 8.0 (SD = Stein and Trabasso (in press). On the first
1.5, range = 5.8 to 9.9). The less skilled readers page the protagonists and social/physical
had a mean reading comprehension score of environment were introduced (the so-called
3.5 (SD = .80; range = 2.3 to 4.7). A t-test setting). Each of the following pages correindicated that these scores were significantly sponded to an element in the realization of the
different, t(15) = 10.46, p = .0001. On the goal: initiating event, internal response,
Coloured Progressive Matrices Test both attempt, obstacle, solution or reaction, consegroups had the same mean score of 29.43, quence, and reaction. The only difference
with a range of 25 to 35. Standard deviations between the inconsistentand consistent stories
were 3.4 and 3.1 for the skilled and less skilled was that the former contained only reactions
to the obstacle rather than an effective way of
readers, respectively.
To make sure that all subjects would be dealing with it (solution).
A second series of stories was formed by:
able to decode the words in the test material,
all subjects read, under supervision, 30 words (a) converting the first two inconsistent
representative of the words they would later stories in the first series into consistent stories,
encounter in the stories. The procedure used accomplished by changing the reaction to the
to select the words was developed by Calfee obstacle (page 6 of the inconsistent story) into
and Calfee (1980). Because the scores in this a sensible solution (page 6 of the consistent
decoding test were high, none of the students story); (b) changing the two consistent stories
tested had to be disqualified from the study. in the first series into inconsistent stories by
reversing the procedure just described. Half
of the subject pairs read the first series and the
Materials
other half read the second series. Alternate
The materials consisted of two series of
of
pairs
(beginning with those who
five stories each. (See Appendix A.) Each had the subjects
lowest intelligence scores) were
story was eight pages long and approximately assigned to each series so that each group
130 words in length. From three of the stories would have the same mean score and there
in each series a portion of the story was would not be an interactionIQ
between group IQ
purposefully omitted to render the stories and series. Two series were used so that the
internally inconsistent. The other two stories results would not be attributable to story
were left intact and consistent. In a pilot effects.
study, all 10 adult expert readers, 5 reading
Children read stories on a Commodore
each series, detected the intended inconsisso that reading times and
microcomputer
tencies in the three problematic stories. With lookbacks could be
automatically and preeach of the two series, the order of administrarecorded. The pages of each story
cisely
tion of the stories was the same: inconsistent,
appeared one at a time. By pressing a button
consistent, inconsistent, consistent, inconsis- the child would advance to the next
page and
tent.
to
a
back
another,
previous
up
Several things were done to make the by pressing
consistent and inconsistentstoriescomparable. page.
All stories were written at the second-grade
readability level. The formula used (Bur- Design
The experiment was designed to assess
meister, 1978) called for 100 words to be
selected from each passage (an entire story in differences between skilled and less skilled
one instance). Average word length was then readers, matched on intelligence, in gist recall
plotted against average sentence length in and comprehension monitoring. The latter

42

READINGRESEARCH
QUARTERLY* Fall 1984

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

XX/1

was assessed by two on-line measures and by


verbal report. The two on-line measures
consisted of: (a) the time it took to read the
two pages following the built-ininconsistencies
as compared to the time it took to read the
equivalent portions of the consistent stories,
and (b) the difference in the number of
lookbacks in the two kinds of stories. As to
verbal report, after reading each story the
subjects were asked if they thought the story
was inconsistent. If they thought it was, they
were also asked to locate the inconsistency,
explain why they thought the story was
inconsistent, and indicate what they would
say to fix the story.
Another objective of the study was to
investigate the relationship between verbal
report of problem detection and (a) gist story
recall, and (b) on-line measures of problem
detection.

Procedure
After all the preliminarytesting was done
(i.e., Comprehensive Test of Basic Skills,
Coloured Test of Progressive Matrices, Decoding Test), the children were tested individually in a session that lasted about 30 minutes
per subject. The sessions were tape recorded
and eventually transcribed.
The procedure consisted of a first pass
through the five stories (initial reading)
followed by a review of the same five stories:
1. The children were told that they
would be reading five stories and that "some
of the stories might have a page missing" and
that "this would mean that a piece of the story
had been left out." They were told that they
would be reading each story to themselves
after which they would be asked if they
thought a page was missing.
2. The subjects were taught how to read
the stories on the computer, how to advance
the pages, and how to back up. They were
specifically told that going back a page was
just like rereading a page in a book. After
practicing the page-turning procedure, they
had to follow the investigator's instructions
correctly-seven commands to advance or
back-up-before they were allowed to read
the experimental stories.

Comparison of comprehension

monitoring

3. After readingeach story to themselves,


the subjects were asked to recall it in their own
words. They were then asked if they thought a
page was missing in the story.
4. If they thought a page was indeed
missing, the subjectswere handed the story on
eight 3 x 5 in. (7.62 x 12.70 cm) cards (each
card corresponding to a page on the computer's screen) and asked where they would
place the missing page, why they thought a
page was missing, and what they thought the
missing page would have to say in order to fix
the story.
5. If the subjects concluded that no page
was missing, they were simply asked to go on
to the next story.
6. The fifth story was read aloud first
from the index cards and then again on the
computer. Here again, the children were
asked if they thought a page was missing and
the procedure outlined under 4 and 5 was then
followed.
7. During the review each child was
probed. Beginning with the first story, they
were all asked to rereadpages 5, 6, and 7 (i.e.,
the obstacle, reaction or solution, and consequence pages) from the computer screen "just
so that the experimenter would be sure that
she got everything right." Having read the
three pages, the subjects were asked to recall
what they had just read. The experimenter
then said that "some kids think there is a page
missing somewhere betweenthese three pages:
What do you think?"
8. If they thought that a page was
missing, they were again asked the three
questions in 4 above.
9. If a subject thought that no page was
missing, he or she would be confronted
directly with the inconsistency (i.e., If David's
car stopped, how was he able to win the
race?). A list of these probes can be found in
Appendix B.
Coding Instrument and Interjudge
Reliability
A coding instrument was developed to
quantify target information from the transcripts of the experimental sessions. In order
to assess interjudge reliability: (a) a primary

AUGUST,FLAVELL,& CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

43

judge coded all 32 transcripts;(b) a subset of


eight transcripts, half of skilled and half of
less skilled readers, was randomly selected
and scored by an informed reliability judge;
(c) for each coding category, the percentageof
interjudge agreement between the primary
judge and the informed reliability judge was
computed. For each category, this was done
by computing how frequently across 8
subjects, three tasks per subject, the two
coders agreed exactly. Interjudge agreement
ranged from .83 to 1.00.

Results
Differences between skilled and less
skilled readers in verbal report of problem
detection and the explanation for these
differences were the central issues of the
study.
Initial Reading
Verbal report of problem detection was
assessed by the children's answers to four
questions asked after each story. Each
question corresponded to a dependent measure: (a) "Do you think a page is missing?"
(missing page detection); (b) "Where would
you put the missing page?" (missing page
placement); (c) "Why do you think a page is
missing?" (explanation of the problem); (d)
"If you were going to fix the story, what
would you make the missing page say?"
(booklet repair). A score of 1 was given for
each correct response. For each subject a
composite verbal report of problem detection
score for each dependent measure was formed

by averaging scores for the three inconsistent


stories. The average composite scores can be
found in Table 1, T-tests were used to
compare problem detection for the two
groups across all three inconsistent stories.
Significant group differences favoring the
skilled readers were found for the dependent
measures:missing page detection, t(30) = 2.84,
p <.01; correct page placement, t(30) = 2.07,
p <.05; correct booklet repair, t(30) = 2.57, p
<.05; the groups did not differ significantly
on correct explanation.
Skilled and less skilled readers were
compared on gist recall of the inconsistent
stories to determine whether this variable
could have accounted for differences in verbal
report of problem detection. To quantify gist
recall, information in the obstacle, reaction,
and consequence pages was partitioned into
idea units. Each unit was a sentence or clause
that referred to only one event. For scoring
purposes, recall was scored as correct if
children reported the exact words or close
synonyms for the subject, verb, and noun
phrase for each sentence or clause. Children
who rememberedall the information from the
obstacle, reaction, and consequence pages
crucial to the creation of the inconsistent
message received a score of 2. Children who
rememberedsome or none of the information
received a score of 1. For each subject, a
composite gist recall score was formed by
averaging scores for the three inconsistent
stories. The average composite scores for each
group can be found in Table 2. T-testsused to
compare the gist recall of skilled and less
skilled readers using the composite scores for
all inconsistent stories revealed no significant

Table 1 Average composite problem detection scores of skilled and less skilled readers-initial reading
Skilled Readers
Measures
Missing page detection
Correct page placement
Correct explanation
Correct booklet repair

44

Less Skilled Readers

SD

SD

2.13
1.81
1.44
1.69

.81
1.05
.90
.95

1.25
1.06
.88
1.00

.93
1.00
.89
.97

* Fall 1984
READINGRESEARCH
QUARTERLY

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

XX/1

Table 2

Average composite recall scores of


skilled and less skilled readers
Skilled Readers

Less Skilled Readers

Stories

SD

SD

Initial
Review

1.73
1.83

.30
.21

1.60
1.69

.39
.28

group differences. It does not appear to be the


case, therefore, that differences in story recall
could account for differences in verbal report
of problem detection.
Three indices were used to find out if
children were detecting the problem at some
level but not reporting it. Two were on-line
measures of problem detection and the third
was a measure that indicated whetherchildren
were inferring information to make the story
consistent. The first on-line measure was a
rank-order reading score. Children received a
score of from 0-4 depending on how often
their reading time on page 7 and 8 of the
inconsistent stories (excluding the fifth story)
was greater than their reading time of pages 7
and 8 of the consistent stories. A score of 4
indicated that they had longer reading times
on pages 7 and 8 of both inconsistent stories.
Other scores corresponded to the following
rank orders (with reading times increasing
from left to right)-consistent, inconsistent,
consistent, inconsistent = 3; consistent, inconsistent, inconsistent, consistent = 2; inconsistent, consistent, consistent, inconsistent = 2;
inconsistent, consistent, inconsistent, consistent = 1;inconsistent, inconsistent, consistent,
consistent = 0. Average group scores are
presented in Table 3 where it can be seen that
skilled readers spent more time on the
Table 3

Average reading time rank order


scores of skilled and less skilled
readers
Skilled Readers

Less Skilled Readers

Stories

SD

SD

Initial

3.25

1.00

2.25

1.24

Comparison of comprehension

monitoring

inconsistent stories than less skilled readers.


A t-test comparing skilled and less skilled
readers revealed significant differences favoring skilled readers, t(30) = 2.51, p <.05.
The second on-line measure was a
lookback difference score. It was assumed
that children who had detected a problem
would look back more often after page 7 or 8
of an inconsistent story than after the same
pages of a consistent story. This score was
formed by subtracting the sum of lookbacks
on an inconsistent story from that of the
immediately preceding consistent story. Each
child received a lookback difference score for
the first pair of stories and one for the second
pair of stories. No statistical tests were
performed because very few subjects had
other than zero difference scores. This was
attributable to the small number of subjects
who looked back at all--9 subjects in the first
set of stories and 10 subjects in the second set.
Overall, 9 of the 32 subjects looked back more
(at least once) in reading the inconsistent
stories. Six of these 9 readers were skilled.
These skilled readers, when confronted with
an inconsistent message, were more likely to
search for information that would clarify
their misunderstanding.
There was a considerable amount of
inference generation to fix-up the story. More
than one-third of the readers(both skilled and
less skilled) made their own inferences in
Stories 1 and/or 3. In these stories, 6 of the 12
skilled readerswho did not detect the missing
page made inferences to fix up the story. Ten
of the 20 less skilled readers who did not
detect the missing page made inferences.
Skilled readers tended to slow down
more at the inconsistency than less skilled
readers. This in part may explain differences
in problem reporting. However, for those
children in both groups who did show signs of
minimal detection as measured by longer
reading times at the inconsistency, lookbacks
at the inconsistency, and/ or inferring,the less
skilled readers failed to report that the story
was inconsistent more often than the skilled
readers (see Table 4). According to a onesided, binomial test for the equality of
proportions (this amounts to halving the
significance level for the usual chi-square test,
taking into account the fact that the effect is

AUGUST,FLAVELL,& CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

45

going in the expected direction), these differences approach significance for Story 1
(vX'2
=1.46, p = .07) and are significant for Story 3
= 1.89, p = .03). For Stories 1 and 3
(I/?
combined, less skilled readers who showed
signs of minimal detection reporteda problem
30% of the time while the skilled readerswho
showed these signs reported a problem 89%of
the time.
Table 4

Problem reporting of skilled and


less skilled readersshowing minimal
signs of problem detection
Skilled Readers

Less Skilled Readers

Story #

Report

No Report

Report

No Report

Onea
Threeb

9
10

6
5

3
4

7
9

"p= .07. bp = .03.


Review
Parallel analyses were also made of the
data collected when the children reread three
pages of the stories and were probed (the
review). Across the three inconsistent stories,
t-tests revealed significant group differences
in missing page detection, t(30) = 2.99,p<.01;
correct explanation, t(30) = 2.57, p <.05; and
correct booklet repair, t(30)= 2.57,p<.05, all
favoring skilled readers (see Table 5). Of
interest is that less skilled readers were not
helped by the probes.
Pearson correlations calculated between
intelligence and missing page detection did
not reveal a significant relationship in the

initial reading.However,therewas a significant


relationship in the review, r = .38, p <.05.

Discussion
To summarize, in the initial reading of
the three inconsistent stories the less skilled
readers were significantly poorer at reporting
the missing page, placing it correctly, and
fixing the story. Differences in intelligence
cannot explain these differences because the
two groups were equated for intelligence and
because intelligence and problem detection
were not significantly correlated. Differences
in decoding cannot explain these differences
because all children who participated in the
study passed a decoding test based on the
words in the story. In addition the stories were
written at the second grade readability level.
Because there were not significant differences
in gist recall of crucial information, recall
cannot explain the differences in problem
reporting. Even though skilled readersslowed
down at the inconsistency more than less
skilled readers,for the children in both groups
who did show minimal detection, as evidenced
by either longer reading time or lookbacks at
the inconsistency and/ or inferring,less skilled
readers failed to report the problem significantly more often than skilled readers. In
addition, the review and probes did not help
these less skilled readers. Less skilled readers
who indicated minimal detection went from a
70% nonreport rate in the initial reading to a
64% nonreport rate in the review.

Table5 Average problem detection scores of skilled and less skilled readers on inconsistent
stories-review reading
Skilled Readers
Measures

2.31
Missing page detectiona
2.13
Correct page placement
1.94
Correct explanationb
Correct booklet repairc
2.19
= 2.99,p <.01. bt(30) = 2.57,p <.05. ct(30)= 2.57,p <.05.
"at(30)

46

Less Skilled Readers

SD

SD

.70
.81
.85
.75

1.31
1.25
1.06
1.31

1.14
1.18
1.06
1.14

* Fall 1984
READINGRESEARCH
QUARTERLY

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

XX/ 1

Although less skilled readers reported


the problem much less frequently, the reasons
for the lack of problem reporting for Stories 1
and 3 appear to be the same for both skilled
and less skilled readers. In both groups, all in
all, inferring accounts for more failure to
report the missing page than any other
variable-62% of the instances of no report of
the less skilled readers and 55% of the
instances of no report of the skilled readers.
Our belief is that inferring that is not marked
as hypothetical by the children may lead them
to believe that a message is consistent. In 25%
of the instances of no report for the skilled
readers and 9% of those for the less skilled
readers, the children did not rememberall the
information needed to createthe inconsistency.
They may have forgotten the problematic
material and perhaps the problem itself at the
time when they had to decide whether the
message was problematic. Twelve percent of
the instances of no report of the skilled
readers cannot be explained by our data.
These children indicated by lookbacks and by
longer reading times that they were detecting
the problem in some sense, at some level.
They were not making inferences to fill in
missing information, thus making the message
consistent. They had perfect recall of the
information crucial to creating the inconsistency. As noted in the introduction, Capelli
(1982) suggests possible explanations for this
phenomenon.
In the introduction two views of why
some readers demonstrate little comprehension monitoring were presented. Essentially,
these views are that these readers do not
sufficiently integrate the text material to
detect problems or they do integrate and
construct information in text, but do not
report detection. There appears to be a subset
of less skilled readers who do not sufficiently
integrate text material and so do not report a
problem with the text. Evic'mnceof this is that
less skilled readers tend to slow down less at
the inconsistency than skilled readers. However, this does not explain all of their failure
to report the problem. Both the less skilled
readers who do show minimal detection as
evidenced by either longer reading time or
lookbacks at the inconsistency and/ or inferComparison of comprehension

monitoring

ring and do not report a problem, as well as


their skilled counterparts, corroborate the
findings of those investigators who contend
that children process information in a constructive fashion but that this processing does
not result in conscious awareness of noncomprehension.
The results of this study suggest four
possible implications for education.
1. A child's ability to recall a text does
not necessarily imply that the text has been
fully understood. In three of the stories (1, 3,
and Review 3) there was no relationship
between gist recall of information and verbal
report of problem detection: As many children
remembered all the crucial information without detecting the missing page (12 in Story 1,
8 in Story 3, 12 in Review Story 3) as
remembered all the information and did
detect it (11 in Story 1, 10 in Story 3, 12 in
Review Story 3). These findings suggest that
rememberingthe pertinent information is not
necessarily indicative of text comprehension.
This suggests that teaching strategies that rely
solely on recall questions as indicators of
student comprehension may not provide the
teacher with sufficient information about the
student's reading abilities. Teachers should be
trained to use a variety of assessment
techniques to find out if a student has fully
understood the text.
2. Children, especially those who are less
skilled, may engage in constructive processing
without reporting a problem in the text. In the
present study, one-half of the children in each
group who made inferences to fill in the
perceived gaps did not report a problem.
Activities such as giving children practice in
discovering missing information in problematic messages, having them generate inconsistent messages and then make them nonproblematic, and helping them distinguish between
inconsistencies that are resolved by the text
and those that are not might help prevent
children from making unwarrantedinferences
and then forgetting they have done so.
Because other possible causes of poor comprehension monitoring are children's insensitivity to such cues as a slow-down in reading
rate, and/or a greater expenditure of effort,
or their failure to use these cues as an

AUGUST,FLAVELL,& CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

47

indication of a problematic message, children


trained to be aware of these feelings may well
become better comprehension monitors. They
could be given obviously inconsistentpassages
to read, passages that cause a "jolt," and
asked to tell why the passage was inconsistent
and how they felt when reading it. Slowly
decreasing the magnitude of these feelings of
noncomprehension by making the inconsistency less obvious might help sensitize
children to the problem.
3. Probing questions do not necessarily
impel a student to question the adequacy of a
message. The probes helped the more intelligent children rather than the skilled or less
skilled, as evidenced by the significant correlation between problem reporting and intelligence on the review data. This implication is
particularly noteworthy if one educational
goal is to foster critical analysis of textual
understanding of reading material. More
explicit instruction may be necessary to foster
adequate comprehension or comprehension
monitoring.
4. Teachers would be well advised to
assume that children who score at or above
grade level on standardized comprehension
tests still need instruction in comprehension
monitoring. Although there were certainly
differences in comprehension monitoring
between our skilled and less skilled readers,
the skilled readers were not at ceiling on our
tasks. This was true in spite of the fact that the
stories were well below grade level and
obviously flawed, and that there were no
subtle distinctions or complex relationships
such as those that exist in some school texts.
REFERENCES
AUGUST,D.L. (1981). Individual differences in the com-

prehension monitoring of inconsistencies in text.


Unpublished manuscript, Stanford University.
L. (1979). Comprehension monitoring: IdentifyBAKER,
ing and coping with text confusions. Journal of
Reading Behavior, 11, 363-374.
L.A. (1978). Reading strategies for middle
BURMEISTER,
and secondary school teachers. Reading, MA:
Addison-Wesley.
CALFEE, R.C., & CALFEE, K.H. (1980). Reading, reading
acquisition. Prepared for the Course Team, Deacon
University Open Campus Program, School of

48

Education, Maryborough,Victoria, Australia:Hedges


& Bell Pty. Ltd.
CANNEY, G., & WINOGRAD,P. (1979, April). Schematic
for reading and reading comprehensionperformance.
(Tech. Rep. No. 120). Urbana: University of Illinois,
Center for the Study of Reading. (ERIC Document
Reproduction Service No. ED 169 520)
CAPELLI, C.A. (1982). Factors affecting children's ability

to evaluate their comprehension. Unpublished manuscript, Stanford University.


CAPELLI, C.A., & MARKMAN, E.M. (1981). Children s sensitivity to incomprehensive material in written text.
Unpublished manuscript, Stanford University.
COMPREHENSIVE TEST OF BASIC SKILLS, LEVEL 2, FORM S.

(1973). Monterey, CA: CTB/ McGraw-Hill.


& AUGUST, D.L.
(198 1). The development of comprehension monitoring and knowledge about communication. Monographs of the Society for Reseach in Child Development, 46(4, Serial No. 191).
R. (1981). Monitoring of passage inconsistency
GARNER,
among poor comprehenders:A preliminarytest of the
"piecemeal processing" explanation. Journal of
Educational Research, 74, 159-162.
J. (1982). Monitoring-ofGARNER,R., & ANDERSON,
understanding research: Inquiry directions, methodological dilemmas. Journal of Experimental Education, 50, 70-76.
FLAVELL, J.H., SPEER, J.R., GREEN, F.L.,

HARRIS, P.L., KRUITHOF, A., TERWOGT, M.M., & VISSER,

T. (1981). Children'sdetection of awareness of textual


anomaly. Journalof Experimental Child Psychology,
31, 212-230.
IRONSMITH, M., & WHITEHURST, G.J. (1978). The development of listener abilities in communication: How
children deal with ambiguous information. Child
Development, 49, 348-352.
E.M.(1977). Realizing that you don't underMARKMAN,
stand: A preliminary investigation. Child Development, 48, 986-992.
MARKMAN,
E.M. (1979). Realizing that you don't understand: Elementary school children's awareness of
inconsistencies. Child Development, 50, 643-655.
T.D. (1977). Telling more than
R.E.,& WILSON,
NISBETT,
we know: Verbal reports of mental processes.
Psychological Review, 34, 231-259.
PATTERSON, C.J., COSGROVE, J.M., & O'BRIEN, R.G. (1980).
Non-verbal indicants of comprehension and noncomprehension in children. Developmental Psychology,
16, 38-48.
J.C.(1962). Coloured Progressive Matrices (Sets
RAVEN,
A, Ab, B). London: H.K. Lewis & Co.
STEIN, N.L., & TRABASSO,T. (in press). What's in a story:

An approach to comprehension and instruction. In


R. Glaser (Ed.), Advances in the psychology of
instruction (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum
Associates, Inc.
WINOGRAD, P., & JOHNSON,P. (1980, January). Comprehension monitoring and the errordetection paradigm
(Tech. Rep. No. 153). Urbana: University of Illinois,
Center for the Study of Reading. (ERIC Document
Reproduction Service No. ED 181 425)

READING

RESEARCH QUARTERLY * Fall 1984

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

XX/

Footnote
The research reported herein was supported in part by
the National Institute of Education under Contract No.
US-NIE-G-81-0116. This report does not necessarily
represent the opinion of NIE or the U.S. Department of
Education.

APPENDIXES

Appendix A: Stories

Series #1: Story 1


Kate lived with her parents in a house by the railroad tracks.
Kate's father worked for the railroad.
One day a terrible storm caused a flood.
The flood washed away the wooden train bridge near Kate's house.
Kate knew that she would have to stop the train before it got to the bridge.
She decided to run to the tracks to warn the engineer.
Kate grabbed a flashlight.
She ran toward the tracks.

Setting
Initiating Event
Internal Response
Attempt

She was about five hundred yards away from the railroad tracks.
Then she fell down hard.
She hurt her left leg.

Obstacle

Kate was worried.


She knew that someone had to stop the train before it got to the bridge.
The train stopped safely before the bridge.

Reaction

Kate was very glad that she had helped.


The railroad gave her a medal for saving the train.
Series #1: Story 2
David was a race car driver.
He drove a red car with big bold stripes.
One day he decided to enter a race.
The winner of the race would be famous.
He wanted to win.
He knew that he would have to drive faster than he had ever driven before.
On the day of the race David drove as fast as he could.
He stayed on the inside track.
David was winning the race.
All of a sudden his engine began to lose power.
His car began to slow down.
It stopped.
Then David saw he was out of gas.
He put more gas in the tank and the car started.

Consequence
Reaction

Setting
Initiating Event
Internal Response
Attempt
Obstacle

Solution

David came in first


and the crowd went wild.

Consequence

David was so proud that he had won his first big race.
He received a beautiful gold trophy.

Reaction

Comparison of comprehension

monitoring

AUGUST,FLAVELL, & CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

49

Series #1: Story 3


Annie was sitting by the window watching the rain.
It had been raining for two weeks.
All of a sudden the dam near the town burst.
Cars and trucks were swept away by the water.

Setting
Initiating Event

Annie knew that she had to get out of the house.


She decided to climb up onto her roof.

Internal Response

She opened the window.


She stood on the wet window ledge.
Annie began to pull herself up to the roof.
She lost her balance and then she slipped.
She fell straight toward the cold water.

Attempt

Annie knew that she was in great danger.


She hoped that she could save herself from drowning.
Annie was safe
on top of the high roof.
Annie was very happy
that she had made it.
She knew soon she would be dry and warm again.

Reaction

Obstacle

Consequence
Reaction

Series #1: Story 4


Allen and Mary lived in Japan.
They owned a boat.
Their mother was going to have a birthday soon.
She would have a big surprise party.
They wanted to get her a wonderful present.
They decided to look for a black pearl.
They took their boat
to a quiet spot.
They lowered themselves into the water.
They began to open oyster shells
to find black pearls.
Inside a large oyster
Allen found the pearl.
Suddenly a large octopus grabbed his foot.
Allen struggled to get free.

Setting
Initiating Event
Internal Response
Attempt

Obstacle

Solution

Mary turned
and saw Allen's problem.
She struck the octopus
and it let him go.
They floated to the surface
with the pearl.
They were happy
about finding the pearl.
Their mother would be very excited
with her gift.

Consequence
Reaction

Series #1: Story 5


Rod and Steve were good friends.
They climbed mountains together.
They decided to climb a very dangerous mountain.
They would be the first Americans
to climb the mountain.

50

Setting
Initiating Event

READING RESEARCH QUARTERLY *

Fall 1984

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

XX/ 1

They wanted to get to the top of the mountain safely.


They only packed the best mountain climbing tools.
On the mountain
Rod climbed first.
Sometimes he pulled Steve
up steep ledges
by a rope.
They were near the mountain top.
Steve was holding onto the rope.
Suddenly the rope broke.
Steve began to slide down the mountain.
Steve was very much afraid.
He hoped that somehow Rod would be able to help him.
They continued climbing toward the top.
When they reached the top
the men were happy.
They stuck a flag in the ground
to prove they made it.
Series #2: Story 1
David was a race car driver.
He drove a red car with big bold stripes.
One day he decided to enter a race.
The winner of the race would be famous.

Internal Response
Attempt

Obstacle

Reaction
Consequence
Reaction

Setting
Initiating Event

He wanted to win.
He knew that he would have to drive faster than he had ever driven before.
On the day of the race
David drove as fast as he could.
He stayed on the inside track.

Internal Response
Attempt

David was winning the race.


All of a sudden
his engine began to lose power.
His car began to slow down.
It stopped.
David knew he had to work fast.
He knew if he did not hurry he would lose the race.

Obstacle

David came in first


and the crowd went wild.

Consequence

David was so proud that he had won his first big race.
He received a beautiful gold trophy.

Reaction

Series #2: Story 2


Kate lived with her parents in a house by the railroad tracks.
Kate's father worked for the railroad.
One day a terrible storm caused a flood.
The flood washed away the wooden train bridge near Kate's house.
Kate knew that she would have to stop the train before it got to the bridge.
She decided to run to the tracks to warn the engineer.
Kate grabbed a flashlight.
She ran toward the tracks.
She was about five hundred yards away from the railroad tracks.
Then she fell down hard.
She hurt her left leg.

Comparison of comprehension

monitoring

AUGUST,FLAVELL,& CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

Reaction

Setting
Initiating Event
Internal Response
Attempt
Obstacle

51

Kate grabbed the flashlight and she began to blink danger.


The engineer saw the light.
The train stopped safely before the bridge.

Solution
Consequence
Reaction

Kate was very glad that she had helped.


The railroad gave her a medal for saving the train.
Series #2: Story 3
Allen and Mary lived in Japan.
They owned a boat.
Their mother was going to have a birthday soon.
She would have a big surprise party.
They wanted to get her a wonderful present.
They decided to look for a black pearl.
They took their boat
to a quiet spot.
They lowered themselves into the water.
They began to open oyster shells
to find black pearls.
Inside a large oyster
Allen found the pearl.
Suddenly a large octopus grabbed his foot.
Allen struggled to get free.

Setting
Initiating Event
Internal Response
Attempt

Obstacle

Allen knew he couldn't call for help.


He knew Mary could not hear him underwater.

Reaction

They floated to the surface with the pearl.


They were happy
about finding the pearl.
Their mother would be very excited
with her gift.

Consequence
Reaction

Series #2: Story 4


Annie was sitting by the window
watching the rain.
It had been raining for two weeks.
All of a sudden
the dam near the town burst.
Cars and trucks were swept away by the water.
Annie knew that she had to get out of the house.
She decided to climb up onto her roof.
She opened the window.
She stood on the wet window ledge.
Annie began to pull herself up to the roof.
She lost her balance
and then she slipped.
She fell straight towards the cold water.

Setting

Initiating Event

Internal Response
Attempt
Obstacle

She landed on the balcony below


and began to climb up to the roof again.
Annie was safe
on top of the high roof.
Annie was very happy
that she had made it.
She knew soon she would be dry and warm again.

52

Solution
Consequence
Reaction

READING

RESEARCH QUARTERLY * Fall

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

1984

XX/ 1

Series #2: Story 5


Rod and Steve were good friends.
They climbed mountains together.
They decided to climb a very dangerous mountain.
They would be the first Americans
to climb the mountain.

Setting
Initiating Event

They wanted to get to the top of the mountain safely.


They only packed the best mountain climbing tools.
On the mountain
Rod climbed first.
Sometimes he pulled Steve
up steep ledges
by a rope.
They were near the mountain top.
Steve was holding onto the rope.
Suddenly the rope broke.
Steve began to slide down the mountain.
Steve was very much afraid.
He hoped that somehow Rod would be able to help him.
They continued climbing toward the top.

Internal Response
Attempt

Obstacle

Reaction
Consequence
Reaction

When they reached the top


the men were happy.
They stuck a flag in the ground
to prove they made it.

Appendix B: Probes
Series #1
Kate fell and hurt her leg. How was she able to help stop the train?
David's car stopped. How was he able to win the race?
Annie slipped and fell towards the water. How did she get up onto the roof?
Allen was grabbed by an octopus. How did he get to the surface?
Steve began to slide down the mountain. How did he get to the top?
Series #2
David's car stopped. How was he able to win the race?
Kate fell and hurt her leg. How was she able to stop the train?
Allen was grabbed by an octopus. How did he get to the surface?
Annie slipped and fell towards the water. How did she get up onto the roof?
Steve began to slide down the mountain. How did he get to the top?

Comparison of comprehension

monitoring

AUGUST,FLAVELL,&

CLIFT

This content downloaded from 14.139.86.99 on Tue, 23 Jun 2015 22:12:29 UTC
All use subject to JSTOR Terms and Conditions

53

You might also like