Learning and Instruction 69 (2020) 101350

Contents lists available at ScienceDirect

Learning and Instruction

journal homepage: www.elsevier.com/locate/learninstruc

Using mobile devices to enhance inquiry-based learning processes☆ T

a,∗ b c a
Sebastian Becker , Pascal Klein , Alexander Gößling , Jochen Kuhn
a
University of Kaiserslautern, Physics / Physics Education Research Group, Kaiserslautern, Germany
b
Georg-August-University Göttingen, Faculty of Physics / Physics Education Research, Germany
c
Institute for School, Educational and Professional Science, Bielefeld, Germany

A R T I C LE I N FO A B S T R A C T

Keywords: Technological advances offer the possibility of using mobile devices to enrich learning environments with
multimedia Earning multimedia content. Although physical experiments play a key role in science learning, little is known about
Multiple representations integrating multimedia learning in physical experimental processes, especially in realistic classroom settings. In
Cognitive load our approach, students use tablets to investigate motion with an application providing multiple representations
Mobile digital devices
of the measurement data. We present the results of a cluster-randomized controlled study with high school
Physics experiments
students (average age = 15.6, N = 286) conducted between 2017 and 2018 in 11 secondary schools in
Germany. Conceptual understanding of physics, cognitive load, and assessment of teacher behavior were sur-
veyed. Multilevel regression analysis revealed that this approach leads to a significant reduction of extraneous
cognitive load and to greater conceptual knowledge. A confirmatory path analysis revealed that these effects can
be traced back to the treatment and are not significantly influenced by teacher behavior.

1. Introduction meta-analysis the positive effects on learning performance of the in-

In the context of science education, laboratory experiences play a
key role in learning (e.g., Haury & Rillero, 1994), but positive learning
outcomes are not guaranteed (Husnaini & Chen, 2019; Kapici, Akcay, &
de Jong, 2019; Wilcox & Lewandowski, 2017). de Jong, Linn, and
Zacharia (2013) argue that laboratory experiences should combine
virtual and physical components in order to promote learning pro-
cesses, as without this combination, both virtual and physical experi-
mentation cannot include all important experimental and conceptual
competencies. In this context, multimedia learning research in science
education has so far focused more on desktop-based environments,
especially with animations (e.g., Al-Balushi, Al-Musawi, Ambusaidi, &
Al-Hajri, 2017; Plass, Homer, & Hayward, 2009; Rieber, 1990 or see
Mayer & Moreno, 2002 for a review on the role of animations in mul-
timedia learning) and simulations (e.g., de Jong, 1991; Rutten, van
Joolingen, & van der Veen, 2012 or see Smetana & Bell, 2012 for a
critical review). However, since tablets and smartphones, in combina-
tion with appropriate applications, have opened up new and exciting
possibilities to enrich learning environments with multimedia content,
it is time to focus more on multimedia learning with digital devices in
science education. While Sung, Chang, and Liu (2016) identified in a
structional use of mobile digital devices, they also concluded that a
blanket statement on how mobile digital devices can be meaningfully
used in the classroom is not possible because of the variety of teaching
and learning scenarios and of media and learning applications used.
Rather, each newly developed digitally supported learning process must
be tested for the fit of the medium and the learning application, as well
as its learning effectiveness. Considering the increasing use of mobile
technologies in everyday life and in educational environments, Zydney
and Warner (2016) claimed in their review report on mobile applica-
tions for science learning that studies are needed to better align un-
derlying theories and outcome measures. This is especially true for
integrating multimedia content in learning processes with physical
experiments (Oliveira et al., 2019), since applications for science
learning offer a variety of novel methods for the automatic acquisition
and visualization of measurement data. In this context, a very pro-
mising approach to harness the capabilities of mobile devices that is
being discussed in the didactic community is inquiry-based learning in
science laboratories (e.g., Kuhn & Vogt, 2015), as it is assumed that
data collection and data visualization technologies can be helpful in
supporting the inquiry-based learning process in particular (e.g., Chiu &
Linn, 2014; Gilbert, 2005; Kwon, 2002).

☆
Author Note: I am a doctoral student preparing a “dissertations by publication”, and the submitted manuscript constitutes a part of my dissertation in the field of
technology-supported learning in physics lessons. The research focus here is on the investigation of the learning effectiveness of experimental digital tools in regular
teaching (compared to traditional, non-digital methods)
∗
Corresponding author. Erwin-Schrödinger-Str. 1 67663, Kaiserslautern, Germany.
E-mail address: s.becker@physik.uni-kl.de (S. Becker).

https://doi.org/10.1016/j.learninstruc.2020.101350
Received 8 August 2019; Received in revised form 29 April 2020; Accepted 5 May 2020
Available online 07 June 2020
0959-4752/ © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
S. Becker, et al.
1.1. Technology-enhanced inquiry-based learning
In contrast to traditional, more teacher-centred teaching, inquiry-
based science education places the student and his or her active and
critical occupation with the learning content at the centre of the
learning process. In this constructivist approach, students are given the
opportunity to ask questions, formulate and test hypotheses, and carry
out experiments on their own, which is expected to increase student
motivation, interest, and development of scientific literacy (Collins,
1997; DeBoer, 1991; Singer, Hilton, & Schwiengruber, 2005). Scientists
have identified several elements of instruction that contribute to the
success of inquiry-based learning. Especially for an experiment-based
learning activity, which is essential in the natural sciences, two ele-
ments of instruction are of crucial importance: the collection of data
(Songer, 1996) and the connections between alternative representa-
tions of the phenomena (White & Frederiksen, 1998). In the study by
Kwon, for example, middle school students used graphing calculators to
acquire time-dependent position and velocity data. Kwon was able to
show that this instructional approach is more effective than traditional
teaching in developing students’ skills of linking realistic contexts of
physical phenomena with graphs. Chiu and Linn showed that a tech-
nology-enhanced inquiry unit using dynamic molecular visualizations
improved the ability of the students to connect molecular, observable,
and symbolic representations of chemical reactions. However, as in
these studies, either data acquisition or data visualization is supported
by technology, but not both combined. For experiment-based activities,
this means that different devices usually need to be used. For example,
the measurement data is first recorded with a sensor, then transferred to
a desktop PC, where it is evaluated and visualized. This can lead to
difficulties for students, since the separation of corresponding in-
formation violates the principle of contiguity (Mayer & Moreno, 2003;
Mayer, 2009; see 1.3.1) and results in the experimental process and the
associated representations no longer being closely linked, which can
have a negative effect on the learning process (Wu & Puntambekar,
2012). Against this background, Wu and Puntambekar (2012) argue
that “pairing is a powerful way to use multiple external representations
(MERs) in science education” (p. 754) and predict that “with the ad-
vancement of science and technology (…) teachers and researchers may
explore other ways of pairing multiple representations and scientific
processes” (p. 760). Mobile devices such as smartphones and tablets are
equipped with sensors that can be read without time delay by specially
developed applications. In this way, data can be digitally measured and
visualized in various forms of representation, such as tables and graphs,
automatically and in real time. In this view, initial studies have shown
positive effects of using mobile devices to enhance inquiry-based ex-
perimental learning with MERs on conceptual learning (e.g., Becker,
Klein, Gößling, & Kuhn, 2019; Becker, Klein, & Kuhn, 2018; Hochberg,
Becker, Louis, Klein, & Kuhn, 2020; Klein, Kuhn, & Müller, 2018) and
motivation (e.g., Hochberg, Kuhn, & Müller, 2018). In support of this,
Sung et al. (2016) were able to demonstrate in a meta-analysis that an
inquiry-based teaching approach that integrates mobile devices has a
positive effect on learning performance with a large effect size. At the
same time, however, the authors note that “more elaborate instruc-
tional design developments are needed to more thoroughly exploit the
educational benefits possible by utilizing mobile devices” (p. 265).
Zydney and Warner (2016) even identify a research gap regarding
mobile applications for science learning and conclude that further
studies are necessary to better align the underlying theories, designs,
and outcome measures.
1.2. Multiple representations in science education
Information can be made available to learners in various forms of
representations, including texts, formulas, diagrams, and illustrations.
The use of more than one representation in learning environments is an
essential and frequently used method of conveying information and
2
Learning and Instruction 69 (2020) 101350
supporting the construction of knowledge in scientific learning (Tytler,
Prain, Hubber, & Waldrip, 2013). The term MER refers to those re-
presentations that have a common reference object. For example, a
motion process can be described physically by a diagram, a data table,
or a formula. All three representations have the same reference object,
the motion process, and provide information about it, but they do so in
different ways. The level of abstraction of a symbolic–mathematical
form of representation, such as a formula, is, for instance, higher than
that of a visual–graphical form of representation, such as a diagram or
table, which can make it difficult for students to relate the formula to
the process in the real world. A combination of different forms of re-
presentation can therefore support the students in the transfer process
from a highly abstract visualization to the real world, and vice versa.
This important role of MERs is well documented for the natural sciences
in general (Tytler et al., 2013) and for physics in particular (Treagust,
Duit, & Fischer, 2017). It is of especially great importance for con-
ceptual understanding (Verschaffel, de Corte, Ton, & Jan 2010), and it
is assumed that learners can gain a deeper understanding if they ab-
stract over MERs (Ainsworth, 2014). A systematic overview of the un-
ique benefits of MERs for learning complex or new scientific content is
provided by the DeFT (Design, Functions, Tasks) framework created by
Ainsworth (2006, 2008).
1.2.1. DeFT framework for learning with multiple representaions
Ainsworth's basic idea is that the presentation of MERs is not ef-
fective in itself unless three key functions are fulfilled. According to the
first function, the presentation of MERs with complementary informa-
tion content can be advantageous if the presentation of all relevant
information in a single form of representation would lead to cognitive
overload. However, even if different representations contain the same
information, they can support the learning process by allowing the
learners to choose from among the different forms of representation the
one(s) they consider most suitable for the specific learning situation.
Treagust et al. (2017) explain this function using an example from
physics lessons: if the concept of constant acceleration is taught, the
teacher could simply present the acceleration as a mathematical ex-
pression. However, students often find it difficult to link this very ab-
stract representation with the real movement process. An additional
presentation of a table with exemplary values for the acceleration of a
vehicle and the corresponding graph can support the students in this
connection. The second function is that the presentation of a more fa-
miliar representation can constrain the interpretation of the less fa-
miliar one. Ainsworth (2006) explains this function through the inter-
pretation of complex graphs, which can be a great challenge for
students, especially in the natural sciences (e.g., Glazer, 2011; Shah &
Hoeffner, 2002). The additional presentation of an explanatory text or a
table with selected data points can help students understand how the
data is presented in the graph and how to interpret it. Ultimately, ac-
cording to the third function, the integration of corresponding in-
formation from MERs supports deeper understanding and thus the
construction of flexible knowledge that can be more easily transferred
to new learning situations. For example, if students have developed a
basic conceptual understanding of a motion with constant velocity
using different representations, they can extend their knowledge to
such representations for a motion with constant acceleration, since the
basic concept is the same for both types of motion.
However, Wu and Puntambekar (2012) rightly point out that “al-
though (…) theories (…) have been developed to explain why the use of
MERs is beneficial to learning, they do not provide much information
on pedagogical issues such as how and in what conditions MERs could
be introduced and used in classroom instruction” (p. 755). Following
the authors, one key success factor for effective use of MERs in class-
room situations is how their use is supported in the classroom. Against
this background, digital technologies for data acquisition and data vi-
sualization offer innovative ways to support students in learning with
and from MERs, especially in experiment-based learning environments.
S. Becker, et al.
Fig. 1. Screenshots from video motion analysis application: video with tracking of object (upper), y(t)-(position-time) graph (middle), vy(t)-(velocity-time) graph
(lower).
One promising instructional approach to promote the integration of
MERs into mechanics lessons is video motion analysis.
1.3. Video motion analysis
The basis of this non-contact measurement method is a video re-
cording of moving objects, from which their velocity and acceleration
can be calculated. The measurement data can be visualized in diagrams,
data tables, or strobe pictures, for which time-synchronous markers are
superimposed on the real video image to indicate the position of the
moving object (see Fig. 1). Tablets now have technically advanced
cameras that can record videos of even (fast-)moving objects with
outstanding quality. Video analysis applications developed especially
for physics lessons create the possibility of combining all the single
video analysis process steps on a single mobile device, from the re-
cording of the moving object in an experiment to the analysis of the
movement to the visualization of the relevant physical quantities by
providing MERs such as data tables, strobe pictures, and diagrams. The
learner can then switch between these MERs as needed and view them
in combination. These multi-representational visualization capabilities
fulfill the three key functions of the DeFT framework and therefore
support student learning with and from MERs, especially in an experi-
ment-based learning environment.
• 1st key function: The learner is offered different forms of re-
presentation for the analysis of the same movement (e.g., strobo-
scopic imaging and the corresponding motion diagrams).
• 2nd key function: Simultaneously with the real motion sequence
with which the learners are familiar, the stroboscopic imaging is
displayed, a form of representation with which the learners are less
familiar.
• 3rd key function: The learner is shown the time-position and time-
velocity graph simultaneously.
Even though MERs have the potential to promote learning
3
Learning and Instruction 69 (2020) 101350
processes, learners may not be able to take advantage of the benefits
that they offer, as they also create complex demands and can even in-
crease cognitive load for the learners (de Jong et al., 1998; Seufert,
2003; van Meter, List, Lombardi, & Kendeou, 2020). Indeed, there are
many studies pointing towards student difficulties with MERs (e.g.,
Ainsworth, 2006; Nieminen, Savinainen, & Viiri, 2010). In this context,
Horz, Schnotz, Plass, Moreno, and Brünken (2009) argue that tech-
nology can help reduce cognitive load when learning with MERs and
therefore facilitate the learning-promoting effect of MERs.
1.4. Managing cognitive load in multimedia learning environments
The reason for the need to reduce the cognitive load in a multi-
representational learning environment is mainly due to the increased
complexity of information presentation. The type of load that can be
influenced by the design of the learning procedure (e.g., “how the in-
formation is presented and what the learner is required to do by the
instructional procedure” (Sweller, van Merriënboer, & Paas, 2019, p.
269)) is called extraneous cognitive load (ECL). One fundamental
learning principle to be considered, especially in a multi-representa-
tional learning environment, is keeping ECL as low as possible during
the learning process (Leppink, 2017; Leppink & van den Heuvel, 2015).
While ECL refers to “extraneous processes not productive for learning”
(Sweller et al., 2019, p. 269), intrinsic cognitive load (ICL) refers to the
complexity of the information the learner must process during the
learning process and is therefore determined by the learning task as
well as the prior knowledge of the learner regarding the learning con-
tent (e.g. Leppink & van den Heuvel, 2015; Sweller et al., 2019). The
last type of load, GCL, refers to the amount of cognitive resources
needed while processing the information in a learning process. It should
be noted, however, that the view of GCL as an independent type of load
has been increasingly criticized in the research community. More and
more researchers demand that GCL no longer be seen as an independent
contribution to cognitive load but rather as a subjective judgment of
learning (e.g., Leppink, Paas, van Gog, van der Vleuten, & van
S. Becker, et al.
Merriënboer, 2014).
1.4.1. Principle of contiguity
This instructional principle aims to reduce ECL by avoiding the
split-attention effect (Mayer & Pilegard, 2014; Sweller et al., 2019).
This means that corresponding information should not be presented
spatially or temporally separated (Mayer & Moreno, 2003; Mayer,
2009). The use of video analysis applications fulfills this principle by
simultaneously presenting the learner with different forms of corre-
sponding representations in combination (e.g., stroboscopic imaging
and the corresponding motion diagrams or time-position and corre-
sponding time–velocity diagram).
1.4.2. Segmentation principle
This principle, also known as the interactivity principle, postulates a
learning-enhancing effect in connection with dynamic visualizations if
the learners can determine the sequence or tempo of the information
presentation themselves (Mayer & Pilegard, 2014). Video analysis ap-
plications allow the learners themselves to control the transition be-
tween the individual forms of representation. For example, to improve
their understanding of the motion diagrams, the students can again call
up the stroboscopic image. According to this principle, such self-control
over the learning process avoids cognitive overload.
1.4.3. Dynamic linking
Ainsworth (2006) argues that dynamic linking is assumed to reduce
the cognitive load of learners as well. Dynamic linking means that when
learners manipulate one representation, the resulting changes can be
directly observed in another representation. The video analysis appli-
cation dynamically links the representation of the coordinate system
and the movement diagrams, because by manipulating the coordinate
system, the motion diagrams are automatically adjusted by the appli-
cation.
1.5. The present study
Conceptual understanding of physics is considered a fundamental
condition factor both for understanding physical learning contents
(Vosniadou, 2007) and the ability to solve physical problems (Rittle-
Johnson, Siegler, & Alibali, 2001). According to Pundak and Rozner
(2007), there is a growing consensus that traditional teaching is in-
sufficiently supportive of students’ understanding of physical concepts.
Looking more closely at traditional teaching methods, one reason for
this may be that they often reach their limits when integrating MERs
into experimental activities, so that learners often do not reap the
benefits of MERs in such learning situations. At this point, this will be
explained in more detail using the example of the physical subject area
kinematics, which focuses on the concept of velocity and acceleration of
moving bodies. In traditional teaching, student experiments usually
start by recording measured values for the time-dependent position of a
moving body and entering them into a measurement table. After the
experiment has been carried out, the experiment setup is first dis-
mantled before the students transfer their data from the table into a
diagram and evaluate it. As a result, the individual forms of re-
presentation are separated from each other as well as from the real-
world process. In order to mentally connect the representations and the
scientific process, the students must strain their cognitive resources due
to the spatial and temporal separation of corresponding information.
This can result in cognitive overload, especially in an experiment-based
learning activity. From the perspective of cognitive load theory, the
avoidance of such overload is an essential condition for successful
learning. Consequently, the cognitive load in a multi-representational
learning environment must be considered and managed carefully. Since
ICL results from the degree of difficulty of the learning content and the
complexity of the learning task, according to common doctrine, ICL
cannot be manipulated by the instructional design for a given learning
4
Learning and Instruction 69 (2020) 101350
content (e.g., Sweller et al., 2019). On the other hand, ECL is de-
termined by how the information is presented and what the learner is
required to do by the instructional procedure (Sweller et al., 2019).
Unlike ICL, ECL can therefore be manipulated by changing instructional
procedures, so this type of load can very well be modified by the in-
structional design. As reported in the introduction, video analysis ap-
plications have the potential to effectively reduce ECL, so that it seems
possible that through this reduction of load, students will achieve better
learning outcomes. Although the learning effectiveness of video motion
analysis has been demonstrated in several studies, especially in the field
of mechanics (e.g. Hockicko, Trpišová, & Ondruš, 2014; Wee, Tan,
Leong, & Tan, 2015), the causal relationship between ECL and learning
performance has not yet been investigated for this method. Thus, the
reasons for the positive learning effects found have not yet been un-
covered. The present study contributes to closing this research gap by
empirically investigating and searching for connections between cog-
nitive load and the learning effectiveness regarding a conceptual un-
derstanding of physics for a specific learning scenario in regular school
lessons.
2. Hypothesis and questions
From the theoretically-founded positive influence of video analysis
applications on the learning process, we derive the following hypoth-
esis:
2.1. Hypothesis 1
Compared to traditional teaching, enhanced experimental learning
environments based on tablet-supported video analysis lead to a re-
duction of ECL.
2.2. Hypothesis 2
Compared to traditional teaching, enhanced experimental learning
environments based on tablet-supported video analysis lead to a better
conceptual understanding of the addressed physical concepts.
In cognitive load theory, it is assumed that with a reduction of ECL,
more cognitive resources are available to the learner for active
knowledge construction. This should lead to a more efficient learning
process resulting in increased learning performance. If differences in
ECL and learning gains are found in favor of technology-supported
compared to traditional teaching sequences, it would be unclear if the
reduction in ECL is really causal for the enhanced learning gains, which
leads to the following research question.
2.3. Question
Can a causal connection between the reduction of ECL and learning
gains be statistically supported?
3. Methodology
3.1. Study design
We conducted a cluster-randomized controlled trial in a pre–post
test design involving high school physics courses, which took place in a
natural teaching–learning situation in which the students are taught
together. Since the spatial separation of the students of one course was
not possible due to the general conditions at the participating schools,
we followed Dreyhaupt, Mayer, Keis, Öchsner, and Muche (2017) and
opted for a cluster randomization and randomly assigned whole courses
as experimental group (tablet group) or control group (calculator
group). The study covered a curricular-valid and essential topic of
mechanics: uniform motion. The lesson sequence focused on the fos-
tering of a conceptual understanding of physics by gaining insights
S. Becker, et al.
through independent, collaborative experimentation in small groups of
two students. The effect of the intervention is captured by the oper-
ationalization of two dependent variables: cognitive load and con-
ceptual understanding. In order for us to investigate the possible in-
fluence of the teacher, the students were also asked post-hoc about their
teacher's behavior during the intervention.
3.2. Participants
The data were collected in 18 courses from 11 secondary schools in
different states in Germany between 2017 and 2018. In each experi-
mental and control condition, questionnaires were completed in the
students’ regular classrooms and in the presence of the associated tea-
cher.
3.2.1. Students
In total, 294 students participated in both test periods, which is
above the desired sample size of 252, as derived from an a priori power-
calculation using the software G*Power (1-β = 0.999) based on the
effect sizes found in our preliminary studies (Authors, 2018; Authors,
2019). Sociodemographic data were evaluated for 286 students, of
whom 94 are female and 191 are male (one did not complete the
question about gender), with an average age of 15.6 (SD = 0.72). To
ensure the comparability of the groups, only advanced physics courses
were selected for the study. The sociodemographic composition of the
population separated according to tablet and calculator group is shown
in Table 4.
3.2.2. Teachers
The seven female and nine male teachers voluntarily participating
in the study are all teachers of physics who are qualified to teach at
secondary schools and have several years of professional experience. By
the beginning of the study, they had spent several weeks teaching
physics to the participating groups of students and integrated the
learning sequence of the study into their regular lessons. Either a single
teacher or two teachers in each participating school were randomly
assigned to either the experimental or the control group, on the premise
that where there are two teachers in one school, one is assigned to the
experimental group and the other to the control group. For organiza-
tional reasons in the participating schools, only one of the teachers
taught under both conditions with different classes.
3.3. Experimental manipulation
Desktop-based video analysis has the disadvantage that the re-
cording and analysis of moving objects within experiments are tem-
porally separated, which interrupts the learning process. Since this
complicates implementation in regular school lessons, as time-on-task is
extended beyond what would be required for conducting the experi-
ment using conventional methods, we decided to use tablets and the
video analysis application Viana.1 While students in the tablet group
conducted and recorded experiments with a tablet, analyzed motion
processes of these experiments, and visualized the measuring data with
different MERs, students in the calculator group conducted, recorded,
analyzed, and visualized experimental motion processes with different
MERs using experimental tools used in traditional school education: a
stopwatch, a tape measure, and a graphing calculator (TI-84 Plus, Texas
Instruments). Fig. 2 shows a comparison of the user interfaces of the
video analysis application and the graphing calculator. Table 1 com-
pares the opportunities for data visualization of the two devices, and
Table 2 specifies the affordances in visualization of the tablet-supported
1
The application Viana is available for iOS for free at https://goo.gl/
4RWv8g, a detailed description of the application is available from Becker,
Klein, et al. (2018).
5
Learning and Instruction 69 (2020) 101350
Table 1
Opportunities for data visualization.
Tablet-supported video Graphing calculator
analysis
Diagram Yes yes
Table No yes
Simultaneous display Yes no
Overlaying Yes no
Change of representation Yes no
Dynamic Linking Yes no
video analysis, whose positive influence on cognitive load has already
been deduced from the learning theories presented in Section 1. In
order to ensure a fair comparison between the two instructional ap-
proaches, the experiments, learning content, time-on-task, and group
size were identical. In particular, the students in both groups used the
same forms of representation for learning: diagram, strobe picture,
table, and formula (Table 3 compares the number of individual forms of
representation used for learning during the intervention for both
groups). The students in both groups were given learning tasks for
conducting the experiments and analyzing the measured data devel-
oped for this study in cooperation with teachers with many years of
professional experience. To illustrate the comparability, Fig. 3 shows an
example learning task that the students of the tablet and the calculator
groups were set. In the 4x45-min lesson sequence, the students ex-
perimented independently in pairs. The experiments aimed at the fos-
tering of conceptual understanding of uniform motion and were de-
signed so that the students could carry them out completely
independently, thus reducing the influence of the teacher as far as
possible. For this reason, the learning gains result from the independent
processing of the learning tasks and not from interaction with the tea-
cher. Moreover, the involved teachers were instructed to take a passive
role during the experimentation process and to avoid actively inter-
vening. However, they were permitted to respond to inquiries regarding
experiment set-up and execution. This was to ensure that all student
groups were able to carry out the experiments successfully.
3.4. Study procedure
3.4.1. Introductory lesson
In the first lesson of the intervention, the students of the tablet
group received standardized instruction in physical video analysis with
the tablet. The guidance included an explanation of the measurement
methodology and the video analysis application used. In addition, the
students were given the opportunity to familiarize themselves with the
video analysis application by analyzing sample videos. The sample vi-
deos are already included in the library of the application and have no
contextual relationship to the subject area of the study. The students of
the calculator group received standardized instruction in data analysis
with the graphing calculator and were given the opportunity to famil-
arize themselves with the functionality of the graphing calculator by
analyzing given data sets, also with no contextual relationship to the
subject area of the study.
3.4.2. Pre-testing
In the lesson after the introductory lesson, sociodemographic data,
as well as preliminary grades in physics, mathematics, and German,
were requested. Next, the students completed the pre-test on conceptual
understanding.
3.4.3. Experiment-based learning sequence
The students conducted two experiments in two lessons each, which
involved set-up, execution, and evaluation. The experiments were de-
signed with two aspects in mind. Firstly, in order to ensure a fair
comparison between the experimental and the control group, the
S. Becker, et al. Learning and Instruction 69 (2020) 101350

Table 2
Affordances in visualization of the tablet-supported video analysis.
Simultaneous Display Students are simultaneously provided with the time-position and the time-velocity graph

Overlaying The stroboscopic imaging is superimposed on the real image of the motion sequence
Change of Representation Students can switch between different representations by a gesture of hand without noticeable time delay
Dynamic Linking Changes of the coordinate system is simultaneousely reflected in the movement diagrams

Table 3 friction between ball and rail during the movement is so small that it
Number of forms of representation used during the intervention. can be neglected for the physical description of the movement. The task
Form of representation Tablet group Calculator group was to experimentally determine the velocity of the sphere by evalu-
ating the time-dependent position data. While the students in the tablet
Diagram 9 8 group used video analysis to determine the position data and visualize
Strobe picture 2 2
the velocity data, the students in the calculator group used a stopwatch,
Table 1 4
Formula 5 5
tape measure, and graphing caluclator.
Experiment 2: The second experiment aimed at the conceptual un-
derstanding of velocity in a back-and-forth uniform motion. To achieve
Table 4 this, the students rolled the steel sphere once in one direction and then
Sociodemographic composition. back again in the other. The idea is that velocity in physics is a vectorial
quantity, which can have a positive or negative sign depending on the
Variable Tablet group (N = 150) Calculator group (N = 136)
reference system and direction of the movement. Again, the students in
Average age 15.6 15.7 the tablet group used video analysis, and the students in the calculator
Female (in %) 34.0 31.9 group used a stopwatch, tape measure, and graphing caluclator to ex-
AC physics (in %) 100.0 100.0
perimentally determine the velocity of the sphere by evaluating the
AC mathematics (in %) 38.9 37.7
time-dependent position data.
Note: Data basis N = 286.

Fig. 2. User Interface of the graphing calculator (left) and the video analysis application (right).

experiments should allow the acquisition of measurement data by 3.4.4. Post-testing

means of both video analysis and traditional experimental tools. Immediately following the learning sequence, a post-test was car-
Secondly, in order to promote transfer into regular school lessons, care ried out. In order to maximize the comparability, the post-test included
was taken to ensure that the students experiment with everyday ma- identical items, but in a different order. Following the post-test, the
terials that can be purchased at low cost. In this case, the materials were students were also asked to complete the questionnaire on cognitive
an aluminum rail and a steel sphere. Thus, in principle, these experi- load and teacher behavior.
ments can be carried out in any regular school without great additional
financial expenditure. The instructions for carrying out the experiments
3.5. Data collection
and the associated learning tasks were given to the students in a pro-
tocol booklet, in which they could also enter their results.
3.5.1. Teacher behavior
Experiment 1: The first experiment aimed at the conceptual under-
The teacher's behavior during the intervention was evaluated on a
standing of velocity regarding a uniform motion in one direction. For
5-item scale. For this purpose, a self-assessement survey (see
this purpose, the students let the steel ball roll over the aluminum rail
Supplementary Material), which has already been evaluated in our
that was lying flat on the table, once quickly and once slowly. The
preliminary study (Authors 2018; Authors 2019), was used. Thus,

6
S. Becker, et al.
Fig. 3. Exemplary learning task for the students of the calculator group (left) and the students of the tablet group (right).
immediately after completion of the intervention, the students were
asked about their subjective assessment of their teacher's behavior. By
means of confirmatory factor analysis (see Appendix), a two-factorial
structure was identified, which implies a splitting of the scale into two
subscales: (willingness to) support and commitment.
3.5.2. Conceptual understanding
In order to determine the learning gains in conceptual under-
standing, a multiple-choice test was used (see Supplementary Material),
which consists of adapted items from validated test instruments (KCT
(Lichtenberger, Wagner, Hofer, Stern, & Vaterlaus, 2017), KiRC (Klein,
Müller, & Kuhn, 2017), and TUG-K (Beichner, 1994)) as well as self-
developed items and has already been evaluated in our preliminary
study (Authors 2018; Authors, 2019). The test measures three sub-di-
mensions: “Velocity as alteration rate” (AltRate), “Velocity as vectorial
quantity” (VectQuant), and “Reference system” (RefSys), which are
approved sub-concepts in physics education. The items (three per sub-
concept) contain the common forms of representation used in kine-
matics: diagram, table, and strobe picture. To confirm the intended
structuring, we performed a factor analysis on the response pattern at
the post-time point (see Appendix). In addition, the students were asked
to rate their confidence for each item on a four-point Likert scale ran-
ging from “very sure” to “guessed.”
3.5.3. Cognitive load
A ten-item subjective survey developed and validated by Leppink,
Paas, van der Vleuten, van Gog, and van Merriënboer (2013) was used
to measure the intervention-induced cognitive load. With this instru-
ment, it is possible to not only measure the overall cognitive load but
also differentiate between the three types of cognitive load (Hadie &
Yusoff, 2016; Zukić, Đapo, & Husremović, 2016): ICL, ECL, and GCL.
Since the original questionnaire was developed for students in a sta-
tistics course, the items had to be adjusted to the specific physical
context in this study (see Supplementary Material) and literally trans-
lated into German. Even though empirical studies have already de-
monstrated that the three-factor model is robust against adaption to the
disciplinary context (Leppink et al., 2014), we decided to verify the
three-factor structure for the adapted questionnaire used in this study
7
Learning and Instruction 69 (2020) 101350
by factor analysis (see Appendix).
3.6. Data analysis techniques
Since it cannot be excluded that the learning process of individual
students is influenced by the social group to which they belong, the
nested data structure must be taken into account when selecting sui-
table analysis methods.
3.6.1. Multilevel regression analysis
Since students are nested within classes, which are nested within
schools, three-level regression analysis (see Hox, Moerbeek, & van de
Schoot, 2018, Chapter 2) was chosen in this study to examine the stu-
dent test scores and the cognitive load data using the R-package lme4
(vs. 1.1.21). Additionally, the R-package sjstats (vs. 0.17.7) was used to
calculate the intraclass correlation (ICC), which indicates how much of
the total variance is explained by the grouping structure at the class and
school level.
3.6.2. Confirmatory path analysis
To examine multivariate causal hypotheses for data sets with a
multilevel structure, Shipley (2000, 2009) introduced the method of
confirmatory path analysis based on applications from graph theory.
With this method, which is also called piecewise structural equation
modeling (e.g., Shipley & Douma, 2019), path models can be tested
even if variables are defined at different levels or are not normally
distributed. As with regression analysis, a three-level model including
student, class, and school level was used. The relationship between
cognitive load and learning gains was thus empirically tested for
causality using the R-package piecewiseSEM (vs. 2.1.0).
4. Results
4.1. Preliminary analysis of covariate balance
In order to examine if the population is balanced at the pre-time
point regarding preliminary grades and prior knowledge, a Wilcoxon-
Mann-Whitney test (U test) was carried out. The results show a
S. Becker, et al. Learning and Instruction 69 (2020) 101350

Table 5 scores is located at the class and the school level: ICCAltRate = 10.1%,
Test for significant group differences. ICCVectQuant = 7.3%, ICCRefSys = 13.4%. Since the performance mea-
Variable Tablet group M Calculator group M U test P surement is a pre–post comparison, it was tested in a next step whether
(SD) (SD) a significant interaction effect exists between time of testing and group
affiliation regarding the different sub-concepts. The results are shown
Preliminary grade
in Table 7. No significant group difference could be detected for the
Physics 2.26 (1.20) 2.63 (1.41) 0.040
Mathematics 2.54 (1.25) 3.01 (1.37) 0.004
sub-concepts AltRate (F(1,257) = 1.66, p = 0.199) and VectQuant (F
German 3.05 (1.09) 3.25 (1.15) 0.100 (1,257) = 0.102, p = 0.750), but for RefSys, this was possible, with
Conceptual understanding high test power (F(1,257) = 7.38, p = 0.007, η2 = 0.015, 1-
Pre-test score 4.66 (2.09) 3.92 (1.99) 0.004 β = 0.784). Since there was one teacher who taught both a tablet and a
calculator class, analysis of the data from these students can indicate
Notes: Data basis N = 286. For marks, 1 = very good, 2 = good, 3 = sa-
whether the observed effect might be related to a teacher effect.
tisfactory, 4 = sufficient, 5 = poor, 6 = deficient.
However, the effect regarding sub-concept RefSys could also be de-
monstrated for this sub-population (see Appendix).
significant difference between the groups for the preliminary mark in
physics and mathematics, as well as the conceptual knowledge re-
4.3. Cognitive load
garding uniform motion (see Table 5).
Since the possibility that they can affect the outcome of the inter-
The mean score of every item was first linearly transformed into a
vention cannot be excluded, a sample balanced in these covariates was
[0, 100%] scale, where 0% meant no affirmation of the statement and
generated through propensity score matching (PSM; Rosenbaum &
100% meant a full affirmation of the statement. For negated items, the
Rubin, 1983; Guo & Fraser, 2010) prior to the following comparative
scale was inverted. The sub-scale scores were then calculated as mean
analysis. PSM allows causal statements to be made about intervention
values of the items of the respective sub-scale. Fig. 5 gives an overview
effects in empirical studies in which complete randomization is not
of the group-dependent scale scores and related standard errors for the
possible or sufficiently successful from the outset (Fan & Nowell, 2011).
different sub-scales. As with the performance variables, only a small
Based on a logistic regression model of all potentially confounded
proportion of the variance is localized at the upper levels: IC-
variables, the propensity score (PS) for each subject is estimated from
CICL = 3.5%, ICCECL = 15.3%, ICCGCL = 8.5%. Subsequently, it was
the total population as probability of belonging to one of the com-
tested whether the sub-scales of cognitive load differ significantly be-
parative groups. Each participant in one group is assigned one or more
tween the tablet and calculator groups (see Table 8). While no sig-
participants of the other group with the same or very similar PS values.
nificant difference could be found for intrinsic load (F(1,233) = 1.56,
Following that, the treatment effect can be estimated for the matched
p = 0.216), a significantly lower intervention-induced extraneous load
population with conventional statistical techniques.
(F(1,233) = 18.42, p < 10−3, η2 = 0.078, 1-β = 1.000) and a sig-
To match the given population, all covariates collected (preliminary
nificantly higher germane load (F(1,233) = 4.35, p = 0.041,
grades, pre-test score) were used to determine the PS. We decided on
η2 = 0.020, 1-β = 0.982) could be proven for the tablet group, both
the matching technique known as “nearest neighbor matching,” which
with high test power. Again, these effects were equally evident in the
matches a student of the tablet group to a student of the calculator
independent analysis of the data of both classes taught by a single
group that is closest in terms of a distance measure estimated by logistic
teacher who did both conditions (see Appendix).
regression. After the matching process, the population is balanced in all
covariates (see Table 6), but the sample size has been reduced from
4.4. Connection between cognitive load and learning gains
N = 286 to N = 262. However, this sample size is still sufficient for
further statistical analysis, and so all further evaluations are based on
To clarify the influence of cognitive load on the learning gains of the
the data set of the balanced sample.
students, the cognitive load variables (ICL, ECL, GCL), the cognitive
performance variables (sub-concepts AltRate, VectQuant & RefSys), and
4.2. Learning gains
the teacher behavior (support & commitment) were fit into a path
model.
Initially, the confidence information was used to rate guessed an-
swers post hoc as incorrect, which contributes to the validity and re-
4.4.1. Genesis of the path model
liability of the test results. Subsequently, correct answers were scored as
Base model: Following the cognitive load theory, there is a direct
1, wrong answers as 0. The total score per sub-concept was calculated
influence of the different sub-scales of cognitive load on learning per-
as the sum of points divided by number of items per sub-concept. Fig. 4
formance. However, it should be noted that ICL, ECL, and GCL are re-
gives an overview of the group-dependent averages and standard errors
lated to learning performance in different ways. According to common
of relative test scores separated according to the different sub-concepts.
doctrine, this relationship is reciprocal for ECL, which means that
In a first step of the multilevel regression analysis, the ICC value was
learning performance is inhibited by a high level of ECL (Sweller, van
determined, which shows that only little variance of the student test
Merriënboer, & Paas, 1998). In contrast, learning performance is en-
hanced by a high germane load (e.g., Sweller et al., 1998). However, no
Table 6
general statement can be made for ICL. On the one hand, a high level of
Test for significant group differences after PSM.
ICL can lead to cognitive overload, but on the other, it can also have an
Variable Tablet group M Calculator group M U test P activating effect and thus lead to better learning performance (Lafleur &
(SD) (SD) Leppink, 2015). Consequently, all three cognitive load variables are
Preliminary grade included in the underlying model. Although the learning sequence was
Physics 2.34 (1.20) 2.61 (1.40) 0.157 designed in such a way that the learning gains results from the in-
Mathematics 2.71 (1.22) 3.00 (1.37) 0.101 dependent processing of the learning tasks and not from interaction
German 3.05 (1.13) 3.24 (1.16) 0.157 with the teacher, an influence of the cognitive load and learning per-
Conceptual understanding
Pre-test score 4.31 (1.94) 3.95 (2.01) 0.198
formance by the teacher cannot be excluded. For this reason, teacher
behavior is also included in the model.
Notes: Data basis N = 262. For marks, 1 = very good, 2 = good, 3 = sa- Alternative model:As reported in Section 1, the interpretation of
tisfactory, 4 = sufficient, 5 = poor, 6 = deficient. GCL as an independent type of load is increasingly criticized. For this

8
S. Becker, et al. Learning and Instruction 69 (2020) 101350

Fig. 4. Relative test scores for sub-concepts.

Table 7 Table 9
Results of multilevel regression analysis. Results of Fisher’s test.
Sub-concept F(1,237) p η2 1-β Model AIC BIC Fischer's C P

AltRate 1.656 0.199 – – alternative 134.804 245.237 70.804

VectQuant 0.102 0.750 – – Base 151.650 272.436 81.650 0.0044
RefSys 7.380 0.007 0.015 0.784
Notes: Data basis N = 233 (base model), N = 237 (alternative model).
Note: Data basis N = 244.
Table 10
Significant path coefficients.
Response Predictor DF se p Β

RefSys ECL 214.323 0.024 0.009 −0.185

VectQuant ICL 219.684 0.014 0.003 −0.195
ECL ICL 225.706 0.2053 < 10−3 0.253
ECL treatment 26.803 0.155 0.002 −0.231
Support commitment 224.973 0.062 < 10−3 0.460

Notes: Data basis N = 237, β = standardized estimate.

with the theory, ICL and ECL significantly impact the test scores, but
only regarding sub-concept RefSys (β = −0.185, p = 0.009) for ECL
and only regarding sub-concept VectQuant (β = −0.195, p = 0.003)
for ICL. As hypothesized, the treatment has a direct impact on ECL
(β = −0.231, p = 0.002). In contrast, no significant influence on
cognitive load or test scores could be demonstrated for teacher support,
which in turn is significantly influenced by teacher commitment
(β = 0.460, p < 10−3). A graphical illustration of the results can be
found in Fig. 6.

5. Discussion
Fig. 5. Intervention-induced cognitive load.
The cluster-randomized controlled trial presented in this work was
Table 8 aimed at the empirical investigation of the effectiveness of enhancing
Results of multilevel regression analysis. the experimental learning process with MERs using tablet-supported
video analysis in a realistic classroom setting. The digital experimental
Sub-scales F(1,233) p η2 1-β
tool was used in regular high school lessons for an essential topic in
ICL 1.558 0.216 – – mechanics instruction, uniform motion. It could be shown that sup-
ECL 18.420 < 10−3 0.078 1.000 porting the learning process with MERs using the tool led, in compar-
GCL 4.351 0.041 0.042 0.982
ison to a control group taught traditionally, to a significant reduction of
Note: Data basis N = 233. ECL and had a positive impact on conceptual understanding of the
description of movement with regard to a reference system. Moreover, a
reason, researchers argue for a two-factorial model that incorporates statistical analysis of causality related to the connection between ECL
only ICL and ECL (e.g., Leppink & van den Heuvel, 2015). To address reduction and increased conceptual understanding empirically sup-
this, an alternative model was generated that contains only ICL and ECL ported the theoretical assumptions of the method's learning effective-
as load variables and compared with the base model regarding the ness.
quality of the fit to the data.
5.1. Effects on cognitive load
4.4.2. Model selection and fit
To decide between the competing models, a Fisher's test was per- While ICL and thus the complexity of the learning content were
formed (see Table 9). The test reveals that the alternative model fits the comparable for both groups, the intervention-induced ECL was sig-
data significantly better, so that this model was chosen for path ana- nificantly lower for the students of the tablet group. Consistent with
lysis. The significant path coefficients are reported in Table 10. In line this, it was found that the germane load was significantly higher in this

9
S. Becker, et al.
Fig. 6. Analyzed path model with color-coded significant paths. (For interpretation of the references to color in this figure legend, the reader is referred to the Web
version of this article.)
group. This supports our research hypothesis that the automatic vi-
sualization of MERs in combination (which fulfills the contiguity
principle and thus avoids the split attention effect), the possibility for
the learners to switch between them as needed without time delay
(which fulfills the segmentation principle) and the dynamic linking of
the coordinate system with the movement diagrams contribute to a
reduction of ECL. This insight also provides an explanation for previous
positive research findings (Becker et al., 2018, 2019; Klein et al., 2018)
and indicates that enhancing the experimental learning process with
MERs using a mobile device can successfully support students during
the experimentation process. In summary, it can be stated that the
positive effects of video motion analysis on cognitive load remain with
the implementation in regular school lessons using a mobile device.
5.2. Effects on conceptual understanding
A positive effect of supporting the experimental learning process
with MERs using video motion analysis on the development of con-
ceptual understanding was found for sub-concept RefSys. This confirms
the results of our own preliminary study (Authors, 2018; Authors,
2019) and the research findings from Hockicko et al. (2014) and Wee
et al. (2015), who have already empirically demonstrated positive ef-
fects of this method on conceptual understanding in other fields of
mechanics. Since sub-concept RefSys refers to analysis of the movement
in dependence on the reference system, the dynamic linking of the re-
presentation of the coordinate with the movement diagrams could be
the cause for the positive effect found. In addition, the video analysis
application enables the learner to determine the origin and spatial or-
ientation of the coordinate system independently and to change it as
needed, which fulfills the segmentation principle and increases the
active interaction with MERs, which is a prerequisite for learners to
exploit the affordances of representations (Goldman, 2003; Wu &
Puntambekar, 2012).
5.3. Connection between cognitive load and conceptual understanding
Using the method of confirmatory path analysis, the causal re-
lationship between cognitive load variables and cognitive performance
variables has been statistically demonstrated. Path modeling delivers a
10
Learning and Instruction 69 (2020) 101350
significant negative path coefficient for ECL with sub-concept RefSys,
for which the significant group difference in favor of the tablet group
occurs, while for ICL with sub-concept AltRate. Following from this, it
can be assumed that learners only need assistance through support with
MERs in understanding sub-concept RefSys and that a reduction of ECL
is actually the cause of enhanced learning gains regarding this sub-
concept. In addition, a direct effect of the treatment, but not of the
teacher behavior, could be detected on ECL. This is also supported by
the fact that the effects found for the entire population could also be
demonstrated for the sub-population taught by a single teacher.
Concluding from this, only the treatment is responsible for the reduc-
tion of ECL and thus for the increased performance of the tablet group.
Thus, the research hypothesis that the learning effectiveness of video
analysis is based on the reduction of extraneous cognitive load can be
empirically verified.
5.4. Practical implications
First, we were able to implement video motion analysis using a
mobile device as a digital experimental tool in regular physics lessons
with little training required for the participating teachers and students.
This should encourage teachers to use this tool in their own lessons.
Moreover, the additional effort is limited, since the investment of only
one lesson is required for the introduction of the tool. Second, in our
study, we have shown that scaffolding learning with and from MERs
using tablet-supported video analysis is particularly conducive to un-
derstanding uniform motion relative to a reference system in an ex-
perimental learning process. This should be considered by teachers
when planning experimental learning environments on this topic. Since
uniform motion is an essential but relatively simple topic, the results of
our study also suggest that the positive effects of video analysis on more
complex topics in mechanics could have an even greater impact on the
effectiveness of the learning process. This opens up new possibilities for
teachers to create innovative experimental learning environments in
physics lessons.
5.5. Limitations
To be able to generalize the results of our study to other contexts,
S. Becker, et al. Learning and Instruction 69 (2020) 101350

we used a large sample size of students and conducted a cluster-ran-
domized controlled trial by randomly assigning complete courses to the
experimental or control group. To enable a fair comparison, the ex-
periments and learning time were identical for both groups. Moreover,
the learning materials, in terms of learning content, representations
used, and level of difficulty, were comparable. However, in interpreting
the study's findings, some limitations must be considered. First, it
should be noted that due to the quasi-experimental design, a complete
randomization could not be achieved, and the sample was unbalanced
regarding the preliminary grades and pre-test score. A balance re-
garding these variables was achieved using the PSM method, but this
reduced the sample size and thus the statistical significance of the
study. Second, the learning-related effect, based on the assessment
used, was demonstrated for a special lesson scenario covering a specific
topic of mechanics. Therefore, it remains unclear whether the positive
effects of the tablet-supported video analysis also hold for a more
complex topic. Given this research deficit and the practical significance
of the use of digital experimental tools, we hope that other researchers
will build on our results and conduct further studies to broaden the
research findings. Third, the students used video analysis for the first
time in class. Moreover, since the learning time was only four lessons,
an influence of the novelty effect (e.g., Clark, 1983) cannot be excluded.
Thus, the results of the study do not allow any conclusion on the
learning effectiveness of the video analysis for long-term teaching.
Again, we hope that our findings lead to further research, since video
analysis may support the experimental learning process in numerous
sub-topics of mechanics in regular school education. Fourth, the tea-
cher's behavior was measured via a retrospective self-report survey,
which is subject to some limitations and biases. Although the survey
was conducted immediately after the intervention, the students had to
remember the teacher's behavior in the past four lessons, and memory
recall errors may have occurred. Furthermore, the emotional state
during the completion of the questionnaire could have had a positive or
negative influence on the subjective assessment. Thus, the student-
teacher relationship may also have had an influence on the assessment
of teacher behavior. For example, if a student had a good relationship
with the teacher at the time of the assessment, he or she may have
perceived the teacher's behavior during the intervention as better than
a student who had a worse relationship with him or her. As a result,
self-report data tend to be biased. Future research should combine self-
reports with more objective sources of information, such as video-
graphy. Fifth, cognitive load was measured individually, but the stu-
dents conducted the experiments in pairs. A possible influence of joint
experimentation on the reduction of cognitive load therefore cannot be
clarified with the available data. Thus, it remains unclear whether the
use of digital learning tool also improves collaboration during experi-
mentation, which could also have an impact on learning performance.
To elucidate this possible effect, collaborative cognitive load
(Kirschner, Sweller, Kirschner, & Zambrano, 2018) could also be cap-
tured in follow-up studies and combined with process data from vi-
deographies.
5.6. Conclusion and outlook
We demonstrated in a study with high ecological validity firstly that
the enhancement of an experimental learning process with MERs using
a mobile device in a realistic classroom setting reduces the learners’ ECL
significantly. Second, we revealed by comparing the learning gains with
those of traditionally-taught students that the support with MERs also
leads to a deeper understanding of an important sub-concept of a cur-
ricularly relevant and essential topic of mechanics, uniform motion.
Third, we statistically verified that the reduction of ECL is the cause of
this enhanced learning gains and that this effect can be attributed to the
intervention and is not due to a teacher or school effect, which suggests
a generalizability of the study findings. The research results presented
in this work thus contribute to explaining the positive effects found in
preliminary studies for the use of the video analysis method on the basis
of fundamental theories of learning and provide important implications
for a transfer of this instructional approach to school practice. We
continue to believe that, especially in experiment-based learning pro-
cesses, in which the students are already cognitively burdened by the
experimental activity itself, the reduction of the extraneous load
through the use of technology can be reduced and thus the learning
success increased. We assume that the more challenging the topic is, the
greater the positive effect will be, implying great potential of the
method to sustainably improve experimental learning processes in
regular school lessons. This opens up new possibilities for teachers to
design innovative experiment-based learning environments using this
technology in physics lessons while also enabling them to meet the
growing demand for the increased use of digital tools in science edu-
cation. Moreover, the video analysis approach offers instructional de-
signers many more possibilities for supporting the learning process, but
in the present study, this has been omitted due to the strong controls to
ensure a fair comparison to traditional teaching. The students could
also, for example, demonstrate the movement process with their own
body, videograph it, and analyze it. This bodily experience could be
helpful for learning based on embodiment theory and might have a
positive effect on cognitive load. That embodied learning can have a
positive influence on the learning of physics concepts has already been
demonstrated, for example, by Brassel (1987) in the context of kine-
matics and by Johnson-Glenberg and Megowan-Romanowicz (2017) in
the context of the electric field. Therefore, future research should also
study the bodily experience when recording and analyzing the video of
a movement in the field of mechanics. Furthermore, the video analysis
method can also be used to increase student agency, as students can
also create, share, analyze, and publish their own film of any movement
in project work. This shows the potential of the methodology to pro-
mote learning processes in realistic teaching scenarios in a variety of
ways. With regard to the increasing digitization in the education sector,
we see the investigation of the learning effectiveness of individual ap-
proaches or their combination for the integration of video analysis in
the classroom learning process as fruitful areas for future research.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.learninstruc.2020.101350.

Appendix

1. Factor analysis of CU performance test: Subsequent to an examination of the necessary conditions (KMO, Bartlett's test of sphericity), we
performed an exploratory factor analysis (EFA) on the CU response pattern at the post-time point by using the R-package psych (vs. 1.8.12). In this
way, we found an underlying factor structure according to the three sub-concepts (Kaiser criterion, Scree-Plot, Parallel Analysis) and tested the
quality of this model by a confirmatory factor analysis (CFA) by using the R-package lavaan (vs. 0.6–3). The resulting model fits the data well, p
(χ2) = 0.251, CFI = 0.994, TLI = 0.991, RMSEA = 0.022, SRMR = 0.038.
2. Factor analysis of CL questionnaire: Subsequent to an examination of the necessary conditions (KMO, Bartlett's test of sphericity), we
performed an EFA with a subsequent CFA. As expected, we confirmed the three-factor structure (Kaiser criterion, Scree-Plot, Parallel Analysis) as
intended by Leppink et al. (2013) for the questionnaire used. Only one item (item CL6) could not be assigned to the factor structure, so we decided to

11
S. Becker, et al. Learning and Instruction 69 (2020) 101350

remove this item from the subsequent analysis. Later, we tested the quality of the model with the remaining nine items by CFA. The resulting model
fits the data well, p(χ2) = 0.111, CFI = 0.993, TLI = 0.990, RMSEA = 0.037, SRMR = 0.023.
3. Factor analysis of teacher behavior questionnaire: Subsequent to an examination of the necessary conditions (KMO, Bartlett's test of
sphericity), we first performed an EFA. As a result, we identified a two-factor structure with the sub-scales support and commitment. We tested the
quality of the two-factorial model by CFA. The resulting model fits the data well, p(χ2) = 0.146, CFI = 0.997, TLI = 0.981, RMSEA = 0.068,
SRMR = 0.015.
4. analysis of the data of both classes, which were taught by a single teacher who did both conditions
4.1 Learning gain

Fig. 1. Relative test scores for sub-concepts.

Table 1
Results of rmANOVA.

Sub-concept F(1,32) P η2 1-β

AltRate 0.034 0.855 – –

VectQuant 0.776 0.385 – –
RefSys 6.747 0.014 0.167 0.730

Note: Data basis NTG = 17, NCG = 18.

4.2 Cognitive load

Fig. 2. Intervention-induced cognitive load.

Table 2
Results of ANOVA.

Sub-scales F(1,32) P η2 1-β

ICL 5.158 0.030 0.139 0.635

ECL 17.430 < 10−3 0.353 0.989
GCL 13.270 < 10−3 0.293 0.959

Note: Data basis NTG = 17, NCG = 18.

12
S. Becker, et al.
References
Ainsworth, S. (2006). DeFT: A conceptual framework for considering learning with multiple
representations, 16(3), 183–198. https://doi.org/10.1016/j.learninstruc.2006.03.001.
Ainsworth, S. (2008). The educational value of multiple representations when learning
complex scientific concepts. In J. K. Gilbert, M. Reiner, & M. Nakhleh (Eds.).
Visualization: Theory and practice in science education (pp. 191–208). Netherlands:
Dordrecht: Springer.
Ainsworth, S. (2014). The multiple representation principle in multimedia learning. In R.
E. Mayer (Ed.). Cambridge handbooks in psychology. The Cambridge handbook of mul-
timedia learning (pp. 464–486). Cambridge University Press. https://doi.org/10.
1017/CBO9781139547369.024.
Al-Balushi, S. M., Al-Musawi, A. S., Ambusaidi, A. K., & Al-Hajri, F. H. (2017). The ef-
fectiveness of interacting with scientific animations in chemistry using mobile de-
vices on grade 12 students' spatial ability and scientific reasoning skills. Journal of
Science Education and Technology, 26(1), 70–81. https://doi.org/10.1007/s10956-
016-9652-2.
Becker, S., Klein, P., Gößling, A., & Kuhn, J. (2019). Förderung von Konzeptverständnis
und Repräsentationskompetenz durch Tablet-PC-gestützte Videoanalyse [Promoting
conceptual understanding and representational competence through tablet PC-based
video analysis]. Zeitschrift Für Didaktik Der Naturwissenschaften, 54(7), 440. https://
doi.org/10.1007/s40573-019-00089-4 426.
Becker, S., Klein, P., & Kuhn, J. (2018a). Promoting students' conceptual knowledge using
video analysis on tablet computers. PERC Proceedings. https://doi.org/10.1119/perc.
2018.pr.Becker.
Becker, S., Klein, P., Kuhn, J., & Wilhelm, T. (2018b). Viana analysiert Bewegungen
[Viana analyzes motions]. Physik in Unserer Zeit, 49(1), 46–47. https://doi.org/10.
1002/piuz.201801502.
Becker, S., Thees, M., & Kuhn, J. (2018c). The dynamics of the magnetic linear accelerator
examined by video motion analysis. The Physics Teacher, 56(7), 484–485. https://doi.
org/10.1119/1.5055338.
Beichner, R. J. (1994). Testing student interpretation of kinematics graphs. American
Journal of Physics, 62(8), 750–762. https://doi.org/10.1119/1.17449.
Brasell, H. (1987). The effect of real-time laboratory graphing on learning graphic re-
presentations of distance and velocity. Journal of Research in Science Teaching, 24,
385–395. https://doi.org/10.1002/tea.3660240409.
Chiu, J. L., & Linn, M. C. (2014). Supporting knowledge integration in chemistry with a
visualization-enhanced inquiry unit. Journal of Science Education and Technology,
23(1), 37–58. https://doi.org/10.1007/s10956-013-9449-5.
Clark, R. E. (1983). Reconsidering research on learning from media. Review of Educational
Research, 53(4), 445–459. https://doi.org/10.3102/00346543053004445.
Collins, A. (1997). National science education standards: Looking backward and forward.
The Elementary School Journal, 97(4), 299–313. https://doi.org/10.1086/461867.
DeBoer, G. E. (1991). A history of ideas in science education. New York: Teachers College
Press.
Dreyhaupt, J., Mayer, B., Keis, O., Öchsner, W., & Muche, R. (2017). Cluster-randomized
studies in educational research: Principles and methodological aspects. GMS Journal
for Medical Education, 34(2), 1–25. https://doi.org/10.3205/zma001103.
Fan, X., Nowell, L., & D (2011). Using propensity score matching in educational research.
Gifted Child Quarterly, 55(1), 74–79. https://doi.org/10.1177/0016986210390635.
Gilbert, J. K. (2005). Visualization: A metacognitive skill in science and science educa-
tion. In J. K. Gilbert (Ed.). Visualization in science education (pp. 9–27). Dordrecht:
Springer.
Glazer, N. (2011). Challenges with graph interpretation: A review of the literature. Studies
in Science Education, 47, 183–210. https://doi.org/10.1080/03057267.2011.605307.
Goldman, S. R. (2003). Learning in complex domains: When and why do multiple re-
presentations help. Learning and Instruction, 13(2), 239–244. https://doi.org/10.
1016/s0959-4752(02)00023-3.
Guo, S., Fraser, W., & M (2010). Propensity score analysis: Statistical methods and applica-
tions (advanced quantitative techniques in the social sciences: Vol. 11). Thousand Oaks,
CA: Sage Publications, Inc.
Hadie, S. N. H., & Yusoff, M. S. B. (2016). Assessing the validity of the cognitive load scale
in a problem-based learning setting. Journal of Taibah University Medical Sciences,
11(3), 194–202. https://doi.org/10.1016/j.jtumed.2016.04.001.
Haury, D. L., & Rillero, P. (1994). Perspectives of hands-on science teaching. Columbus, OH:
ERIC Clearinghouse for Science, Mathematics and Environmental Education.
Hochberg, K., Becker, S., Louis, M., Klein, P., & Kuhn, J. (2020). Using smartphones as
experimental tools—a follow-up: Cognitive effects by video analysis and reduction of
cognitive load by multiple representations. Journal of Science Education and
Technology, 29(2), 303–317. https://doi.org/10.1007/s10956-020-09816-w.
Hochberg, K., Kuhn, J., & Müller, A. (2018). Using smartphones as experimental tools-
effects on interest, curiosity, and learning in physics education. Journal of Science
Education and Technology, 27(5), 385–403. https://doi.org/10.1007/s10956-018-
9731-7.
Hockicko, P., Trpišová, B., & Ondruš, J. (2014). Correcting students' misconceptions
about automobile braking distances and video analysis using interactive program
Tracker. Journal of Science Education and Technology, 23(6), 763–776. https://doi.
org/10.1007/s10956-014-9510-z.
Horz, H., Schnotz, W., Plass, J. L., Moreno, R., & Brünken, R. (2009). Cognitive load in
learning with multiple representations. In J. Plass, R. Moreno, & R. Brünken (Eds.).
Cognitive load theory (pp. 229–252). Cambridge: Cambridge University Press.
Hox, J. J., Moerbeek, M., & an de Schoot, R. (2018). Multilevel analysis: Techniques and
applications (3rd ed.). New York, NY, US: Routledgehttps://doi.org/10.4324/
9781315650982.
13
Learning and Instruction 69 (2020) 101350
Husnaini, S. J., & Chen, S. (2019). Effects of guided inquiry virtual and physical la-
boratories on conceptual understanding, inquiry performance, scientific inquiry self-
efficacy, and enjoyment. Physical Review Physics Education Research, 15(1), https://
doi.org/10.1103/physrevphyseducres.15.010119.
Johnson-Glenberg, M. C., & Megowan-Romanowicz, C. (2017). Embodied science and
mixed reality: How gesture and motion capture affect physics education. Cognitive
Research: Principles and Implications, 2(1), 24. https://doi.org/10.1186/s41235-017-
0060-9.
de Jong, T. (1991). Learning and instruction with computer simulations. Education and
Computing, 6(3), 217–229. https://doi.org/10.1016/0167-9287(91)80002-F.
de Jong, T., Ainsworth, S., Dobson, M., van der Hulst, A., Levonen, J., Reimann, P., &
Swaak, J. (1998). Acquiring knowledge in science and mathematics: The use of
multiple representations in technology-based learning environments. In M. W. van
Someren (Ed.). Learning with multiple representations (Advances in learning and in-
struction series (pp. 9–41). Oxford: Elsevier Science.
de Jong, T., Linn, M. C., & Zacharia, Z. C. (2013). Physical and virtual laboratories in
science and engineering education. Science, 340(6130), 305–308. https://doi.org/10.
1126/science.1230579.
Kapici, H. O., Akcay, H., & de Jong, T. (2019). Using hands-on and virtual laboratories
alone or together: Which works better for acquiring knowledge and skills? Journal of
Science Education and Technology, 28(3), 231–250. https://doi.org/10.1007/s10956-
018-9762-0.
Kirschner, P. A., Sweller, J., Kirschner, F., & Zambrano, R. J. (2018). From cognitive load
theory to collaborative cognitive load theory. International Journal of Computer-
Supported Collaborative Learning, 13(2), 213–233. https://doi.org/10.1007/s11412-
018-9277-y.
Klein, P., Kuhn, J., & Müller, A. (2018). Förderung von Repräsentationskompetenz und
Experimentbezug in den vorlesungsbegleitenden Übungen zur Experimentalphysik
[Promotion of representational competence and experiment reference in the lecture-
accompanying exercises on experimental physics]. Zeitschrift für Didaktik der
Naturwissenschaften. 24(1), 17–34. https://doi.org/10.1007/s40573-018-0070-2.
Klein, P., Müller, A., & Kuhn, J. (2017). Assessment of representational competence in
kinematics. Physical Review Physics Education Research, 13(1), https://doi.org/10.
1103/PhysRevPhysEducRes.13.010132 10132.
Kuhn, J., & Vogt, P. (2015). Smartphones & co. in physics education: Effects of learning
with new media experimental tools in acoustics. In W. Schnotz, A. Kauertz, H.
Ludwig, A. Müller, & J. Pretsch (Eds.). Multidisciplinary research on teaching and
learningLondon: Palgrave Macmillan. https://doi.org/10.1057/9781137467744 UK.
Kwon, O. N. (2002). The effect of calculator-based ranger activities on students' graphing
ability. School Science & Mathematics, 102(2), 57–67. https://doi.org/10.1111/j.
1949-8594.2002.tb17895.x.
Lafleur, A., Côté, L., & Leppink, J. (2015). Influences of OSCE design on students' diag-
nostic reasoning. Medical Education, 49, 203–214. https://doi.org/10.1111/medu.
12635.
Leppink, J. (2017). Cognitive load theory: Practical implications and an important
challenge. Journal of Taibah University Medical Sciences, 12(1), 1–7. https://doi.org/
10.1016/j.jtumed.2016.08.007.
Leppink, J., Paas, F., van Gog, T., van der Vleuten, C. P. M., & van Merriënboer, J. J. G.
(2014). Effects of pairs of problems and examples on task performance and different
types of cognitive load. Learning and Instruction, 30(c), 32–42. https://doi.org/10.
1016/j.learninstruc.2013.12.001.
Leppink, J., Paas, F., van der Vleuten, C. P. M., van Gog, T., & van Merriënboer, J. J. G.
(2013). Development of an instrument for measuring different types of cognitive
load. Behavior Research Methods, 45(4), 1058–1072. https://doi.org/10.3758/
s13428-013-0334-1.
Leppink, J., & van den Heuvel, A. (2015). The evolution of cognitive load theory and its
application to medical education. Perspectives on Medical Education, 4(3), 119–127.
https://doi.org/10.1007/s40037-015-0192-x.
Lichtenberger, A., Wagner, C., Hofer, S. I., Stern, E., & Vaterlaus, A. (2017). Validation
and structural analysis of the kinematics concept test. Physical Review Physics
Education Research, 13(1), https://doi.org/10.1103/PhysRevPhysEducRes.13.010115
010115.
Mayer, R. E. (2009). Multimedia learning (2nd ed.). New York, NY: Cambridge University
Presshttps://doi.org/10.1017/CBO9780511811678.
Mayer, R. E., & Moreno, R. (2002). Animation as an aid to multimedia learning.
Educational Psychology Review, 14(1), 87–99. https://doi.org/10.1023/
A:1013184611077.
Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive load in multimedia
learning. Educational Psychologist, 38, 43–52. https://doi.org/10.1207/
S15326985EP3801_6.
Mayer, R. E., & Pilegard, C. (2014). Principles for managing essential processing in
multimedia learning: Segmenting, pre-training, and modality principles. In R. Mayer
(Ed.). The Cambridge handbook of multimedia learning (Cambridge handbooks in psy-
chology (pp. 316–344). Cambridge: Cambridge University Press. https://doi.org/10.
1017/CBO9781139547369.016.
van Meter, P., List, A., Lombardi, D., & Kendeou, P. (2020). Handbook of learning from
multiple representations and perspectives. New York, NY: Routledgehttps://doi.org/10.
4324/9780429443961.
Nieminen, P., Savinainen, A., & Viiri, J. (2010). Force concept inventory-based multiple-
choice test for investigating students' representational consistency. Physical Review
Special Topics - Physics Education Research, 6. https://doi.org/10.1103/
PhysRevSTPER.6.020109 020109.
Oliveira, A., Behnagh, R. F., Ni, L., Mohsinah, A. A., Burgess, K. J., & Guo, L. (2019).
Emerging technologies as pedagogical tools for teaching and learning science: A lit-
erature review [special issue]. Human Behavior and Emerging Technologies, 1(2),
S. Becker, et al.
149–160. https://doi.org/10.1002/hbe2.141.
Plass, J. L., Homer, B. D., & Hayward, E. O. (2009). Design factors for educationally
effective animations and simulations. Journal of Computing in Higher Education, 21(1),
31–61. https://doi.org/10.1007/s12528-009-9011-x.
Pundak, D., & Rozner, S. (2007). Empowering engineering college staff to adopt active
learning methods. Journal of Science Education and Technology, 17(2), 152–163.
https://doi.org/10.1007/s10956-007-9057-3.
Rieber, L. P. (1990). Animation in computer-based instruction. Educational Technology
Research & Development, 38(1), 77–86. https://doi.org/10.1007/BF02298250.
Rittle-Johnson, B., Siegler, R. S., & Alibali, M. W. (2001). Developing conceptual un-
derstanding and procedural skill in mathematics: An iterative process. Journal of
Educational Psychology, 93(2), 346–362. https://doi.org/10.1037/0022-0663.93.2.
346.
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in
observational studies for causal effects,. Biometrika, 70(1), 41–55. https://doi.org/10.
1093/biomet/70.1.41.
Rutten, N., van Joolingen, W. R., & van der Veen, J. T. (2012). The learning effects of
computer simulations in science education. Computers & Education, 58(1), 136–153.
https://doi.org/10.1016/j.compedu.2011.07.017.
Seufert, T. (2003). Supporting coherence formation in learning from multiple re-
presentations. Learning and Instruction, 13(2), 227–237. https://doi.org/10.1016/
S0959-4752(02)00022-1.
Shah, P., & Hoeffner, J. (2002). Review of graph comprehension research: Implications
for instruction. Educational Psychology Review, 14(47), https://doi.org/10.1023/
A:1013180410169.
Shipley, B. (2000). Cause and correlation in biology: A user's guide to path analysis, structural
equations, and causal inference. Oxford: Oxford University Presshttps://doi.org/10.
1017/CBO9780511605949.
Shipley, B. (2009). Confirmatory path analysis in a generalized multilevel context.
Ecology, 90(2), 363–368. https://doi.org/10.1890/08-1034.1.
Shipley, B., & Douma, J. C. (2019). Generalized AIC and chi-squared statistics for path
models consistent with directed acyclic graphs. Ecology. https://doi.org/10.1002/
ecy.2960.
Singer, S. R., Hilton, M. L., & Schwiengruber (2005). America's lab report: Investigations in
school scienceWashington, DC: National Academies Press.
Smetana, L. K., & Bell, R. L. (2012). Computer simulations to support science instruction
and learning: A critical review of the literature. International Journal of Science
14
Learning and Instruction 69 (2020) 101350
Education, 34(9), 1337–1370. https://doi.org/10.1080/09500693.2011.605182.
Songer, N. B. (1996). Exploring learning opportunities in coordinated network-enhanced
classrooms: A case of kids as global scientists. The Journal of the Learning Sciences,
5(4), 297–327. https://doi.org/10.1207/s15327809jls0504_1.
Sung, Y.-T., Chang, K.-E., & Liu, T.-C. (2016). The effects of integrating mobile devices
with teaching and learning on student's learning performance: A meta-analysis and
research synthesis. Computers & Education, 94, 252–275. https://doi.org/10.1016/j.
compedu.2015.11.008.
Sweller, J., van Merriënboer, J. J. G., & Paas, F. (1998). Cognitive architecture and in-
structional design. Educational Psychology Review, 10(3), 251–296. https://doi.org/
10.1023/A:1022193728205.
Sweller, J., van Merriënboer, J. J. G., & Paas, F. (2019). Cognitive architecture and in-
structional design: 20 years later. Educational Psychology Review, 31(2), 261–292.
https://doi.org/10.1007/s10648-019-09465-5.
Treagust, D., Duit, R., & Fischer, H. (2017). Multiple representations in physics education.
Dordrecht: Springerhttps://doi.org/10.1007/978-3-319-58914-5.
Tytler, R., Prain, V., Hubber, P., & Waldrip, B. (2013). Constructing representations to learn
in science. Rotterdam: Sense Publishershttps://doi.org/10.1007/978-94-6209-203-7.
Verschaffel, L., de Corte, E., Ton, D. J., & Jan, E. (2010). Use of representations in reasoning
and problem solving. New York: Routledgehttps://doi.org/10.1119/1.1286662.
Vosniadou, S. (2007). Conceptual change and education. Human Development, 50(1),
47–54. https://doi.org/10.4324/9780203847824.
Wee, L. K., Tan, K. K., Leong, T. K., & Tan, C. (2015). Using tracker to understand toss up
and free fall motion: A case study. Physics Education, 50(4), 436–442. https://doi.org/
10.1088/0031-9120/50/4/436.
White, B. Y., & Frederiksen, J. R. (1998). Inquiry, modeling, and metacognition: Making
science accessible to all students. Cognition and Instruction, 16(1), 3–118. https://doi.
org/10.1207/s1532690xci1601_2.
Wilcox, B. R., & Lewandowski, H. J. (2017). Students' views about the nature of experi-
mental physics. Physical Review Physics Education Research, 13. https://doi.org/10.
1103/PhysRevPhysEducRes.13.020110 020110.
Wu, H.-K., & Puntambekar, S. (2012). Pedagogical affordances of multiple external re-
presentations in scientific processes. Journal of Science Education and Technology,
21(6), 754–767. https://doi.org/10.1007/s10956-011-9363-7.
Zydney, J. M., & Warner, Z. (2016). Mobile apps for science learning: Review of research.
Computers & Education, 94(c), 1–17. https://doi.org/10.1016/j.compedu.2015.11.
001.