Instr Sci

DOI 10.1007/s11251-008-9090-5

Prior knowledge activation: how different concept

mapping tasks lead to substantial differences in cognitive
processes, learning outcomes, and perceived self-efficacy

Johannes Gurlitt Æ Alexander Renkl

Received: 3 April 2008 / Accepted: 16 December 2008

Springer Science+Business Media B.V. 2009

Abstract Two experiments investigated the effects of characteristic features of concept

mapping used for prior knowledge activation. Characteristic demands of concept mapping
include connecting lines representing the relationships between concepts and labeling these
lines, specifying the type of the semantic relationships. In the first experiment, employing a
within-subjects design, 20 psychology students completed a label-provided-lines eco-
nomics mapping task and then a create-and-label-lines meteorology mapping task or vice
versa. The analysis of 40 think-aloud protocols indicated more elaboration processes for
the label-provided-lines task than for the create-and-label-lines task. On the other hand, the
protocols indicated more model-construction and organization processes in the create-and-
label-lines task. The second experiment used the same variation but focused on learning
outcomes and perceived self-efficacy as dependent measures. Forty-two psychology stu-
dents were randomly assigned to either a label-provided-lines mapping task or a create-
and-label-lines mapping task. Subsequently, both groups completed a learning phase in a
hypertext environment and a posttest. Results showed substantial differences in learning
outcomes and perceived self-efficacy in favor of the label-provided-lines prior knowledge
activation task. The findings are congruent with coherence effects found in text-compre-
hension research and support the position that concept mapping should not be seen as a
unitary method but be differentiated according to the specific tasks to be completed.

Keywords Prior knowledge activation
Concept mapping
Coherence

Mental set
Hypertext

Is prior knowledge activation only the activation of specific concepts in long-term memory
or does the specific task used for prior knowledge activation matter? Is concept mapping a
unitary instructional method, or does a variation of characteristic task features make a
substantial difference for cognitive and metacognitive processes, learning outcomes, and

J. Gurlitt (&)
A. Renkl

University of Freiburg, Freiburg, Germany
e-mail: johannes.gurlitt@psychologie.uni-freiburg.de

123
 J. Gurlitt, A. Renkl

perceived self-efficacy? Shedding more light on these important questions, the following
studies examine whether prior knowledge activation with high- versus low-coherence
mapping tasks elicits different processes, learning outcomes, and self-efficacy beliefs. In
the following, we first outline the importance of prior knowledge and then turn to the
question of how prior knowledge can be activated using an instructional method that
focuses on the macro-structure of the contents to-be-learned.

The importance of prior knowledge for learning

Prior knowledge is conceptualized as the learner’s content knowledge related to the

domain studied, which is present before the implementation of a particular instruction (for
a detailed discussion and conceptualization of prior knowledge see Dochy 1992). Research
suggests that prior knowledge is one of the most important prerequisites for learning (e.g.,
Ausubel 1968; Schneider et al. 1989; Weinert and Helmke 1998). Investigating the role of
domain-specific prior knowledge, Weinert and Helmke showed that correlations between
prior knowledge and mathematics performance remained strong, even when intelligence
was partialed out. Their longitudinal study also showed that the influence of prior
knowledge on performance improvement increased, while the influence of intelligence as a
source of performance differences decreased. Schneider et al. showed that prior knowledge
compensated for low aptitude, while high aptitude could not compensate for low prior
knowledge. On the basis of dozens of studies, Dochy (1992) concluded that a learner’s
prior knowledge overrules all other variables and explains between 30% and 60% of the
variance in learning outcomes. Thus, he emphasized that more attention should be paid to
this crucial prerequisite of learning.
Prior knowledge plays a prominent place in the theory of multimedia learning (Mayer
1997), in the construction–integration theory of text comprehension (Kintsch 1988), in the
constructivist theory about narrative text comprehension (Graesser et al. 1994), and in the
assimilation theory of meaningful learning (Ausubel 1968). In the context of multimedia
learning, Mayer (1997) proposed that meaningful learning occurs when students select,
organize, and integrate relevant visual and verbal information. On a general level, prior
knowledge can focus the learner’s attention on relevant information and helps to avoid
distractions, thereby saving information-processing resources. Prior knowledge also pro-
vides a framework through which new information may be organized and assimilated. This
reduces the amount of information chunks to be recalled and provides association cues for
accessing information from the long-term memory. Furthermore, an organized and inte-
grated knowledge base allows for the reconstruction of information that cannot be directly
retrieved.
The integration of new information into prior knowledge also plays a prominent role in
the construction–integration theory of text comprehension (Kintsch 1988). In this theory,
comprehension is achieved through different processes. Starting with character and word
decoding and the subsequent construction of propositions stated in the text, links are formed
between nodes, for example, due to shared elements of different propositions. Related
information as well as schemata may also be activated automatically by the explicit content
of the text (Kintsch 2005). During a proposed integration phase, activated knowledge
components serve as constraints in a spreading activation process where highly intercon-
nected nodes accumulate activation, whereas less well-connected nodes lose activation and
may drop from the network. Previously activated schemata may function as control units,
resolving contradictions or ambiguities. Therefore, the construction–integration model

123
Prior knowledge activation

claims that prior knowledge is activated more or less automatically through words and
propositions in the text. The integration phase, however, explicitly considers the initial
activation of certain nodes in earlier cycles. This latter assumption suggests that there might
be some value of additional instructional prior knowledge activation in the sense that it
increases the a priori activation of specific nodes.
The constructivist theory about narrative text comprehension (Graesser et al. 1994)
assumes that specific inferences (e.g., referential, causal antecedent, superordinate, the-
matic), are constructed online and under most conditions of comprehension. The core of
this theory is the ‘‘search after meaning principle’’: Readers have goals and aim to con-
struct a coherent and meaningful representation, which addresses these goals and explains
why actions and events mentioned in the text occur. At a first glance, this may be taken as
theoretical evidence against the value of instructional prior knowledge activation. How-
ever, prior knowledge may play an important role in constraining the goals to be pursued
during reading. While some goals may be activated more or less automatically, especially
in narrative text comprehension, others may be strengthened or established by instructional
prior knowledge activation.
While all the previously mentioned theories more or less acknowledge the importance
of prior knowledge, only the assimilation theory of meaningful learning (Ausubel 1968)
places prior knowledge at its core. According to this theory, meaningful learning occurs
when learners connect new information to the existing knowledge structure. This learner-
controlled process is called assimilation. According to this theory, meaningful learning has
at least three components—the learners’ prior knowledge, meaningful material, and the
learners’ intent and ability to use their prior knowledge.

Concept maps as an instructional method

Although Ausubel (1968) emphasized the importance of prior knowledge and developed
the idea of the advance organizer, he did not provide educators with simple functional tools
to assess and activate prior-knowledge (Novak and Gowin 1984). Based on Ausubel’s
work, Novak and Gowin recommended concept maps as a tool for getting students to
examine their prior-knowledge before studying new materials. Concept maps are diagrams
that represent ideas as node-link assemblies (Nesbit and Adesope 2006). Concept maps
provide an external network-like representation of knowledge structures and consist of
spatially grouped nodes with key words representing concepts, connecting lines repre-
senting the semantic connection of concepts, and labels on the lines to specify the kind of
the semantic relation. Many studies have examined the benefits of studying concept maps,
comparing them to traditional text, or the benefits of using concept maps as a follow-up
strategy in learning from text as compared to using traditional study techniques (e.g.,
outlining). Results indicate that maps facilitate learning in a variety of instructional con-
ditions including different topics and educational levels (for a recent meta-analysis see
Nesbit and Adesope 2006). One of the main benefits of maps is highlighting the learning
topic’s macrostructure (for review, see O’Donnell et al. 2002).
Concept maps may be created entirely by the student. Alternatively, educators may
prepare incomplete maps and leave specific activities such as creating and labeling con-
necting lines for the learners. Chang et al. (2001) compared computer-based ‘free-
construction’ and ‘construct-on scaffold,’ as well as ‘paper-pencil free-construction’
mapping tasks. The ‘free-construction’ (in the original article called ‘construct-by-self’)
groups had to construct concept maps by themselves without construction scaffolding.

123
 J. Gurlitt, A. Renkl

The ‘construct-on scaffold’ group received an incomplete expert concept map within
which some nodes and links had to be supplemented. All groups were provided with a
concept list and a relation list, which contained all the concepts and relations to be used.
Both computer-based groups contained a system-based evaluation with corresponding
hints for feedback. This automated map evaluation was performed on demand (pressing an
evaluation button) and compared the constructed concept map with an expert concept map,
providing a similarity score as well as indicating what nodes and links differed. The
posttest showed that the computer based ‘construct-on-scaffold’ group outperformed the
computer based ‘free-construction’ and ‘paper–pencil free-construction’ group, with no
differences between the computer based and the paper based ‘free-construction’ groups.
Thus, the findings showed that different mapping tasks lead to differences in learning
outcomes. The results of Hauser et al. (2006) also confirmed this conclusion.
Research about concept mapping tasks used for assessment indicates that different
mapping tasks evoke different cognitive processes (Ruiz-Primo et al. 2001). Exploring the
validity of different concept mapping tasks for science assessment, Ruiz-Primo et al.
compared think-aloud protocols of eight participants in a repeated measurement design. In
the first task, learners constructed a map from scratch using provided concepts; in the
second task, learners filled in the nodes of a skeleton map with concepts provided; in the
third task, learners filled in the lines on a skeleton map with a list of the relations provided
for each concept pair. Results indicated that the three mapping tasks provide different
pictures of the learners’ knowledge. It seems fruitful to build on this idea in subsequent
research by increasing the sample size and using different topics.
In an earlier, but related review, Ruiz-Primo and Shavelson (1996) encouraged more
research about cognitive and metacognitive effects of different concept mapping tasks, and
also underlined the lack of an integrative cognitive theory to limit the variation in concept-
mapping techniques. Similarly, pointing out high-priority areas for further research, Nesbit
and Adesope (2006) concluded that research should investigate the processes by which
students learn with concept maps.
Although a new integrative cognitive theory is beyond the scope of this article, we
propose a bridging reference to research about text comprehension, more specifically text
coherence, to theoretically elucidate relevant processes. While ‘‘text coherence is the extent
to which the relationships between ideas in a text are explicit’’ (McNamara 2001, p. 51),
map coherence can be defined as the extent to which relationships between the concepts in
the map are made explicit. To built on this proposed relation, the following paragraph
briefly exemplifies research on coherence effects in text-comprehension research.
One of the first experiments investigating coherence effects was conducted by Britton
and Gulgoz (1991). After identifying the location of coherence gaps in the text, using
Kintsch’s computational model (Miller and Kintsch 1980, based on Kintsch and van Dijk’s
1978, theory of text comprehension) they increased the coherence of a textbook chapter by
inserting linking words for better argument overlap, making important references explicit,
and rearranging sentences in a way that learners encountered information they already
knew prior to the new information to be connected. Thus, they eliminated the need for the
readers to make the basic connections by themselves by establishing more ordered and
explicit links between the arguments in the text. Results showed that the undergraduate
students’ performance in free recall and inference questions was better after reading the
revised high-coherence text compared to reading the low-coherence text. Similar results
showing the positive impact of high-coherence text were obtained by related studies such
as McKeown et al. (1992) and Ainsworth and Burcham (2007). Others indicated that only
texts designed to facilitate causal-explanatory inferences improved inferential learning

123
Prior knowledge activation

outcomes. Vidal-Abarca et al. (2000) used an 8th grade school text on history. The first
revised text was constructed similarly to the above-mentioned procedure, by inserting
linking words for better argument overlap, but without adding any explanatory information
to clarify causal relationships. Their second revised text was designed to facilitate causal
inferences by providing additional information to trigger the reader’s causal antecedent and
superordinate goal inferences. Results showed, that only the text designed to facilitate
causal inferences enhanced learning outcomes for inference questions. Surprisingly, a third
revised text that combined argument overlap with causal explanatory information only had
an effect on recall and not inference questions. Similar results showing the positive impact
of facilitating causal-explanatory inferences through an increase of causal coherence were
also obtained by Gilabert et al. (2005).
Showing that the advantages of text coherence are not uniform, but depend upon the
characteristics of learners, McNamara et al. (1996) showed that high-knowledge readers
learned more after reading a low-coherence text, while low-knowledge readers benefited
more from a high-coherence text (see also McNamara and Kintsch 1996). In a recent study,
O’Reilly and McNamara (2007) showed that high-coherence texts were better for low-
knowledge readers and high-knowledge readers with good learning-strategy skills, whereas
only high-knowledge readers with poor learning-strategy skills benefited more from the
low-coherence text.
Summarizing this research about text comprehension, there is a tendency that increasing
coherence increases learning outcomes. However, research by O’Reilly and McNamara
(2007) showed that in some cases, such as high-knowledge readers with poor learning-
strategy skills, learners may benefit more from low-coherence text. In addition, research by
Vidal-Abarca et al. (2000) indicates that facilitating causal inferences is particularly
important.
While it is beyond the scope of this article to investigate the effects of all the above
mentioned variations for high-coherence versus low-coherence mapping tasks, our first
experiment investigates, as a first step, cognitive and metacognitive effects for learners
with low to medium prior knowledge (e.g., psychology students in the domain of eco-
nomics or meteorology). The second experiment builds on the first experiment, using the
same variation but focusing on learning outcomes and perceived self-efficacy as dependent
measures. Thus, both experiments explore in detail the effects of low-coherence versus
high-coherence prior knowledge activation with concept maps on cognitive and meta-
cognitive processes, learning outcomes, and self-efficacy.

Experiment 1

The first experiment is a think-aloud study to investigate how characteristic affordances of

concept maps may be used for prior knowledge activation. Characteristic affordances of
concept maps are connecting lines representing the relationships between concepts and
labels of these lines specifying the type of the semantic relationships. As outlined in the
previous section, map coherence can be defined as the extent to which relationships
between the concepts in the map are made explicit. Therefore, we regard the mapping task
of creating and labeling lines as low-coherence and the mapping task of labeling provided
lines as high-coherence prior knowledge activation.
As mentioned in the introduction about the importance of prior knowledge for learning,
integrative processes to relate new information to concepts, experiences, or schemas
already stored in memory are key processes of new learning (see also Weinstein and Mayer

123
 J. Gurlitt, A. Renkl

1986). Therefore, the so-called elaboration processes should play a crucial role for prior
knowledge activation tasks and deserve special attention in the to-be-collected think-aloud
protocols. Organization and negative monitoring are other processes found to be useful in
learning strategy research (Weinstein and Mayer 1986). Organization processes may be
used to cluster and reduce the new knowledge of the key features and to relate new
knowledge pieces to each other, while monitoring concerns processes in which learners
‘‘look at themselves over their own shoulder’’ and find out what they know or understand
and what they do not know or understand. Unfortunately, these processes often do not
occur spontaneously, therefore it seems to be fruitful to investigate possible effects elicited
through high-coherence versus low-coherence prior knowledge activation. In addition,
differences in these processes would provide support for the hypothesis that concept
mapping should be differentiated, based on the specific demand of the task.
Based on these considerations, we addressed the following research question: Are
elicited cognitive and metacognitive processes different for high-coherence versus low-
coherence mapping tasks? Considering that acceptance of an instructional method is an
important factor for future learning, we are also interested in the following question: How
do learners rate their acceptance of each mapping task for prior knowledge activation?

Method

Participants and design

Twenty German psychology students (age M = 26.5 years) volunteered to participate in

this study. As a reward for participating, the students could choose between 7 EUR or the
mapping-software used in this study (Easy-Mapping Tool, see www.cognitive-tools.com).
The students had not studied economics or meteorology and were therefore considered to
be low to medium knowledge participants.
We used a protocol analysis method in a one-factorial repeated measurement design.
The within-subject factor ‘‘prior knowledge activation task’’ included a low-coherence and
a high-coherence prior knowledge activation condition. Learners completed either a high-
coherence economics mapping task (see Fig. 1) and then a low-coherence meteorology
mapping task (as in Fig. 1 but without connecting lines) or vice versa. To control for
sequence-effects of the topic and mapping task, four types of maps were created. The task
and topic order were counterbalanced across the participants. Participants were randomly
assigned to conditions. The different domains were chosen to minimize possible carry-over
effects: learning about economics should not influence learning about meteorology to a
substantial degree or vice versa.

Procedure and materials

First, the participants received an introduction familiarizing them with the mapping tasks.
To introduce a meaningful setting, we asked participants to imagine being in a meteo-
rology class or economics class (depending on the mapping topic) and to imagine that,
before the lesson started, they were given the opportunity to reflect on what they know and
do not know about the respective mapping task. Participants were asked to think aloud
while they worked on their mapping task. If participants did not think aloud for more than
5 s, the experimenter would prompt them to continue talking. The whole experiment lasted
about 50 min while each of the two mapping tasks took 10 min.

123
Prior knowledge activation

Export Wage Agreements

Euro-Strength
Oil Price Oil Price

Inflation Revenue of Companies Consumption Consumer Confidence

Interest Rate Bond Rate NYSE

Fig. 1 Map on economic relationships used for the high-coherence prior knowledge activation. [Translated
by Johannes Gurlitt]

Instruments and measures

Since we investigated prior knowledge activation effects, we did not use a pretest. Such a
test would have also been a type of prior knowledge activation interfering with the
experimental variation.
Think-aloud protocols were transcribed and separated into verbal units (Ruiz-Primo
et al. 2001). One verbal unit was defined as one coherent semantic block, that is, thinking
about a single concept. Verbal units were then rated blind to the experimental conditions
for elaboration, organization, model construction, definitions, and negative monitoring (see
Table 1). Elaboration was defined as processes that connect new knowledge with existing
knowledge structures (Weinstein and Mayer 1986). Organization was defined as thinking
processes dealing with relations between two concepts. Model construction was defined as
drawing a conclusion, or relating more than two concepts to gain an understanding of the
system. Verbal units were rated as definitions if learners clarified the meaning of a word.
Negative monitoring was applied to verbal units in which the learners ‘‘looked themselves
over their shoulder’’ and explained what they did not know or understand. Twenty-five
percent of the protocols were co-rated by a second rater (10 of the 40 protocols). Interrater
agreement was high (Cohen’s Kappa was .97 for elaboration, .84 for organization, .78 for
definition, .74 for model construction, and .90 for negative monitoring). Thus, only one
rater coded the rest of the protocols.

Results and discussion

We used a one-factorial repeated measures ANOVA to analyze the data. An alpha level of
.05 was used for all statistical tests. As an effect size measure, we used partial g2, quali-
fying values \.06 as small effects, values in the range between .06 and .13 as medium
effects, and values[.13 as large effects (see Cohen 1988). Table 2 provides an overview of
the results.
ANOVA revealed that the label-the-lines task (high-coherence prior knowledge acti-
vation) elicited significantly more elaborations on how two concept were interrelated in the
think-aloud protocols than the create-and-label-the-lines task, F(1,19) = 6.57, p \ .021,

123
 J. Gurlitt, A. Renkl

Table 1 Categories used to rate think aloud protocols

Category Description Examples from think aloud protocols

Elaboration Elaborations are processes that connect ‘‘Rising oil price has a negative influence
new knowledge with existing on the export of most companies. Or,
knowledge structures. Indications of yes, let me think, if you need more
elaborations are the generation of money for the production of some
examples, the imagination of a picture, plastic then the product will be more
or other cognitive processes that expensive and it will be more difficult
enhance the given information to sell it on the international markets.
(Weinstein and Mayer 1986). Yes, this has a negative influence.’’
Organization These strategies are used to cluster and ‘‘The higher the temperature, the higher
reduce the new knowledge of the key the evaporation.’’
features and to relate new knowledge
pieces to each other (Weinstein and
Mayer 1986). As the task did not
include the rearrangement or reduction
of the learning content, we
concentrated on the second aspect:
thinking processes dealing with
relations between two concepts.
Definitions The learner clarifies the meaning of a ‘‘NYSE, what is NYSE, New York Stock
word. Exchange.’’
Model construction Do learners draw a conclusion; do they ‘‘It seems that many paths lead to the cash
relate more than two concepts and gain flow. The cash-flow seems to be related
an understanding of the ‘‘larger with almost all other concepts. So, this
picture’’? is the view of the business-owner …’’
Negative Negative monitoring concerns processes, ‘‘For example, I’m not sure about the
monitoring where the learner ‘‘looks her/himself concepts interest rate and bond rate, but
over her/his shoulder’’ and finds out, I’m sure I did hear them before. Still, to
what she/he did not know, or where relate them is very difficult for me.’’
she/he is not sure.

Table 2 Means and standard deviations (in parentheses) of dependent variables

Label provided Create and F-valuea g2 p
lines label lines

Elaboration 7.1 (3.42) 5.2 (3.70) 6.57 .25 .021*

Organization 21.9 (5.11) 25.1 (7.15) 6.04 .24 .024*
Definitions 1.65 (1.56) 1.65 (1.46) .00 .00 [.99
Model construction 2.1 (2.63) 3.6 (4.45) 5.51 .23 .030*
Negative monitoring 11.1 (4.49) 12.7 (5.86) 1.20 .06 .287
Acceptance (6-point rating-scale) 4.7 (0.97) 4.8 (0.72) .01 .00 .915
Favorite mapping task 10 10 – – –
(number of participants)
a
df = 1,19; all tests were two-sided, N = 20
* Statistically significant

g2 = .25 (large effect). However, ANOVA also revealed that the create-and-label-the-lines
task (low-coherence prior knowledge activation) elicited significantly more organizational
and model construction processes referring to whether two or more concepts are

123
Prior knowledge activation

interrelated than the label-the-lines task, F(1,19) = 6.04, p \ .024, g2 = .24 and
F(1,19) = 5.51, p \ .03, g2 = .23 (large effects), respectively. The conditions did not
significantly differ in the number of definitions or the number of negative monitoring
statements.
Acceptance rated after each mapping task on a 6-point rating-scale (from 1 = not
helpful to 6 = very helpful) did not differ. It was 4.7 in the label-provided-lines condition,
and 4.8 in the create-and-label-lines condition. The choice of the reward for participation
indicated that participants enjoyed working with this instructional tool and that they will
use concept mapping in the future: 17 of the 20 participants chose the software (Easy-
Mapping Tool), only three participants opted for the monetary compensation (7 EUR).
Thus, for the low-coherence prior knowledge activation, participants focused on the
organization of their prior knowledge, whereas for the high-coherence prior knowledge
activation, participants focused more on their ability to justify or elaborate on the depicted
relationships. We propose the conjecture that the high-coherence task of only labeling the
lines evoked more elaborative processes because learners did not have to complete search
processes for semantic connections between concepts. On the other hand, constraining the
task in this way had the trade-off of leading to fewer organization and model construction
processes. A restriction of this study is that we did not assess learning outcomes. Hence, we
conducted a further experiment.

Experiment 2

The value of instructional prior knowledge for learning should increase the more control of
the learning process is left up to the learner, since external instructional guidance for these
processes is low. Although the first experiment showed that high-coherence versus low-
coherence mapping tasks elicit different cognitive processes, it remains an open question
whether these differences have effects on learning outcomes and perceived self-efficacy.
Thus, the second experiment uses the same variation but focuses on learning outcomes and
perceived self-efficacy as dependent measures in a self-regulated learning environment.
Hypertexts are an example of learning environments in which learners have to self-
regulate their learning to a substantial degree. Hypertexts contain information pages with
multiple links between information units (Rouet and Levonen 1996). These structures
provide ‘‘information diets’’ by allowing different paths, offering learners navigational
choices and enabling them to decide which information to access. However, hypertexts
often fall short of the expectations directed to the learners’ constructive knowledge
building (Unz and Hesse 1999; for a review see Dillon and Gabbard 1998; Shapiro and
Niederhauser 2004). Learners’ access to and control over the text with which they interact
is only beneficial if they know what information they need (Dee-Lucas and Larkin 1995;
Jonassen 1986). Otherwise, learners may stumble through the ‘‘information jungle’’ and get
disoriented or ‘‘lost in hyperspace.’’ Several studies showed the positive influence of prior
knowledge on learning outcomes with hypertext environments (e.g., Gerjets et al. 2000;
Gerjets and Scheiter 2003; Potelle and Rouet 2003). Other studies looked at the strategies
learners use (e.g., Hill and Hannafin 1997; Azevedo et al. 2004), the effects of training self-
regulated learning (e.g., Azevedo and Cromley 2004), and the value of scaffolds during the
learning phase (e.g., Shapiro 1999; Potelle and Rouet 2003; Azevedo et al. 2004). The
findings of these studies indicated that learners often do not activate relevant prior
knowledge spontaneously, have difficulties organizing related ‘‘pieces’’ of information,
and hardly plan or engage in metacognitive monitoring of what they already know and do

123
 J. Gurlitt, A. Renkl

not know. Therefore, hypertexts seem to be a type of learning environment that offers
enough ‘‘space’’ to use elicited processes for learning and to examine potential effects of
different prior knowledge activation on learning outcomes.
In addition to learning outcomes, the structure of the prior knowledge activation task
may also influence the perceived self-efficacy, an important variable for future learning. As
a judgment of one’s ability to organize and execute actions related to a specific context,
Bandura (1977) introduced self-efficacy as a key component in social cognitive theory. He
defined perceived self-efficacy as people’s beliefs in their capabilities to organize and
execute the courses of action required to produce given attainments (Bandura 1997, 2006).
Considering possible effects of low-coherence versus high-coherence prior knowledge
activation on learning outcomes, we formulate two competing hypotheses and one
exploratory research question:
1. Linking the current research about low-coherence versus high-coherence maps to
research about text comprehension (e.g., Britton and Gulgoz 1991; Ainsworth and
Burcham 2007; O’Reilly and McNamara 2007), learners most likely benefit more from
activating their prior knowledge with the high-coherence maps as compared to the
low-coherence maps.
2. Bearing in mind the results from Experiment 1 and qualitative studies about the effects
of different concept mapping tasks (see also Ruiz-Primo et al. 2001), differential
effects are possible. High-coherence prior knowledge activation may be beneficial for
the elaboration of the learning content, while low-coherence prior knowledge
activation may be advantageous to construct an understanding on the model level.
3. Considering that self-efficacy is an important variable for future learning and
application, we also ask the research question whether low- versus high-coherence
prior knowledge activation leads to differences in self-efficacy beliefs.

Method

Participants and design

This experiment was conducted at the same German university as Experiment 1. Forty-two
German psychology university students (age: M = 23.0 years) participated in this study on
learning economics. Participants received credits for participation as part of their psy-
chology classes. The students had not studied economics and were, therefore, considered to
be low to medium knowledge participants.
We used a one-factorial design. The between-subject factor ‘‘prior knowledge activation
task’’ included a low-coherence and a high-coherence prior knowledge activation condition
similar to Experiment 1. Learners were randomly assigned to one of the two conditions: (1)
the high-coherence group activated prior knowledge by a label-provided-lines task (see
Fig. 1). (2) The low-coherence group activated prior knowledge through a create-and-
label-lines task (as in Fig. 1 but without connecting lines).

Procedure and materials

Until the beginning of the learning phase, the procedure was identical to Experiment 1.
After activating their prior knowledge about economics, learners entered a learning phase
in a hypertext environment, followed by a posttest and self-efficacy ratings.

123
Prior knowledge activation

The learners studied the hypertext for 20 min. It contained definitions, relations, and
examples of economics. All concepts and relations presented in the prior knowledge
activation tasks were included as headings and in the content of the hypertext. Altogether,
the hypertext consisted of 65 short pages. The navigation bar on the top was segmented
into task, definitions, factors influencing a certain concept, specific relations, and examples.
The task segment of the hypertext described on one page the goal of the learning phase,
that is, to close knowledge gaps and to deepen existing knowledge. It also included a short
description of the navigation and described the possibility to take notes in a provided text-
field. The hypertext’s definition segment included short pages with definitions of all the
concepts that had been part of the prior knowledge activation phase. The segment about
factors influencing certain concepts explained, for example, factors influencing the oil
price. The segment about specific relations explained the relations between two concepts,
for example, the relation between bond rates and company revenues. The ‘‘example seg-
ment’’ included short newspaper articles on topics such as the Euro reaching highs against
the US$. A clock counting down from 20 min was provided inside the learning environ-
ment. One minute before the end of the learning phase, learners were reminded to come to
an end. When the time was up, learners were told to stop reading and work on the posttest.

Instruments and measures

Since we investigated prior knowledge activation effects, we did not use a pretest (which
would have also been a type of prior knowledge activation interfering with the experi-
mental variation). To verify the successfulness of randomization and to rule out that
learners in the two groups differed prior to the experimental intervention, we compared the
final school grade of participants for each condition (we used the A-level grade, which is a
mean of the final exams and subjects learners took during their final two high-school
years). A grade of 1.0 is the best grade, while a grade of 5 and 6 means failing; therefore
the possible range is between 1.0 and 4.4, as learners that failed their A-level exams do not
get admitted to university (only learners with a very good final grade are accepted to study
psychology in Germany).
The posttest included nine open format questions and three multiple-choice items. The
first open question asked participants to write an economic forecast, based on a given
scenario that provided a number of economic parameters. The other eight open questions
asked participants about the possible influences, effects, and explanations, for example
‘How does an interest rate increase affect the stock market? Please also explain why.’
Three to eight minutes were provided to answer each of these open format questions. The
available time for answering the open posttest questions was visualized and controlled by
the computer (the posttest was implemented as a Java Applet; the timer was placed near
the question). To automatically remind learners to finish up, 30 s before the allotted time
had expired, the timer turned from black to red. The multiple-choice questions asked about
specific relationships, for example, the higher the oil price, the … the inflation. Students
had to decide between ‘greater,’ ‘lesser’’ ‘equal,’ ‘not enough information to decide’ and ‘I
don’t know’.
Multiple-choice questions were scored either correct or incorrect. All open questions
were graded with scores between 1 and 5 by two raters. Each rater assigned three separate
grades to each open format question: First, raters assessed the elaboration, second, the
model understanding, and, third, they assigned a holistic grade, which combined the two
sub-scores but also considered the overall quality of the answer. Elaboration and model
understanding were defined analogous to the categories of Experiment 1: For elaboration,

123
 J. Gurlitt, A. Renkl

the maximum score (five) was assigned for answers that showed an integrated under-
standing including examples from everyday life and other relations to prior knowledge
enhancing the given information. For model construction, the maximum score (five) was
assigned for a logical and clear argumentation chain considering relations and the relations
between relations. The minimum score (one) was assigned if the answer did not show an
understanding of any relation between concepts. Scores from each open format answer
were aggregated separately for each dimension and divided by the number of open format
questions, so each learner was assigned one mean elaboration score, one mean model-
construction score, and one overall grade. The reliabilities of the aggregated elaboration
score (Cronbach’s a = .90), of the model-understanding score (Cronbach’s a = .87), and
of the overall grade (Cronbach’s a = .89) were good. Interrater agreement for the scores,
shown by the intraclass correlation coefficient (ICC) was also good (elaboration scores:
ICC2,2 = .97; model-understanding: ICC2,2 = .91; overall grade: ICC2,2 = .92).
Lastly, participants also completed six self-efficacy items. Based on Bandura’s method
for measuring self-efficacy beliefs (Bandura 2006), learners were presented with items
portraying different levels of task demands. Learners rated their efficacy beliefs on a 100-
point scale, ranging in 10-unit intervals from 0% (‘‘Cannot do at all’’), through interme-
diate degrees of assurance, 50% (‘‘Moderately certain can do’’), to 100% (‘‘highly certain
can do’’). For example, one item asked participants to self-assess whether they can explain
the relation between interest rates and the stock market to a friend, while another asked
whether they can give a presentation about economic relationships. The reliability of the
self-efficacy scale was good (Cronbach’s a = .92).

Results and discussion

We used a one-factorial ANOVA to analyze the data. An alpha level of .05 was used for all
statistical tests. As an effect size measure, we used partial g2, qualifying values \.06 as
small effects, values in the range between .06 and .13 as medium effects, and values [.13
as large effects (see Cohen 1988). Learners in the low-coherence condition (M = 1.38,
SD = 0.52) and the high-coherence condition (M = 1.44, SD = 0.46) did not differ in
their A-level grades F(1,40) \ 1, which indicates that learners in the two groups did not
differ on this measure of academic achievement prior to the intervention. Effects of low-
coherence and high-coherence prior knowledge activation on learning outcomes and self-
efficacy beliefs are reported. In addition, correlations between the grade, elaboration and
model-understanding scores, which were assigned to the open format questions, are
described. Table 3 provides an overview about the influence of different prior knowledge
activation tasks on learning outcomes and self-efficacy beliefs.
ANOVA revealed a significant main effect of prior knowledge activation tasks for all
measures of learning outcomes in favour for the high-coherence condition: Learners in the
high-coherence condition outperformed their peers in the low-coherence condition
regarding the grade assigned to open format questions F(1,40) = 11.55, p \ .05, g2 = .22
(large effect). A similar result in favour of the high-coherence condition was obtained by
analyzing the open format questions for elaboration, F(1,40) = 32.41, p \ .05, g2 = .45
(large effect), and model-understanding F(1,40) = 4.15, p \ .05, g2 = .09 (medium
effect). Correspondingly, learners in the high-coherence condition also outperformed their
peers in the low-coherence condition in multiple-choice questions F(1,40) = 8.66,
p \ .05, g2 = .18 (large effect).
Elaboration and model-construction ratings were highly correlated with each other
(r = .82, p \ .001). Both elaboration and model ratings were also highly related to the

123
Prior knowledge activation

Table 3 Means and standard deviations (in parentheses) of the learning outcomes and perceived self-
efficacy
Label provided Create and F-value g2 p
lines label lines

Overall grade 3.93 (.51) 3.16 (.90) 11.55a .22 \.001*

Elaboration 3.73 (.61) 2.46 (.82) 32.41a .45 \.001*
Model 3.67 (.41) 3.25 (.86) 4.15a .09 .048*
Multiple-choice 2.21 (.86) 1.43 (.87) 8.66a .18 \.001*
Self-efficacy 61.39 (16.39) 46.59 (22.15) 5.47b .13 .025*
a b
df = 1,40; df = 1,37; all tests were two-sided, N = 44
* Statistically significant

overall grade (relaboration, grade = .88, p \ .001; rmodel, grade = .91, p \ .001). Based on
these results, it seems reasonable to conclude that multiple choice and open format
questions did not show a differential effect. Therefore, the results supported the first
hypothesis about overall positive coherence effects instead of the competing hypothesis
about differential effects of prior knowledge activation on learning outcomes: Learners
benefited substantially more from the high-coherence prior knowledge activation as
compared to the low-coherence alternative.
This advantage for the high-coherence prior knowledge activation was also supported
by the analysis of the self-efficacy ratings. For the analysis of self-efficacy ratings, three
participants had to be excluded because their files were not recorded due to technical
difficulties. ANOVA revealed a significant main effect of different prior knowledge
activation tasks for self-efficacy ratings in favour of the high-coherence condition
F(1,37) = 5.47, p \ .05, g2 = .13 (medium effect). In a nutshell, results showed sub-
stantial learning outcome and self-efficacy differences in favour of the high-coherence
prior knowledge activation.

General discussion

Summarizing both studies, small variations in the task used for prior knowledge activation
lead to a substantial difference for cognitive processes, learning outcomes, and perceived
self-efficacy. Think-aloud protocols of Experiment 1 showed the differential effect that the
high-coherence condition elicited more elaboration processes, but fewer model construc-
tion and organization processes than the low-coherence condition. The results of
Experiment 2 showed an advantage of the high-coherence prior knowledge activation for
learning outcomes and perceived self-efficacy. Yet, while we found a differential effect on
cognitive processes, we did not find a differential effect on learning outcomes. One pos-
sible explanation is that participants in the low-coherence condition were not able to use
organization and model construction processes to compensate for less structure in the map
and less elaboration processes during prior knowledge activation. Related research using
the same tasks in the domain of physics supports this view (Gurlitt and Renkl 2008). The
latter research showed that although learners in the low-coherence condition asked more
model-construction questions, only learners with a higher level of expertise (physics major
students versus high-school students) were able to benefit from the low-coherence prior
knowledge activation. Conversely, as in this experiment, learners with a lower level of
expertise were better in the high-coherence condition. A different explanation is that

123
 J. Gurlitt, A. Renkl

elaborations are key processes for prior knowledge activation and outrival possible effects
of the other processes. This would be in accord with the Ausubel’s (1968) assimilation
theory which emphasizes the integration of new knowledge into existing knowledge as an
essential process for meaningful learning (see also Novak and Gowin 1984). A third
explanation may be that the multiple-choice and open-format post-test measures were not
fine-grained enough to yield possible differential effects. Other post-test measures may
include for example concept maps (e.g., Hoz et al. 2001) or card sorting tasks (e.g.,
McNamara et al. 1996).
Linking the results of this study to research about text comprehension, the results
support an extension of positive text-coherence effects (e.g., Britton and Gulgoz 1991;
O’Reilly and McNamara 2007) to prior knowledge activation with concept maps. Yet, the
think-aloud protocols from Experiment 1 indicate that it is not a difference in active
processing (as proposed by McNamara et al. 1996), but rather different cognitive processes
leading to these results. However, the reference to text comprehension research has also
some restrictions: First, reading text includes a higher amount of low-level processes such
as character or word decoding. But, if learners do not use their prior knowledge to integrate
the new information, the information processing may be characterized as rote learning
(Novak and Gowin 1984) leading at most to a memorized reproduction of the text. Inte-
grating the text with prior knowledge and constructing an integrated model goes beyond
the textual representation and may be quite similar to processes such as the elaboration
found in the concept mapping tasks. Second, one may advance the criticism that concept
maps with given connections do not only make relations explicit (as in high-coherence
texts) but represent new information to learners which, especially the low-knowledge
learners, cannot infer themselves. However, this may also be true for text comprehension
research, as without background knowledge learners may not be able to infer links between
different pieces of information. For example, in the previously mentioned study from
Britton and Gulgoz (1991), for learners who did not know that Hanoi was the capital of
North Vietnam, no connection existed between the heading and the following sentence
(which is explicitly stated in the high-coherence text):

Original text:

Air War in the North, 1965

By the Fall of 1964, Americans in both Saigon and Washington had begun to focus on
Hanoi as the source of the continuing problem in the South.

Revised, high-coherence text:

Air War in North Vietnam, 1965

By the beginning of 1965, Americans in both Saigon and Washington had begun to
focus on Hanoi, capital of North Vietnam, as the source of the continuing problem in the
South.
A third important difference of this study compared to coherence variations used in text-
comprehension research is that the mapping tasks were used to activate prior knowledge
and to facilitate subsequent learning, whereas the coherence gaps in text comprehension
research were part of the learning phase. It remains an open question whether low-
coherence texts used for prior knowledge activation, or low-coherence maps used during

123
Prior knowledge activation

learning, have similar effects. Further research may clarify whether the proposed bridging
reference to coherence-effects found in text-comprehension research is fruitful for new
hypotheses and insights into diagrammatic representations.
From a practical point of view, the size of the coherence effect on learning outcomes
and perceived self-efficacy is surprising: Rosenthal et al. (2000) recommend the binomial
effect size to illustrate the practical importance of effects. This effect size estimates the
success rate of an implementation. Success is defined by a median cutoff for the dependent
variable, with half of the population succeeding and half of the population failing. Illus-
trating the practical importance of the effects, ten minutes of the ‘‘right’’ prior knowledge
activation, instead of the sub-optimal one, amounts to the following differences between
rates of success: Choosing the high-coherence versus the low-coherence prior knowledge
activation increased ‘‘success rates’’ from about 27% to about 73% for the open format
questions (overall grade), from 28% to 71% for the multiple-choice questions, and from
32% to 68% for the perceived self-efficacy.
Educators using prior knowledge activation tasks should be aware that there is still a lot
of unknown territory to be explored. For example, possible interactions with different kinds
of learners (e.g., level of expertise, learning-strategy skills) can be expected. In this con-
text, it is of value to consider related research about learning from text. While high-
coherence text seems to be more beneficial for most learners, low-coherence text has been
shown to be more beneficial for highly knowledgeable learners with poor learning-strategy
skills (O’Reilly and McNamara 2007). Empirical evidence has shown that prior knowledge
usually overrules all other variables (e.g., Dochy 1992) and taking into account the sub-
stantial differences found for the different prior knowledge activation tasks, it seems worth
investigating possible interactions as well as other valuable educational procedures
explicitly considering prior knowledge activation. Further research may examine the
influence of feedback on the constructed maps and possible interactions with different
learning material such as instruction explicitly dealing with students’ misconceptions
(Uzuntiryaki and Geban 2005). Further research should also investigate which processes or
which combination of cognitive and metacognitive processes during prior knowledge
activation are most beneficial for learning outcomes.
In conclusion, we have shown that a small variation in the mapping task used for prior
knowledge activation leads to substantial differences in cognitive processes, learning
outcomes, and perceived self-efficacy. With respect to cognitive-processes we found more
elaboration processes for the high-coherence prior knowledge activation, but less model-
construction and organization-processes, than in the low-coherence condition. With respect
to learning outcomes, the ‘‘right’’ prior knowledge activation increased success rates
substantially. Therefore, concept mapping is not a unitary instructional method; instead it
should be differentiated based on the specific affordances and tasks that have to be
completed. As both tasks used the same concepts, prior knowledge activation is more than
only the activation of specific concepts in long-term memory. Lastly, theoretical and
empirical links between research about concept maps and coherence effects found in
research about text-comprehension are suggested.

References

Ainsworth, S., & Burcham, S. (2007). The impact of text coherence on learning by self-explanation.
Learning and Instruction, 17, 286–303. doi:10.1016/j.learninstruc.2007.02.004.
Ausubel, D. P. (1968). Educational psychology: A cognitive view. New York: Holt, Rinehart and Winston.

123
 J. Gurlitt, A. Renkl

Azevedo, R., & Cromley, J. G. (2004). The role of self-regulated learning in fostering students’ under-
standing of complex systems with hypermedia. Journal of Educational Psychology, 96, 87. doi:
10.1037/0022-0663.96.3.523.
Azevedo, R., Cromley, J. G., & Seibert, D. (2004a). Does adaptive scaffolding facilitate students’ ability to
regulate their learning with hypermedia. Contemporary Educational Psychology, 29, 344–370. doi:
10.1016/j.cedpsych.2003.09.002.
Azevedo, R., Guthrie, J. T., & Seibert, D. (2004b). The role of self-regulated learning in fostering students’
conceptual understanding of complex systems with hypermedia. Journal of Educational Computing
Research, 30, 87–111. doi:10.2190/DVWX-GM1T-6THQ-5WC7.
Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman and Company.
Bandura, A. (2006). Guide for constructing self-efficacy scales. In F. Pajares & T. Urdan (Eds.), Adoles-
cence and education, Vol. 4: Self-efficacy beliefs of adolescents (pp. 307–337). Greenwich, CT:
Information Age Publishing.
Britton, B., & Gulgoz, S. (1991). Using Kintsch’s computational model to improve instructional text: Effects
of repairing inference calls on recall and cognitive structures. Journal of Educational Psychology, 83,
329–345. doi:10.1037/0022-0663.83.3.329.
Chang, K. E., Sung, Y. T., & Chen, S. F. (2001). Learning through computer-based concept mapping with
scaffolding aid. Journal of Computer Assisted Learning, 17, 21–33. doi:10.1046/j.1365-2729.2001.
00156.x.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum
Associates, Inc.
Dee-Lucas, D., & Larkin, J. H. (1995). Learning from electronic texts: Effects of interactive overviews for
information access. Cognition and Instruction, 13, 431–468. doi:10.1207/s1532690xci1303_4.
Dillon, A., & Gabbard, R. (1998). Hypermedia as an educational technology: A review of the quantitative
research literature on learner comprehension, control, and style. Review of Educational Research, 68,
322–349.
Dochy, F. J. R. C. (1992). Assessment of prior knowledge as a determinant for future learning: The use of
prior knowledge state tests and knowledge profiles. Utrecht: Lemma B.V.
Gerjets, P., & Scheiter, K. (2003). Goal configurations and processing strategies as moderators between
instructional design and cognitive load: Evidence from hypertext-based instruction. Educational
Psychologist, 38, 33–41. doi:10.1207/S15326985EP3801_5.
Gerjets, P., Scheiter, K., & Tack, W. H. (2000). Resource-adaptive selection of strategies in learning from
worked-out examples. Paper presented at the 22nd Annual Conference of the Cognitive Science
Society. Mahwah, NJ: Erlbaum.
Gilabert, R., Martinez, G., & Vidal-Abarca, E. (2005). Some good texts are always better: Text revision to
foster inferences of readers with high and low prior background knowledge. Learning and Instruction,
15, 45. doi:10.1016/j.learninstruc.2004.12.003.
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative text com-
prehension. Psychological Review, 101, 371–395. doi:10.1037/0033-295X.101.3.371.
Gurlitt, J., & Renkl, A. (2008). Are high-coherent concept maps better for prior knowledge activation?
Differential effects of concept mapping tasks on high school vs. university students. Journal of
Computer Assisted Learning, 24, 407–419.
Hauser, S., Nückles, M., & Renkl, A. (2006). Supporting concept mapping for learning from text. In S. A.
Barab, K. E. Hay, & D. T. Hickey (Eds.), Proceedings of the 7th International Conference of the
Learning Sciences (pp. 243–249). Mahwah, NJ: Erlbaum.
Hill, J. R., & Hannafin, M. J. (1997). Cognitive strategies and learning from the World Wide Web.
Educational Technology Research and Development, 45, 37–64. doi:10.1007/BF02299682.
Hoz, R., Bowman, D., & Kozminsky, E. (2001). The differential effects of prior knowledge on learning: A
study of two consecutive courses in earth sciences. Instructional Science, 29, 187–211. doi:
10.1023/A:1017528513130.
Jonassen, D. H. (1986). Hypertext principles for text and courseware design. Educational Psychologist, 21,
269–292. doi:10.1207/s15326985ep2104_3.
Kintsch, W. (1988). The role of knowledge in discourse comprehension: A construction–integration model.
Psychological Review, 95, 163–182. doi:10.1037/0033-295X.95.2.163.
Kintsch, W. (2005). An overview of top-down and bottom-up effects in comprehension: The CI perspective.
Discourse Processes, 39, 125–128. doi:10.1207/s15326950dp3902&3_2.
Kintsch, W., & van Dijk, T. A. (1978). Toward a model of text comprehension and production. Psycho-
logical Review, 85, 363–394. doi:10.1037/0033-295X.85.5.363.

123
Prior knowledge activation

Mayer, R. E. (1997). Multimedia learning: Are we asking the right questions? Educational Psychologist, 32,
1–19. doi:10.1207/s15326985ep3201_1.
McKeown, M. G., Beck, I. L., Sinatra, G. M., & Loxterman, J. A. (1992). The contribution of prior knowledge
and coherent text to comprehension. Reading Research Quarterly, 27, 78. doi:10.2307/747834.
McNamara, D. S. (2001). Reading both high-coherence and low-coherence texts: Effects of text sequence and
prior knowledge. Canadian Journal of Experimental Psychology, 55, 51–62. doi:10.1037/h0087352.
McNamara, D. S., & Kintsch, W. (1996). Learning from texts: Effects of prior knowledge and text
coherence. Discourse Processes, 22, 247–288.
McNamara, D. S., Kintsch, E., Songer, N. B., & Kintsch, W. (1996). Are good texts always better?
Interactions of text coherence, background knowledge, and levels of understanding in learning from
text. Cognition and Instruction, 14, 1–43. doi:10.1207/s1532690xci1401_1.
Miller, J. R., & Kintsch, W. (1980). Readability and recall of short prose passages: A theoretical analysis.
Journal of Experimental Psychology Human Learning and Memory, 6, 335–354. doi:10.1037/0278-
7393.6.4.335.
Nesbit, J. C., & Adesope, O. O. (2006). Learning with concept and knowledge maps: A meta-analysis.
Review of Educational Research, 76, 413–448. doi:10.3102/00346543076003413.
Novak, J. D., & Gowin, D. B. (1984). Learning how to learn. New York: Cambridge University Press.
O’Donnell, A. M., Dansereau, D. F., & Hall, R. H. (2002). Knowledge maps as scaffolds for cognitive
processing. Educational Psychology Review, 14, 71–86. doi:10.1023/A:1013132527007.
O’Reilly, T., & McNamara, D. S. (2007). Reversing the reverse cohesion effect: Good texts can be better
for strategic, high-knowledge readers. Discourse Processes, 43, 121–152. doi:10.1207/s15326950
dp4302_2.
Potelle, H., & Rouet, J.-F. (2003). Effects of content representation and readers’ prior knowledge on the
comprehension of hypertext. International Journal of Human-Computer Studies, 58, 327–345. doi:
10.1016/S1071-5819(03)00016-8.
Rosenthal, R., Rosnow, R. L., & Rubin, D. B. (2000). Contrasts and effect sizes in behavioral research. New
York: Cambridge University Press.
Rouet, J.-F., & Levonen, J. J. (1996). Studying and learning with hypertext: Empirical studies and their
implications. In J.-F. Rouet, J. J. Levonen, A. Dillon, & R. J. Spiro (Eds.), Hypertext and cognition
(pp. 9–23). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Ruiz-Primo, M. A., & Shavelson, R. J. (1996). Problems and issues in the use of concept maps in science
assessment. Journal of Research in Science Teaching, 33, 569–600. doi:10.1002/(SICI)1098-
2736(199608)33:6\569::AID-TEA1[3.0.CO;2-M.
Ruiz-Primo, M. A., Shavelson, R. J., Li, M., & Schultz, S. E. (2001). On the validity of cognitive inter-
pretations of scores from alternative concept-mapping techniques. Educational Assessment, 7, 99–141.
doi:10.1207/S15326977EA0702_2.
Schneider, W., Körkel, J., & Weinert, F. E. (1989). Domain-specific knowledge and memory performance:
A comparison of high- and low-aptitude children. Journal of Educational Psychology, 81, 306–312.
doi:10.1037/0022-0663.81.3.306.
Shapiro, A. M. (1999). The relevance of hierarchies to learning biology from hypertext. Journal of the
Learning Sciences, 8, 215–243. doi:10.1207/s15327809jls0802_2.
Shapiro, A., & Niederhauser, D. (2004). Learning from hypertext: Research issues and findings. Mahwah,
NJ: Lawrence Erlbaum Associates.
Unz, D. C., & Hesse, F. W. (1999). The use of hypertext for learning. Journal of Educational Computing
Research, 20, 279–295.
Uzuntiryaki, E., & Geban, Ö. (2005). Effect of conceptual change approach accompanied with concept
mapping on understanding of solution concepts. Instructional Science, 33, 311–339. doi:
10.1007/s11251-005-2812-z.
Vidal-Abarca, E., Martinez, G., & Gilabert, R. (2000). Two procedures to improve instructional text: Effects
on memory and learning. Journal of Educational Psychology, 92, 107. doi:10.1037/0022-0663.
92.1.107.
Weinert, F. E., & Helmke, A. (1998). The neglected role of individual differences in theoretical models of
cognitive development. Learning and Instruction, 8, 309–323. doi:10.1016/S0959-4752(97)00024-8.
Weinstein, C. E., & Mayer, R. E. (1986). The teaching of learning strategies. In C. M. Wittrock (Ed.),
Handbook of research in teaching (pp. 315–327). New York: Macmillan Publishing Company.

123