Memory & Cognition

2007, 35 (7), 1567-1577

The effects of prior knowledge and text structure

on comprehension processes during
reading of scientific texts
Panayiota Kendeou
McGill University, Montreal, Quebec, Canada
and

Paul van den Broek

University of Minnesota, Minneapolis, Minnesota

The aim of the present study was to investigate the effects of prior knowledge and text structure on cogni-
tive processes during comprehension of scientific texts. To investigate the processes online, we used a think-
aloud methodology in Experiment 1 and a reading time methodology in Experiment 2. In both experiments,
we obtained offline comprehension measures and measures of individual differences in working memory and
need for cognition. Across the two experiments, the results indicated that readers adjust their processing as a
function of the interaction between prior knowledge and text structure. In particular, adjustments in the actual
processes that take place during reading were observed for readers who had erroneous prior knowledge, but
only when they read a text that was structured to explicitly refute this prior knowledge. Furthermore, the results
showed that readers’ memory for the text was affected by differences in their prior knowledge, independently
of text structure. These findings contribute to our understanding of the relation between different factors that
are associated with comprehension of scientific texts and have implications for theories of comprehension and
conceptual change.

Many factors affect success or failure during compre- focus on readers’ quality of knowledge. Quality refers to
hension of expository texts. Reader characteristics, text the accuracy of one’s knowledge and has been investigated
properties, and the instructional context in which reading mostly with respect to students’ inaccurate ideas in sci-
takes place are just a few of them (Goldman & Bisanz, ence (Kendeou, Rapp, & van den Broek, 2004; Kendeou
2002; Jenkins, 1979; van den Broek, Fletcher, & Risden, & van den Broek, 2005). Inaccurate ideas, or misconcep-
1993; van den Broek & Kremer, 1999). Although these tions, have been found to interfere with the acquisition of
factors have often been studied in isolation, it is their in- new, related knowledge from texts (Alvermann, Smith,
teractions and interdependencies that provide important & Readence, 1985; Diakidoy & Kendeou, 2001; Lipson,
information about naturalistic text comprehension (Rapp 1982; Maria & MacGinitie, 1987; Peeck, van den Bosch,
& van den Broek, 2005). In the present study, we focus on & Kreupeling, 1982). The evidence for interference with
two factors: reader characteristics and text properties. The learning comes from offline studies—that is, studies in
goal is to investigate the possible interactions between which the end product of reading is assessed by tasks such
readers’ prior knowledge and the structure of the text in as recall or question answering, but the presumption is
the comprehension processes that occur during reading of that the interference occurs during the reading process it-
scientific texts. self. How the processes during reading are influenced by
There is ample evidence that the quantity or amount of misconceptions has not been studied directly, however,
readers’ prior knowledge influences comprehension of sci- and therefore the details of how misconceptions affect
entific texts (see, e.g., Chi, 1978; Chi, Feltovich, & Glaser, reading are not known.
1981). Indeed, there is a well-documented advantage in A second factor that influences comprehension of in-
comprehension of texts for readers with high knowledge formation in scientific texts is the structure of the text.
on the topics described by the texts over readers with low There are different ways for an author to organize the
knowledge (e.g., Bartlett, 1932; Chiesi, Spilich, & Voss, ideas in texts in general or in scientific texts in particular,
1979; Dochy, Segers, & Buehl, 1999; Means & Voss, resulting in different structures (Graesser, Leon, & Otero,
1985; Recht & Leslie, 1988). In the present study, we 2002). Meyer and colleagues (Meyer, 1975, 1999; Meyer

P. Kendeou, kend0040@umn.edu

1567 Copyright 2007 Psychonomic Society, Inc.

1568     Kendeou and van den Broek

& Freedle, 1984) demonstrated that readers’ comprehen- Bennett, & Rice, 1996). Thus, it is likely that readers’ need
sion is influenced by the text structure used to convey for cognition plays a role when they need to resolve incon-
the information. In the present study, we focus on a text sistencies between their misconceptions and information
structure that is often used in organizing the material in presented in the text during comprehension (Kardash &
scientific texts, refutation texts. Refutation texts are used Scholes, 1996).
primarily to persuade students to change prior beliefs by To address these issues, we conducted two experiments.
explicitly identifying misconceptions and explaining the In Experiment 1, we used a think-aloud methodology to
correct ideas (Alvermann & Hague, 1989; Chambliss, investigate cognitive processes online. This methodology
2002; Guzzetti, Snyder, Glass, & Gamas, 1993). Such allows the consideration of a variety of readers’ responses
texts have been tested in science education literature and (Ericsson & Simon, 1993; Pressley & Afflerbach, 1995;
have been found to be effective in helping students change Trabasso & Suh, 1993) and has received extensive valida-
their misconceptions and engage in conceptual change tion as a tool to reveal comprehension processes in read-
(Alvermann & Hynd, 1989; Diakidoy, Kendeou, & Ioan- ing (Afflerbach, 2002; Coté & Goldman, 1999; Magliano
nides, 2003; Maria & MacGinitie, 1987). & Graesser, 1991; Magliano & Millis, 2003; Magliano,
It is likely that in most instances, prior knowledge and Trabasso, & Graesser, 1999; Zwaan & Brown, 1996).
text structure interact in their effects on readers’ com- In Experiment 2, we used a reading time methodology
prehension processes. Indeed, there is evidence that the to investigate the cognitive processes online, because it
interaction between readers’ prior knowledge and text is unobtrusive and reflects both strategic and automatic
characteristics influences comprehension once reading is processes. Thus, the methods used in the two experi-
completed. For example, using offline tasks such as recall ments complemented each other and provided the basis
and inferential and problem-solving questions, McNamara for acquiring convergent evidence for readers’ cognitive
and colleagues have demonstrated that high-knowledge processing. In both experiments, we obtained offline com-
readers’ comprehension was better for low- than for high- prehension measures and measures of readers’ individual
cohesion texts, whereas low-knowledge readers’ compre- differences in prior knowledge, working memory, and
hension was better for high- than for low-cohesion texts need for cognition.
(McNamara, 2001; McNamara & Kintsch, 1996; McNa-
mara, Kintsch, Songer, & Kintsch, 1996). The findings Experiment 1
of these studies suggest that the interaction between prior Think-Aloud Methodology
knowledge and text characteristics affects comprehen-
sion, at least when it is measured after reading has been The objective of Experiment 1 was to investigate the
completed. The implicit assumption is, however, that the effects of prior knowledge and text structure online, using
interaction between prior knowledge and text character- a think-aloud methodology. In this experiment, partici-
istics affects the offline product of reading by affecting pants with and without misconceptions in physics were
the online processes that take place during reading (e.g., asked to read a refutation and a nonrefutation scientific
Goldman & Varma, 1995; Kintsch, 1988; Trabasso & Suh, text while performing a think-aloud task. Participants
1993; van den Broek, Rapp, & Kendeou, 2005; van den were also asked to recall the texts they read and to com-
Broek, Young, Tzeng, & Linderholm, 1999; Zwaan & plete measures of working memory capacity and need for
Singer, 2003). In the present study, we directly investigate cognition.
this assumption. Given the evidence from offline tasks that refutation
The aim of our study is to investigate the effects of texts are particularly effective in bringing about learning
readers’ prior knowledge, as well as text structure, and in readers with misconceptions (e.g., Guzzetti et al., 1993),
the possible interaction of the two, online—that is, during we expected that the interaction between readers’ prior
comprehension of text. We focus particularly on cogni- knowledge and text structure would affect readers’ online
tive processing of refutation and nonrefutation scientific comprehension processes. First, we expected that readers
texts by readers with and without misconceptions related with misconceptions would engage in conceptual change
to the topics of the text. We also take into account other processes during reading of the refutation text structure
individual differences, such as working memory capacity more so than during reading of nonrefutation text, and
(Baddeley & Hitch, 1974; Budd, Whitney, & Turley, 1995) more so than their counterparts without misconceptions
and need for cognition (Cacioppo & Petty, 1982), because for either text structure. Second, we expected that readers’
these differences are commonly considered important in prior knowledge would affect their online comprehension
many reading situations. For instance, reading a text while processes, particularly the content of those processes that
having misconceptions may result in interference, because involve prior knowledge. Thus, we expected that read-
readers need to reconcile the contradiction between what ers with misconceptions would generate more incorrect
they already know and what is presented in the text. In knowledge-based inferences than would readers without
tasks that involve interference and the need to reconcile misconceptions, whereas readers without misconceptions
different sources of information, working memory is would generate more correct knowledge-based inferences
likely to play a role (Engle, 2002; Engle & Conway, 1998). than would readers with misconceptions. A third predic-
Also, enjoyment of thinking and engagement—the two tion pertains to the connection between online reading pro-
main characteristics of need for cognition—are strongly cesses and offline reading outcome. Given that the online
related to reading comprehension (Guthrie, McGough, processes form the basis for the construction of a memory
 Scientific Texts     1569

representation (Goldman &Varma, 1995; Kintsch, 1988; refutation format. The nonrefutation text was a shorter version of the
Langston & Trabasso, 1999; Myers & O’Brien, 1998; van original textbook section on Newton’s laws of motion. It included an
den Broek et al., 1999; Zwaan, 1999), we predicted that introduction, a definition of each law of motion, and several every-
day examples and explanations related to each law. The other version
the patterns of online processes that we obtained would
followed a refutation format. The refutation text was also a shorter
be reflected in the patterns of results for the offline mem- version of the original textbook section on Newton’s laws of motion
ory task. That is, we expected that higher order cognitive and included an introduction, a definition of each law, a few every-
processing (e.g., inference making) would result in better day examples and explanations related to each law, and refutations
comprehension (e.g., higher recall). of certain misconceptions related to each law. Seven experts in sci-
ence (four faculty members in science education and three graduate
students in chemical engineering) reviewed and provided feedback
Method for improving the initial versions of all four texts. The texts were
Participants finalized after accommodating the experts’ suggestions. Texts were
A total of 86 University of Minnesota undergraduates enrolled equated in word length (approximately 550 words each) and con-
in introductory psychology or physics courses participated in this ventional readability indices. The average Flesch–Kincaid reading
study. Students received extra course credit or a gift certificate for grade level was 8.3.
their participation. The data of 6 students were eliminated from the
study because of equipment failure (2 students), or because they Procedure
were nonnative English speakers (4 students). Of the remaining par- The participants were tested individually in a single session last-
ticipants, 41 were female and 39 were male, with an age range of ing approximately 1 h 40 min. The participants were first adminis-
18–48 years. tered the Force Concept Inventory. They were asked to read each
question carefully and to select the correct answer to the best of
Materials their knowledge. Next, the participants were informed that they were
Identification of misconceptions. To diagnose readers’ mis- going to read some texts and think aloud after every sentence (see
conceptions in Newtonian mechanics, we used the Force Concept Ericsson & Simon, 1993, for a detailed description of the think-aloud
Inventory (Hestenes, Wells, & Swackhamer, 1992). The inventory methodology). The experimenter explained that thinking aloud was
includes a total of 30 multiple-choice questions in six categories: a process during which one is asked to state aloud his/her thoughts
kinematics, Newton’s first law, Newton’s second law, Newton’s third while reading a text. The participants were given the practice text
law, superposition, and kinds of force. Each question has one cor- and were asked to read each sentence on the card aloud and talk
rect answer, with the remainder of the answers corresponding to about their thoughts. They were also instructed to make sure they
commonsense misconceptions of Newtonian concepts. The par- understood what they were reading. If the participant had difficulty
ticipants were divided into two groups, misconception and nonmis- talking aloud, the experimenter posed the following question: “What
conception, following the procedures recommended by Hestenes are you thinking right now?” When the practice text was completed,
and colleagues (Hestenes & Halloun, 1995; Hestenes et al., 1992). the participants were administered the first experimental text, and
The misconception group was made up of participants who scored they were asked to read and think aloud after every sentence. They
60% or less on the inventory (30 females, 10 males). Participants were also asked to make sure they understood what they were read-
in this range have been described as holding incoherent ideas and ing, because they would later be asked to recall the text. When the
misconceptions related to Newtonian mechanics (Hestenes et al., participants finished reading and thinking aloud, they were asked
1992). The nonmisconception group was made up of participants to complete a distractor task consisting of 10 math problems. After
who scored above 60% on the inventory (11 females, 29 males). In the distractor task, the participants were given a recall sheet and
prior research (Hestenes et al., 1992), participants in this range have were asked to write down everything they could remember from
been determined to hold coherent ideas, including the concepts of the text they had read. The same procedure was followed with a
velocity, acceleration, and force. second experimental text. The participants were administered one
Working memory capacity. To assess readers’ working memory nonrefutation and one refutation text, in random order. Next, the
capacity, we used a reading span test (Singer, Andrusiak, Reisdorf, participants completed the reading span test on a computer. Finally,
& Black, 1992). In this test, participants are asked to read sets of the participants completed the Need for Cognition Scale.
unrelated sentences one at a time on the computer. The sets begin in
groupings of two sentences and increase to five sentences per set. Scoring
Each sentence in a set appears for 7 sec on the screen. After reading Prior knowledge tests. The Force Concept Inventory was scored
each set, participants are prompted by a chime to write down the dichotomously. Each correct response received a score of 1, whereas
last word of each sentence they read. Finally, they complete a cloze each incorrect response received a score of 0. The observed total
task by writing two missing words from one of the sentences. The scores ranged from 3 to 30 (possible range 0–30).
cloze task is intended to prevent readers from adopting a strategy of Think-aloud protocols. Students’ responses during the think-
focusing on and memorizing only the final words. The total number aloud procedure were transcribed. Two raters independently parsed
of final words recalled correctly constitutes the reader’s working and coded 25% of the protocols into clauses, and their agreement was
memory capacity score. The maximum score a participant could reliably high for both parsing (K 5 .90, p , .01) and coding (K 5
achieve was 46. .87, p , .01). Disagreements were resolved through discussion, and
Need for Cognition Scale. To assess readers’ enjoyment of the remaining protocols were divided among the two raters for them
thinking and engagement, we used the Need for Cognition Scale to parse and code them independently. Each clause was categorized
(Cacioppo & Petty, 1982). The instrument consists of 18 statements, on the basis of a coding scheme adapted from previous research
for which students rate the degree to which each statement char- (Kendeou & van den Broek, 2005; Linderholm & van den Broek,
acterizes them (using a 1–5 Likert scale, ranging from extremely 2002; Pritchard, 1990). The coding scheme identified the cognitive
uncharacteristic to extremely characteristic). For instance, students processes in which readers engaged during reading. It consisted of
are asked to rate the degree to which they “prefer complex to simple the following eight categories (example responses are included in
problems” or they “find satisfaction in deliberating for long hours.” parentheses). Comprehension monitoring included responses that
Texts. Four texts were used that focused on two topics: Newton’s showed (1) readers’ awareness of how much progress they were
first law of motion and Newton’s third law of motion. The texts were making, (2) any problems readers encountered, and (3) whether the
adapted from a college-level physics textbook (Hewitt, 2002). We readers understood or failed to understand (e.g., “This makes sense
constructed two versions for each topic. One version followed a non- regarding inertia”). Associations included responses that showed
1570     Kendeou and van den Broek

that readers activated prior knowledge related or unrelated to the text were used as covariates in all subsequent analyses. Also,
material (e.g., “I have seen this done before, where the tablecloth is because we had two texts, with different topics for each
removed quickly and the dishes remain on the table”). Intrasentential structure, we included in the analysis a variable reflecting
connections included readers’ attempts to develop an understanding
the topic of the text.
of the current sentence by rereading or paraphrasing it (e.g., “All
objects continue in a state of rest unless acted on by another force”).
Intersentential connections included readers’ attempts to relate the Online Comprehension Processes
current text to other sentences in the text (e.g., “Continues is also an We conducted separate repeated measures ANOVAs,
important word, in addition to change”). Correct inferences included with prior knowledge (misconception, nonmisconception),
correct explanatory, forward, and other inferences that were based text structure (refutation, nonrefutation), and text topic
on readers’ prior knowledge and text (e.g., “So, if there is no change
in forces, there will be no change in motion”). Incorrect inferences (Newton’s first law, Newton’s third law) as independent
included incorrect explanatory, forward, and other inferences that variables and frequency for each of the seven think-aloud
were based on readers’ prior knowledge and the text (e.g., “So, iner- response categories (in the two texts and across partici-
tia is the force that changes an object’s motion”). Conceptual change pants) as a dependent variable. Reading span and Need for
strategies included responses that showed that readers were engag- Cognition variables were entered as covariates. The means
ing in conceptual change, such as experiencing cognitive conflict, for each response category are provided in Table 1.
responding to conflict, and contrasting information (e.g., “Greater
mass certainly leads me to believe that the forces are more forceful
There were significant differences as a function of
in my mind, but apparently that is not true”). Responses that did not prior knowledge with respect to correct [F(1,71) 5 13,
fall into any of the categories above were categorized as other (e.g., p 5 .001, h2 5 .15] and incorrect [F(1,71) 5 35.5, p 5
“We will see next”). .0001, h2 5 .33] inferences. In particular, the misconcep-
Recall protocols. Students’ written recall protocols were parsed tion group generated significantly fewer correct infer-
into clauses. Each clause was matched to the text sentences accord- ences (M 5 7.5) than did the nonmisconception group
ing to a gist criterion. Two raters independently parsed and coded
(M 5 13.9) [t(71) 5 3.6, p , .001] and significantly more
25% of the recall protocols, and their agreement was reliably high,
for both parsing (K 5 .95, p , .01) and coding (K 5 .90, p , .01). incorrect inferences (M 5 3.8) than did the nonmiscon-
Disagreements were resolved through discussion, and the remaining ception group (M 5 0.5) [t(71) 5 5.9, p , .001]. There
protocols were divided among the two raters for them to parse and were also significant prior knowledge differences with
code independently. regard to conceptual change strategies [F(1,71) 5 50.1,
p 5 .0001, h2 5 .41]. The misconception group generated
Results significantly more conceptual change strategies (M 5
We performed preliminary analyses to explore whether 3.5) than did the nonmisconception group (M 5 0.2)
the misconception and nonmisconception groups dif- [t(71) 5 7.1, p , .001]. This main effect was qualified by
fered with regard to the individual difference measures. an interaction between prior knowledge and text structure
The results show that there was no significant difference [F(1,71) 5 50.8, p 5 .0001, h2 5 .42]. Post hoc compari-
between the reading span scores for the misconception sons showed that the misconception group included more
(M 5 38.3, SD 5 4.9) and nonmisconception (M 5 37.5, conceptual change strategies than did the nonmisconcep-
SD 5 6.7) groups [F(1,78) , 1, p 5 .5]. There was a sig- tion group, but only during their reading of the refutation
nificant difference, however, in the Need for Cognition text [t(71) 5 4.5, p , .001].
score, with the misconception group scoring significantly Finally, there were no significant prior knowledge dif-
lower (M 5 65.8, SD 5 10.1) than the nonmisconception ferences with respect to the remaining four types of pro-
group (M 5 72.9, SD 5 9.4) [F(1,78) 5 8.9, p 5 .004, cesses: comprehension monitoring [F(1,71) , 1, p 5 .5],
h2 5 .10]. To eliminate the possibility of a confound due associations [F(1,71) 5 2.1, p 5 .1], intrasentential con-
to individual differences in the reading span and Need for nections [F(1,71) , 1, p 5 .5], and intersentential con-
Cognition scores, individuals’ scores on these measures nections [F(1,71) , 1, p 5 .8]. No other main effects or

Table 1
Mean Frequencies and Standard Deviations of the Types of Processes During the
Think-Aloud Phase, As a Function of Prior Knowledge
(Misconception, Nonmisconception) and
Text Structure (Refutation, Nonrefutation) in Experiment 1
Misconception Nonmisconception
Refutation Nonrefutation Refutation Nonrefutation
Processes M SD M SD M SD M SD
Comprehension monitoring 15.3 9.6 15.1 8.8 14.1 8.3 14.2 7.8
Associations 12.9 9.1 17.9 8.9 13.6 7.8 16.5 8.1
Intrasentential connections 11.1 7.7 13.7 7.3 9.3 8.3 10.9 11.4
Intersentential connections 3.2 2.5 4.0 2.4 4.0 4.0 3.2 2.7
Correct inferences* 10.0 7.3 5.4 4.2 16.3 11.0 10.9 7.6
Incorrect inferences* 3.9 3.4 4.0 4.0 0.5 1.5 0.3 0.7
Conceptual change strategies*† 6.2 4.6 0.9 1.5 0.1 0.4 0.1 0.6
Note—The means are adjusted with reference to the covariates.  *p , .05, main effect of prior knowl-
edge.  †p , .05, interaction effect of prior knowledge and text structure.
 Scientific Texts     1571

interactions were significant for the seven processes iden- Table 2

tified during thinking aloud (all ps . .2). Average Number of Statements Recalled As a Function of
Prior Knowledge (Misconception, Nonmisconception) and Text
Structure (Refutation, Nonrefutation) in Experiments 1 and 2
Offline Recall
Misconception Nonmisconception
We conducted a repeated measures ANOVA, with prior
knowledge (misconception, nonmisconception), text struc- Text Structure M SD M SD
ture (refutation, nonrefutation), and text topic (Newton’s Experiment 1
  Refutation* 11.5 3.4 14.0 4.6
first law, Newton’s third law) as independent variables,   Nonrefutation* 11.0 4.5 13.5 5.2
and frequency of recall for the texts as the dependent vari- Experiment 2
able. Reading span and Need for Cognition variables were   Refutation* 14.0 3.5 16.2 3.7
entered as covariates. This analysis revealed a significant   Nonrefutation* 11.6 3.6 14.2 4.7
effect of prior knowledge [F(1,74) 5 5.8, p 5 .02, h2 5 Note—The means are adjusted with reference to the covariates.  *p , .05,
.08]. As can be seen in Table 2, readers with misconcep- main effect of prior knowledge.
tions recalled less textual information than did readers
with no misconceptions [t(74) 5 2.4, p , . 05], regardless The offline results show that readers’ memory for both
of structure or topic. No other main effects or interactions refutation and nonrefutation texts is affected by prior
were significant (all ps . .2). knowledge. These results are consistent with those ob-
tained in prior research demonstrating that misconcep-
Connecting Online Processes and Offline Recall
tions have an intrusive effect on students’ memory repre-
We explored whether the patterns of think-aloud re-
sentations of text (Guzzetti et al., 1993; Kendeou & van
sponses during reading were related to subsequent text
den Broek, 2005).
recall. We performed a hierarchical multiple regression
The findings also show that the online processes were
analysis for each group (misconception, nonmisconcep-
reflected in memory once reading was completed. In par-
tion), with the proportion of sentences recalled as the de-
ticular, for readers with misconceptions, the number of cor-
pendent variable and, as independent variables, a dichoto-
rect inferences generated during the think-aloud phase was
mous text structure variable (refutation, nonrefutation) in
associated with higher recall for the text. This was not the
the first step and the number of each of the seven types
case for readers without misconceptions. These findings
of think-aloud responses (associations, comprehension
suggest that the two groups of readers differentially ben-
monitoring, intrasentential connections, intersentential
efited from constructing correct inferences during reading.1
connections, correct inferences, incorrect inferences, and One possible explanation for this interesting finding relates
conceptual change) in the second step. For the misconcep- to the function of correct inferences. It is conceivable that
tion group, the regression model for text structure (first the generation of correct inferences increased the compat-
step) did not account for any significant variance in text ibility of readers’ prior knowledge and the text. So, in the
recall [F(1,75) 5 2.2, p . .7]. The regression model for case of readers with misconceptions, more correct infer-
the think-aloud responses (second step), however, ac- ences increased the coherence between their knowledge
counted for a significant 18% of the variance in text recall and the text, which in turn increased their text recall. For
[F(7,68) 5 2.2, p 5 .04]. Specifically, the number of cor- readers without misconceptions, the existing coherence be-
rect inferences that occurred during the think-aloud phase tween readers’ prior knowledge and the text was already at a
strongly predicted the proportion of sentences recalled high level, so correct inferences had no additional benefit in
[b 5 .37, t(68) 5 2.4, p , .05]: The greater the number of that direction. An alternative explanation relates to the pres-
correct inferences an individual produced, the more that ence of a mediating variable—that of conceptual change
individual recalled. None of the other six think-aloud pre- processes. It is likely that the more conceptual change pro-
dictors were significant (all ps . .2). For the nonmiscon- cesses readers produced, the more correct inferences they
ception group, neither regression model (for text structure produced, which in turn led to better recall. In this case,
and think-aloud responses, respectively) accounted for generating correct inference alone (as in the case of readers
any significant variance in text recall (both ps . .4). without misconceptions) made no contribution to readers’
text recall. We explored this possibility in our data, and bi-
Discussion variate correlations showed that, indeed, in the misconcep-
The online results show that the processes in which tion group, correct inferences and conceptual change pro-
readers with misconceptions and readers without miscon- cesses were significantly related (r 5 .45, p , .01), correct
ceptions engage during their reading of scientific texts inferences and recall were also significantly related (r 5
are similar in some respects, yet different in others. In .42, p , .01), but conceptual change processes and recall
particular, readers with misconceptions engage in more were not directly related (r 5 2.04, p . .8). In the non-
conceptual change strategies than do readers without mis- misconception group, there were no significant relations
conceptions during the reading of a refutation text, but between any of these variables. Although further research is
not during reading of a nonrefutation text. In addition, in needed to investigate these issues, these tentative explana-
both text structures, readers with misconceptions generate tions can provide some insight into why correct inferences
more incorrect and fewer correct inferences than do read- contributed to better recall for readers with misconceptions
ers without misconceptions. and did not for readers without misconceptions.
1572     Kendeou and van den Broek

In this experiment, we used a think-aloud methodology Finally, we expected a convergence between the online
to investigate online processing and to obtain a variety of results obtained in Experiment 1 and those obtained in
students’ responses. Think-aloud procedures provide a Experiment 2, with the think-aloud processes revealed
rich source of data, but they also have several limitations. during reading by one group of readers predicting reading
One limitation is that this methodology reveals only a times of another group of readers during silent reading.
subset of cognitive processes during reading—namely, Validating the think-aloud data with independent behav-
those of which readers are aware. However, readers also ioral measures such as reading times helps strengthen any
engage in processes of which they are not aware. For in- conclusions that might be drawn from either data set alone
stance, when readers encounter information in a text that (Fletcher, 1986; Magliano & Graesser, 1991; Magliano
is inconsistent with information in the preceding text, et al., 1999).
they slow down without necessarily being aware of doing
so (Albrecht & O’Brien, 1993; Myers, O’Brien, Albrecht, Method
& Mason, 1994). A second limitation of the think-aloud Participants
methodology is that it may alter readers’ spontaneous A total of 69 University of Minnesota undergraduates enrolled
processing of the text, resulting in changes in compre- in introductory psychology or physics courses participated in this
hension (Fletcher, 1986). Because of these limitations, study. The participants received extra course credit or a gift certifi-
it is important to obtain converging evidence by using cate for participating. The data of 9 students were eliminated from
a less intrusive and more automatic measure (Magliano the study because they were nonnative English speakers. Of the re-
maining participants, 30 were female and 30 were male, with an age
& Graesser, 1991). Thus, in Experiment 2, we obtained range of 18–32 years. Thirty participants (20 female, 10 male) made
reading time measures for the text sentences. Reading up the misconception group (students who scored 60% or less on
times at the word, sentence, and text levels have been the Force Concept Inventory). The remaining 30 participants (who
used extensively in psychological research as an indicator scored above 60% on the Force Concept Inventory) made up the
of processing (Lorch & van den Broek, 1997; O’Brien, nonmisconception group (10 female, 20 male).
1995; Rapp, Gerrig, & Prentice, 2001; Zwaan & Singer,
Materials
2003). The materials were the same as those in Experiment 1.

Experiment 2 Procedure
Reading Times Methodology The participants were seen individually in a single session. The
session lasted approximately 1 h 10 min. The procedure was similar
The objective of Experiment 2 was to investigate the to that of Experiment 1, with one difference: Instead of reading the
interaction between prior knowledge and text structure two texts using the think-aloud methodology, the participants simply
read the texts silently on a computer screen one sentence at a time,
online, using a reading time methodology. In this experi- at their own speed. The participants advanced from one sentence to
ment, participants with and without misconceptions were the next by pressing the space bar.
asked to read a refutation and a nonrefutation scientific
text on the computer, one sentence at a time. We investi- Scoring
gated reading times for sentences that were relevant for Individual difference tests. The Force Concept Inventory, the
understanding the scientific laws described in each text— Need for Cognition Scale, and the reading span test were scored as
namely, the sentences stating Newton’s laws and those that in Experiment 1.
Recall protocols. Students’ written recall protocols were parsed
explained these laws. As in Experiment 1, participants and coded as in Experiment 1. Two raters independently parsed and
were also asked to recall the texts they read and to com- coded 25% of the recall protocols, and their agreement was reli-
plete the measures of working memory capacity and the ably high, for both parsing (K 5 .98, p , .01) and coding (K 5 .94,
Need for Cognition Scale. p , .01). Disagreements were resolved through discussion, and the
We made several predictions with respect to the out- remaining protocols were divided among the two raters for them to
comes of this experiment. First, as in Experiment 1, parse and code independently.
we expected that the interaction between readers’ prior
knowledge and text structure would affect readers’ online Results
comprehension processes. In particular, given the findings We performed preliminary analyses to explore whether
of Experiment 1, we expected that readers with miscon- misconception and nonmisconception groups differed
ceptions would spend more time reading sentences that with regard to the individual difference measures. The
contradicted their prior knowledge (Newton’s laws and analysis showed that there were no significant differences
explanations) than would readers without misconceptions. in reading span [F(1,56) , 1, p 5 .9] or in Need for Cog-
Given the interaction effect of text structure and prior nition [F(1,56) 5 2.2, p 5 .1] scores between miscon-
knowledge observed in Experiment 1, this effect would ception (reading span, M 5 39.2, SD 5 3.9; Need for
most obviously obtain when participants read refutation Cognition, M 5 65.3, SD 5 12.9) and nonmisconception
text structures. Second, we expected to replicate the off- (reading span, M 5 38.9, SD 5 4.3; Need for Cognition,
line results obtained in Experiment 1—that readers with M 5 69.3, SD 5 7.5) groups. As in Experiment 1, the
misconceptions would remember less information from reading span and Need for Cognition scores for each indi-
the texts than would readers without misconceptions. This vidual were used as covariates in all subsequent analyses,
result would suggest that the think-aloud methodology in to eliminate possible confounding of text structure and
Experiment 1 did not alter readers’ normal processing. prior knowledge effects with these factors. Also, because
 Scientific Texts     1573

we had two texts, with different topics for each structure, Offline Recall
we included in the analysis a variable reflecting the topic We conducted a repeated measures ANOVA, with prior
of the text. knowledge (misconception, nonmisconception), text struc-
ture (refutation, nonrefutation), and text topic (Newton’s
Online Reading Times first law, Newton’s third law) as the independent variables
We performed an analysis on the reading times per syl- and frequency of recall for the two texts as the dependent
lable (to adjust for variability in sentence length) for tar- variable. Reading span and Need for Cognition variables
get Newton’s laws and explanation sentences from each were entered as covariates. This analysis revealed a sig-
text. Newton’s law sentences explicitly stated the correct nificant effect of prior knowledge [F(1,54) 5 6.6, p 5 .01,
information about Newton’s laws of motion, whereas ex- h2 5 .11]. As can be seen in Table 2, readers with miscon-
planation sentences provided evidence for the refutations ceptions recalled less text information than did readers
in the refutation texts and for the supporting examples with no misconceptions [t(54) 5 2.5, p , .05]. No other
in the nonrefutation texts (see Appendix for examples). main effects or interactions were significant (all ps . .2).
Explanation sentences differed in content across topics
and text structures, so they may be compared only with Connecting Think-Aloud Responses
caution. We removed outlying reading time data using (Experiment 1) and Reading Times
Tukey’s (1977) hinge criterion. This eliminated approxi- (Experiment 2)
mately 5.34% of the data. We conducted separate repeated We assessed the relation between the thoughts produced
measures ANOVAs, with prior knowledge (misconcep- during the think-aloud phase in Experiment 1 and the
tion, nonmisconception), text structure (refutation, non- sentence reading times in Experiment 2, using the proce-
refutation), and text topic (Newton’s first law, Newton’s dures suggested by Magliano and colleagues (Magliano &
third law) as the independent variables and average read- Graesser, 1991; Magliano et al., 1999). To do so, we tested
ing times for each selected category of target sentences whether the frequency of think-aloud processes produced
as the dependent variables. Reading span and Need for in Experiment 1 predicted the reading times during silent
Cognition variables were entered as covariates. The means reading obtained in Experiment 2. Following Magliano
are provided in Table 3. et al., we performed multiple regression analysis in which
For reading times of Newton’s law sentences, the the predictor variables consisted of the average frequency
analysis revealed a significant effect of text structure of each of the seven types of cognitive processes for each
[F(1,44) 5 16.5, p 5 .001, h2 5 .27]. Post hoc analyses sentence produced in Experiment 1, and the dependent
showed that readers read Newton’s law sentences faster in variable consisted of the average reading time for each
the refutation (M 5 201) than in the nonrefutation (M 5 sentence recorded in Experiment 2. We also included three
295) [t(44) 5 24.1, p , .01] texts. This main effect was variables to control for (1) the number of syllables in each
qualified by an interaction between text structure and sentence, (2) text structure (refutation, nonrefutation), and
prior knowledge [F(1,44) 5 8.3, p 5 .006, h2 5 .16]. (3) prior knowledge (misconception, nonmisconception).
Post hoc analyses showed that readers with misconcep- As we have seen from the previous analyses, text structure
tions spent more time reading the law sentences than did and prior knowledge significantly influence think-aloud
readers without misconceptions in the refutation texts responses and reading times, so these were included as
[t(44) 5 4.3, p , .01], but not in the nonrefutation texts dichotomous variables.
[t(44) , 1, p . .05]. The regression model accounted for a significant 74% of
For the explanation sentences, the analysis also revealed the variance in sentence reading times [F(10,271) 5 76.9,
a significant effect of text structure [F(1,52) 5 11.8, p 5 p 5 .001]. Two factors were responsible for this effect.
.001, h2 5 .18]. Post hoc analyses showed that readers Sentence reading times increased as a function of the num-
read the explanation sentences faster in the refutation ber of syllables in a sentence [b 5 .84, t(271) 5 26, p ,
(M 5 206) than in the nonrefutation (M 5 253) [t(52) 5 .001] and as a function of the number of associations that
23.4, p , .01] texts. No other main effects or interactions occurred during the think-aloud phase [b 5 .07, t(271) 5
were significant (all ps . .3). 2.3, p , .05]: The longer the sentence and the more asso-

Table 3
Average Reading Times (Per Syllable) of Target Sentences in Experiment 2
As a Function of Prior Knowledge (Misconception, Nonmisconception)
and Text Structure (Refutation, Nonrefutation)
Misconception Nonmisconception
Refutation Nonrefutation Refutation Nonrefutation
Sentences M SD M SD M S M SD
Laws*† 252 84 280 138 148 84 310 138
Explanations* 217 72 254 82 195 72 253 86
Note— The means are adjusted with reference to the covariates.  *p , .05, main effect of text
structure.  †p , .05, interaction effect of prior knowledge and text structure.
1574     Kendeou and van den Broek

ciations a sentence elicited, the longer the reading time. No related to the text material. The observation that such ac-
other predictors were significant (all ps . .2). tivation requires additional time is consistent with current
theories of reading comprehension, according to which
Discussion readers routinely activate prior knowledge from memory
The online reading times suggest that when readers when it is required for comprehension, through memory-
with misconceptions read refutation texts, they detect the based and constructionist processes (Graesser, Singer,
contradiction between their faulty prior beliefs and the & Trabasso, 1994; McKoon & Ratcliff, 1992; Myers &
information presented in the text. Newton’s law sentences O’Brien, 1998). Whereas activation of prior knowledge
were read more slowly by readers with misconceptions through memory-based processes may occur without a
about the laws than by readers who had no such miscon- processing cost (van den Broek et al., 2005), activation
ceptions, but only when these sentences were included in a of prior knowledge using strategy-based processes—the
refutation text. With respect to the explanation sentences, ones likely to be reflected in think-aloud protocols—
the results showed that there were no differences in the would require additional time.
reading times of the two groups. Although we expected
that readers with misconceptions would spend more time General Discussion
reading the explanation sentences in the refutation than
in the nonrefutation texts because the explanations would The aim of this set of experiments was to investigate the
assist them in reconciling the inconsistent information, effects of text structure and readers’ prior knowledge—
the lack of differences may be due in part to the fact that and their possible interaction—on the comprehension pro-
explanations differed in content across texts; hence, these cesses that take place during reading of scientific texts. We
findings should be interpreted with caution. focused particularly on the processing of refutation and
Interestingly, all readers read the target sentences faster nonrefutation scientific texts by readers with and without
in the refutation texts than in the nonrefutation texts. One misconceptions related to the topics of the text. We also
possibility is that refutation texts were on average easier to took into account readers’ individual differences in work-
read than nonrefutation texts. All texts, however, were con- ing memory capacity and Need for Cognition scores as
trolled with respect to conventional readability measures, possible factors that might influence processing in reading
including word length, grade level, and ease of reading. So situations of this nature.
there is no indication that the refutation texts were in fact The online results provide direct evidence for the inter-
easier. Another possibility is that refutation texts result in active effects of readers’ prior knowledge and text struc-
a more coherent representation because they elicit an ex- ture during reading. Readers with misconceptions engaged
plicit integration between readers’ potential background in different processes than did readers without miscon-
knowledge and the to-be-learned material and, hence, ceptions, but only during reading of text with the refuta-
greater reading speed. This remains an empirical question tion structure. Specifically, readers with misconceptions
to be addressed in the future. It is important to note that engaged in conceptual change strategies more than did
even though we observed a text structure effect on reading readers without misconceptions when reading a refutation
times in this experiment, no corresponding effect was ob- text (Experiment 1). These conceptual change strategies
served on conceptual change processes in Experiment 1. included statements indicating that readers were experi-
Given that readers without misconceptions did not engage encing conflict, responding to conflict, and contrasting
in conceptual change processes in Experiment 1, the latter information. Furthermore, readers with misconceptions
finding most likely reflects a floor effect. spent more time reading sentences describing Newton’s
The offline findings demonstrate that readers’ memory laws that directly contradicted their misconceptions about
for both text structures (refutation, nonrefutation) was a those laws than did readers without misconceptions, but
function of their prior knowledge. Readers with miscon- only when those sentences were included in a refutation
ceptions recalled less textual information than did read- text (Experiment 2). Thus, our findings provide evidence
ers with no misconceptions across the two text structures. that readers adjusted their processing of the text when
These findings are consistent with those obtained in Ex- prior knowledge conflicted with textual information dur-
periment 1. Thus, they provide convergent evidence for ing comprehension of the refutation texts. This pattern of
our conclusions about online processes and, moreover, results was not present when readers—with and without
suggest that the think-aloud methodology did not alter misconceptions—read the nonrefutation texts.
readers’ normal reading. With regard to other comprehension processes (i.e.,
The findings also demonstrate that there is a direct associations, establishing connections within and across
connection between the online processes in the two ex- sentences, generating explanatory, predictive, and other
periments. Indeed, the think-aloud responses obtained in inferences, and comprehension monitoring), the results
Experiment 1 predicted reading times for sentences in Ex- indicate that the frequency of these processes was not af-
periment 2. Specifically, the number of associations read- fected by readers’ prior knowledge or by text structure but
ers generated significantly predicted reading times of sen- that the content of some of them was (Experiment 1). In
tences during silent reading. Note that in the present study, particular, readers with misconceptions generated more
we defined associations as responses that reflected activa- incorrect inferences and fewer correct inferences than did
tion of prior knowledge, including knowledge that was readers without misconceptions.
 Scientific Texts     1575

These findings add to our understanding of the role of beliefs (Dole, 2000; Posner, Strike, Hewson, & Gertzog,
text structure and readers’ misconceptions during online 1982). Alternatively, it is possible that readers with mis-
text processing of scientific texts by demonstrating that conceptions successfully revised their incorrect ideas but
the actual reading processes themselves, as they unfold that the recall measure was not sensitive enough to cap-
during reading of the text, are influenced by both mis- ture this effect. Evidence for revising prior knowledge
conceptions and text structure. On the one hand, readers is most likely to be reflected on tasks that assess read-
with misconceptions proceed through nonrefutation texts ers’ learning from text (as opposed to memory for the
at the same rate and with the same types of processes as do text)—that is, on tasks that require readers to apply the
readers without misconceptions. When the text calls for acquired knowledge in a new situation (Guzzetti et al.,
it, the readers with misconceptions activate and integrate 1993; Kintsch, 1998).
their background knowledge with the textual information The described results are independent of individual
as frequently as do readers without misconceptions (see differences in working memory and Need for Cognition.
also Hannon & Daneman, 2001; Kintsch, 1988). But the On average, readers with and readers without misconcep-
content of knowledge-based inferences (e.g., explanatory, tions did not differ in their working memory capacity in
predictive, and other inferences) reflects their misconcep- either experiment. In Experiment 1, readers with miscon-
tions, resulting in more incorrect inferences and fewer ceptions on average had lower Need for Cognition scores
correct ones. On the other hand, readers with misconcep- than did their counterparts without misconceptions, sug-
tions proceed through refutation texts at a slower rate and gesting that readers with misconceptions may maintain
rely on different processes than do readers without mis- lower standards for coherence (van den Broek, Risden, &
conceptions. When the text contradicts their prior knowl- Husebye-Hartmann, 1995). Although the latter difference
edge, they spend more time reading and they engage in was not repeated in Experiment 2, we included both vari-
conceptual change strategies. ables as covariates in all analyses, to eliminate any possible
Previous research in reading and science education has confounds. In none of the analyses did either the working
suggested that refutation texts are effective means of con- memory capacity or the Need for Cognition score account
ceptual change learning (Guzzetti et al., 1993). This con- for a significant portion of the variance. Thus, the observed
clusion stemmed primarily from results obtained with off- effects of readers’ prior knowledge and text structure are
line tasks, in which students’ learning was measured using not the result of individual differences in working memory
application questions or problem-solving activities. Such capacity or Need for Cognition.
offline measures provide limited insight into what happens Different types of online data were obtained in the two
during reading, however, and restrict claims with respect experiments. Although these data were collected from
to readers’ cognitive processing. The results of the present different participants, comparison of the two types of in-
study allow us to go beyond these findings and consider formation about online processes may suggest interesting
how the actual reading processes themselves unfold. They connections2 (Fletcher, 1986; Magliano & Graesser, 1991;
provide evidence that during online reading of refutation Magliano et al., 1999). In particular, in light of the focus
text, readers coactivate and integrate prior knowledge and of the present study on the role of background knowledge,
text information, which, in turn, allows them to detect we explored whether the activation of such information
the inconsistency between their knowledge and the text. was related to the speed of reading. Indeed, the number
Detecting the inconsistency enables readers to engage in of associations that participants produced in response to
additional processing in an attempt to establish coherence different sentences (Experiment 1) significantly predicted
(Glenberg, Wilkinson, & Epstein, 1982; Graesser et al., reading times for the same sentences (Experiment 2). No
1994; McNamara & Kintsch, 1996) or reconcile the in- other think-aloud process significantly predicted reading
consistent information (e.g., Hakala & O’Brien, 1995; times, including knowledge-based inferences. The find-
Linderholm, Virtue, Tzeng, & van den Broek, 2004). The ing that associations did but knowledge-based inferences
present findings indicate that refutation texts help readers did not predict processing times suggests that, consistent
engage in conceptual change learning by eliciting coacti- with current theories of reading comprehension, activa-
vation, thereby facilitating inconsistency detection and the tion of prior knowledge may occur with (Graesser et al.,
establishment of coherence. 1994) and without (McKoon & Ratcliff, 1992; Myers &
With regard to the offline text representation, readers’ O’Brien, 1998) a processing cost.
misconceptions affected their memory for the text in both In conclusion, the findings of this study contribute to
experiments in the same way: Regardless of text struc- our theoretical understanding of the effects that reader
ture, readers with misconceptions remembered less infor- characteristics, such as background knowledge, and text
mation than did readers without misconceptions. Thus, properties have on readers’ cognitive processes during
the conceptual change processes elicited by refutation comprehension. This understanding is particularly impor-
texts in readers with misconceptions do not inevitably tant in the context of comprehension of scientific texts, be-
result in successful revision of their incorrect ideas. This cause aspects of the reading situation besides the content
may reflect the fact that successful revision is a function itself—such as readers’ individual differences in domain
of many factors, including the sufficiency of an alterna- knowledge, strategies, attention, and motivation, as well
tive explanation, the plausibility of the new scientific in- as text properties such as structure—strongly codetermine
formation, and readers’ commitment to their preexisting the likelihood of acquisition of the content in such texts.
1576     Kendeou and van den Broek

AUTHOR NOTE Engle, R. W., & Conway, A. R. A. (1998). Working memory and com-
prehension. In R. H. Logie & K. J. Gilhooly (Eds.), Working memory
This project was supported by a Robert and Corrie Beck Fellowship and thinking (pp. 67-91). East Sussex, U.K.: Psychology Press.
from the College of Education and Human Development at the Univer- Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal re-
sity of Minnesota and by a dissertation research grant from the Graduate ports as data. Cambridge, MA: MIT Press.
School at the University of Minnesota to P. K., by a Golestan and Lorentz Fletcher, C. R. (1986). Strategies for the allocation of short-term
fellowship at the Netherlands Institute for Advanced Study in the Hu- memory during comprehension. Journal of Memory & Language, 25,
manities and Social Sciences to P. v. d. B., and by the Center for Cogni- 43-58.
tive Sciences at the University of Minnesota through Grant HD-07151 Glenberg, A. M., Wilkinson, A. C., & Epstein, W. (1982). The illu-
from the National Institute of Child Health and Human Development. sion of knowing: Failure in the self-assessment of comprehension.
We thank Joe Magliano, David N. Rapp, and two anonymous review- Memory & Cognition, 10, 597-602.
ers for their advice and comments on the manuscript. Correspondence Goldman, S. R., & Bisanz, G. L. (2002). Toward a functional analy-
concerning this article should be addressed to P. Kendeou, Faculty of sis of scientific genres: Implications for understanding and learn-
Education, McGill University, 3700 McTavish Avenue, Montreal, QC, ing processes. In J. Otero, J. A. Leon, & A. C. Graesser (Eds.), The
H3A 1Y2 Canada (e-mail: panayiota.kendeou@mcgill.ca). psychology of science text comprehension (pp. 19-50). Mahwah, NJ:
Erlbaum.
References Goldman, S. R., & Varma, S. (1995). CAPing the construction–­
integration model of discourse comprehension. In C. A. Weaver,
Afflerbach, P. (2002). Verbal reports and protocol analysis. In M. L. S. Mannes, & C. R. Fletcher (Eds.), Discourse comprehension: Essays
Kamil, P. B. Mosenthal, P. D. Pearson, & R. Barr (Eds.), Handbook of in honor of Walter Kintsch (pp. 337-358). Hillsdale, NJ: Erlbaum.
reading research (Vol. III, pp. 87-103). Mahwah, NJ: Erlbaum. Graesser, A. C., Leon, J. A., & Otero, J. (2002). Introduction to the
Albrecht, J. E., & O’Brien, E. J. (1993). Updating a mental model: psychology of science text comprehension. In J. Otero, J. A. Leon, &
Maintaining both local and global coherence. Journal of Experimental A. C. Graesser (Eds.), The psychology of science text comprehension
Psychology: Learning, Memory, & Cognition, 19, 1061-1069. (pp. 1-15). Mahwah, NJ: Erlbaum.
Alvermann, D. E., & Hague, S. A. (1989). Comprehension of coun- Graesser, A., Singer, M., & Trabasso, T. (1994). Constructing infer-
terintuitive science text: Effects of prior knowledge and text structure. ences during narrative comprehension. Psychological Review, 101,
Journal of Educational Research, 82, 197-202. 371-395.
Alvermann, D. E., & Hynd, C. (1989). Effects of prior knowledge acti- Guthrie, J. T., McGough, K., Bennett, L., & Rice, M. E. (1996).
vation modes and text structure on nonscience majors’ comprehension Concept-oriented reading instruction: An integrated curriculum to
of physics. Journal of Educational Research, 83, 97-102. develop motivations and strategies for reading. In L. Baker, P. Affler-
Alvermann, D. E., Smith, L. C., & Readence, J. E. (1985). Prior bach, & D. Reinking (Eds.), Developing engaged readers in school
knowledge activation and the comprehension of compatible and in- and home communities (pp. 165-190). Hillsdale, NJ: Erlbaum.
compatible text. Reading Research Quarterly, 20, 420-436. Guzzetti, B. J., Snyder, T. E., Glass, G. V., & Gamas, W. S. (1993).
Baddeley, A., & Hitch, G. (1974). Working memory. Psychology of Promoting conceptual change in science: A comparative meta-­analysis
Learning & Motivation, 8, 47-90. of instructional interventions from reading education and science edu-
Bartlett, F. C. (1932). Remembering: A study in experimental and cation. Reading Research Quarterly, 28, 117-159.
social psychology. Cambridge: Cambridge University Press. Hakala, C. M., & O’Brien, E. J. (1995). Strategies for resolving coher-
Budd, D., Whitney, P., & Turley, K. J. (1995). Individual differences ence breaks in reading. Discourse Processes, 20, 167-185.
in working memory strategies for reading expository text. Memory & Hannon, B., & Daneman, M. (2001). A new tool for measuring and
Cognition, 23, 735-748. understanding individual differences in the component processes of
Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal reading comprehension. Journal of Educational Psychology, 93, 103-
of Personality & Social Psychology, 42, 116-131. 128.
Chambliss, M. J. (2002). The characteristics of well-designed sci- Hestenes, D., & Halloun, I. (1995). Interpreting the Force Concept
ence textbooks. In J. Otero, J. A. Leon, & A. C. Graesser (Eds.), The Inventory. The Physics Teacher, 33, 502-506.
psychology of science text comprehension (pp. 1-15). Mahwah, NJ: Hestenes, D., Wells, M., & Swackhamer, G. (1992). Force Concept
Erlbaum. Inventory. The Physics Teacher, 30, 141-158.
Chi, M. T. H. (1978). Knowledge structures and memory. In R. S. Siegler Hewitt, P. G. (2002). Touch this! Conceptual physics for everyone.
(Ed.), Children’s thinking: What develops? (pp. 73-96). Hillsdale, NJ: Glenview, IL: Pearson Education.
Erlbaum. Jenkins, J. J. (1979). Four points to remember: A tetrahedral model
Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization of memory experiments. In L. S. Cermak & F. I. M. Craik (Eds.),
and representation of physics problems by experts and novices. Cogni- Levels of processing in human memory (pp. 429-446). Hillsdale, NJ:
tive Science, 5, 121-152. Erlbaum.
Chiesi, H. L., Spilich, G. J., & Voss, J. F. (1979). Acquisition of Kardash, C. M., & Scholes, R. J. (1996). Effects of pre-existing be-
­domain-related information in relation to high and low domain knowl- liefs, epistemological beliefs, and need for cognition on interpreta-
edge. Journal of Verbal Learning & Verbal Behavior, 18, 257-273. tion of controversial issues. Journal of Educational Psychology, 88,
Coté, N., & Goldman, S. R. (1999). Building representations of in- 260-271.
formational text: Evidence from children’s think-aloud protocols. Kendeou, P., Rapp, D. N., & van den Broek, P. (2004). The influence
In H. van Oostendorp & S. R. Goldman (Eds.), The construction of of readers’ prior knowledge on text comprehension and learning from
mental representations during reading (pp. 169-193). Mahwah, NJ: text. In R. Nata (Ed.), Progress in education (Vol. 13, pp. 189-209).
Erlbaum. New York: Nova Science.
Diakidoy, I. N., & Kendeou, P. (2001). Facilitating conceptual change Kendeou, P., & van den Broek, P. (2005). The effects of readers’ mis-
in astronomy: A comparison of the effectiveness of two instructional conceptions on comprehension of scientific text. Journal of Educa-
approaches. Learning & Instruction, 11, 1-20. tional Psychology, 97, 235-245.
Diakidoy, I. N., Kendeou, P., & Ioannides, C. (2003). Reading about Kintsch, W. (1988). The role of knowledge in discourse comprehen-
energy: The effects of text structure in science learning and conceptual sion: A construction–integration model. Psychological Review, 95,
change. Contemporary Educational Psychology, 28, 335-356. 163-182.
Dochy, F., Segers, M., & Buehl, M. M. (1999). The relation between Kintsch, W. (1998). Comprehension: A paradigm for cognition. Cam-
assessment practices and outcomes of studies: The case of research on bridge: Cambridge University Press.
prior knowledge. Review of Educational Research, 69, 145-186. Langston, M. C., & Trabasso, T. (1999). Modeling causal integration
Dole, J. A. (2000). Readers, texts and conceptual change learning. and availability of information during comprehension of narrative
Reading & Writing Quarterly, 16, 99-118. texts. In H. van Oostendorp & S. R. Goldman (Eds.), The construc-
Engle, R. W. (2002). Working memory capacity as executive attention. tion of mental representations during reading (pp. 29-69). Mahwah,
Current Directions in Psychological Science, 11, 19-23. NJ: Erlbaum.
 Scientific Texts     1577

Linderholm, T., & van den Broek, P. (2002). The effects of reading modation of a scientific conception: Toward a theory of conceptual
purpose and working memory capacity on the processing of exposi- change. Science Education, 66, 211-227.
tory text. Journal of Educational Psychology, 94, 778-784. Pressley, M., & Afflerbach, P. (1995). Verbal protocols of read-
Linderholm, T., Virtue, S., Tzeng, Y., & van den Broek, P. (2004). ing: The nature of constructively responsive reading. Hillsdale, NJ:
Fluctuations in the availability of information during reading: Captur- Erlbaum.
ing cognitive processes using the Landscape Model. Discourse Pro- Pritchard, R. (1990). The effects of cultural schemata on reading pro-
cesses, 37, 165-186. cessing strategies. Reading Research Quarterly, 24, 273-295.
Lipson, M. Y. (1982). Learning new information from text: The role of Rapp, D. N., Gerrig, R. J., & Prentice, D. A. (2001). Readers’ trait-
prior knowledge and reading ability. Journal of Reading Behavior, based models of characters in narrative comprehension. Journal of
14, 243-261. Memory & Language, 45, 737-750.
Lorch, R. F., Jr., & van den Broek, P. (1997). Understanding reading Rapp, D. N., & van den Broek, P. (2005). Dynamic text comprehen-
comprehension: Current and future contributions to cognitive science. sion. Current Directions in Psychological Science, 14, 276-279.
Contemporary Educational Psychology, 22, 213-246. Recht, D. R., & Leslie, L. (1988). Effect of prior knowledge on good
Magliano, J. P., & Graesser, A. C. (1991). A three-pronged method and poor readers’ memory of text. Journal of Educational Psychol-
for studying inference generation in literary text. Poetics, 20, 193- ogy, 80, 16-20.
232. Singer, M., Andrusiak, P., Reisdorf, P., & Black, N. L. (1992). Indi-
Magliano, J. P., & Millis, K. K. (2003). Assessing reading skill with vidual differences in bridging inference processes. Memory & Cogni-
a think-aloud procedure and latent semantic analysis. Cognition & tion, 20, 539-548.
Instruction, 21, 251-283. Trabasso, T., & Suh, S. (1993). Understanding text: Achieving explan-
Magliano, J. P., Trabasso, T., & Graesser, A. C. (1999). Strategic atory coherence through on-line inferences and mental operations in
processes during comprehension. Journal of Educational Psychology, working memory. Discourse Processes, 16, 3-34.
91, 615-629. Tukey, J. W. (1977). Exploratory data analysis. Reading, MA: Addison-
Maria, K., & MacGinitie, W. (1987). Learning from texts that refute Wesley.
the reader’s prior knowledge. Reading Research & Instruction, 26, van den Broek, P., Fletcher, C. R., & Risden, K. (1993). Investiga-
222-238. tions of inferential processes in reading: A theoretical and method-
McKoon, G., & Ratcliff, R. (1992). Inference during reading. Psycho- ological integration. Discourse Processes, 16, 169-180.
logical Review, 99, 440-466. van den Broek, P., & Kremer, K. E. (1999). The mind in action: What
McNamara, D. S. (2001). Reading both high-coherence and low- it means to comprehend during reading. In B. Taylor, M. Graves, &
­coherence texts: Effects of text sequence and prior knowledge. Cana- P. van den Broek (Eds.), Reading for meaning (pp. 1-31). New York:
dian Journal of Experimental Psychology, 55, 51-62. Teacher’s College Press.
McNamara, D. S., & Kintsch, W. (1996). Learning from texts: Ef- van den Broek, P., Rapp, D. N., & Kendeou, P. (2005). Integrating
fects of prior knowledge and text coherence. Discourse Processes, memory-based and constructionist processes in accounts of reading
22, 247-288. comprehension. Discourse Processes, 39, 299-316.
McNamara, D. S., Kintsch, E., Songer, N. B., & Kintsch, W. (1996). van den Broek, P., Risden, K., & Husebye-Hartmann, E. (1995). The
Are good texts always better? Text coherence, background knowledge, role of readers’ standards of coherence in the generation of inferences
and levels of understanding in learning from text. Cognition & In- during reading. In R. F. Lorch, Jr., & E. J. O’Brien (Eds.), Sources of
struction, 14, 1-43. coherence in reading (pp. 353-373). Hillsdale, NJ: Erlbaum.
Means, M. L., & Voss, J. F. (1985). Star Wars: A developmental study of van den Broek, P., Young, M., Tzeng, Y., & Linderholm, T. (1999).
expert and novice knowledge structures. Journal of Memory & Lan- The Landscape model of reading: Inferences and the online construc-
guage, 24, 746-757. tion of memory representation. In H. van Oosterndorp & S. R. Gold-
Meyer, B. J. F. (1975). The organization of prose and its effects on man (Eds.), The construction of mental representations during read-
memory. Amsterdam: North-Holland. ing (pp. 71-98). Mahwah, NJ: Erlbaum.
Meyer, B. J. F. (1999). Importance of text structure in everyday read- Zwaan, R. A. (1999). Five dimensions of narrative comprehension: The
ing. In A. Ram & K. Moorman (Eds.), Understanding language un- events indexing model. In S. R. Goldman, A. C. Graesser, & P. van
derstanding: Computational models of reading (pp. 227-252). Cam- den Broek (Eds.), Narrative comprehension, causality, and coherence
bridge, MA: MIT Press. (pp. 93-110). Mahwah, NJ: Erlbaum.
Meyer, B. J. F., & Freedle, R. O. (1984). Effects of discourse type on Zwaan, R. A., & Brown, C. M. (1996). The influence of language pro-
recall. American Educational Research Journal, 21, 121-143. ficiency and comprehension skill on situation-model construction.
Myers, J. L., & O’Brien, E. J. (1998). Accessing the discourse repre- Discourse Processes, 21, 289-327.
sentation during reading. Discourse Processes, 26, 131-157. Zwaan, R. A., & Singer, M. (2003). Text comprehension. In A. C.
Myers, J. L., O’Brien, E. J., Albrecht, J. E., & Mason, R. A. (1994). Graesser, M. A. Gernsbacher, & S. R. Goldman (Eds.), Handbook of
Maintaining global coherence during reading. Journal of Experimen- discourse processes (pp. 83-121). Hillsdale, NJ: Erlbaum.
tal Psychology: Learning, Memory, & Cognition, 20, 876-886.
O’Brien, E. J. (1995). Automatic components of discourse comprehen- Notes
sion. In R. F. Lorch, Jr., & E. J. O’Brien (Eds.), Sources of coherence
in reading (pp. 159-176). Hillsdale, NJ: Erlbaum. 1. We thank an anonymous reviewer for pointing this out.
Peeck, J., van den Bosch, A. B., & Kreupeling, W. J. (1982). Effect 2. We thank Joe Magliano for this suggestion.
of mobilizing prior knowledge on learning from text. Journal of Edu-
cational Psychology, 74, 771-777. (Manuscript received March 22, 2006;
Posner, G., Strike, K., Hewson, P., & Gertzog, W. (1982). Accom- revision accepted for publication December 19, 2006.)