SPECIALSECTION

to be inconsistent with the text? Relatively little

REVIEW psycholinguistic research has been conducted to
answer these questions, but findings on text

Using Texts in Science Education: comprehension provide insights into processes

and mechanisms involved in learning from text.

Cognitive Processes and Attention Allocation

and Inference Generation
Knowledge Representation Reading as a landscape of activations. The
situation model depends on the identification of
meaningful relations between text elements and
Paul van den Broek1,2 between those elements and background knowl-
edge. The processes by which readers identify
Texts form a powerful tool in teaching concepts and principles in science. How do readers relations, or fail to identify important relations,
extract information from a text, and what are the limitations in this process? Central to are captured by psychological models such as
comprehension of and learning from a text is the construction of a coherent mental representation the Landscape Model (5, 8). These models have
that integrates the textual information and relevant background knowledge. This representation been developed primarily in the context of
engenders learning if it expands the reader’s existing knowledge base or if it corrects narrative reading (fiction and storytelling) but
misconceptions in this knowledge base. The Landscape Model captures the reading process and have been found to apply to science (informa-
the influences of reader characteristics (such as working-memory capacity, reading goal, prior tional and factual) texts as well (9–11). The
knowledge, and inferential skills) and text characteristics (such as content/structure of presented Landscape Model identifies the reading process
information, processing demands, and textual cues). The model suggests factors that can as a balancing act between the reader’s limited
optimize—or jeopardize—learning science from text. attention or working memory and the need for
coherence. At any point during reading, the reader
exts support the acquisition of scientific background knowledge becomes the object of can only attend to a subset of all the elements in

T knowledge. Schools, colleges, and univer-

sities depend on texts for their science
instruction. The use of texts is not simply a mat-
change. How are the reader’s existing knowl-
edge structures updated to incorporate new
concepts? How are they modified when found
a text or all the relevant background knowledge
(2). As the reader proceeds through the text, the
contents of working memory are continually
ter of convenience: The subject matter usually is
so abstract that detailed verbal descriptions—
possibly supplemented by illustrations, videos, young
demonstrations, and so on—are indispensable. married
thankful
Given the importance of texts in science edu- freed
cation, how is it that people learn science from sword
killed Activation
text, and how can such learning be optimized? fire
scorched 5
armor 4
Comprehension and Learning Concepts lifedeath 3
2
fought 1
The processing of texts in the context of learn- hurriedafter
ing requires both comprehension and learning— love
wantmarry
the incorporation of comprehended information wantfree
in the reader’s background knowledge. Inves- princess
beautiful
tigations of text processing have focused pre- kidnapping
dominantly on the comprehension component. appeared
dragon
Most theoretical models of comprehension agree country
that comprehension entails the construction by unfamiliar
trees
the reader of a coherent mental representation forest
that captures the intended meaning of the text horse
rode
(1–6). In a successful representation, the indi- knight
vidual elements in the text (e.g., concepts and
facts about the concepts) are connected to each
other by meaningful relations. The reader re-
1
cruits relevant background knowledge into the 2
3
representation as well. Together, the elements 4
5
and relations create a “situation model,” an in- 6
7
8
terpreted description of the information in the Sentences
9
10
11
text (7). For comprehension, the reader’s back- 12
13
ground knowledge supports the process of inter-
preting and representing the text. For learning, Fig. 1. Computer simulation of hypothetical activations during reading of a narrative text about a
1
knight, a princess, and a dragon. The simulation displays the activation (vertical dimension) of text and
Department of Education and Child Studies, University of background-knowledge elements (“concepts”) over the course of reading (sentences 1 to 13). Concepts
Leiden, Wassenaarseweg 52, 2333 AK Leiden, Netherlands.
E-mail: broekpwvanden@fsw.leidenuniv.nl 2Center for Cog- that are activated simultaneously are connected in the memory representation, with the strength of the
nitive Sciences, University of Minnesota, Minneapolis, MN resulting relation a function of the degree of activation of each concept. [Based on (5), used with
55455, USA. permission]

www.sciencemag.org SCIENCE VOL 328 23 APRIL 2010 453

 Science, Language, and Literacy
refreshed, with some elements remaining the
focus of attention and others being replaced by
new ones. As reading progresses, the landscape
of activations of individual concepts fluctuates
(Fig. 1).
Establishing coherence. The exact nature of
the landscape of fluctuating activations has im-
portant consequences for the establishment of
coherence because a direct relation between two
elements is most likely to be detected if the two
elements are simultaneously in the focus of at-
tention of the reader and unlikely to be detected
if the elements are not simultaneously activated.
The limits imposed by working memory capac-
ity and the importance of simultaneous activa-
tion for inferring relations make the reader’s
allocation of attention of prime importance for
comprehension. Relations between elements
presented far apart in the text may not be de-
tected effectively, and the reader’s misconcep-
tions may impede comprehension because they
may support incorrect relations.
Text and reader characteristics. The alloca-
tion of attention is influenced by characteristics
of both text and reader. Demands on attention
are affected by such text characteristics as or-
ganization, verbal complexity, and the linguistic
markers or typographical prompts that cue up
relations. The optimal format of texts depends
on the properties of the audience. For example,
readers with relevant background knowledge
or good reading skills remember more from a
somewhat incoherent text than from a very co-
herent text, whereas readers with little background
knowledge or poor reading skills show the op-
posite pattern (9, 12).
The reader’s attention is influenced by the
availability of working memory and of relevant
background knowledge (12). Motivational factors
such as the goal for reading the text (9) and skill
factors such as reading and reasoning strategies
(13) also determine what attention the reader
can bring to bear on the text. These strategies
and skills enable the reader to control attentional
focus, to determine what types of relations exist
between activated elements, and to decide wheth-
er further analysis is needed before further
reading. Readers bring individual differences to
reading, differences that can be measured in eye
movements, speeded responses to probes, or
think-aloud protocols that can predict infer-
ence generation (14, 15) (Fig. 2).
Standards of coherence. The tension between
limited attentional resources and the need for
coherence is negotiated by the reader’s stan-
dards of coherence, which reflect the reader’s
understanding of what sorts of relations are
needed to comprehend the text (9). These stan-
dards influence when and with what strategies
the reader will try to establish coherence (16).
Strategies invoked might be to reactivate in-
formation from earlier text, to slow down, or to
search background knowledge. In a particular
454 23 APRIL 2010
reading situation, a reader’s standards depend on chronological relations in a history text or spa-
knowledge of standards in general, the reading tial relations in a geometry book) or if the
goals, and the type of text. Different types of reader’s interests focus on those relations [for
text may depend on different types of relations example, a homebuyer and a burglar identify
(for example, some types of relations may be different relations when reading a description of
more central to narrative than to informational a house (19)].
texts and vice versa). Relations may be “signaled” by the text. In
Although many types of relations can exist the example, the implicit causality conveyed by
in a text, two types of relations crucial for just the verb “contributed” directs the reader toward
detecting a particular semantic rela-
tion; similar functions can be per-
formed by other means such as
the use connectives (for instance,
“thus”). The example also illus-
trates that the identification of a
relation often requires background
knowledge. The text indicates that
a causal relation exists, but without
background knowledge (for in-
stance, the clouds caused by the
explosion may lead to a drop in
temperature) the exact nature of
the relation will elude the reader.
Indeed, having or recruiting the
wrong background knowledge
(for instance, that the explosion
killed by direct impact) may inter-
fere with proper comprehension.
Recruiting incorrect information can
interfere with detection of relations.
Texts that offer relations sep-
arated by paragraphs or pages or
Fig. 2. Cognitive processes during reading can be investigated that require coordination of multiple
by eye-tracking devices. [Photo courtesy of S. Carlson, University pieces of information, features com-
of Minnesota] mon to informational texts, are
particularly challenging for readers.
about any reading situation are referential and As a reader progresses through a text, a
causal/logical relations (17, 18). Referential landscape of activations unfolds. In successful
relations provide connections of identity, relating comprehension, relations are detected between
entities (e.g., persons, objects, concepts) across coactivated concepts, leading to the gradual
sentences, whereas causal/logical relations indi- emergence of a mental representation of the
cate how facts and events described in one text, relevant background knowledge, and their
sentence cause or lead to facts and events in semantic relations. Most texts outpace most
another sentence. Consider, for example, the readers’ attentional resources. Readers with ef-
following pair of sentences: fective and efficient strategies for selective al-
The explosion caused by the giant meteor location of attention are more likely to identify the
was enormous. relations that matter, and texts that offer helpful
It contributed to the extinction of many guideposts optimize their ability to engender
species. comprehension.
In the second sentence, the pro-
noun “it” refers to “explosion” in the
first sentence, establishing referential
coherence. The event described in the
second sentence, the extinction of spe-
cies, is (at least partly) caused by the
fact described in the first sentence, the
explosion and its aftermath.
Referential and causal/logical rela-
tions such as these are common in most
texts. Other types of relations may also
be recognized if the text makes them
particularly salient or they are uniquely
important to the topic (for example,
VOL 328 SCIENCE www.sciencemag.org

Learning Science from Text
The processes involved in comprehending while
reading a text have direct bearing on the reader’s
learning from the text. Both when reading a text
and when learning, informational elements and
their relations are incorporated in a representa-
tion of the text (in the case of comprehension)
and of background knowledge (in the case of
learning).
Expanding the reader’s knowledge. One form
of learning from text is the addition of new in-
formation to a reader’s background knowledge.
The “new information” may be informational
elements (facts, events, or concepts) or may be
new relations between elements that the reader
already knew. The design of science texts affects
the likelihood that new concepts or relations can
be added to the reader’s background knowledge.
Texts that promote effective attention allocation
support learning. Relations are more easily de-
tected if the to-be-connected information is
presented close together in the text, which may
entail repeating previously stated information
(20). Relation construction can be facilitated by
linguistic and other markers. In the case of ref-
erential relations, for example, consistent use of
the same terms (rather than synonyms) for the
same concepts and unambiguous use of pro-
nouns can assist the reader. For causal/logical
relations, the use of connectives such as “thus,”
“as a result of,” and so on facilitates relation
identification. Text signals such as headers,
underlining, italics, and so on likewise direct
attention (21). Finally, the use of straightforward
grammar and familiar vocabulary can minimize
the amount of attention that the reader needs to
divert from learning.
These aspects of the design of a science
text must be balanced by the purpose of the
text and the practical constraints of the pub-
lication medium. For example, if the purpose
is to expand a reader’s knowledge of various
nomenclatures for brain regions or gene types,
then referential simplification may be counter-
productive. With regard to practical constraints,
where space is at a premium (as in many sci-
entific journals), repetition of concepts to fa-
cilitate relation identification may be a limited
option. In such cases, one must rely on the
motivation and prior knowledge of the readers
to allow them to handle the increased process-
ing demands.
Modifying the reader’s knowledge. A com-
mon, albeit less obvious, form of learning from
text entails the modification of existing knowl-
edge. A reader’s misconceptions are especially
problematic. Not only do new elements and
relations need to be added to the reader’s knowl-
edge base but conceptual change also needs to
take place. Existing elements and relations need
to be removed or adjusted (22). Modification of
existing knowledge is often more difficult than
acquisition of new knowledge (11, 23). One
www.sciencemag.org
strategy is to present the correct information and
assume the reader will adjust any misconcep-
tions. Unfortunately, misconceptions are resil-
ient to this kind of strategy. Instead, conceptual
change is most likely when the correct and the
incorrect information are pre-
sented conjointly and identi-
fied as such (11, 23).
Learning from text may
entail both expansion and
modification of the reader’s
knowledge base (Fig. 3). The
processing of a particular text
by a particular reader is the re-
sult of an interaction between
text and reader properties. No
text is perfect for all readers.
Coherent versus correct
representations. The estab-
lishment of coherence is a
necessary step toward learning
from a text. However, coherence—the perception
that the text “hangs together” and is consistent
with one’s background knowledge—does not en-
sure that correct learning will take place, because
it is possible for a reader to create an internally
coherent representation that nevertheless is an in-
accurate representation of the facts (22).
Examples from the history of science are
heliocentrism and electrical current as the flow
of charged particles. Thus, the student’s assess-
ment of understanding serves as an unreliable
2 24
1
4 5
3 attain the desired goal of increasing motivation,
7
6 8
9 “seductive detail” effect. This undesired effect can
10
11 12
13
14
15
18
20 19 17
23 21 22
Fig. 3. Learning from text. Expansion and mod-
ification of a hypothetical reader’s knowledge
structure. Represented are correct prior knowl-
edge (black), newly acquired knowledge (blue),
and corrected misconceptions (red).
SCIENCE VOL 328
SPECIALSECTION
guide by which to judge whether or not learning
occurred as intended by the author.
Using nontextual materials to supplement
texts. Science texts are often supplemented by
nontextual materials such as figures, diagrams,
movies, or hands-on experience.
Sometimes such materials are
used to draw the reader into the
text or to lessen the formidabil-
ity of a dense page of text. At
other times, they are used to
support directly the expansion
and modification of the reader’s
knowledge—often interrupt-
ing the processing of the text.
To support learning, supple-
mental materials and activities
must increase the likelihood of
simultaneous activation of the
to-be-related information, there-
by supporting the establishment
of important relations. When used judiciously,
supplemental materials and activities can greatly
assist science learning. However, materials and
activities that are more of a distraction than an
enhancement will exact a cost on comprehen-
sion. This is likely to occur, for example, when
the simultaneous processing of information from
multiple input channels increases cognitive load
or when the nontextual information is irrelevant
to the meaning of the text, as may happen when
pictures are included solely to increase the attrac-
tiveness of the text.
Seductive details. Another frequently used
method to increase a reader’s motivation to read
the science text at hand is to provide an anecdote
designed to elicit the personal interest of the
reader in the topic of the text. As with the use of
nontextual materials, such anecdotes may indeed
but they pose a risk if precious attentional re-
sources flow to processing the motivating infor-
mation and away from the conceptually central
information (24, 25), a phenomenon known as the
be diminished by limiting and demarcating
anecdotes.
Learning science from texts, learning reading
16 from science texts. The emphasis in this article
is on the transmittal of knowledge from science
texts to the reader. Concurrently, the act of read-
ing contributes to the development of reading
comprehension skills. The attention allocation
and inferential skills that skilled readers em-
ploy during reading are enhanced by exposure
to texts that vary in structure, difficulty, and con-
tent area. Science texts differ greatly from nar-
ratives in their demands on working memory
management, comprehension strategies, and
the use of background knowledge. Thus, prac-
tice with science and narrative texts broadens
the cognitive “toolbox” that a reader can bring
to any text.
23 APRIL 2010 455
 Science, Language, and Literacy
Unresolved Issues
Investigation of cognitive processes during read-
ing has provided important insights into con-
ditions that promote or hinder the acquisition of
new knowledge from science texts. However,
many unresolved issues remain. One such issue
is that mere coactivation of elements in working
memory cannot fully account for the learning of
all scientific concepts, many of which are chains
of conceptual relations too long or complex to
reside in working memory under the best of cir-
cumstances. It seems likely that direct relations
between a few concepts accumulate into clusters
of knowledge characterized by conceptual prox-
imity. A second issue concerns the conditions
that determine whether textual information over-
rides a misconception that the reader holds.
What does it take for a single text to modify a
reader’s prior knowledge? A third issue con-
cerns the variety of types of texts. As noted,
much reading research has been executed in the
realm of narrative texts, less with informational
texts (26). It is likely that the essential toolbox
of cognitive processes (working memory ca-
pacity, comprehension strategies, and standards
of coherence) applies to all types of text, but the
specific implementation may vary as a function
of text type.
Using Texts to Teach Science
Texts are frequent and powerful tools for con-
veying scientific facts, principles, and explana-
tions. To be effective, however, science texts
need to be designed to optimize the likelihood
that learning will occur. Central to comprehen-
sion of and learning from science texts is the
identification of relations among the elements
in the text and between these elements and the
reader’s prior knowledge, processes that occur
while reading. Optimally designed science texts
direct the reader’s landscape of activations dur-
ing reading in such a way that elements that
should be connected do indeed get connected.
When that happens, science texts are among the
most effective tools we have available to teach
science, to expand readers’ knowledge of scien-
tific topics, and to correct misconceptions.
References and Notes
1. A. C. Graesser, M. Singer, T. Trabasso, Psychol. Rev. 101,
371 (1994).
2. W. Kintsch, Psychol. Rev. 95, 163 (1988).
3. E. J. O’Brien, J. L. Myers, in Narrative Comprehension,
Causality, and Coherence: Essays in Honor of Tom Trabasso,
S. R. Goldman, A. C. Graesser, P. W. van den Broek, Eds.
(Erlbaum, Mahwah, NJ, 1999), pp. 35–53.
4. T. Trabasso, T. Secco, P. W. van den Broek, in Learning and
Comprehension of Text, H. Mandl, N. L. Stein, T. Trabasso,
Eds. (Erlbaum, Mahwah, NJ, 1984), pp. 83–111.
5. P. W. van den Broek, M. Young, Y. Tzeng, T. Linderholm,
in The Construction of Mental Representations During
Reading, H. van Oostendorp, S. R. Goldman, Eds.
(Erlbaum, Mahwah, NJ, 1999), pp. 71–98.
6. R. A. Zwaan, D. N. Rapp, in Handbook of
Psycholinguistics, M. A. Gernsbacher, M. J. Traxler, Eds.
(Elsevier, San Diego, CA, 2006), pp. 725–764.
456 23 APRIL 2010
7. W. Kintsch, Comprehension: A Paradigm for Cognition. 17. P. W. van den Broek, in Handbook of Psycholinguistics,
(Cambridge Univ. Press, Cambridge, 1998). M. A. Gernsbacher, Ed. (Academic Press, London, 1994),
8. The Landscape Model (computational implementation pp. 539–588.
available at www.landscapemodel.leiden.edu). 18. J. A. León, G. E. Peñalba, in The Psychology of Science
9. T. Linderholm et al., Cogn. Instr. 18, 525 (2000). Text Comprehension, J. Otero, J. A. León, A. C. Graesser,
10. P. W. van den Broek, S. Virtue, M. Everson, Y. Tzeng, Eds. (Erlbaum, Mahwah, NJ, 2002), pp. 155–178.
Y. Sung, in The Psychology of Science Text Comprehension, 19. J. W. Pichert, R. C. Anderson, J. Educ. Psychol. 69, 309
J. Otero, J. Leon, A. C. Graesser, Eds. (Erlbaum, (1977).
Mahwah, NJ, 2002), pp. 131–154. 20. R. F. Lorch Jr., J. Exp. Psychol. Learn. Mem. Cogn. 19,
11. P. W. van den Broek, P. Kendeou, Appl. Cogn. Psychol. 1071 (1993).
22, 335 (2008). 21. J. Hyönä, R. F. Lorch, Learn. Instr. 14, 131
12. D. S. McNamara, E. Kintsch, N. B. Songer, W. Kintsch, (2004).
Cogn. Instr. 14, 1 (1996). 22. S. Vosniadou, in Intentional Conceptual Change,
13. P. W. van den Broek, M. J. White, P. Kendeou, S. Carlson, G. M. Sinatra, P. R. Printrich, Eds. (Lawrence Erlbaum,
in Beyond Decoding: The Behavioral and Biological Mahwah, NJ, 2003), pp. 377–406.
Foundations of Reading Comprehension, R. Wagner, 23. P. Kendeou, P. W. van den Broek, Mem. Cognit. 35, 1567
C. Schatschneider, C. Phythian-Sence, Eds. (Guilford, (2007).
New York, 2009), pp. 107–123. 24. G. Schraw, J. Educ. Psychol. 90, 3 (1998).
14. D. N. Rapp, P. W. van den Broek, K. L. McMaster, 25. R. F. Lorch Jr., E. P. Lorch, M. A. Klusewitz, Contemp.
P. Kendeou, C. A. Espin, Sci. Stud. Read. 11, 289 (2007). Educ. Psychol. 20, 51 (1995).
15. K. Cain, J. Oakhill, in Children’s Comprehension Problems 26. S. R. Goldman, G. L. Bisanz, in The Psychology of Science
in Oral and Written Language: A Cognitive Perspective, Text Comprehension, J. Otero, J. A. León, A. C. Graesser,
K. Cain, J. Oakhill, Eds. (Guilford, New York, 2007), Eds. (L Erlbaum, Mahwah, NJ, 2002), pp. 19–50.
pp. 41–76. 27. I thank P. Kendeou, R. F. Lorch Jr., and two anonymous
16. J. Dunlosky, K. A. Rawson, D. J. Hacker, in The Psychology reviewers for their thoughtful comments.
of Science Text Comprehension, J. Otero, J. A. León,
A. C. Graesser, Eds. (Lawrence Erlbaum, Mahwah, NJ,
2002), pp. 255–280. 10.1126/science.1182594
REVIEW
Supporting Students in Developing
Literacy in Science
Joseph S. Krajcik1,2*† and LeeAnn M. Sutherland1*
Reading, writing, and oral communication are critical literacy practices for participation in a
global society. In the context of science inquiry, literacy practices support learners by enabling
them to grapple with ideas, share their thoughts, enrich understanding, and solve problems. Here
we suggest five instructional and curricular features that can support students in developing
literacy in the context of science: (i) linking new ideas to prior knowledge and experiences, (ii)
anchoring learning in questions that are meaningful in the lives of students, (iii) connecting
multiple representations, (iv) providing opportunities for students to use science ideas, and (v)
supporting students’ engagement with the discourses of science. These five features will
promote students’ ability to read, write, and communicate about science so that they can
engage in inquiry throughout their lives.
ystematic investigation of meaningful ques- in decision-making that is personal or that af-
S tions about natural phenomena and the
development of evidence-based explana-
tions form the foundations of science inquiry (1, 2).
fects others in a global community. In addition,
the Standards state that scientific literacy re-
quires the ability to critique the quality of evi-
In classrooms that emphasize such inquiry, funda- dence or validity of conclusions about science
mental literacy practices such as reading, writing, in various media, including newspapers, mag-
and oral discourse are essential to developing azines, television, and the Internet. The Amer-
an understanding of the core ideas of science. ican Association for the Advancement of Science
The National Science Education Standards stresses the importance of scientific literacy for
(1) define scientific literacy as the understanding citizens’ ability to participate with others in a
of science content and scientific practices and global society (3). Underlying all of these def-
the ability to use that knowledge to participate initions is the understanding that students must
read, write, and communicate effectively to make
1
School of Education, University of Michigan, 610 East decisions as informed citizens and engage in
University Avenue, Ann Arbor, MI 48109–1259, USA. 2Ewha the critical thinking that active science learning
Womans University, Institute for Global Science, Technology
and Society Education, Seoul, South Korea. requires. We have selected five aspects from the
*These authors contributed equally to this work.
many features of literacy that are important to
†To whom correspondence should be addressed. E-mail: embed in inquiry science: (i) linking new ideas
krajcik@umich.edu to prior knowledge and experiences, (ii) anchor-
VOL 328 SCIENCE www.sciencemag.org