Learning differences and eye fixation patterns in virtual and physical science laboratories

Computers & Education 82 (2015) 191e201

Contents lists available at ScienceDirect

Computers & Education journal homepage: www.elsevier.com/locate/compedu

Learning differences and eye fixation patterns in virtual and physical science laboratories Kuei-Pin Chien a, Cheng-Yue Tsai b, Hsiu-Ling Chen b, Wen-Hua Chang c, Sufen Chen b, * a

Department of Athletics, National Taiwan University of Science and Technology, 43, Sec.4, Keelung Rd., Taipei 106, Taiwan, ROC Graduate Institute of Digital Learning and Education, National Taiwan University of Science and Technology, 43, Sec.4, Keelung Rd., Taipei 106, Taiwan, ROC c Graduate Institute of Science Education, National Taiwan Normal University, 88. Sec. 4, Tingchou Rd., Taipei 116, Taiwan, ROC b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 25 September 2014 Received in revised form 25 November 2014 Accepted 26 November 2014 Available online 4 December 2014

This project analyzed high school students' performance and eye movement while learning in a simulation-based laboratory (SBL) and a microcomputer-based laboratory (MBL). Although the SBL and the MBL both used computers to collect, graph, and analyze data, the MBL involved manual manipulation of concrete materials, whereas the SBL displayed everything on a monitor. Fifty senior high school students at three urban public high schools in Taipei were randomly assigned to the MBL and SBL settings. The participants conducted the Boyle's Law experiment with an accompanying worksheet and completed pre- and post-conceptual tests. FaceLAB and ASL MobileEye were used to record each participant's eye movements in the SBL and MBL settings, respectively. The results showed that lower achievers improved significantly from the pre-to post-conceptual tests. The SBL group tended to carry out more experiments. Moreover, the MBL group's performance on the worksheet was moderately correlated with their post-test. However, this correlation was not found for the SBL group. Furthermore, at the beginning of the laboratories, the SBL group had a higher percentage of fixations with longer fixation duration, which implies more attention to and deeper cognitive processing of the equipment and running experiments, while the MBL group focused on the worksheet. This study concludes that, for elearning like SBLs, students tend to start off doing an experiment, and then think about the questions on the worksheets, whereas for physical laboratories like MBLs, they tend to think before doing. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Simulations Applications in subject areas Secondary education

1. Introduction This project investigates high school students' conceptual learning and eye fixation patterns in physical and virtual laboratories. In this study, we use the term physical laboratory to refer to a microcomputer-based laboratory (MBL) in which students manipulate concrete objects with their hands, and collect and analyze data using sensors and handheld computers. On the other hand, virtual laboratory refers to a simulation-based laboratory (SBL) in which students manipulate objects and variables to conduct experiments on desktop computers. MBLs and SBLs, two alternatives to traditional laboratories, are technology-enhanced. Both support quick data acquisition and real-time graphing. Multiple trials can be carried out within a certain period of time. Moreover, more time is saved for higher-order cognitive tasks such as designing experiments, justifying decisions, evaluating experimental results, interpreting data and graphs, and building models. Therefore, they make authentic experimentation feasible in a typical class period. MBLs and SBLs are similar in many respects. The major difference is the interaction with concrete versus virtual materials. In an MBL, students measure properties such as temperature, light, force, etc. with probes. The probes output data in voltages. With analog-to-digital converters, students can record and analyze data on computers or handheld devices such as tablets and smart phones. Previous studies have revealed that MBLs are more effective in terms of promoting conceptual understanding than traditional laboratories (Brassell, 1987; Nakhleh & Krajcik, 1994; Nicolaou, Nicolaidou, Zacharia, & Constantinou, 2007). MBLs have several features that promote

* Corresponding author. Tel.: þ886 22737 6362; fax: þ886 22737 6433. E-mail address: [email protected] (S. Chen). http://dx.doi.org/10.1016/j.compedu.2014.11.023 0360-1315/© 2014 Elsevier Ltd. All rights reserved.

192

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

learning. The first distinguishing feature of MBLs is that they graph data instantaneously (Nicolaou et al., 2007; Pierri, Karatrantou, & Panagiotakopoulos, 2008). The handheld device can display the graph in real time. Real-time graphing lessens short term memory loading (Brassell, 1987) and enhances students' ability to interpret graphs (Bisdikian & Psillos, 2002; Linn & Songer, 1991). As a result, students can better link phenomena to scientific theories (Bisdikian & Psillos, 2002). Furthermore, MBLs save an extensive amount of time spent on tedious work such as data recording and organizing, allowing the students to repeat the experiments (Russell, Lucas, & McRobbie, 2004) and giving them time (Hucke & Fischer, 2002) and cognitive resources for abstract cognitive thinking. Consequently, the students' ability to understand scientific concepts is enhanced. An MBL involves the hands-on feature of traditional laboratories, i.e., physical manipulation of equipment, and also some of the strengths of SBLs. When a traditional laboratory is replaced with an MBL, the number of experiments conducted during the same period of time is greatly enhanced and would be close to the number of experiments done in an SBL (Chen, 2014). Moreover, the real-time display of graphs also brings MBLs closer to the dynamic visualization of simulations. In short, the replacement of traditional laboratories with MBLs in the present study makes the experimental (virtual laboratory) and control (physical laboratory) groups comparable in terms of quick data acquisition, dynamic visualization of data, and error rates. With these extraneous variables controlled, this experimental design provides robust evidence of the efficiency of the SBL and physical manipulatives. Our previous study revealed that they are equally effective in terms of conceptual learning (Chen, 2014). The present study goes further in examining students' attention and cognitive processing in these environments. This study pays particular attention to students' eye fixation patterns while conducting the Boyle's Law experiment. An often quoted proverb regarding learning is that “I hear and I forget, I see and I remember, I do and I understand.” Visual attention is an important issue in , 1987), mechanisms for directing attention to a learning. According to the premotor theory of attention (Rizzolatti, Riggio, Dascola, & Umilta location might be similar to mechanisms for preparing an eye movement. Experiments have also found overlapping neural systems for covert spatial attention with overt oculomotor shifts (Nobre, Gitelman, Dias, & Mesulam, 2000). Thus, this project detects learners' eye movements to study their attention. The results will contribute to the understanding of learning in laboratories, and the tradeoffs between the two types of laboratories. This knowledge will help science educators to develop laboratory curricula and integrate physical and virtual laboratories. 2. Students' learning in virtual and physical laboratories 2.1. Theoretical base Researchers have proposed different theories to explain learning in virtual and physical laboratories. On the one hand, SBLs can involve multiple representations, especially dynamic visualizations and analogies for abstract and unobservable phenomena, which help learners integrate theories, symbolic equations, graphs, processes, and phenomena (Olympiou, Zacharias, & de Jong, 2013; Ploetzner, Lippitsch, Galmbacher, Heuer, & Scherrer, 2009; Trey & Khan, 2008; Trindade, Fiolhais, & Almeida, 2002). Learners receive rich information and support. Furthermore, they can make clear observations (Renken & Nunez, 2013). For example, time and space may be rescaled for macroand micro-phenomena; graphs can be shown in real time to link phenomena to scientific theories; and data are clean and matched with scientific theories. Clear observations help to expose cognitive conflicts between the observed evidence and learners' alternative conceptions, and thus promote conceptual change. On the other hand, the grounded/embodied cognition theory suggests that physical laboratories enhance learning. The theory considers cognition as interactive, embodied, and embedded (Calvo & Gomila, 2008). “Cognition depends upon the kind of bodily experiences we have as we interact through our perceptual and motor systems with the environment” (Bivall, Ainsworth, & Tibell, 2011, p. 702). Thought and knowledge emerge from physical and social interactions between learners and the surrounding environment. Consequently, physical manipulatives are important for laboratory learning. The abovementioned theories support SBLs and physical laboratories from different perspectives of learning context/environment and information delivery (multiple representations versus multiple modalities). They provide a basis for understanding learning outcomes and processes in SBLs and physical laboratories. 2.2. Empirical evidence A few researchers have succeeded in comparing virtual and physical laboratories under conscious control of extraneous variables. Most studies reveal that they are equally effective in terms of learning concepts, learning the control of variables, and fostering confidence (de Jong, Linn, & Zacharia, 2013). For example, Klahr and colleagues have conducted a series of experiments on this issue. Their studies show no statistical difference in knowledge of controlling variables and causal factors and confidence between students manipulating physical materials and computer simulations. Triona and Klahr (2003) compared 46 fourth and fifth graders manipulating physical materials with an equal number of students working on computer simulations. Both groups received the same instruction and practiced the control of variables strategy to design and interpret simple unconfounded experiments. The appearance of the materials was featured as closely as possible in both groups. The two groups were not significantly different in any of their test scores, including for their pre- and post-test knowledge of the control of variables as well as for a far transfer test given 7 months later. The results suggest that a well-designed virtual environment that preserves the essential experience of manipulation and experimentation is as effective as hands-on manipulation of physical materials. Next, Klahr, Triona, and Williams (2007) tested a different type of instruction and a different type of knowledge with an older age group (7th and 8th grades). Instead of focusing on domain-general knowledge of controlling variables, domain-specific knowledge about features of good running cars was spotlighted. The visual and tactile information presented in the virtual condition was substantially different from the physical condition; cartoon-like cars, rather than photographs or video images, were used. The pre-/post-test of knowledge of causal factors, the ability to design optimal cars, and confidence in their knowledge revealed that the students learned equally well using physical

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

193

and virtual materials in conditions with either a fixed number of cars allowed to be assembled and tested or a fixed time duration. Tactile information is clearly not essential for learning in laboratories. Similarly, through careful control of the curriculum, instruction method, resource accessibility and time for experiments, Zacharia and Constantinou (2008) compared undergraduate students' understanding of temperature and change in temperature in virtual and physical laboratories. They found that pre-service elementary teachers performed equally well in both modes of experimentation. No significant difference was observed in concept tests. Likewise, Wiesner and Lan (2004) found that engineering students performed similarly for conceptual understanding in both physical and virtual laboratories. The students, however, preferred to have some simulations, rather than being totally computer-based in their experimentation. Recent research with undergraduate students has suggested that, for learning science concepts, the best approach is a combination of virtual and physical laboratories. Zacharia (2007) found that, for pre-service elementary school teachers, a virtual laboratory alone or in combination with a physical laboratory enhanced their understanding of electric circuits more than the physical laboratory alone. Zacharia stated possible reasons for this improvement. Some phenomena such as charge flow were made visible in the virtual laboratory. This abstract representation in addition to concrete objects contributed to their understanding. Moreover, participants in the virtual laboratory could perform more experiments within the same period of time and receive immediate feedback. Zacharia, Olympiou, and Papaevripidou (2008) carried out the same research protocol with 66 undergraduates on the temperature and heat unit of the Physics by Inquiry curriculum. They attained the same findings, i.e., the virtual alone and virtual-physical combined settings yielded better results in terms of conceptual understanding. However, their further research on the same topic with undergraduates found no significant difference in the concept tests for virtual-only, physical-only, and combinations of the two modes of experimentation (Zacharia & Constantinou, 2008; Zacharia & Olympiou, 2011). In another study, a significant effect was revealed in the combination of SBL and physical manipulation for the light and color unit (Olympiou & Zacharia, 2012). Overall, the results of most studies favor a combination of virtual and physical laboratories for college-level students. Additionally, Chini, Madsen, Gire, Rebello, and Puntambekar (2012) remarked that the physical-virtual and virtual-physical sequences of experimentation do not differ in their effects on conceptual learning. For elementary students, a combination of simulation and laboratory activities also generated a more positive effect on understanding of simple electricity, compared with using the simulation or hands-on laboratory alone (Jaakkola & Nurmi, 2008). Jaakkola and Nurmi (2008) proposed that, on the one hand, simulations promote an understanding of the theoretical principles of electricity. On the other hand, physical experimentation plays an essential role in conceptual change by demonstrating the application of the principles in the real world. The different roles of virtual and physical experimentation may be more discernible in young children because of less prior experience of physical manipulation of objects. For example, Zacharia, Loizou, and Papaevripidou (2012) showed that kindergarteners who had incorrect prior knowledge of the use of a beam balance needed physical manipulation of the objects to learn to use the beam balance to compare and differentiate objects according to their weight. The results of studies with undergraduate students and young learners suggest that the effectiveness of SBLs is regulated by at least two factors: availability of the abstract representations, and prior experience of manipulating the related physical objects. Abstract representations such as electron flow and light rays promote conceptual understanding in a simulation learning environment. Moreover, SBLs alone may be adequate if students have already previously manipulated the concrete objects such as thermometers, electronic circuits, running cars, etc. Both factors were controlled in the current study. The comparison groups were offered similar representations and had not conducted related laboratories before. In addition to prior experience of physical manipulatives, students' learning in a simulation is also affected by their prior knowledge of the associated science concepts. Students with more accurate prior knowledge could construct more solid mental models (Liu, Andre, & Greenbowe, 2008). Prior knowledge influences mental modeling and information processing, which lead to conceptual understanding. For example, Olympiou et al. (2013) revealed that, for less complicated phenomena, high achievers could construct abstract concepts in a simulation without representations of abstract objects. Nevertheless, representations of abstract objects benefit high achievers' learning of complicated phenomena. Gijlers and de Jong (2005) also found that domain-specific prior knowledge is crucial for knowledge development and interpretation of experimental results in computer simulations. Moreover, students' number of on-task utterances is moderately correlated with the amount of accurate prior knowledge. Prior knowledge may mediate the learning effects and, thus, is taken into account in the current study. SBLs are not advocated without caution, in particular if educational objects beyond conceptual understanding are concerned. Chen (2010) has pointed out that most SBLs convey an oversimplified view of scientific inquiry, and deviate from the common objectives of science education, such as cultivating students' ability to conduct authentic scientific inquiry and to solve everyday problems, or promoting students' scientific literacy. Likewise, Renken and Nunez (2013) indicated that clear observation characterized in SBLs is insufficient for conceptual change. Adequate experiments (control of variables and multiple trials of experiments) are determinant for conceptual understanding. Students had significantly better performance in controlling variables in a physical laboratory, whereas they ran more trials in an SBL. They concluded that “isolated physical science computer simulations are no better in informing conceptual knowledge than are reallife experiments, and in fact, they may be worse in encouraging students' careful experimental design” (p. 20). In sum, virtual and physical laboratories may play different roles in learning. More research is needed to understand the functions and roles of SBLs for students who have not had physical experiences relevant to a topic, and for cultivating a wide scope of learning objectives. In-depth analyses of the learning processes in virtual and physical laboratories would help educators to realize their optimal use. 3. This study The research questions of this study include how low and high achievers differ in the following aspects in the SBL and MBL settings: (1) performance, (2) learning behaviors, (3) attention distribution, and (4) cognitive processing. First of all, their performance was measured by pre- and post-tests of concepts and a worksheet which assessed the participants' inquiry practices during the laboratory. Secondly, learning behaviors could be broad. We focused on the number of experimental trials they conducted in the SBL or MBL, and the total time spent on tasks. Finally, their attention distribution and cognitive processing were measured by eye movement. The following section discusses the measurement and interpretation of eye movement.

194

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

3.1. Eye movement Much of a student's cognitive processing at work may not be expressed explicitly or verbally (Garber & Goldin-Meadow, 2002; Reiner & Gilbert, 2000). Researchers have developed different methods to supplement test, verbal, and self-reported data to understand participants' learning and reasoning. Eye movement tracking is an appropriate method in terms of our research context and purposes. Rayner (1998) indicates that eye movement tracking is an objective assessment measure and provides quantitative data on reading processes and visual and attentional abilities. It has been used to analyze students' learning with complex animations (Boucheix & Lowe, 2010), graphics (Canham & Hegarty, 2010), multimedia (She & Chen, 2009), and problem solving (Tsai, Hou, Lai, Liu, & Yang, 2012), to name a few. This process-related measure could advance research on comparison between learning in virtual and physical laboratories by analyzing the perceptual and cognitive characteristics and by providing the psychological basis for laboratory design and instruction. Eye movements and attention are closely linked for complex information processes (Rayner, 1998). Although it is possible to move attention without eye movement (Posner, 1980), in a complex task such as reading or lab work, it is more efficient to move the eyes than to move attention (He & Kowler, 1992). Inhoff and Radach (1998) also pointed out that oculomotor measures are sensitive to subjects' perceptual and cognitive processes and can be taken in relatively natural conditions. Eye movement data are typically interpreted based on the eye-mind assumption (Just & Carpenter, 1980). A fixation means the attention and cognitive processing at that location. Fixation duration was determined by the perceptual and cognitive analysis of the information at that location. Eye fixations reflect attention, i.e., more fixations indicate more attention distribution, and fixation durations reflect cognitive processing, i.e., longer fixation durations indicate deeper processing or greater processing difficulties (Just & Carpenter, 1980; Rayner, 1998; Tsai et al., 2012). 3.2. Research context Boyle's Law describes the relationship between the volume (V) and pressure (P) of ideal gas. The classical Boyle's experiment uses a Ushaped glass tube and mercury. In this study, the equipment was replaced with a syringe and a pressure sensor connected by a plastic tube. In the SBL, all objects were virtual whereas, in the MBL, they were real. Fig. 1 displays the SBL and MBL settings. The SBL group collected data on Internet PCs. To take a data point, they clicked an action to pull or push the piston of the syringe to a position and then clicked “Take data”;

Fig. 1. The SBL and MBL settings and look zones.

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

195

the computer recorded both the V and P. In the MBL, data were collected by a handheld device (GLX Explore from PASCO). To take a data point, the students manually pulled or pushed the piston to a position, keyed in the V, and the GLX recorded the P simultaneously. The SBL was programmed as closely to the MBL as possible. The range of data for the MBL was normally between the pressure of 30 kpa and 400 kpa, depending on how much a student could push or pull the piston. Therefore, the range for the SBL was set between 30 kpa and 400 kpa. The error rate was also similar for both laboratories. It included systematic and random errors and intermolecular forces due to non-ideal gas. The SBL involved errors so that the students could improve the experiments as was also possible in the MBL. Finally, both laboratories automatically created real-time graphs. Students could choose to observe VeP, 1/VeP, or Ve1/P graphs one at a time. A detailed comparison between the two environments can be found in Chen (2014). 4. Methods 4.1. Subjects The participants were fifty 11th grade students (aged 16e17) from three urban high schools. Senior high schools in Taiwan are stratified according to their students' scores on the entrance examination. Two of the participating schools were at the average level. The other school was a vocational high school, which was less academically-oriented as the students normally have lower academic achievement than those of senior high schools. The participants had received lectures on Boyle's Law in the previous semester. They were randomly assigned to the SBL and MBL groups, each consisting of 72% males. 4.2. Procedures The participants conducted the laboratory one at a time. Each of them was given the pre-test to measure their prior knowledge related to Boyle's Law. Then the experiment and the measurement and calibration of eye movement were explained in detail to the participant. In either the SBL or the MBL, they could choose to start with the experiment, the materials, or the worksheet. They were told to spend as much time and to perform as many trials of the experiments as they wanted. They could pause at any time; otherwise a 10 min break was given for every 20 min of experimentation. Moreover, each participant worked with a teacher and a peer, who were graduate research assistants and trained to give only social and technical support, such as encouraging to move on, echoing ideas proposed by the participant, reminding (but not forcing) to complete all tasks on the worksheet, holding a syringe or handheld devices as requested by the participant, etc. Scripts were provided for them to respond to various questions we had collected from a pilot test with three classes. They basically avoided giving any solutions regarding the tasks on the worksheet. The teacher gave different verbal instructions to the SBL and MBL groups regarding usage of the equipment, i.e., point and click selections on a monitor vs. GLX handheld devices before the experiments. No formal instruction was given during the experiments. Finally, the participants were assessed using the post-test on Boyle's Law. From the pre-to the post-test took about 4 h. 4.3. Worksheet The SBL and MBL groups had an identical worksheet. The outline and tasks are displayed in Table 1. It was presented alongside the experiment window in the SBL and in printouts in the MBL. The worksheet engaged several features that are distinctive for authentic Table 1 Outline of the worksheet. Section

Goal

Section 1/Time 1

1. Review of Boyle's Law concept 2. Planning: making decisions about materials and the amount of data needed 3. Experimenting: conducting experiments, recording data and observing graphs 4. Evaluating: thinking about possible experimental flaws 5. Improving experiment: designing an experiment to improve the study 6. Experimenting: conducting own experiments 7. Evaluating and concluding: evaluating results and interpreting graph

Section 2/Time 2

Section 3/Time 3

8. Designing experiment: generating own research questions and designing experiments

Task Reflecting on 4 questions about thermodynamics

1. Think about what size syringe you will use. Why? 2. Think about how many pairs of volume and pressure you need to record so that you will convince others that your result is trustworthy. Why? 1. Drawing VeP graph 2. Drawing Ve1/P graph 3. Recording data and calculating Is the P  V at each point close to a constant? Compared with the P  V in the beginning, the experimental error happens at the time when the volume of the gas is (, large or , small). Why? Think about how to improve the experiment. Please use text or figures to describe your experimental design.

1. 2. 1. 2. 3.

Recording data Decreasing errors Look at the Ve1/P diagram, is it a straight line? What do you think the slope means? Does it pass the origin point? If not, what is the meaning of the Y-axis intercept? In this experiment, which are the controlled variables that do not change? (,P ,V ,n ,R ,T) Which are the independent variables and dependent variables that change? (,P ,V ,n ,R ,T) 4. Do you think the controlled variable is well controlled? Give an example and explain. The pressure, volume, number of molecules and gas temperature interact with each other. Choose any two of the variables and design an experiment to investigate their relationship. Use text or figures to describe your ideas.

196

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

inquiry: generating research questions, designing studies such as selecting variables, controlling variables and planning measures, conducting experiments, and explaining results such as transforming data, finding experimental flaws, and indirect reasoning (Chen, 2012; Chinn & Malhotra, 2002). The experimental procedures were not specified for the students. Many decisions, such as the experimental materials, number of data points to take, and the approaches to solving problems, were left for them to make themselves. The worksheet was composed of three sections. The first focused on an introduction to Boyle's Law and used four multiple choice questions to review the concept. The second section focused on planning, conducting and improving the experiments, and evaluating the results. For the planning, students chose the materials from different sizes of syringes and tubes and decided the amount of data, i.e., the number of pairs of V and P, to take. They were asked to justify their choices and decisions. Although the PC and GLX recorded data and graphed in real-time, the students were asked to graph, fill in the data table, and do the necessary calculations on the worksheet, so that we could identify which run of the experiments they decided to collect and which set of data and graphs they worked on for the subsequent tasks. Then they evaluated the results and examined possible experimental flaws. Based on the evaluation, they tried to improve the experiments. Following the improved experiments, they encountered advanced questions regarding graph interpretations. In the third section, they generated their own research questions about gas properties, and designed experiments accordingly. The worksheet involved critical components of scientific inquiry as previously mentioned. Thus, the students' responses on it were scored as a measure of their inquiry performance. It consisted of 8 multiple choice questions, two data recording tasks, and 8 open-ended questions (Table 1). Each of the multiple choice questions was scored 2 for responses that were correct or consistent with the data, 1 for incorrect or data-inconsistent responses, and 0 for no response. The data recording and open-ended questions were evaluated with rubrics. A response was rated 0e4 points depending on the completeness of the recording/description, the correctness of the science concepts, the specifics and workability of the design, and the clarity of the explanation (see more details and examples in Chen, Chang, Lai, and Tsai (2014)). The worksheet had been tested in two high schools and had been revised by two high school teachers before being applied in the present study. The interrater reliability of the rubrics measured by Cohen's Kappa between the two raters was .90. 4.4. Conceptual test The same conceptual test was administered before and after the laboratories to assess the students' ability to graph, explain Boyle's Law, analyze and solve problems, evaluate the value of ideas, and identify experimental errors. There were totally 10 open-ended items. They were scored blind to the students' condition. A 0e4 point rubric was developed based on the correctness of the concepts, the calculation, and the reasoning: 0 meant missing or irrelevant responses; 1 denoted concepts that were mostly incorrect/irrelevant; 2 indicated partially incorrect or correct answers but with no reason/calculation; 3 meant mostly correct or correct answers with underdeveloped reasons/ miscalculation; and 4 denoted correct in both answers and reasoning/calculation (see more details and examples in Chen et al. (2014)). The interrater reliability of the rubric among three physics teachers, the primary researcher and a research assistant was .85 as measured by the Kendall's coefficient of concordance. The maximal score was 40. The items were validated by three high school physics teachers and two professors. Each expert evaluated the appropriateness of the test content, the clarity of the test items, and our interpretation of the test items for face and content validity. Then the items were piloted with 127 11th graders. The internal consistency of the whole test was high, with a Cronbach's alpha of .92. 4.5. Eye movement tracking Seeing Machines Company faceLAB and Applied Science Laboratories (ASL) MobileEye were used to record each participant's eye movements in the SBL and MBL settings, respectively. They are non-intrusive, automatic tracking systems which allow free head motion. The SBL was displayed on a 48 cm  30 cm widescreen monitor. The MBL was conducted on a 70 cm  82 cm desk. The faceLAB was calibrated on the screen, whereas the ASL MobileEye was on a 70.5 cm  86 cm cardboard, which covered the desk and was about twice the size of the gesture space for a sitting person, to ensure that the fixations on objects were accurately captured. The spatial accuracy of both systems was ±0.50. The eye trackers sampled the position of the users' eyes at the frequency of 30 Hz. Videos were taken for subsequent analysis. 5. Results 5.1. Data analysis The students were further divided into high and low achievers based on the pre-test. Kelley (1939) has suggested the upper and lower 27% as the selection criteria for the divide. Cureton (1957) commented on a 33% rule. Nevertheless, many empirical studies use the mean or median to split the participants. In this study, due to the limitation of the small sample size, we compromised the rule as the upper and the lower 40% on the pre-test. As a result, there were 11 high achievers and 12 low achievers in the SBL group, and 11 high achievers and 9 low achievers in the MBL group. Statistical comparisons among the groups were tested at the .05 significance level. 5.1.1. Worksheet and pre- and post-tests Two-way MANOVA (2 achievement levels  2 lab types) was used to compare the four groups' inquiry performance (i.e., the worksheet scores), initial knowledge of Boyle's Law (i.e., the pre-test), and conceptual gains in the post-test. Paired t-tests were applied to compare the pre- and post-tests to understand each groups' improvement in concepts. Pearson's correlation was carried out to examine the correlation between the worksheet and conceptual tests. The effect sizes were estimated by partial eta squared h2p and Cohen's d. 5.1.2. Number of experimental trials and time on the laboratory From the videos, we estimated individual students' time spent on the tasks and the number of experimental trials, which Renken and Nunez (2013) have identified as a crucial factor for conceptual understanding. Time on the tasks was the duration from the beginning to the submission of their worksheet. An experimental trial was defined as a complete action of setting the equipment and taking data. Taking data

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

197

without resetting the equipment, such as repeating or adding more data to an existing trial, was not counted as a new trial. For both SBL and MBL, the set up and data collection could be completed in less than 2 min. Two researchers coded the trials on the videos. Disagreement was resolved through discussion. Likewise, two-way ANOVAs and Pearson's correlation were applied to compare the groups and the relationships between variables. 5.1.3. Eye movement The eye movement data were analyzed by the GazeTracker software. Fixation in this study was identified as a gaze point that lasted for at least 60 ms, which is the minimum time for a stimulus to travel from the retina to the center of the brain that processes information (Ishida & Ikeda, 1989; Rayner, 1998). Percentage of fixation was defined as the number of fixations in a look zone divided by the total number of fixations. Fixation duration was calculated as the average length of fixations in a look zone. We were particularly interested in their attention to and cognitive processing of the worksheet and the experiment (and material) zones (as shown in Fig. 1). Outside the worksheet and the experimental zones was defined as Others. Fixations on the worksheet, calculators for calculating values on the worksheet, and pop-up windows for drawing were counted as being in the worksheet zone. Fixations on the experiment materials and real-time graphs and tables were counted as being in the experiment zone. In other words, experiments included not only data collection, but also the inspection of equipment, graphs, and data tables. Once they decided to accept a set of data, they graphed and filled in the data table on the worksheet. This was then counted as fixations in the worksheet zone. The SBL students were free to move the windows, and the MBL students changed their visual space and moved objects from moment to moment. Thus, the look zones were adjusted manually frame by frame (30 frames per second) to derive an accurate result. The frame by frame analysis required tremendous manpower. Thus, instead of analyzing the complete eye movement data, we sampled three 5-min long sets of eye tracking data, represented by Time 1, Time 2 and Time 3. These consisted of 27,000 frames per subject. The three 5-min samples were taken from three sections of the laboratories to ensure that the important components of the inquiry were involved in the analysis of the data from both groups. Time 1 was the first 5 min of the laboratories. It demonstrated how the students started a laboratory. If they followed the worksheet, Time 1 would fall in the period from the beginning to where they responded with the amount of data they were going to collect (Table 1). Nevertheless, they were free to explore the experiments or start from the worksheet. Time 2 was taken from the beginning of Section 2 on the worksheet, where they planned, conducted, and improved their experiments as well as interpreting the results. Time 3 was taken from the beginning of Section 3, which focused on the task of designing a new experiment related to gas law. It revealed their attention while generating research questions and designing new experiments. Times 1e3 covered crucial components in scientific inquiry, including messing about, reading, planning and conducting experiments, generating research questions, and designing new experiments. The data were exported to Excel and transformed to SPSS for ANOVA and t-tests. 5.2. Performance For the pre- and post-tests and the worksheet, the two-way MANOVA revealed a main effect for the achievement level, Wilk's F (3, 37) ¼ 38.88, p ¼ .001, h2p ¼ .76, but not for the laboratory type, Wilk's F (3, 37) ¼ .48, p ¼ .70, h2p ¼ .04, nor for the interaction between the two variables, Wilk's F (3, 37) ¼ .50, p ¼ .68, h2p ¼ .04. The following paragraphs analyze the differences between the four groups using two-way ANOVA and within the groups using pair-wise t-tests. 5.2.1. Conceptual tests The mean scores of each group in the pre- and post-tests are displayed in Table 2. The pre-test scores were consistent across the experimental conditions. The high achievement groups were significantly higher than the low achievement groups, Fachievement (1, 39) ¼ 121.98 for the pre-test and 24.63 for the post-test, p ¼ .001, h2p ¼ .76 and .39. However, the laboratory type made no significant difference, Flab type (1, 39) ¼ .34 and .79, p ¼ .56 and .38, h2p ¼ .01 and .02 for the pre- and post-tests, respectively. Students' concepts and learning of Boyle's Law were similar in the two settings. There was no interaction effect between achievement and laboratory type, F (1, 39) ¼ .45 and .24, p ¼ .51 and .63, h2p ¼ .01 and .01 for the pre- and post-tests, respectively. The low achievers in both groups improved considerably from the pre-to the post-tests with large effect sizes, t (11) ¼ 3.70, p ¼ .004, d ¼ 1.07 for the SBL group and t (8) ¼ 2.48, p ¼ .038, d ¼ .83 for the MBL group. However, the SBL and MBL high achievers did not improve in the post-test, t (10) ¼ 1.93 and 1.24, p ¼ .58 and .37, d ¼ .58 and .37 (Table 2). The laboratories might not improve the high achievers' conceptual learning. 5.2.2. Worksheet Similar to the conceptual tests, the laboratory type had no effect, F (1, 39) ¼ .32, p ¼ .58, h2p ¼ .01, while a significant main effect was found between the high and low achievement groups, F (1, 39) ¼ 5.52, p ¼ .024, h2p ¼ .12. The MBL high achievers (M ¼ 55.27, SD ¼ 1.95, Table 2 Means, standard deviations, and paired-samples t-values for pre- and post-tests. Group

SBL High achiever Low achiever MBL High achiever Low achiever Note: Highest possible scores were 40.

n

Pre-test

Post-test

M (SD)

M (SD)

t-values(p)

d

11 12

32.91 (2.02) 22.08 (4.14)

31.73 (3.04) 26.67 (3.39)

1.93 (.083) 3.70 (.004)

0.58 1.07

11 9

33.00 (2.37) 20.78 (4.55)

31.27 (3.50) 25.11 (4.86)

1.24 (.243) 2.48 (.038)

0.37 0.83

198

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

CI95 ¼ [52.19, 58.36]) performed better than the MBL low achievers (M ¼ 49.78, SD ¼ 1.69, CI95 ¼ [46.37, 53.19]) and the SBL low achievers (M ¼ 50.75, SD ¼ 1.46, CI95 ¼ [47.80, 53.71]). The SBL high achievers (M ¼ 52.55, SD ¼ 1.53, CI95 ¼ [49.46, 55.63]) showed no statistical difference from the other groups. In the SBL group, the students' scores on the post-test were not correlated with their performance on the worksheet, r ¼ .14, p ¼ .528. On the contrary, in the MBL group, the two were moderately correlated, r ¼ .53, p ¼ .016. The better the MBL students performed on the worksheet, the higher scores they gained on the post-laboratory conceptual test. 5.3. Learning behaviors 5.3.1. Number of experimental trials Table 3 shows the average number of trials by section that each group conducted. Section 2 was when they were asked to collect data, and therefore had the highest average number of trials. Unlike the MBL students, those in the SBL tended to try out the simulations in Section 1. Overall, the SBL low achievers ran significantly more trials than the SBL high achievers, who also performed significantly more trials than the MBL high/low achievers, F (1, 39) ¼ 27.98, p ¼ .001, h2p ¼ .42. In this study, there was no time constraint, so all the participants could run the experiments as many times as they wanted. Each run in either the SBL or MBL could be completed in less than 2 min. However, the results reveal that the MBL group did not run as many experiments as the SBL group did. Furthermore, in the SBL group, the low achievers ran experiments many more times than the high achievers (16.42 vs. 9.55 times). In fact, the lower their pre-test score was, the more times they tried the experiments, r ¼ .50, p ¼ 0.16. It should be noted that the low achievers' scores significantly improved from the pre-to the post-tests. Multiple trials might have contributed to their conceptual learning. In contrast, in the MBL, the low achievers did not try more experiments than the high achievers (2.44 vs. 2.82 times). 5.3.2. Time spent on the laboratories The analysis of time on the tasks yielded a significant main effect of laboratory type F (1, 39) ¼ 11.00, p ¼ .002, h2p ¼ .22, with the SBL group spending more time (M ¼ 82.04 min) on average than the MBL group (M ¼ 58.05 min). An interaction between laboratory type and achievement, F (1, 39) ¼ 5.09, p ¼ .030, h2p ¼ .12, was also found. In the SBL, the low achievers (M ¼ 90.08) worked significantly longer than the high achievers (M ¼ 73.27), whereas in the MBL, the low achievers worked for a shorter time (M ¼ 49.00 vs. 65.46). In short, the MBL group had adequate time, but did not run their experiments repeatedly. The low achievers in the SBL setting spent almost twice as long as those in the MBL setting. 5.4. Attention distribution We used eye tracking techniques to collect each participant's eye movements throughout the entire experiment. There were two indicators of eye movement used in this study, which were percentage of fixation and average fixation duration, measuring attention and cognitive processing, respectively. We focused on two of the zones that were particularly relevant to the tasks: the worksheet and the experiment zones. Both zones were important for laboratory learning. Throughout Time 1 to Time 3, the SBL group had a consistently higher percentage of fixations on the two task-relevant zones, with the subtotal ranging from 87% to 92%, than the MBL group, whose subtotal ranged from 73% to 87% (Table 4). They could concentrate better on the task-relevant zones. During the first 5 min (Time 1), a main effect for laboratory type was observed in the analysis of attention in the experiment zone, F (1, 39) ¼ 9.81, p ¼ .003, h2p ¼ .20, and a main effect for achievement was identified for the worksheet zone, F (1, 39) ¼ 4.81, p ¼ .034, h2p ¼ .11. Specifically, 34% of the SBL low achievers' fixations were located in the experiment zone, which was higher than the SBL high achievers (19.37%), the MBL low achievers (11.59%), and the MBL high achievers (9.86%). The SBL low achievers also had a high percentage (42%) of choosing to manipulate the materials and run the experiment before working on the worksheet. In contrast to the SBL low achievers, the MBL high achievers paid the most attention to the worksheet zone. All groups increased their attention to the experiment zone at Time 2, and then to the worksheet zone at Time 3. At both Times 2 and 3, the SBL group had a significantly higher percentage of fixations on the worksheet zone than the MBL group, Flab type (1, 39) ¼ 6.59 and 4.39, p ¼ .014 and .043, h2p ¼ .14 and .10. The analyses showed no difference in the laboratory type in the experiment zone and no effect for the achievement level in either zone. 5.5. Cognitive processing In the beginning (Time 1), the SBL group had longer average fixation durations on the experiment zone, Flab type (1, 39) ¼ 5.68, p ¼ .022, h2p ¼ .13, whereas the MBL group had longer average fixation durations on the worksheet zone, Flab type (1, 39) ¼ 7.66, p ¼ .009, h2p ¼ .16. Up to halfway through the laboratory (Time 2), their average fixation durations in either experiment or worksheet zone were not statistically different, Flab type (1, 39) ¼ .02, 1.27, p ¼ .88, .27, h2p ¼ .001, .03. Toward the end of the laboratory (Time 3), the MBL group's fixation durations in the worksheet zone were longer than those of the SBL group, Flab type (1, 39) ¼ 8.64, p ¼ .005, h2p ¼ .18. This implies that initially the SBL Table 3 Means (standard deviations) for experimental trials by section. Group SBL High achiever Low achiever MBL High achiever Low achiever

n

Section 1

Section 2

Section 3

Total

11 12

3.00 (1.84) 2.25 (2.83)

4.82 (2.32) 8.25 (6.38)

1.73 (1.94) 5.92 (11.49)

9.55 (2.25) 16.42 (11.76)

11 9

.27 (.47) .11 (.33)

2.09 (.30) 2.00 (.00)

.45 (.93) .33 (.71)

2.82 (1.08) 2.44 (.73)

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

199

Table 4 Means and standard deviations for fixation and fixation duration. SBL (n ¼ 23)

Time 1 Worksheet Exp. Zone Subtotal Time 2 Worksheet Exp. Zone Subtotal Time 3 Worksheet Exp. Zone Subtotal

MBL (n ¼ 20)

High achiever (n ¼ 11)

Low achiever (n ¼ 12)

High achiever (n ¼ 11)

Low achiever (n ¼ 9)

Fixation (%)

Duration (ms)

Fixation (%)

Duration (ms)

Fixation (%)

Duration (ms)

Fixation (%)

Duration (ms)

M (SD)

M (SD)

M (SD)

M (SD)

M (SD)

M (SD)

M (SD)

M (SD)

68.09 (17.67) 19.37 (16.07) 87.46

239 (37) 281 (55)

52.90 (26.93) 34.00 (23.90) 86.90

225 (53) 289 (97)

77.08 (10.82) 9.86 (6.02) 86.94

293 (86) 232 (105)

65.80 (18.79) 11.59 (13.86) 77.39

274 (60) 205 (96)

50.23 (20.01) 37.46 (23.56) 87.69

259 (56) 261 (40)

53.97 (15.21) 37.40 (18.10) 91.37

245 (59) 265 (59)

37.74 (23.34) 35.88 (17.96) 73.62

266 (72) 238 (98)

38.56 (19.42) 34.38 (13.00) 72.94

288 (99) 295 (84)

80.58 (13.20) 11.64 (8.02) 92.22

287 (89) 214 (80)

79.52 (19.59) 11.67 (16.88) 91.19

255 (72) 248 (84)

63.61 (21.99) 14.08 (10.77) 77.69

351 (139) 307 (171)

72.63 (18.15) 8.26 (6.99) 80.89

422 (196) 273 (98)

group had deeper cognitive processing of the experiment zone, while the MBL group processed the worksheet more. At Time 3, when they were asked to design a new experiment to answer their own research question, the MBL students might have experienced difficulties in the task as shown by their extremely long fixation durations on the worksheet. Furthermore, their fixation durations on the experiment zone were also longer than those of the SBL students, signifying their usage of information from the experiment zone to help them complete the task. Unlike the analyses of the performance, where the main effect for the achievement level was identified, most analyses of eye movement data revealed a main effect for the laboratory type. Although the high and low achievement subgroups in the SBL and MBL groups had sample sizes too small to observe statistically significant effects in all aspects except for the percentage of total fixation at Time 1 in the worksheet zone, the data did reveal some tendencies. Within SBL, throughout the time span, the high achievers' fixation durations on the worksheet were longer than those of the low achievers, whereas the low achievers' fixation durations on the experimental zone were longer than those of the high achievers. For the MBL, both the high and low achievers often had longer fixation durations on the worksheet than on the experiment zone. Moreover, the high achievers' fixation durations on both zones were longer than those of the low achievers in the beginning. Then, at Time 2 and Time 3, the low achievers' fixation durations on both zones became longer than those of the high achievers. The MBL low achievers might have faced cognitive difficulties in Time 2 and Time 3.

6. Discussion and conclusion In undertaking this study, we explored students' performance of conducting experiments of Boyle's Law and their learning process by using eye tracking techniques in both the SBL and MBL settings. We also examined their learning achievements from the pre-to post-tests as well as their performance in the inquiry tasks. This study found that the achievement level had a significant main effect on conceptual learning and inquiry performance, while the laboratory type had a highly significant main effect on learning behaviors, attention, and cognitive processing. The latter suggests different learning styles in the SBL and MBL. In the following paragraphs, we discuss the results of their conceptual understanding and the differences in their learning processes, and then conclude with a discussion of their learning styles. Firstly, in either SBL or MBL, low achievers can significantly improve their conceptual learning. This finding is consistent with previous studies that virtual and physical laboratories are equally effective (Klahr et al., 2007; Triona & Klahr, 2003; Wiesner & Lan, 2004; Zacharia & Constantinou, 2008). Moreover, our laboratories focused on Boyle's Law, for which they had not previously had physical manipulatives, and were accompanied with an MBL. We have added to the literature by finding the same result in topics alien to learners using an MBL. Perhaps due to the fairly simple equipment, the virtual and physical manipulatives did not make any difference. Furthermore, the higher achievement groups did not improve in the post-test in either the SBL or the MBL settings. As found by Olympiou et al. (2013), prior knowledge influences mental modeling for conceptual understanding. The laboratories might not have contributed to the high achievers' mental modeling as much as they did to that of the low achievers. Moreover, the high achievers did not have as much room for improvement because their pre-test scores were already relatively high. Secondly, concerning the learning process, the students could concentrate better on the task-relevant zones in the virtual laboratory. This echoes Finkelstein et al.'s (2005) finding that students in physical laboratories are more likely to be distracted by external factors. Nevertheless, there were no differences in their performance in the concept tests and worksheet. Zacharia et al.'s (2012) study with kindergartners has demonstrated that multiple modalities such as texts, sound, and tactile sensations, can deliver information for constructing a stronger representation than a single modality. The MBL group, who exploited the tactile sensory channel, was supposed to have better achievements. However, the effect might have been balanced by the concentrated attention on the task-relevant zones held by the SBL group. Thirdly, the moderate positive correlation between the MBL group's worksheet and post-test indicates that their learning is associated with the worksheet. Unlike the MBL group, such a correlation was not found in the SBL group. Moreover, the eye movement data also evidenced that the MBL students had deeper cognitive processing of the worksheet zone. The SBL group did not present deep cognitive processing of the worksheet until the last task about generating a new research question and designing a new experiment. Fourthly, from the number of experimental trials, we observed that, contradictory to the MBL condition, the SBL group tended to try out their experiments repeatedly. Particularly, their low achievers ran the experiments many more times. The number of experimental trials was moderately negatively correlated with the pre-test. This correlation was not found in the MBL group. The eye movement data also revealed

200

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

that the SBL students and low achievers particularly paid relatively more attention to and performed more cognitive processing in the experiment zone than the MBL group. After all, the low achievers improved significantly in the post-test. These findings indicate that the SBL group learned from conducting experiments repeatedly. Renken and Nunez (2013) revealed that students are less likely to control variables and are more likely to run additional trials in simulations, compared with physical experiments. Researchers have attributed these findings to the entertaining nature of computer simulations, which is considered not constructive for conceptual understanding. The present study provides an alternative explanation. As mentioned above, no matter what types of experiments students conduct, this study concludes that SBLs are as effective as MBLs in learning physics concepts. When we examined the SBL students' learning patternsdmore trials and more attention to the experiment zone, we concluded that they might have learned from doing the experiments repeatedly. On the contrary, the MBL students spent most of their time working on the worksheets. Their scores on the worksheet were positively correlated with their post-test scores. They learned by thinking about the questions on the worksheets. 7. Implications In terms of the learning outcomes, the two conditions are indistinguishable and fairly effective for low achievers. We recommend that low achievers be given adequate laboratory practice. Moreover, the learning objectives of laboratories are more than learning outcomes. Students tend to use different learning strategies in SBLs and MBLs. Both thinking about and running adequate experiments are crucial for laboratory learning. In MBLs, besides thinking, the worksheet may encourage students to try more experiments. In SBLs, besides running experiments, salient features or prompts may be used to draw students' attention to reflect on questions and promote thinking. Moreover, based on the patterns, we suggest that teachers might implement MBLs with inquiry-based worksheets in school and use SBLs for extracurricular learning if the class periods do not allow having both. It is generally considered that using SBLs in instruction and learning is efficient and time saving. However, the findings of this study reveal that the SBL group spent considerably more time to reach the same level of conceptual understanding as measured by the post-test, and inquiry performance as indicated by the worksheet, as the MBL group. In other words, traditional physical laboratories may not be as efficient as SBLs. However, technology-enhanced physical laboratories like MBLs could be more efficient than SBLs. Students in SBLs tend to try out experiments more times, which implies that SBLs are more suitable for situations that do not have time constraints and in which students can explore freely, such as after-school time. In addition to the issue of time constraints, another reason for using MBLs in school is that students in MBLs are likely to think before acting or planning experiments. As a result, they may experience heavy cognitive load as shown by the long fixation duration of the MBL low achievers during Times 2 and 3. The presence of a teacher in the classroom may provide timely help. Although students in SBLs may take much time to try out experiments, they can obtain science concepts as effectively as they can in MBLs. Teachers can encourage them to explore more using SBLs in their after-school time. As a result, students may feel less stress in science learning, and the learning will no longer be restricted to the classroom. 8. Limitation and future studies We have made our conclusions based on the condition that everything is equal or similardthe worksheet, dynamic representation, measurement error, time needed for each trial, and studentestudent interaction. However, we did not take into account some SBL advantages such as multiple representations (Olympiou et al., 2013) and customized instructional cues (Butcher & Sumner, 2011). Moreover, this study inferred attention distribution and cognitive processing from eye fixations and fixation durations. Future research may apply other techniques such as think aloud and brainwave and physiology sensors to reveal more details of cognitive load and processing in laboratory learning, such as the observation, reasoning, and interpretation processes, as highlighted by Renken and Nunez (2013). Furthermore, each student was given pre- and post-tests at very short intervals. There might be practice effects. The students might have grown weary of the retest. Moreover, this study did not look into the details of the components in the laboratories, such as graphs, tables, and equipment. Instead, we used two general look zones to explore students' learning behaviors. The results have shown some interesting patterns and the roles the worksheet played in the SBL and MBL. These can serve as a foundation for further, finer investigation specifically targeting students' attention during an experiment. Future studies may examine students' attention to specific components and sequence of eye movement, which would provide evidence for information processing and integration. Future studies may also investigate whether the thinking before acting habit in a physical laboratory leads to better mental modeling, and thus promotes conceptual understanding. Would it lead to more self-regulating and self-monitoring? Even though both the SBL and the MBL are useful for students to learn the concepts of physics, future studies could take into account students' traits in order to explore what kinds of traits are more favorable for physical, virtual, or a combination of physical and virtual laboratories. For example, students who feel stressed when learning physics would probably like to conduct experiments in SBLs where they can manipulate freely and with less pressure. Learners may choose laboratory types that match their learning traits for optimal results. In sum, with the extraneous variables well controlled, this study provides clear evidence regarding the similarities and differences in students' performance, learning behaviors, attention, and cognitive processing in the SBL and MBL. The students in the two laboratories had similar performance in terms of conceptual understanding, which they obtained from different strategies, i.e., high attention on the taskrelevant zones, carrying out multiple experimental trials, and more cognitive processing in the experimental zone in the SBL versus usage of multiple modalities and deep cognitive processing of the questions on the worksheet in the MBL. With the understanding of students' learning in the different technology-enhanced laboratories, different teaching approaches are suggested to maximize students' conceptual learning. Acknowledgments This research project was supported by grants from the Ministry of Science and Technology Taiwan (NSC 98-2628-S-011-001-MY3 and NSC 102-2628-S-011-001-MY4).

K.-P. Chien et al. / Computers & Education 82 (2015) 191e201

201

References Bisdikian, G., & Psillos, D. (2002). Enhancing the linking of theoretical knowledge to physical phenomena by real-time graphing. In D. Psillos, & H. Niedderer (Eds.), Teaching and learning in the science laboratory (pp. 193e204). Dordrecht, The Netherlands: Kluwer Academic Publishers. Bivall, P., Ainsworth, S., & Tibell, L. A. E. (2011). Do haptic representations help complex molecular learning? Science Education, 95, 700e719. Boucheix, J.-M., & Lowe, R. K. (2010). An eye tracking comparison of external pointing cues and internal continuous cues in learning with complex animations. Learning and Instruction, 20, 123e135. Brassell, H. (1987). The effect of real-time laboratory graphing on learning graphic representations of distance and velocity. Journal of Research in Science Teaching, 24, 385e395. Butcher, K. R., & Sumner, T. (2011). Self-directed learning and the sensemaking paradox. Human-Computer Interaction, 26(1e2), 123e159. Calvo, P., & Gomila, T. (2008). Handbook of cognitive science: An embodied approach. NY: Elsevier. Canham, M., & Hegarty, M. (2010). Effects of knowledge and display design on comprehension of complex graphics. Learning and Instruction, 20, 155e166. Chen, S. (2010). The view of scientific inquiry conveyed by simulation-based virtual laboratories. Computers & Education, 55(3), 1123e1130. Chen, S., Lo, H.-C., Lin, J.-W., Liang, J.-C., Chang, H.-Y., Huang, F.-K., et al. (2012). Development and implications of technology in reform-based physics laboratories. Physical Review Special Topics-Physics Education Research, 8. http://dx.doi.org/10.1103/PhysRevSTPER.8.020113, 020113-1-12. Chen, S., Chang, W.-H., Lai, C.-H., & Tsai, C.-Y. (2014). A comparison of students’ approaches to inquiry, conceptual learning, and attitudes in simulation-based and microcomputer-based laboratories. Science Education, 98(5), 905e935. Chini, J. J., Madsen, A., Gire, E., Rebello, N. S., & Puntambekar, S. (2012). Exploration of factors that affect the comparative effectiveness of physical and virtual manipulatives in an undergraduate laboratory. Physical Review Special TopicsdPhysics Education Research, 8(1), 010113. http://dx.doi.org/10.1103/PhysRevSTPER.8.010113. Chinn, C. A., & Malhotra, B. A. (2002). Epistemologically authentic inquiry in schools: a theoretical framework for evaluating inquiry tasks. Science Education, 86, 175e218. Cureton, E. (1957). The upper and lower twenty-seven per cent rule. Psychometrika, 22(3), 293e296. Finkelstein, N. D., Adams, W. K., Keller, C. J., Kohl, P. B., Perkins, K. K., Podolefsky, N. S., et al. (2005). When learning about the real world is better done virtually: a study of substituting computer simulations for laboratory equipment. Physical Review Special Topics e Physics Education Research, 1(1), 010103. http://dx.doi.org/10.1103/ PhysRevSTPER.1.010103. Garber, P., & Goldin-Meadow, S. (2002). Gesture offers insight into problem-solving in adults and children. Cognitive Science, 26, 817e831. Gijlers, H., & de Jong, T. (2005). The relation between prior knowledge and students' collaborative discovery learning processes. Journal of Research in Science Teaching, 42, 264e282. He, E., & Kowler, E. (1992). The role of saccades in the perception of texture patterns. Vision Research, 32, 2151e2163. Hucke, L., & Fischer, H. E. (2002). The linking of theory and practice in traditional and in computer-based university laboratory experiments. In D. Psillos, & H. Niedderer (Eds.), Teaching and learning in the science laboratory (pp. 205e218). Boston: Kluwer Academic Publishers. Inhoff, A. W., & Radach, R. (1998). Definition and computation of oculomotor measures in the study of cognitive processes. In G. D. M. Underwood (Ed.), Eye guidance in reading and scene perception (pp. 29e53). Oxford: Elsevier Science. Ishida, T., & Ikeda, M. (1989). Temporal properties of information extraction in reading studied by a text-mask replacement technique. Journal of the Optical Society A: Optics and Image Science, 6, 1624e1632. Jaakkola, T., & Nurmi, S. (2008). Fostering elementary school students' understanding of simple electricity by combining simulation and laboratory activities. Journal of Computer Assisted Learning, 24, 271e283. de Jong, T., Linn, M. C., & Zacharia, Z. C. (2013). Physical and virtual laboratories in science and engineering education. Science, 340, 305e308. http://dx.doi.org/10.1126/ science.1230579. Just, M. A., & Carpenter, P. A. (1980). A theory of reading: from eye fixations to comprehension. Psychological Review, 87, 329e354. Kelley, T. L. (1939). The selection of upper and lower groups for the validation of test items. Journal of Educational Psychology, 30, 17e24. Klahr, D., Triona, L. M., & Williams, C. (2007). Hands on what? The relative effectiveness of physical versus virtual materials in an engineering design project by middle school children. Journal of Research in Science Teaching, 44, 183e203. Liu, H.-C., Andre, T., & Greenbowe, T. (2008). The impact of learner's prior knowledge on their use of chemistry computer simulations: a case study. Journal of Science Education and Technology, 17, 466e482. Linn, M. C., & Songer, N. B. (1991). Teaching thermodynamics to middle school students: What are appropriate cognitive demands? Journal of Research in Science Teaching, 28, 885e918. Nakhleh, M. B., & Krajcik, J. S. (1994). Influence of levels of information as presented by different technologies on students' understanding of acid, base, and pH concepts. Journal of Research in Science Teaching, 31, 1077e1096. Nicolaou, C. T., Nicolaidou, I., Zacharia, Z., & Constantinou, C. P. (2007). Enhancing fourth graders' ability to interpret graphical representations through the use of microcomputer-based labs implemented within an inquiry-based activity sequence. Journal of Computers in Mathematics and Science Teaching, 26(1), 75e99. Nobre, A. C., Gitelman, D. R., Dias, E. C., & Mesulam, M. M. (2000). Covert visual spatial orienting and saccades: overlapping neural systems. Neuroimage, 11, 210e216. Olympiou, G., & Zacharia, Z. C. (2012). Blending physical and virtual manipulatives: an effort to improve students' conceptual understanding through science laboratory experimentation. Science Education, 96, 21e47. Olympiou, G., Zacharias, Z., & de Jong, T. (2013). Making the invisible visible: enhancing students' conceptual understanding by introducing representations of abstract objects in a simulation. Instructional Science, 41, 575e596. Pierri, E., Karatrantou, A., & Panagiotakopoulos, C. (2008). Exploring the phenomenon of ‘change of phase’ of pure substances using the microcomputer-based-laboratory (MBL) system. Chemistry Education Research and Practice, 9, 234e239. Ploetzner, R., Lippitsch, S., Galmbacher, M., Heuer, D., & Scherrer, S. (2009). Students' difficulties in learning from dynamic visualisations and how they may be overcome. Computers in Human Behavior, 25, 56e65. Posner, M. I. (1980). Orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3e25. Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124, 372e422. Reiner, M., & Gilbert, J. (2000). Epistemological resources for thought experimentation in science learning. International Journal of Science Education, 22, 489e506. Renken, M. D., & Nunez, N. (2013). Computer simulations and clear observations do not guarantee conceptual understanding. Learning and Instruction, 23(1), 10e23. , C. (1987). Reorienting attention across the horizontal and vertical meridians: evidence in favor of a premotor theory of attention. Rizzolatti, G., Riggio, L., Dascola, I., & Umilta Neuropsychologia, 25, 31e40. Russell, D. W., Lucas, K. B., & McRobbie, C. J. (2004). Role of the microcomputer-based laboratory display in supporting the construction of new understandings in thermal physics. Journal of Research in Science Teaching, 41, 165e185. She, H.-C., & Chen, Y.-Z. (2009). The impact of multimedia effect on science learning: evidence from eye movements. Computers & Education, 53, 1297e1307. Trey, L., & Khan, S. (2008). How science students can learn about unobservable phenomena using computer-based analogies. Computers & Education, 51, 519e529. Trindade, J., Fiolhais, C., & Almeida, L. (2002). Science learning in virtual environments: a descriptive study. British Journal of Educational Technology, 33, 471e488. Triona, L. M., & Klahr, D. (2003). Point and click or grab and heft: comparing the influence of physical and virtual instructional materials on elementary school students' ability to design experiments. Cognition and Instruction, 21, 149e173. Tsai, M.-J., Hou, H.-T., Lai, M.-L., Liu, W.-Y., & Yang, F.-Y. (2012). Visual attention for solving multiple-choice science problem: an eye-tracking analysis. Computers & Education, 58, 375e385. Wiesner, T. F., & Lan, W. (2004). Comparison of student learning in physical and simulated unit operations experiments. Journal of Engineering Education, 93(3), 195e204. http://dx.doi.org/10.1002/j.2168-9830.2004.tb00806.x. Zacharia, Z. C. (2007). Comparing and combining real and virtual experimentation: an effort to enhance students' conceptual understanding of electric circuits. Journal of Computer Assisted Learning, 23, 120e132. Zacharia, Z. C., & Constantinou, C. P. (2008). Comparing the influence of physical and virtual manipulatives in the context of the Physics by Inquiry curriculum: the case of undergraduate students' conceptual understanding of heat and temperature. American Journal of Physics, 76, 425e430. Zacharia, Z. C., Loizou, E., & Papaevripidou, M. (2012). Is physicality an important aspect of learning through science experimentation among kindergarten students? Early Childhood Research Quarterly, 27(3), 447e457. Zacharia, Z. C., & Olympiou, G. (2011). Physical versus virtual manipulative experimentation in physics learning. Learning and Instruction, 21, 317e331. Zacharia, Z. C., Olympiou, G., & Papaevripidou, M. (2008). Effects of experimenting with physical and virtual manipulatives on students' conceptual understanding in heat and temperature. Journal of Research in Science Teaching, 45, 1021e1035.