EFFECTS OF PASSAGE TYPE ON COMPREHENSION

Journal of Reading Behavior
1992, Volume XXIV, No. 3
EFFECTS OF PASSAGE TYPE ON COMPREHENSION
MONITORING AND RECALL IN GOOD AND POOR READERS
Karen Zabrucky
Georgia State University
Hilary Horn Ratner
Wayne State University
ABSTRACT
An error detection paradigm was used to assess the effects of passage type on
students' comprehension monitoring and recall of texts. Good and poor readers in
the sixth grade (M age = 1 2 years, 0 months) read inconsistent expository and
narrative passages presented one sentence at a time. On- and off-line measures
(reading times and verbal reports of passage inconsistencies) were used to examine
students' comprehension evaluation, and look-backs to inconsistencies during
reading were used to measure comprehension regulation. Although all students
detected problems on-line, good readers were better able to verbally report on
passage consistency following reading. Passage type affected regulation of
understanding with students more likely to look back at inconsistencies in
narratives but not expository passages. Students were more likely to reread
expository passages, however, when passages did not contain problems and were
less able to recall expository passage information. Implications for instruction and
directions for future research are discussed.
During the last several years there has been a great deal of interest in the
development of students' comprehension monitoring. Comprehension monitoring
consists of two component skills, evaluation and regulation of understanding
(Baker, 1985). Evaluation involves assessment of understanding during reading,
whereas regulation involves the use of a strategy such as rereading to correct a
comprehension failure. Both skills are critical for successful text comprehension
and preparation for additional study.
Poor readers, in particular, are at risk for both evaluation and regulation failures (Zabrucky & Moore, 1989; Zabrucky & Ratner, 1989). However, it is not
373
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
374
Journal of Reading Behavior
clear whether evaluation and regulation skills are uniformly better for good than
poor readers, whether evaluation and regulation problems are equally likely for
poor readers, and whether evaluation and regulation performance is equivalent
across task situations for either good or poor readers. Most critically, it is important
to determine if task structure influences the frequency of poor readers' monitoring
failures.
In this study we examined evaluation and regulation skills of good and poor
readers for two types of passages, narrative and expository texts. To date, little, if
any, research has been conducted to examine the role of passage type in good and
poor readers' evaluation and regulation of understanding. Comparisons across these
passage types are of interest because narrative and expository texts vary in cohesiveness (Petros, Norgaard, Olson, & Tabor, 1989; Tun, 1989), which, in turn, affects
comprehension and memory. Thus, ease of evaluation and regulation of understanding may differ as a result of passage structure, especially for poor readers who
may need the support of better structured texts to integrate information and make
inferences (Bridge & Winograd, 1982).
Components of Comprehension Monitoring
Evaluation of understanding. In studies of comprehension monitoring, students' evaluation of understanding has most often been examined. Evaluation is
usually measured using an error detection paradigm. In this paradigm, a problem
of some sort is introduced into an otherwise intact passage. Failure to detect
the problem indicates a failure to evaluate comprehension. Many types of measures have been used to assess ability to detect passage problems, including
performance-based measures (e.g., underlining problems), on-line measures, and
verbal report responses following reading. When verbal report or performance
measures are used, good readers are found to evaluate understanding better than
poor readers (Zabrucky & Moore, 1989).
Although relatively rare in the literature, studies using on-line measures of
error detection, such as reading times, indicate that such measures are the most
sensitive to rudimentary evaluation skills in students, especially when performance
across age is compared. For example, Zabrucky and Ratner (1986) have shown
that although both younger and older students slow down while reading problematic
sentences, younger students are less able to verbally report the problem following
reading. This indicates that younger students' verbal reports of error detection
following reading underestimate problem identification. The on-line measure revealed quite clearly that problems are noticed during reading because reading time
increased only when the problem was encountered. Although on-line measures of
error detection have been used infrequently to assess the effects of reading proficiency on evaluation, the developmental findings suggest that reading times are
useful measures of these effects.
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
375
Another important consideration when examining students' evaluation is the
type of problem students are presented. Some problems are more difficult than
others to detect, and these differences can be used to gauge students' skills at using
various standards of evaluation (Baker, 1985). For example, students' detection of
nonsense words reflects their ability to use a lexical standard of evaluation during
reading, their detection of falsehoods reflects their use of an external consistency
standard, and their detection of inconsistencies reflects their use of a standard
internal to the passage itself. Lexical and external consistency problems are easier
for poor readers to detect than are internal consistency problems. Detection of
internal consistency problems requires more thorough and integrated processing of
text information than does detection of other problems, exacerbating differences
between good and poor readers (Zabrucky & Moore, 1989).
In the present study, we used on-line and verbal report measures to assess
good and poor readers' use of an internal consistency standard of evaluation. This
allowed us to more accurately assess error detection for poor readers and to determine if poor readers were sensitive to internal consistency problems.
Regulation of understanding. Although students' evaluation of understanding
has been explored most often, evaluation and regulation may not develop concurrently as a single skill. Students who do not evaluate understanding cannot regulate
it; however, evaluation may occur in the absence of regulation. Thus, some students
identified as being able to monitor their comprehension on the basis of evaluation
skills (such as good readers) may still not have developed successful strategies for
regulating comprehension. Cohesiveness of text information, and the nature of the
resulting memory representation, may influence regulation as well as evaluation;
however, it is not clear whether these factors will affect evaluation and regulation
similarly.
Thus far much of our knowledge of strategy use in students is indirect, frequently coming from students' self-reports (e.g., Garner & Kraus, 1981-1982).
However, students' self-reports of strategy use and their actual strategy use may
be unrelated (Tobias, 1987). Furthermore, self-reports do not reveal information
regarding strategy effectiveness (Spring, 1985). Thus, on-line techniques provide
not only a sensitive assessment of students' evaluation, but also a more direct
examination of students' regulation strategies during reading. In the present study,
we examined good and poor students' use of rereading to regulate their understanding.
Influence of Passage Type on Monitoring and Memory
Both evaluation and regulation may vary in response to text characteristics.
Just as type of problem influences error detection, so may type of text hinder or
facilitate the ease of evaluation, regulation, or both. Narrative and expository texts
provide a useful comparison because they vary in propositional coherence. Readers'
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
376
Journal of Reading Behavior
success in comprehending and integrating information across propositions in a text
depends in large part on the coherence of the information read. Although propositions in both narrative and expository texts can be integrated through referential
connections, narrative propositions are easier to integrate because of additional
causal connections (Black & Bern, 1981; Fletcher, Hummel, & Marsolek, 1990;
Perros et al., 1989; Tun, 1989). Propositions containing causal connections are
associated with faster reading times (Tun, 1989) and better memory among adults
(Black & Bern, 1981; Myers, O'Brien, Balota, & Toyofuku, 1984; Petros et al.,
1989; Tun, 1989). Researchers have suggested that causal connections may affect
both encoding and retrieval of text information (Myers, Shinjo, & Duffy, 1987;
Tun, 1989). Because of the increased ease of sentence integration in narratives, we
expected passage type to influence students' detection of inconsistent information in
texts and their need to reread passage information to regulate understanding.
Although investigators have used narrative and expository texts (Zabrucky &
Moore, 1989; Zabrucky & Ratner, 1986) in studies of students' comprehension
monitoring, the effects of passage type on students' monitoring skills has not been
examined directly. There have been several studies, however, in which memory
for narrative and expository texts has been compared in adult subjects. In these
studies, adults have been found to recall more information from narrative than
expository texts (e.g., Petros et al., 1989; Tun, 1989), supporting the notion that
narrative texts are more causally cohesive.
Although direct comparisons of students' memory for narrative and expository
texts are rare, we do know that students' comprehension and recall are affected by
passage structure. For example, students are sensitive to narrative structure and
recall more text information after reading well-formed (i.e., more causally connected) than poorly formed narrative texts (Brennan, Bridge, & Winograd, 1986).
Students are also better able to answer comprehension questions after reading narrative than expository texts (Goelman, 1982), perhaps because they experience
greater difficulty in discovering logical relations in expository texts than in narrative
texts (Geva & Ryan, 1985).
The Present Study
In the present study, we explored the role of passage type in good and poor
readers' ability to monitor their understanding and in their ability to recall text
information. Evaluation and regulation of understanding might be increased in
narrative texts because these texts allow easier integration of information. The
influence of passage type on comprehension monitoring might be more evident in
less proficient readers because they experience particular difficulty integrating textual information in expository passages (Bridge & Winograd, 1982).
We used on-line measures to examine the effects of passage type on comprehension monitoring in good and poor readers. To examine evaluation and regulation
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
377
of understanding, we used a computer program that allows sentence by sentence
presentation of texts and allows readers to move forwards or backwards in the text
easily (see Nason & Zabrucky, 1988). Students read a series of narrative and
expository passages, some of which contained a pair of inconsistent sentences.
Inconsistencies either appeared adjacent in texts, or nonadjacent, with four sentences interposed between sentences. When information was separated, students
were required to retrieve earlier presented information to detect a problem because
this information was no longer in working memory. The additional memory processing, a result of the distance between inconsistent sentences, may affect error
detection rates (Garner & Kraus, 1981-1982) and the likelihood of regulation activities. That is, evaluation may be more difficult and regulation more needed when
text information is less accessible.
To assess evaluation of understanding, we examined students' reading times
for problematic information and students' ability to verbally report passage problems following reading. To assess regulation of understanding during reading,
students' look-backs to problematic text information were examined. If passage
type affects students' ability to monitor their understanding, greater evidence of
evaluation and regulation of understanding for passage inconsistencies (specifically
more errors detected and more look-backs to inconsistent information) should be
found for narrative than for expository passages. However, if expository texts are
generally more difficult for students, then overall passage look-backs should be
higher for expository passages. If poor readers have particular difficulty processing
expository text information, reading ability differences in evaluation and regulation
should increase for expository passages. If on-line measures of evaluation are more
sensitive to evaluation skills in poor readers (as they are in younger students),
reading ability should affect the verbal report measure of evaluation but not the
on-line measure of evaluation, namely reading times of inconsistent information.
Students' overall reading times, look-backs, and memory for narrative and
expository texts were also assessed. Of particular interest was whether students
would adjust their reading speed and number of look-backs when processing expository texts and whether students would recall expository texts more poorly than
narrative texts. The structure of the expository passages should make comprehension and recall more difficult, necessitating greater regulation efforts.
METHOD
Subjects
Subjects were 32 sixth-grade students (M age= 12 years, 0 months), classified
as poor comprehenders (M percentile = 42, range = 29-51) or good comprehenders
(M percentile = 88, range = 76-99 + ) on the basis of scores on the reading compre-
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
378
Journal of Reading Behavior
hension scale of the Stanford Diagnostic Reading Test. For both good and poor
comprehenders, there were 8 males and 8 females in the sample. Students were
given the reading test and participated in the study at the end of the academic year.
Materials
Materials were eight narrative and eight expository passages written at a
fourth-grade level (Fry index). These passages were written at a grade appropriate
level for even our poorest readers, who read at approximately a 4.5 gradeequivalent level. All texts were 11 sentences long and the mean number of words
per text was 129 words for narrative and 112 words for expository passages.
Consistent with previous research, narrative and expository passages were distinguished by the number of causal connections between propositions. Causal connections in narrative texts (M = 6.9) were more than twice as frequent than in expository passages (M=3.0).
Each passage was written in four versions. Across versions, all sentences
except one, the context sentence, remained the same. In two versions the context
was Incongruent with a later target sentence. In remaining versions the context and
target were Congruent. The mean numbers of words and syllables (in parentheses)
in narrative and expository targets were 12.9 (17.0) and 12.8 (16.8), respectively.
The mean numbers of words and syllables in Congruent and Incongruent context
sentences were 12.6 (16.3) and 12.6 (16.1) for narrative texts and 12.4 (15.6) and
12.4 (15.5) for expository texts.
In two versions, the context and target were adjacent (Close), occurring in the
9th and 10th sentence positions in the text, respectively. In remaining versions they
were nonadjacent (Far), occurring in the 5th and 10th positions, respectively. When
the context and target were in the Far conditions, the interposing sentences changed
the topic, producing a breakdown in explicit referential coherence between context
and target information (see Lesgold, Roth, & Curtis, 1979; Zabrucky, 1986). Example passages appear in Appendix A.
The experimental passages were selected from a larger set of 20 passages
pretested to validate the experimenters' judgments of passage inconsistencies. The
20 passages were presented to 75 college students. For each student, each passage
was presented in either the Incongruent Close, Incongruent Far, or Congruent
Close version. Students were instructed to underline any inconsistent sentences they
detected. Passages for which inconsistencies were detected by at least 85% of the
sample were retained. Twelve passages met this criterion. Four of the remaining
eight passages were rewritten and presented in either the Incongruent Close or
Incongruent Far version to 30 more college students. Inconsistencies were detected
by at least 85% of the subjects in three of the four passages. The fourth passage
was rewritten once more and presented in either the Incongruent Close or Incongruent Far version to 26 additional college students. Over 85% of this sample identified
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
379
the inconsistency contained in the passage. The mean error detection rates for the
passages used in the study were 96.4% for narrative and 96.3% for expository
texts.
Procedure
Students were tested individually in two sessions, occurring one week apart.
In each session, students read either eight narrative or eight expository texts; the
order of presentation of narrative and expository passages was counterbalanced
across subjects. In each session, students received two passages in the Congruent
Close version, two in the Congruent Far version, two in the Incongruent Close
version, and two in the Incongruent Far version. Within each session, the order of
passages was presented randomly. Across subjects, passage version was counterbalanced across passages.
In the first session, we told students that we were interested in how students
their age understand stories. All passages were referred to as stories in the instructions to avoid alerting students that they would be reading different passage types.
Students were instructed that they would be reading eight stories and would come
back a week later to read eight more. They were asked to read each sentence in a
passage until they understood it "the best they could" and were told that they
could reread passage sentences if desired. Instructions encouraged reading for understanding and discouraged speed reading as well as memorization of passage
information.
Passages were presented one sentence at a time on a Macintosh SE computer.
The computer program we used was adapted from one developed by Nason and
Zabrucky (1988) for comprehension monitoring research. A short title preceded
each passage. Following the last sentence in each passage, a screen page indicated
that the passage had ended. The program was designed so that the screen appeared
as a page in a notebook. Students could read or reread sentences by simply clicking
the Macintosh mouse on a forward or backward arrow in the upper right hand
corner of the screen. When a forward or backward arrow was clicked, the notebook
page appeared to turn forward or backward, removing the current sentence and
displaying either the next sentence or the previous sentence.
When students finished reading a passage, they were asked to tell the experimenter all about the passage in their own words. Following passage recall, we
asked a Verbal Report Question to examine students' ability to verbally report text
inconsistencies about the passage ("Did everything in the passage make sense?").
If students responded " N o " to the Verbal Report Question, they were asked to
explain their answer. All recall and Verbal Report responses were recorded with a
portable cassette tape recorder.
Before reading the experimental passages in the first session, students were
shown how to use the Macintosh mouse to read and reread sentences and practiced
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
380
Journal of Reading Behavior
the reading and recall task with a practice passage. The instructions included a set
of questions designed to ensure that there were no misunderstandings concerning
the procedure. The practice passage, presented in the Congruent Close version,
was one of the passages used in the pretesting session that had not been selected
for use in the experiment.
Although our poor readers scored low on a standardized test of reading comprehension, we wanted to determine if they had decoding problems as well. Such
problems would seriously inflate the reading times of poor readers. This would be
particularly problematic in interpreting reading times of target sentences. At the
end of the second session, the 16 target sentences were presented, one at a time
and in random order, and students were asked to read them aloud. Oral decoding
was tape-recorded and scored for errors. Virtually no errors occurred for either good
or poor readers (AÍ numbers of decoding errors were 0.016 and 0.066, respectively).
Scoring
Reading times and look-backs. The computer program we used recorded reading times for individual sentences (to within sixtieths of seconds) as well as the
number of look-backs or rereadings for individual sentences. In coding look-backs
we needed to decide which reading times would be viewed as look-backs and which
would be viewed as passes back through the passage to locate other sentences. To
ensure that we did not exclude look-backs during which students had enough time
to pick up some of the words or information they were scanning, we chose 2.5
seconds (s) to discriminate look-backs from passes. This number was slightly lower
than the mean of students' minimum reading times (2.79 s) for all sentences in the
study.
Recall. Each of the 16 passages was broken down into a list of idea units,
expressing the main ideas of the passage. An idea unit was defined as a sequence
of words or a word (e.g., a modifier or connective) that conveyed a single idea.
This method of scoring recall protocols was thus propositionally based. The mean
numbers of idea units for narrative and expository passages, respectively, were
24.6 and 21.5 units. Because target sentences in Congruent and Incongruent passage versions were identical, so were the numbers of idea units. The mean numbers
of target sentence idea units were 3.4 for narrative and 3.8 for expository texts.
Context sentences were the only sentences that changed across passage versions.
In all passages but one, changes required to make sentences Congruent in some
passages and Incongruent in others did not result in differing numbers of idea units
(in one passage, an Incongruent context sentence had one fewer idea unit than in
the Congruent version). The mean numbers of idea units were 3.3 and 3.1 for
Congruent and Incongruent narrative contexts and 3.1 each for Congruent and
Incongruent expository contexts. Recall data were analyzed for context idea units
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
381
recalled, target idea units recalled, and total idea units recalled (minus context and
target units). For all recall analyses, the dependent measure was the proportion of
correct idea units.
Recall protocols were scored by two coders and disagreements between them
were resolved by discussion. Interrater agreement between the two coders was over
90% for a subsample of the passages scored (a set of passages from approximately
one-third of the sample).
Verbal report question. For the Verbal Report Question, a correct response
for Congruent passages was the answer yes; all other responses were considered
incorrect. For Incongruent passages, a correct response was the answer no and an
explanation that cited the incompatibility of the context and target; all other responses were considered incorrect.
RESULTS
Each of the dependent measures was analyzed i n a 2 x 2 x 2 x 2 (Reading
Ability X Passage Type X Congruence X Distance) mixed factorial ANOVA
unless otherwise noted. Only significant (p<.05) effects are reported, unless otherwise noted.
On-line Comprehension Monitoring Measures
Evaluation of comprehension: Reading times for target sentences. Reading
times were first examined for outliers. Two subjects had reading times over three
standard deviations above a condition mean and their data were eliminated from
the analysis. The analysis revealed a significant main effect for Reading Ability,
F(l, 28)= 10.45, p < . 0 1 , indicating that target sentence reading times were longer
for poor readers (M= 17.03s, SD = 3.46s) than for good readers (M= 13.45s,
SD = 261s). The main effect for Congruence was also significant, F(l, 28) = 7.02,
p<.02, indicating that reading times for Incongruent targets (M= 15.98s,
SD = 4.35s) were longer than for Congruent targets (M= 14.41s, SD = 3.17s).
Reading Ability did not interact with Congruence, revealing that both good and
poor readers slowed down while reading passage inconsistencies. Passage type did
not affect students' ability to detect errors on-line.
The analysis of target sentence reading times also revealed a significant but
uninteresting interaction between Reading Ability and Distance, F(l, 28) = 6.81,
p<.02. Reading times for target sentences were somewhat (although not significantly) longer in Far (M= 13.77s, SD = 2.62s) than Close (M= 13.12s, SD =
2.94s) conditions for good readers. For poor readers, reading times were somewhat
(although not significantly) longer for Close than Far conditions (Ms = 17.78s,
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
382
Journal of Reading Behavior
SD = 3.74s and 16.52s, SD = 3.61s, respectively). Poor readers did spend significantly more time reading target sentences than good readers for both Close and Far
passage versions.
The longer reading times for target sentences for poor readers were indicative
of longer reading times in general for these students. An analysis of overall passage
reading time (not including rereadings) revealed only a main effect for Reading
Ability,F(l, 29) = 14.45, p<.001. Poor readers (M = 164.02s; 5D = 27.91s) spent
more time reading passages than did good readers (Af = 130.25s; SD = 21.32s).
Regulation of comprehension: Look-backs. We analyzed the number of lookbacks students employed during reading to determine if students regulated comprehension by rereading problematic information. Information appearing contradictory
(i.e., target sentences) always occurred in the tenth sentence position; however,
the context information rendering target sentences contradictory could occur in
either the fifth or ninth sentence positions, and we analyzed look-backs for these
sentence positions, separately.
An analysis of look-backs to the target sentence (Sentence 10) revealed only
a main effect of Congruence, F ( l , 29) = 4.21, p<.05. There were more lookbacks to target sentences in Incongruent (A/=1.06; SD= 1.12) than Congruent
(M=0.87; SD = 0.98) passages.
A 2 x 2 X 2 (Reading Ability x Passage Type x Conguence) mixed
factorial ANOVA conducted on look-backs for sentence position 5 (in the Far
conditions) failed to reveal any significant effects, although there was a trend
(p<.10) for good readers to look back at sentences more than poor readers. There
were far more look-backs to sentence position 9 (in the Close conditions) than to
Sentence 5, and a similar analysis conducted on look-backs for sentence 9 revealed
several significant effects. The main effects of Congruence, F(l, 29) = 4.63,
p<.04, and Passage Type, F ( l , 29) = 4.28, p<.05, were significant as were the
Passage Type X Congruence, F ( l , 29) = 5.44, p<.03, and Reading Ability x
Passage Type interactions, F(l, 29) = 4.28,/> < .05.
The Passage Type x Congruence interaction is shown in the top of Table 1.
In expository texts, the number of look-backs in Congruent and Incongruent passages was identical. In narrative texts, there were significantly more look-backs in
Incongruent than Congruent passages. For Congruent passages, there were significantly more look-backs to expository than narrative passages. The number of lookbacks was not influenced by passage type for Incongruent passages.
The Reading Ability x Passage Type interaction is shown in the bottom of
Table 1. Poor readers did not differ in the number of look-backs in narratives and
expository texts. Good readers, however, had significantly more look-backs in
expository than in narrative texts. For narrative passages, there was no difference
in the number of look-backs for good and poor readers. For expository passages,
good readers had significantly more look-backs than poor readers.
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
383
Table 1
Mean Number (and Standard Deviation) of Students' Look-Backs to Close
Context Sentences as a Function of Passage Type
Passage Type
Congruent
Incongruent
Good Readers
Poor Readers
Narrative
Expository
0.53(1.02)
1.13(1.21)
1.22(1.52)
1.22(1.54)
' 0.84(0.83)
0.81(1.17)
1.59(1.55)
0.84(1.19)
An analysis was also conducted on total passage look-backs, which included
look-backs for all passage sentences. This analysis revealed a significant Passage
Type x Congruence interaction, F ( l , 29) = 4.42, p<.05, which is shown in Table
2. For expository texts, the number of total look-backs was similar for Congruent
and Incongruent passages. For narrative texts, there were significantly more lookbacks for Incongruent than Congruent passages. For Congruent passages, there
were significantly more look-backs to expository than narrative passages. For Incongruent passages, the number of look-backs was unaffected by passage type.
Verbal Report Question
An analysis of responses to the Verbal Report Question revealed a significant
main effect for Reading Ability, F(l, 30) = 9.22, p<.0l, indicating that good
readers gave more correct responses to the question {M- 1.34, SD = 0.30) than did
poor readers (M= 1.04, SD = 0.26). The main effect of Passage Type was marginally significant, F(l, 30) = 4.02, p<.054, revealing that there were somewhat more
correct responses for narratives (M=1.27, SD = 0.46) than for expository texts
( A / = l . l l , S D = 0.30).
Table 2
Mean Number (and Standard Deviation) of Students' Overall Look-Backs as a
Function of Passage Type
Passage Type
Congruent
Incongruent
Narrative
Expository
6.16(8.31)
7.95 (9.43)
9.53(11.15)
9.45 (10.84)
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
384
Journal of Reading Behavior
Recall of Context and Target Information
An analysis of context idea units recalled revealed a significant main effect for
Reading Ability, F ( l , 30) = 20.78, p<.001, indicating that good readers recalled
more idea units (M=0.54, SD = 0A9) than poor readers (M = 0.31, SD = 0.21).
The main effect of Passage Type, F(l, 30) = 47.79, p<.001, as well as the Passage
Type x Congruence interaction, F ( l , 30) = 4.50, p<.05, were also significant.
The significant interaction was due to the fact that although recall of context information was greater for narrative than expository texts for both Congruent (M = 0.53;
SD = 0.37 vs. M = 0.32; SD = 0.31) and Incongruent (A/=0.59; 5Z> = 0.33 vs.
M = 0.26; 5 0 = 0.31) information, the difference was somewhat larger when information was Incongruent. An analysis of target idea units recalled revealed a significant main effect for Reading Ability only, F ( l , 30) = 24.17, p<.001; good
readers recalled more idea units (M = 0.55; SD = 0.19) than did poor readers
(M=0.32;5D = 0.18).
Overall Passage Recall
An analysis of total passage recall also revealed a significant main effect
of Reading Ability, F(l, 30)= 15.88, p<.001, indicating that good readers recalled more idea units overall (M = 0.56; SD = 0.10) than poor readers (M =
0.43; SD = 0.16). The main effect of Passage Type was also significant,
F(l, 30) = 188.672, p<.00l. Recall was greater for narrative (Ai = 0.60; SD =
0.16) than for expository texts (M = 0.38; SD = 0.18) for all students.
DISCUSSION
The present study addressed two issues. One was whether on- and off-line
measures would reveal differences in good and poor comprehenders' evaluation
and regulation of understanding during reading. The second was whether passage
type would affect evaluation and regulation of understanding and memory in students.
Was there any evidence that poor readers were less likely than good readers
to monitor understanding during reading? Although poor readers were less able
than good readers to accurately comment on passage consistency following reading,
our on-line measure of evaluation revealed that poor readers detected inconsistencies during reading, just like good readers. This finding is noteworthy because
detecting passage inconsistencies requires students to use a complex internal consistency standard of evaluation (Baker, 1985), and indicates that poor readers have
some skills that rival those of good readers. The sensitivity of the on-line measure
is similar to that found by Zabrucky and Ratner (1986; 1989) for younger students
and students identified as lower ability readers, defined on the basis of a median
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
385
split. The finding indicates that on-line measures are appropriate for a wide variety
of groups.
We also found some evidence that poor readers were less likely to regulate
understanding when reading difficult to understand textual information. Although
good readers used a look-back strategy more often when reading context sentences
in expository than narrative texts, passage type had no effect on poor readers' use
of look-backs to these sentences. Thus, good readers did look back more often to
critical sentences, when the sentences were embedded in the more difficult expository texts. These data are consistent with the results of a study by Kletzien (1991)
who found that differences in strategy use between good and poor readers are more
likely to emerge for difficult texts.
Overall, however, large differences in the use of rereading strategies by good
and poor readers were not found here. Generally, all students reread problematic
text information caused by difficult text genre or the presence of passage inconsistencies. Despite poor readers' rereading attempts, however, their recall of text
information was hindered relative to that for good readers. Thus, poor readers' use
of regulation strategies was less effective in promoting better text memory. Additional research is needed to more fully explore the types of regulation strategies
used on-line by good and poor readers and to explore individual differences in
cognitive processes that may contribute to poor readers' text comprehension and
memory problems.
A primary focus of the study was whether passage type would affect students'
comprehension monitoring and recall of discourse information. Investigators interested in discourse comprehension and recall have focused their research on the
processes by which readers establish connections across text propositions. Indeed,
factors influencing the ease of sentence comprehension and integration have been
a principle thrust of theories of reading comprehension (Kintsch & van Dijk, 1978).
As discussed previously, causal connections across text propositions have been
found to facilitate encoding and retrieval skills (e.g., Myers et al., 1987) involved
in comprehension monitoring, and passages containing more of these links would
be expected to be processed more effectively.
There were several sources of information that suggested that passage type did
indeed affect students' comprehension monitoring. The most compelling evidence
comes from our on-line measure of regulation of understanding. Students' lookbacks to inconsistent information relative to control information increased in narratives but not in expository passages. Futhermore, students had higher overall lookback scores for narrative passages containing inconsistencies but not for expository
passages containing inconsistencies. As already noted, researchers (Black & Bern,
1981; Fletcher et al., 1990; Petros et al., 1989; Tun, 1989) have suggested that
increased causal connections in narrative passages may lead to easier integration
of sentences for these passages, and our results do suggest that this is the case for
students. Students seemed better able to focus their rereading attempts when reading
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
386
Journal of Reading Behavior
narrative passages, and the inconsistencies appeared to be more salient against a
background of sentences that may have been easier to integrate. It is also noteworthy that students generally reread problematic text information more often when
context and target sentences were adjacent and thus accessible in memory. Thus,
manipulations proposed by reading theorists to influence the accessibility of information in memory (Cirilo, 1981; Lesgold et al., 1979; Kintsch & van Dijk, 1978)
aided students' monitoring attempts.
It was less clear whether evaluation is sensitive to passage type. Although
students' reading times to inconsistent information did not differ as a function of
passage type, students were better able to verbally report on passage consistency
after reading narrative than expository texts. It should be noted, however, that this
effect was only marginally significant.
Although expository passages have been linked to slower speeds of reading
for adults (Tun, 1989), we found no difference in the amount of time students
spent reading expository relative to narrative texts when students' initial reading
times were examined. When passages did not contain an intentionally embedded
problem, however, students generally reread expository passages more frequently
than narrative passages. Indeed, in congruent conditions the number of rereadings
in expository passages was strikingly similar to that for narrative passages containing problems. Students' rereading attempts for expository passages revealed that
they found these passages problematic and adjusted their look-backs accordingly.
Although students adjusted their look-backs in response to more difficult textual
information, this increased regulation activity did not prevent expository texts from
being more poorly recalled. Thus, rereading in the present study was related to
students' comprehension difficulties, but rereading did not completely resolve their
comprehension problems.
Limitations of the Present Study and Suggestions for Future Research
In the present study we used on-line methods to examine evaluation and regulation in good and poor readers. Given the sensitivity of on-line measures (Zabrucky
& Rattier, 1986; 1989) and problems with students' verbal reports (Winograd &
Johnston, 1982; Zabrucky & Moore, 1989), these measures provide informative
and unique data not readily available from students' reports, including actual strategy use and effectiveness of stategy use in readers. As for any experimental paradigm, however, these procedures may have influenced students' performance and
our results. First, computer presentation of the passages may have affected comprehension or memory processes. Students may have been more likely to notice errors
inserted into the passages because passages were presented sentence by sentence.
Alternatively, because of the computer actions required for page movement, students' reading and rereading may have been slowed. Slower access to information
may have required greater memory processing and hindered students' attempts to
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
387
integrate sentences. It will be important in future research to directly compare the
effects of stimulus presentation on students' monitoring and recall of text information.
Second, the error detection paradigm may have changed students' reading
patterns. Although the on-line measures we used circumvented many of the problems first noted by Winograd and Johnston (1982) regarding the error detection
paradigm, other limitations remained. For example, a detection task undoubtedly
alerted students to the possibility that texts contained problems and as a result
students may have been more likely to read for detection. Additionally, the problems inserted into the present texts were fairly obvious. Students may have greater
difficulty detecting comprehension failures when presented more subtle text problems or when reading unaltered texts. There is a growing body of research on adult
readers' ability to calibrate comprehension using unaltered texts (e.g., Glenberg &
Epstein, 1985; Weaver, 1990). A similar approach with students would be informative and would allow useful comparisons of findings across two different paradigms
for examining self-assessment of comprehension.
In the present study, narrative passages contained more causal connections
than expository passages. It should be noted, however, that narrative and expository
texts may vary along other dimensions not systematically examined here. For example, narrative texts may be more interesting than expository texts because they tell
a story and involve characters. Future research is needed to further delineate the
critical variables between passage types and explore the role of these variables in
students' text comprehension and memory.
We also used passages that were written at a grade appropriate level for even
the poorest readers in our study. Although this was necessary to compare performance across the two groups of readers on identical texts, the passages we used
were no doubt easier for poor readers than those they typically encounter and easier
still for good readers. For this reason, the performance of poor readers found here
may be more similar to that of good readers than had more difficult materials been
used. In future research, it will be important to examine good and poor readers'
performance with text materials that vary in difficulty level.
Different approaches to assessing regulation are also needed. Although rereading has been shown to increase comprehension of text information (Alessi, Anderson, & Goetz, 1979; Bernhardt, 1983; Flood, 1986) and may be necessary for
developing well-integrated memory representations in naturally occurring texts,
which are often inconsiderate (Armbruster, 1984), it is only one of several strategies that students may employ to comprehend texts. In future studies, students'
use of other strategies important for sentence integration such as inferencing and
reinstatement of previously presented information into working memory should be
explored.
Finally, although we examined the effects of students' reading proficiency and
passage type on monitoring, there are other variables that may affect students'
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
388
Journal of Reading Behavior
skills. For example, students' prior knowledge, their interest level, and their motivation have all been linked to their reading performance (Taylor, Harris, & Pearson,
1988) and may have as much or more impact on monitoring as the variables
investigated here. There has been little attention paid in the comprehension monitoring literature to these topics, and future research is necessary to explore the effects
of these variables on students' evaluation and regulation. For example, students
may have developed expectations for certain types of texts, independent of the
content or structure of a particular passage, which they bring to reading tasks. In
our tasks, for instance, students may have reread expository passages more because
they had come to understand that texts like these are typically more difficult.
Particular content knowledge may also influence performance. Prior topic knowledge may improve students' error detection, even when passage cohesion is poor,
and promote more selective rereading when comprehension fails. Similarly, high
interest and motivation may enhance not only students' evaluation of understanding
but also their persistence in regulating comprehension.
Conclusions and Implications for Classroom Practice
Passage type affected students' evaluation and regulation of understanding as
well as their memory of discourse information. Generally, factors that promote
easier integration of information such as text genre (Fletcher et al., 1990; Petras
et al., 1989; Tun, 1989) or distance of propositions to be integrated (Cirilo, 1978;
Lesgold et al., 1979; Kintsch & van Dijk, 1978) affect students' monitoring attempts. Although students frequently reread sentences in expository passages, they
were still less able to integrate expository propositions across sentences and form
coherent memory representations for this type of text. These results are disturbing,
given that exposure to narratives declines in the early school years and exposure
to expository texts increases (Geva & Ryan, 1985). Failure to adjust reading activity
to the extent needed for accurate recall may reflect a very specific type of metacognition problem for both good and poor readers.
Several investigators have found a positive relation between students' awareness of expository text structure and comprehension and recall of text information
(see Armbruster, Anderson, & Ostertag, 1987). There have been fewer investigations, however, which have focused on students' expectations when reading narrative and expository texts, their knowledge of the differences between passage types,
or their use of different strategies in response to the varying demands that passage
types may place on comprehension and memory processes. Such research would
have direct implications for the classroom. Results from the present study suggest
that teachers should not only be concerned with increasing students' awareness of
different text structures but also with informing students of the impact these structures have on evaluation, regulation, and memory. Students need to know how to
adjust their reading and rereading skills and to assess readiness for recall and
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
389
test-taking when text information varies in difficulty. Generalizing metacognitive
skills to the more difficult expository passages after training with narrative passages
may be particularly effective.
REFERENCES
Alessi, S. M., Anderson, T. H., & Goetz, E. T. (1979). An investigation of lookbacks during studying.
Discourse Processes, 2, 197-212.
Armbruster, B. B. (1984). The problem of "inconsiderate text." In G. G. Duffy, L. A. Roehler, & J.
Mason (Eds.), Comprehension instruction: Perspectives and suggestions (pp. 202-217). New
York: Longman.
Armbruster, B. B., Anderson, T. H., & Ostertag, J. (1987). Does text structure/summarization instruction facilitate learning from expository text? Reading Research Quarterly, 22, 331-346.
Baker, L. (1985). How do we know when we don't understand? Standards for evaluating comprehension. In D. L. Forrest, G. E. MacKinnon, & T. G. Waller (Eds.), Metacognition, cognition, and
human performance (pp. 155-205). New York: Academic Press.
Bernhardt, E. B. (1983). Three approaches to reading comprehension in intermediate German. Modern
Language Journal, 67, 111-115.
Black, J. B., & Bern, H. (1981). Causal coherence and memory for events in narratives. Journal of
Verbal Learning and Verbal Behavior, 20, 267-275.
Brennan, A. D., Bridge, C. A., & Winograd, P. N. (1986). The effects of structural variation on
children's recall of basal reader stories. Reading Research Quarterly, 21, 91-103.
Bridge, C. A., & Winograd, P. N. (1982). Reader's awareness of cohesive relationships during cloze
comprehension. Journal of Reading Behavior, 14, 299-312.
Cirilo, R. K. (1981). Referential coherence and text structure in story comprehension. Journal of Verbal
Learning and Verbal Behavior, 20, 358-367.
Fletcher, C. R., Hummel, J. E., & Marsolek, C. J. (1990). Causality and the allocation of attention
during comprehension. Journal of Experimental Psychology: Learning, Memory, and Cognition,
16, 233-240.
Flood, J. (1986). The text, the student, and the teacher: Learning from exposition in middle schools.
The Reading Teacher, 39, 784-691.
Garner, R., & Kraus, C. (1981-1982). Good and poor comprehender differences in knowing and
regulating reading behavior. Educational Research Quarterly, 6, 5-12.
Geva, E., & Ryan, E. B. (1985). Use of conjunctions in expository texts by skilled and less skilled
readers. Journal of Reading Behavior, 27, 331-346.
Glenberg, A. M., & Espstein, W. (1985). Calibration of comprehension. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 11, 702-718.
Goelman, H. (1982). Selective attention in language comprehension: Children's processing of expository
and narrative discourse. Discourse Processes, 5, 53-72.
Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production.
Psychological Review, 85, 363-394.
Kletzien, S. B. (1991). Strategy used by good and poor comprehenders reading expository text of
differing levels. Reading Research Quarterly, 26, 67-85.
Lesgold, A. M., Roth, S. F., & Curtis, M. E. (1979). Foregrounding effects in discourse comprehension. Journal of Verbal Learning and Verbal Behavior, 18, 291-308.
Myers, J. L., O'Brien, E. J., Balota, D. A., & Toyofuku, M. (1984). Memory search without interference: The role of integration. Cognitive Psychology, 16, 217-242.
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
390
Journal of Reading Behavior
Myers, J. L., Shinjo, M., & Duffy, S. A. (1987). Degree of causal relatedness and memory. Journal
of Memory and Language, 26, 453-465.
Nason, S., & Zabrucky, K. (1988). A program for comprehension monitoring of text using HyperCard
for the Macintosh. Behavior Research Methods, Instruments, & Computers, 20, 499-502.
Petros, T. V., Norgaard, L., Olson, K., & Tabor, L. (1989). Effects of text genre and verbal ability
on adult age differences in sensitivity to text structure. Psychology and Aging, 4, 247-250.
Spring, C. (1985). Comprehension and study strategies reported by university freshmen who are good
and poor readers. Instructional Science, 14, 157-167.
Taylor, B., Harris, L. A., & Pearson, P. D. (1988). Reading difficulties: Instruction and assessment.
New York: Random House.
Tobias, S. (1987). Mandatory text review and interaction with student characteristics. Journal of Educational Psychology, 79, 154-161.
Tun, P. A. (1989). Age differences in processing expository and narrative text. Journal of Gerontology:
Psychological Sciences, 44, 9-15.
Weaver, C. A. (1990). Constraining factors in calibration of comprehension. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 16, 214-222.
Winograd, P., & Johnston, P. (1982). Comprehension monitoring and the error detection paradigm.
Journal of Reading Behavior, 14, 61—76.
Zabrucky, K. (1986). The role of factual coherence in discourse comprehension. Discourse Processes,
9, 197-220.
Zabrucky, K., & Moore, D. (1989). Children's ability to use three standards to evaluate their comprehension of text. Reading Research Quarterly, 24, 336-352.
Zabrucky, K., & Ratner, H. H. (1986). Children's comprehension monitoring and recall of inconsistent
stories. Child Development, 57, 1401-1418.
Zabrucky, K., & Ratner, H. H. (1989). Effects of reading ability on children's comprehension evaluation and regulation. Journal of Reading Behavior, 21, 69-83.
APPENDIX A
Example Narrative and Expository Texts
Narrative Text: The Family Trip
Cindy and her family were driving to Boston to see her aunt.
It seemed like days to her since they had stopped to eat.
She was hungry and asked her mother if they could stop for her favorite lunch.
Her mom saw a sign for a restaurant and said they could.
(Far Context Position)
Just then a beautiful deer came out of the woods.
She had never seen a deer so close except at the zoo.
A big truck honked loudly and the deer ran back into the woods.
Cindy and her family were happy the deer was safe.
(Close Context Position)
Target: After Cindv had eaten the burger for lunch, she felt much better.
She wondered what time they would get to Boston.
CONGRUENT CONTEXT: The restaurant was open so Cindy could order a juicy burger.
INCONGRUENT CONTEXT: The restaurant was closed so Cindy couldn't order a juicy
burger.
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016
Comprehension Monitoring
391
Expository Text: Early Settlers
Long ago people shipped goods across the country in many ways.
One way to carry their goods was to use rivers.
The rivers also were used to bring crops to market.
Big rivers were busy with people going west and south.
(Far Context Position)
People also used roads to carry their goods.
One of the important roads was the National Road.
It was the first road built by our government.
Roads were soon used more than rivers.
(Close Context Position)
Target: Flatboats were so common on rivers, they were quite familiar to people.
Using rivers and roads to carry goods helped the country grow.
CONGRUENT CONTEXT: The most common vehicles on the busy rivers were slow flatboats.
INCONGRUENT CONTEXT: The least common vehicles on the busy rivers were slow
flatboats.
(Note: Congruent and Incongruent contexts appear below passages. Close and Far context
positions are given in parentheses within texts; the target sentence is underlined.)
Downloaded from jlr.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016