View in EDS HTML Full Text PDF Full Text

Remote Vocational Learning Opportunities--A Comparative Eye-Tracking Investigation of Educational 2D Videos versus 360° Videos for Car Mechanics

Saved in:

Bibliographic Details
Title:	Remote Vocational Learning Opportunities--A Comparative Eye-Tracking Investigation of Educational 2D Videos versus 360° Videos for Car Mechanics
Language:	English
Authors:	Hekele, Felix, Spilski, Jan, Bender, Simon, Lachmann, Thomas
Source:	British Journal of Educational Technology. Mar 2022 53(2):248-268.
Availability:	Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
Peer Reviewed:	Y
Page Count:	21
Publication Date:	2022
Document Type:	Journal Articles Reports - Research
Descriptors:	Vocational Education, Distance Education, Video Technology, Eye Movements, Auto Mechanics, Technology Uses in Education, Learning Modalities, Computer Simulation, Instructional Effectiveness
DOI:	10.1111/bjet.13162
ISSN:	0007-1013
Abstract:	This study utilises a novel approach to investigate the effectiveness of different learning modalities by combining video-based learning with eye-tracking. An excerpt taken from a vocational education instruction for car mechanics was videotaped using two different cameras: a standard 2D video camera and a professional 360° camera. The video recorded with the 2D camera was presented on a tablet, with a fixed angle, whereas the video recorded with the 360° camera was presented as non-interactive 3DoF virtual reality (nVR) environment using a head-mounted display. In both conditions, participants' fixation patterns were recorded and analysed in conjunction with a set of standardised questionnaires. Participants (N = 48) were randomly assigned to either the 2D-video group or the nVR group, with 23 participants in the 2D-video and 25 participants in the nVR group. The task of the participants in both groups was to watch the educational video while wearing an eye-tracker and then complete a standardised test on the presented content. The eye-tracking data indicated that participants in the nVR group showed longer total fixation durations on the instructor, but not other areas of interest, compared to the 2D video group. The standardised test indicated no differences in learning outcome between the groups. Implications from the current study as well as limitations and a outlook for further research will be discussed.
Abstractor:	As Provided
Entry Date:	2022
Accession Number:	EJ1326762
Database:	ERIC
Full text is not displayed to guests. Login for full access.

FullText	Links: – Type: pdflink Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwEWgD6jPrrRltH3aRzXKTtoAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDD4cWvwP50jze2zukAIBEICBmvCl5zj15ZzcIXatHpGDhYVGOYtv5P5t9oJxIQMwV_I9-BG-jHfFEqjDkM2JNYV2JqyXbXsuZQRuAIwMyRqlky0xiQyDycmeHE2ow3sdTRFn47SJaexO0NClbjiUTFtHsXBmzpbHcXR_CrU3shZV1l_DRCPxu5iyDb-0TqTqJn0mpWGmwRGD7bvmk8w1cTJSOoxttXYSxcaChfU= Text: Availability: 1 Value: <anid>AN0155323232;58i01mar.22;2022Feb21.07:04;v2.2.500</anid> <title id="AN0155323232-1">Remote vocational learning opportunities—A comparative eye‐tracking investigation of educational 2D videos versus 360° videos for car mechanics </title> <p>This study utilises a novel approach to investigate the effectiveness of different learning modalities by combining video‐based learning with eye‐tracking. An excerpt taken from a vocational education instruction for car mechanics was videotaped using two different cameras: a standard 2D video camera and a professional 360° camera. The video recorded with the 2D camera was presented on a tablet, with a fixed angle, whereas the video recorded with the 360° camera was presented as non‐interactive 3DoF virtual reality (nVR) environment using a head‐mounted display. In both conditions, participants' fixation patterns were recorded and analysed in conjunction with a set of standardised questionnaires. Participants (N = 48) were randomly assigned to either the 2D‐video group or the nVR group, with 23 participants in the 2D‐video and 25 participants in the nVR group. The task of the participants in both groups was to watch the educational video while wearing an eye‐tracker and then complete a standardised test on the presented content. The eye‐tracking data indicated that participants in the nVR group showed longer total fixation durations on the instructor, but not other areas of interest, compared to the 2D video group. The standardised test indicated no differences in learning outcome between the groups. Implications from the current study as well as limitations and a outlook for further research will be discussed. Practitioner notesWhat is already known about this topic Virtual reality (VR) technology is getting more commonplace in educational settings.It is unclear whether VR‐based learning holds clear benefits over more traditional approaches such as hands‐on‐training or video‐based learning.What this paper adds This paper adds a comparison of two common video‐based teaching techniques and compares them in respect to learning outcome and learner's attention.The novelty of the current paper is the addition of eye‐tracking in both tablet‐based and HMD‐based learning scenarios to investigate learners' visual attention.Describes potential attention benefits of 360° videos compared to 2D videos despite no differences in learning outcome directly.Implications for practice and/or policy Results highlight the potential of placing more focus on social factors in 360° video material.Future 360° videos and VR environments can add additional focus on instructors or other socially relevant aspects to foster student learning and engagement.</p> <p>Keywords: eye‐tracking; video‐based learning; virtual reality; vocational education</p> <hd id="AN0155323232-2">INTRODUCTION</hd> <p>The rapid advancement of the development of virtual reality (VR) technology quickly expanded beyond the entertainment sector and now holds promise to revolutionise the education sector as well. While a growing body of VR projects exist for primary and secondary education (eg, Morales et al., 2013; Nobrega &amp; Rozenfeld, 2019) as well as university level education (Dyer et al., 2018), less focus has been placed on the use of VR in vocational education. While this varies between countries, in Germany the vocational sector remains one of the main employers for young adults. The German dual vocational education system (VET) interweaves hands‐on practice in the industry with decentralised education. The education is mostly reliant on frontal teaching in dedicated VET schools mixed with practical hands‐on experience at the place of work (Baethge et al., 2007). Contrary to universities and trade schools, vocational institutions are often far smaller in size, with nearly half of the institutions employing &lt;50 people (Autorengruppe Bildungsberichterstattung, 2018). The small size of the individual institutions combined with the variety of vocational jobs on the market pose major challenges to the development and introduction of expensive equipment such as VR‐ready learning environments. The push for digitalisation does not necessitate the exclusive use of VR technology however and alternatives should also be considered.</p> <p>With the greatly accelerated push for remote learning opportunities caused by the COVID‐19 outbreak, it is crucial to bring modern teaching techniques already present in other educational systems to vocational education as well. In this paper, the focus will be more specifically on the material generated for the use of vocational education in car mechanics. Given the smaller size of many institutions and the variety of topics and skills covered in vocational training, additional points need to be considered, such as alternatives to fully immersive VR environments.</p> <p>The development and maintenance of fully interactive VR environments proves difficult given the variety of tasks a car mechanic has to learn over the course of their three‐year long education. As one potential solution to the above, non‐interactive VR environments generated from 360° video recordings of lessons should be considered. We understand non‐interactive VR as an environment requiring only 3° of freedom (3‐DoF, head movement on the <emph>x</emph> and <emph>y</emph> as well as body turning) without controllers, for example a 360° video viewed on a Samsung Gear VR headset. On the other end of the spectrum are fully interactive 6‐DoF environments featuring environmental manipulation such as surgical training programs (eg, Frederiksen et al., 2020) or human‐in‐the‐loop driving simulators (Reinhard et al., 2019). Due to higher degrees of freedom provided by controllers and room‐scale tracking, 6‐DoF environments are usually created with interactivity in mind, which holds great promise in future interactive vocational learning environments, since experienced immersion will be higher (Slater, 2003, 2018). However, a major downside is the complexity of content creation leading to a current scarcity in content (Jensen &amp; Konradsen, 2018).</p> <p>Previous research highlighted the potential of educational 360° videos to increase student engagement (Violante et al., 2019) and leading to similar learning outcomes as traditional face‐to‐face or 2D video‐based education (Ulrich et al., 2021). At the point of writing, the technology to have multiple points of view within a 360° video is still mostly hypothetical, although this could change in the near future (see eg, Jeong et al., 2020) Compared to fully interactive VR environment, 360° video recorded material is more cost effective for vocational institutions since they can simply be played on aforementioned 3DoF head‐mounted display (HMDs) such as the Samsung Gear, Oculus Quest or the Google Cardboard instead of requiring expensive room‐based 6DoF setups.</p> <hd id="AN0155323232-3">Immersion and presence</hd> <p>In addition to the level of interaction, other concepts need to be considered in VR research as well. The first is the level of immersion and sense of presence in virtual environments. While these concepts are often used interchangeably (Grassini &amp; Laumann, 2020), immersion is commonly seen as more related to technological aspects of VR, such as the field of view or realistic audio environments (Slater, 2003, 2018). Slater (2018) noted that immersion should be seen on a scale instead of a binary immersive/non‐immersive decision, where contingency between actions of the participant and its effect in the world correlate—for example, being able to manipulate an object or observe it from multiple angles (see Huang et al., 2021) or maintaining eye‐contact with an interactive avatar (see Albus et al., 2021) should lead to higher feelings of immersion. Presence on the other hand is a mental state based on a more subjective experience—the 'feeling of being in' a virtual environment (Grassini &amp; Laumann, 2020; Schubert, Friedmann, &amp; Regenbrecht, 1999, 2001). This subjective experience is commonly measured using physiological correlates such as EEG, fMRI or the galvanic skin response, or a variety of questionnaires (for an overview, see Grassini &amp; Laumann, 2020). In the context of this study, subjective sense of presence was measured by using one of the standard questionnaires, the iGroup presence questionnaire (IPQ, Schubert et al., 2001). The sense of presence measured by the IPQ is primarily focussed on the individual's perception of being part of the virtual environment (Grassini &amp; Laumann, 2020), and the present paper will use of the term 'sense of presence' in the same way.</p> <hd id="AN0155323232-4">Novelty effect in VR</hd> <p>One relatively recent notion in VR‐based research, especially in the context of education, is the novelty effect of the utilised software and hardware. At the time of writing, a significant portion of students in both vocational facilities as well as the university reported having little to no prior experience with VR, with direct exposure usually limited to large events such as fairs. This led to high motivation to inquire about VR as well as participate in any VR‐related research, irrespective of content. Whether this high motivation leads to a sustainable effect of VR‐based learning is under debate. Merchant and colleagues (2014) reviewed a total of 69 studies utilising different forms of virtual learning environments in higher education and found that while virtual environments increased learning outcome, the effects started to deteriorate with repeated exposure to the virtual environment. Huang (2020) noted that this could be attributed to the diminishing novelty effect and added a first line of research where longitudinal learning in 3DoF and 6DoF VR was investigated. Notable results from Huang (2020) include that, while novelty alone did not increase learning outcome, it decayed slower than presumed by Merchant and colleagues (2014) and novelty also positively influenced learner's motivation. In a recent study, Huang and colleagues (2021) reported that student engagement and immersion remained high throughout several sessions, with task performance being comparable between groups using 3DoF ('moderate immersion') and 6DoF ('higher immersion') setups. This highlights the viability of less interactive presentation of content such as 360° videos for learners.</p> <hd id="AN0155323232-5">Cognitive load</hd> <p>The impact of VR on cognitive load has been under investigation since the technology began to become more widespread. Cognitive load as a construct can be measured in a variety of ways, with correlates ranging the subjectively experienced load measured by questionnaires such as the NASA task load index (TLX, Hart, 2006; Hart &amp; Staveland, 1988) to physiological correlates such as electrodermal activity and changes in pupil size. For an overview of measures for cognitive load specifically in VR see Armougum et al. (2019). Due to the novelty and complexity of VR‐based systems in educational contexts correlates introduced by VR could be highly relevant for cognitive load research as well since the danger of cognitive overload needs to be considered (Albus et al., 2021). In a surgical training context, Frederiksen and colleagues (2020) found that novices in an immersive virtual environment using a VR headset were subject to significantly higher cognitive load and consequently worse task performance compared to training with a conventional virtual setup using a screen and joysticks.</p> <p>When investigating differences in expertise for a spatial navigation task, however, the physiological and subjective measures of cognitive load were mostly affected by expertise, with no differences between real‐world and virtual navigation (Armougum et al., 2019).</p> <p>A looming question is whether the introduction of VR content could prove to be a beneficial addition for vocational learners and learners in general. Empirical evidence in this respect is mixed. There is a general agreement about the potential of VR technology as a viable catalyst to improve education by digitalisation (Freina &amp; Ott, 2015). A recent meta‐analysis by Moro and colleagues (2021) found comparable effects of VR training to traditional methods in four reviewed studies (Moro et al., 2021). Chen and colleagues (2018) investigated the relationship between the use of VR technology and different factors of learning in automotive vocational students and found positive effects of VR on both learning satisfaction and learning outcome. On the contrary, a recent review by Jensen and Konradsen (2018) about the state of VR in education indicated that, in the majority of reviewed papers, the use of VR environments yielded no advantage in either the cognitive or psychomotor domain. Jensen and Konradsen highlighted that one of the biggest shortcomings of educational VR remains the scarcity of content. This shortcoming also extends to the German vocational education system, where traditionally craftsmen with years of professional experience are teaching the next generation with traditional methods such as frontal classes and hands‐on lesions, mixed with some older educational 2D videos (Baethge et al., 2007). At the time of writing, no affordable 'out‐of‐the‐box' solution for generating VR content on a large scale exists outside of research projects and single educational projects. However, as noted before, recent research by Huang (2020), Huang and colleagues (2021) and the meta‐analysis of Moro and colleagues (2021) showcased the viability of 3DoF setups for learning environments, which would be much easier to implement compared to full VR environments which often are made from scratch. Apart from the learning outcome, it is however also important to look at other factors of the learning experience, such as where the learner shifts their attention to.</p> <p>Due to recent technological advances, eye‐tracking became feasible in VR setups. This allows the comparison of participants' gaze pattern in the real versus the virtual world as well as a more direct measurement of attention through gaze tracking. So far, there is only a limited body of research investigating the effects of different learning modalities using eye‐tracking. Reichenberger et al. (2020) note a possible for gaze tracking as a measure for attention in clinical treatments for social anxiety disorders, whereas Cheng and Huang (2012) investigated the possibility of VR to foster joint attention in children with pervasive developmental disorder. In educational contexts, a study by Meppelink and Bol (2015) concluded that participants who were not knowledgeable about a topic showed higher recall of information related to that topic when they spent more time (total fixation duration) on picture‐based information compared to text‐based only. Therefore, for this study, we recruited participants with little to no previous knowledge on the educational content presented. Whereas, however, the learning material used in Meppelink and Bol (2015) was website‐based, this study focussed on the differences between traditional 2D‐video‐based versus 3D‐VR‐video instruction.</p> <hd id="AN0155323232-6">Research question &amp; hypotheses</hd> <p>The learning material used in this study was primarily visual with added verbal instructions from the instructor. Learners are therefore expected to pay attention on the relevant objects and areas in the instructional video in both the 2D video and nVR conditions by fixating on them. According to Yarbus (1967), both the average duration of single fixations as well as the number of fixations on each object or area of interest are ways to measure human visual attention. By an extension of this, the sum of the durations of all fixations, or total fixation duration, on a single area of interest reflects the amount of time an individual actively paid attention to an object or person in a video. Bhoir et al. (2015) and Jeelani et al. (2018) used these measurements, among others, to investigate the attention of construction workers and the handling of personalised safety instructions, respectively. Instead of presenting learners with different versions of video‐based content, this study will focus on strictly on differences stemming from video presentation—whether instructional content leads to differences in visual attention and learning outcome when it is presented as a 360° video on an HMD compared to the same content presented as a 2D video on a tablet. Previous research highlighted different factors influencing learning outcome and learner engagement such as novelty (Huang, 2020, 2021), immersion and sense of presence (see Grassini &amp; Laumann, 2020), this study aims to investigate learner's attention in different video media.</p> <p>The primary research question this study aimed to answer was whether the presentation modality of a non‐interactive 3DoF 360° video (nVR) compared to a 2D video of the same content influences the learners' visual attention. It is predicted that there are differences in attentional focus, indicated by differences in total fixation duration on areas of interest.</p> <p>Additionally, this study will investigate whether the presentation of content in form of an educational 360° video on an HMD will improve the learning outcome compared to presenting the same content in form of a 2D video on a tablet. If that is the case, participants watching educational nVR content using a VR headset will show increased task‐relevant performance, indicated by a higher test score in a standardised knowledge recall test compared to participants who watch the same video in 2D using a conventional tablet.</p> <p>Lastly, since an immersive 360° video is presented on an HMD in one condition, differences in immersion and sense of presence are predicted. This serves as a manipulation check and will be indicated by higher self‐report scores on a standardised presence and immersion questionnaire, with participants in the nVR condition reporting increased sense of presence compared to the participants of the 2D video group.</p> <hd id="AN0155323232-7">METHOD</hd> <p></p> <hd id="AN0155323232-8">Participants</hd> <p>A total of 50 participants, all students at the University of Kaiserslautern, Germany, aged 20–34 (<emph>M</emph> = 25.28; <emph>SD</emph> = 2.53), were invited to be tested individually in a VR cabin and randomly assigned to either the 2D video or the nVR group. None of the participants reported prior knowledge or background of car mechanics. No participant reported dizziness or other symptoms during the experiment, which would have warranted the termination of the experiment. Two participants from the 2D video group had to be excluded due to technical problems or equipment failure during data collection. Therefore, the final sample consisted of 48 participants: 23 (4 female) in the 2D video group, aged 20–30 (<emph>M</emph> = 25.13; <emph>SD</emph> = 2.14) and 25 (9 female) in the nVR group, aged 21–34 (<emph>M</emph> = 25.36; <emph>SD</emph> = 2.96). All participants had normal or corrected‐to‐normal vision. Participants received course credit as compensation for their participation.</p> <hd id="AN0155323232-9">Material</hd> <p></p> <hd id="AN0155323232-10">Questionnaires</hd> <p>A set of standardised questionnaires were deployed to assess different aspects of participants' introspection and a standardised test was used to assess task‐related performance.</p> <p>Cognitive load was measured using the NASA TLX (Hart, 2006; Hart &amp; Staveland, 1988). This questionnaire consists of six items measuring different aspects of cognitive load experienced while performing different tasks, for example: 'How mentally demanding was the task?' Each item is rated on a 21‐point scale ranging from 0 to 20. In this study, the TLX was used to investigate potential differences in the mental effort participants experienced in the 2D video versus the nVR group. The TLX was chosen due to its widespread use as a quick, standardised measure of subjective cognitive load (see eg, Armougum et al., 2019).</p> <p>To assess the sense of presence, the German version of the IPQ (Schubert et al., 2001) was administered after the educational video in printed form. Due to the non‐interactive nature of both video conditions, the version used in this study contained 13 out of the original 14 items, with the item 'How much did your experience in the virtual environment seem consistent with your real‐world experience?' excluded. All questions were answered by participants on a six‐point scale ranging from 'Completely disagree' to 'Completely agree' or contextual variations thereof.</p> <p>Knowledge acquisition was assessed using a 16‐item single‐choice test which was created in cooperation with a vocational institute in North Rhine‐Westphalia, Germany. The test was held in close similarity to written vocational exams conducted as part of the curriculum for car mechanics and was designed to assess acquired task specific knowledge based on the information presented in the instructional video. Each item had a single correct (target) and three incorrect answers (lures). One point per correctly marked answer and per correctly unmarked answer was given, leading to a minimum of 32 and a maximum of 64 points. Before the study, the test was given to seven naïve participants with no prior knowledge of car mechanics, and it was evident that the test needed mechanical education to reach a sufficient score.</p> <hd id="AN0155323232-11">Video material &amp; presentation</hd> <p>The 2D and the 360° video were both recorded in a mechanics garage of the partnered vocational institute for the purpose of this study. The videos were recorded using a double camera setup with a 360° camera on a tripod and a conventional camera fixated below to create equal points of view for both the 2D and 360° videos. While working, the instructor regularly faces the camera to provide instructions and in addition also verbalises the current step he is working on. The content of the video was identical between the 2D and 360° versions, with the difference that the 360° video allowed to survey the entire workshop, although no other activity or distraction took place during recording. Content represents one full exempt of the vocational education curriculum, showing the disassembly and reassembly of an exemplary Otto engine to switch out the intake bridge. An experienced vocational instructor demonstrates all work steps one‐by‐one while giving additional verbal instructions, such as the advice on tool usage. Total video length was 9 min, 52 s for both videos.</p> <p>The presented 2D video was recorded in a resolution of 3,840 × 2,160 pixels at 30 frames per second. It was presented on a fifth generation Microsoft Surface Pro tablet (Microsoft Corp., Redmond, WA) with a resolution of 2,736 × 1,824 at an aspect ratio of 3:2. The tablet was placed in front of the participant ~75 cm away from the face with the screen at an angle of 65°.</p> <p>For the nVR condition, a 360° video was recorded in a resolution of 7,680 × 4,320 at 60 frames per second, downsized to 3,840 × 1,920/60 for presentation during the study. The HTC Vive HMD (High Tech Computer Corporation, New Taipei, Taiwan) used for presentation was a tethered 2016 version and connected to a laptop with a NVIDIA GeForce GTX 1,060 (Santa Clara, CA) via HDMI cable. The HMD had a resolution of 1,080 × 1,200 pixels per eye (2,160 × 1,200 combined) at a refresh rate of 90 Hz with an 110° field of view.</p> <hd id="AN0155323232-12">Apparatus</hd> <p>For measuring eye movements in the 2D‐video group, Tobii Glasses 2 (Tobii Pro, Danderyd, Sweden) were used. This binocular eye‐tracker records eye movements from both eyes with a sampling rate of 100 Hz using two cameras and six infrared illuminators per eye. The recorded gaze patterns are later superimposed on a video recording captured by an integrated scene camera. The camera recorded 25 frames per second with a resolution of 1,920 × 1,080 pixels and a FoV of 90°. For measuring eye movements in the 3D‐video group, the VR implemented eye‐tracker Tobii Pro VR Integration, a retrofitted 2016 version of the HTC Vive, was used. The integrated eye‐tracker recorded both eyes with a sampling rate of 120 Hz using one eye‐tracking sensor and 10 infrared illuminators per eye. The trackable field of view for eye movements is 110°.</p> <p>Eye‐Tracking data in the 2D condition was recorded using the Tobii Glasses Controller v1.108 (Tobii AB, Danderyd, Sweden); 360° data collection was conducted in Tobii Pro Lab 360VR v.118. All eye‐tracking data analyses in both groups were conducted using Tobii Pro Lab 360VR v1.118. Statistical analyses were conducted primarily using IBM SPSS 25 (IBM, Armonk, New York).</p> <p>Calibration was done via Tobii Pro Lab VR's integrated five‐point calibration for participants wearing the Tobii Pro VR Integration, while participants wearing the Tobii Glasses 2 completed the single‐point calibration provided by Tobii Glasses Controller's single‐point calibration. Calibration was successfully completed by all participants described in this study.</p> <hd id="AN0155323232-13">Design</hd> <p>A single‐factor between‐subjects design was used (2D video‐based vs. non‐interactive 360° [nVR] instruction). Participants were randomly assigned to either experimental condition.</p> <hd id="AN0155323232-14">Procedure</hd> <p>Participants were individually invited to the lab. Upon arrival, participants were greeted by two experimenters, shown the lab before receiving information about the purpose and the procedure of the study from the experimenters. There was sufficient time given to ask questions. Afterwards, the participants were then asked to give informed consent and to fill out a brief questionnaire concerning their demographic information as well as their previous experience with modern technology including VR, and car mechanics.</p> <p>Participants of the 3D‐video group were then shown the VR equipment, while participants of the 2D‐video group were shown the head‐mounted eye‐tracker before being both groups were seated and received task‐specific instructions. These instructions did not differ between the groups. When the participant had understood the instructions and any open questions were adequately answered, they were asked to put on the VR headset or eye‐tracker and calibration was initiated.</p> <p>After completing calibration, the participant watched the video assigned to his or her experimental group. Time‐stamped markers were recorded when participants started and finished the video—this happened automatically in the VR group and remotely in the 2D video condition. Since for the 2D‐video group time‐stamped recordings were collected from all participants, accurate markers were added on an individual basis for analysis.</p> <p>When participants finished watching the educational video, recording of the eye‐tracking was stopped. The experimenters then helped the participants to take off the eye‐tracker or the VR HMD before handing them the second set of questionnaires. Participants were then fully debriefed and received course credit at the end of the study. This study was approved by the Ethics committee of the University of Kaiserslautern (Application #72018).</p> <hd id="AN0155323232-15">DATA ANALYSIS</hd> <p></p> <hd id="AN0155323232-16">Selection of areas of interest</hd> <p>The video material presented in this study was recorded in the context of a larger project on digitalisation in the vocational education of car mechanics. The videos showcase a standardised step‐by‐step guide on how to disassemble a common type of engine, where each step is shown and explained by an expert. The content of the videos would usually be demonstrated in person by vocational instructors; therefore, the areas of interest were able to be pre‐defined based on expert opinions from these vocational instructors.</p> <p>The scene of the video is static in both conditions, while head movement is possible in the 3D video the participants are instructed to focus on the content, which is placed directly in front of them when the stimulus presentation starts. There were four relevant stimuli in total: The engine block in the centre of the video, the instructor who was positioned either to the right or left of the engine and hence represented by two areas of interest, a table with tools on the right side as well as an assembly trolley to the left where spare parts were placed. The engine was mounted on a carriage and remained stationary throughout the recording and was defined as the first primary area of interest. To account for movement throughout the video, two counterbalanced areas of interests were defined for the instructor—Instructor (Left) and Instructor (Right). At the beginning of the video, the instructor stands to the right of the engine but moves over the course of the video between both spots.</p> <p>In line with the expert opinions of the collaborating vocational instructors, it was assumed participants would focus primarily on the instructor and the engine. The table with tools was seen as potentially relevant due to the instructor's explanation which tools are utilised in each work step. The assembly trolley was included as a secondary area of interest since the instructor places spare parts throughout the disassembly on it, although this happens rather infrequently.</p> <p>After initial data screening, one of these five areas of interest was excluded, which was the assembly trolley. No participant spent more than a few seconds on it and further analyses suggested most fixations on the trolley were visiting fixations or glances, with few full fixations overall. Thus, the four other areas of interest were analysed in this study: The instructor in two positions to account for movement over the course of the video, the engine, and the table. An exemplary frame of the 360° video, which shows all areas of interest as well as the full 360° video environment, can be seen in Figure 1. The areas of interest and field of view in the 2D group were similar to the lower picture in Figure 1.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0001.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0001.jpg" title="1 Snapshot of the 360° recording with added areas of interest in full 360° view (top) and focal area (bottom)" /> </p> <p></p> <hd id="AN0155323232-18">Eye‐tracking data acquisition</hd> <p>In the 2D video group, automatic mapping on snapshot basis as provided by Tobii Pro Lab was conducted individually for every participant using a single keyframe chosen from each recording. The keyframe for each participant was marked with pre‐defined areas of interest (AoI) based on the relevant objects in the presented video as described above. To conduct AoI‐based analyses, each participants' recording was rendered over the chosen keyframe. Gaze data and pupil data were also recorded but not analysed for this study. In addition to the automatic mapping, participant recordings were checked at least once per second to identify potential automated mislabelling of data.</p> <p>For the 3D‐video group, the data acquisition was conducted similarly with the same selection of areas of interest defined a priori but in addition the 360° video was pre‐rendered before conducting the experiment since there were no individual differences in viewing angle or distance due to the use of a stationary placement of the participant in the video. In effect, this pre‐rendering enables automatic mapping of participant gaze patterns onto areas of interest. Similar to the 2D video group, recorded eye‐movements of each participant were superimposed within pre‐determined temporal markers onto the 360° video. As with the 2D recordings, participant recordings were checked at least once per second to identify potential automated mislabelling of data.</p> <hd id="AN0155323232-19">Data quality</hd> <p>The overall quality of the collected data was satisfactory in both groups. One participant had to be excluded due to a potential calibration issue which resulted in no AoI‐based fixations despite otherwise normal gaze patterns. One participant had to be excluded due to a battery related problem which resulted in an incomplete recording. All other recordings from both groups were within the expected parameters. In two participant cases, recordings were split into two or three parts by the automatic mapping algorithm, further investigation revealed no anomalous eye‐tracking data, both recordings in question were complete. It is presumed that these interruptions might have been caused by participants closing or averting their eyes for a few seconds.</p> <p>It should be noted that presentation of the 360° video was hard‐coded in a self‐running environment within the Tobii Pro Lab software, whereas the 2D‐video was manually started by one experimenter, after manually starting the recording of the eye‐tracking glasses. Therefore, while video length was held constant between groups, the recording duration slightly varied between the groups and within the 2D video group. Overall data matching was working as intended with a single keyframe allowing gaze mapping on average 97.9% of recording duration in the 2D group, and 99.9% in the VR group due to the automatic rendering as described above. Figure 2 shows the distribution of recording durations for the 2D group (<emph>N</emph> = 23) and the VR group (<emph>N</emph> = 25).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0002.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0002.jpg" title="2 Distribution of recording durations, separated by group. Group mean indicated by triangle marker. Error bars indicate standard deviation" /> </p> <p></p> <p>The classification of recorded eye‐tracking data was done in accordance with Tobii Pro's white paper on the fixation filter in Tobii Pro Lab (Olsen, 2012; Tobii Technology, 2012). This study used the Tobii I‐VT fixation filter with two changes—the addition of a gap fill‐in interpolation with a maximum gap length of 75 ms to account for losses of data due to technical issues of up to three frames (see Olsen, 2012). The second change was setting the threshold for maximum gaze velocity calculation from 30°/s to 20°/s since participants in neither experimental condition were expected to make sudden eye‐movements (see Olsen, 2012). This change was made to decrease false‐positive merging of adjacent fixations. All other settings were kept in line with Tobii recommendations, notably fixations shorter than 60 ms were excluded from analysis and adjacent fixations were merged when they were &lt;0,5° apart (Tobii Technology, 2012).</p> <hd id="AN0155323232-21">RESULTS</hd> <p></p> <hd id="AN0155323232-22">Manipulation check: Sense of presence</hd> <p>There was one primary presumption which had to be accounted for. HMD‐based VR is seen as one of the most immersive experimental modalities, therefore participants in the 3D‐video group, using HMD, are expected to report more immersion and feelings of presence (Schubert et al., 1999, 2001) compared to the 2D‐video group using tablets. The manipulation check hypothesis was therefore that there are differences in perceived presence, indicated by higher self‐report scores on the IPQ, with participants in the 3D‐video group reporting increased presence compared to the 2D‐video group. To test this hypothesis, a univariate analysis of variance (ANOVA) with the factor 'Group' and the dependent variable 'IPQ Grand Mean' was conducted. Levene's test was not significant (<emph>F</emph>(<reflink idref="bib1" id="ref1">1</reflink>, 46) = 1.54, <emph>p</emph> = 0.220), indicating equal error variance equal across groups. The IPQ Grand Mean was calculated as the grand average of all 13 used items of the adapted German version of the IPQ used in this study. An unrotated factorial analysis was conducted and results were mostly in accordance with the pre‐existing three factor structure in the original questionnaire. There was a strong tendency towards a single factor solution which explained 44.69% of total variance. To investigate levels of immersion and presence reported by participants, this single factor solution with an Eigenvalue of 6.26 was used. The ANOVA revealed a significant difference (<emph>F</emph>(<reflink idref="bib1" id="ref2">1</reflink>, 46) = 14,75, <emph>p &lt;</emph> 0.001, <emph>η</emph>² = 0.243) in perceived immersion and presence, with the nVR group reporting significantly higher scores on the IPQ (<emph>M</emph> = 4.03, <emph>SD</emph> = 0.65) compared to the 2D group (<emph>M</emph> = 3.18, <emph>SD</emph> = 0.87). The results are displayed in Figure 3.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0003.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0003.jpg" title="3 Averaged Grand Sum IPQ scores, separated by group. Error bars indicate 95% confidence intervals" /> </p> <p></p> <hd id="AN0155323232-24">Hypothesis 1: Differences in AoI‐based total fixation duration</hd> <p>The primary research question this paper aims to answer was concerned with the attentional focus, indicated by differences in total fixation duration on relevant areas of interest, between the two experimental groups. It was assumed that there should be attentional differences between HMD‐based 360° and traditional video‐based learning. For the purpose of this study, the total fixation duration within pre‐defined areas of interest is seen as representative for attentional focus (see eg, Bhoir et al., 2015; Jeelani et al., 2018; Yarbus, 1967). As the literature so far provides evidence in either direction and the viability of eye‐tracking in educational VR is relatively novel, this hypothesis was undirected.</p> <p>For every area of interest, a univariate ANOVA with the variable 'Total Fixation Duration' was conducted. The Total Fixation Duration is the sum of all individual fixations over a given time of interest. In this study, the time of interest was the full duration of the video (<emph>T</emph> = 592 s) for each group. As detailed above, the areas of interest under consideration are the instructor, the engine block and the table with tools. Levene's tests were conducted for each ANOVA, with the total fixation duration data for the instructor not being normally distributed between groups (<emph>F</emph>(<reflink idref="bib1" id="ref3">1</reflink>, 46) = 4.612, <emph>p</emph> = 0.037). Therefore, a Kruskal–Wallis test was conducted for the instructor data. Both the engine (<emph>F</emph>(<reflink idref="bib1" id="ref4">1</reflink>, 46) = 3.230, <emph>p</emph> = 0.079) and table (<emph>F</emph>(<reflink idref="bib1" id="ref5">1</reflink>, 46) = 0.932, <emph>p</emph> = 0.339) fixation data did not differ significantly from normal distribution, hence no non‐parametric tests were necessary.</p> <hd id="AN0155323232-25">Part 1: Attentional focus on the instructor</hd> <p>To investigate whether the groups showed differences in their attentional focus on the instructor, a univariate ANOVA with the factor 'Group' and the dependent variable 'Total Fixation Duration: Instructor' was conducted. There was a significant main effect (<emph>F</emph>(<reflink idref="bib1" id="ref6">1</reflink>, 46) = 89.75, <emph>p &lt;</emph> 0.001, <emph>η</emph>² = 0.661) in total fixation time on the instructor, with the 3D‐video group fixating more on the instructor (<emph>M</emph> = 98.55 s, <emph>SD</emph> = 34.24 s) compared to the 2D‐video group (<emph>M</emph> = 24.93 s, <emph>SD</emph> = 15.28 s). The results can be seen in Figure 4.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0004.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0004.jpg" title="4 Sum of all fixations on areas of interest, in seconds, separated by group. Error bars indicate 95% confidence intervals" /> </p> <p></p> <p>A Kruskal–Wallis Test was conducted to account for the not normally distributed data. The previous results could be confirmed, there was a significant difference in total fixation duration on the instructor, <emph>H</emph>(<reflink idref="bib1" id="ref7">1</reflink>) = 34.231, <emph>p</emph> &lt; 0.001. Participants in the 3D Video group (<emph>Mdn</emph> = 96.78 s) fixated significantly longer on the instructor compared to participants in the 2D Video group (<emph>Mdn</emph> = 18.40 s).</p> <hd id="AN0155323232-27">Part 2: Attentional focus on the engine</hd> <p>The second area of interest was the engine block itself. A univariate ANOVA on 'Total Fixation Duration: Engine' revealed no difference between participants in the 2D (<emph>M</emph> = 150.95 s, <emph>SD</emph> = 89.81 s) and the 3D‐video (<emph>M</emph> = 189.53 s, <emph>SD</emph> = 62.43 s) group in their attentional focus on the engine, <emph>F</emph>(<reflink idref="bib1" id="ref8">1</reflink>, 46) = 3.03, <emph>p =</emph> 0.089, <emph>η</emph><sups>2</sups> = 0.062. The results can be seen in Figure 4.</p> <hd id="AN0155323232-28">Part 3: Attentional focus on the table</hd> <p>Due to the research question of this study, the presented eye‐tracking data primarily focussed on the engine and instructor. The table has been included as a minor area of interest since there is educational context related to it presented in the video. A univariate ANOVA on 'Total Fixation Duration: Table' revealed no difference between participants in the 2D (<emph>M</emph> = 8.11 s, <emph>SD</emph> = 6.09 s) and the 3D‐video (<emph>M</emph> = 10.70 s, <emph>SD</emph> = 5.28 s) group in their attentional focus on the engine, <emph>F</emph>(<reflink idref="bib1" id="ref9">1</reflink>, 46) = 2.50, <emph>p</emph> = 0.121, <emph>η</emph><sups>2</sups> = 0.051. Results can be seen in Figure 4.</p> <p>In summary, Hypothesis 1 could be partially confirmed, with participants in the nVR group fixating more on the instructor, but not the engine or table compared to the 2D video group.</p> <hd id="AN0155323232-29">Hypothesis 2: Learning outcome</hd> <p>The first experimental hypothesis involved the learning outcome. Based on the literature review, it was predicted that participants in the 360° group should perform better compared to participants in the video condition. Performance was measured by a standardised 16 item exam. Each of the 16 items had between 1 and 4 correct responses with each correctly marked or correctly unmarked response awarding one point. Thus, the highest achievable score was 64 points.</p> <p>To investigate whether the participants in the experimental conditions showed differences in the learning outcome, a univariate ANOVA with the factor 'Group' and the dependent variable 'Learning Outcome' revealed no significant differences between groups.</p> <p>However, Levene's test revealed significant differences (<emph>F</emph>(<reflink idref="bib1" id="ref10">1</reflink>, 46) = 5.26, <emph>p</emph> = 0.026) in the error variances between the two experimental groups, hence a Kruskal–Wallis test with the grouping variable 'Group' and the dependent variable 'Learning Outcome' was conducted. No significant difference was found, <emph>H</emph>(<reflink idref="bib1" id="ref11">1</reflink>) = 0.063, <emph>p</emph> = 0.801, the learning outcome of participants in the nVR group (<emph>Mdn</emph> = 59.00) did not differ from participants in the 2D video group (<emph>Mdn</emph> = 59.00). The means and 95% confidence intervals for each group are displayed in Figure 5, whereas the distribution of the individual values for each group is visualised in Figure 6.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0005.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0005.jpg" title="5 Comparison of test scores on the 16‐item knowledge test, separated by group. Error bars indicate 95% confidence intervals" /> </p> <p></p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0006.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0006.jpg" title="6 Distribution of test score averages on the 16‐item, separated by group. Group mean indicated by triangle marker. Error bars indicate standard deviation" /> </p> <p></p> <p>In short, Hypothesis 2 could not be confirmed, no differences in learning outcome could be attributed to the learning modality.</p> <hd id="AN0155323232-32">Additional analyses: Cognitive load</hd> <p>In addition to the results above participants were asked to fill out the NASA TLX after watching the educational video in either group as subjective measure for experienced cognitive load.</p> <p>Previous literature (see eg, Armougum et al., 2019) found a decrease in cognitive load when participants grew familiar with navigation in VR compared to unfamiliar participants. This study recruited primarily participants who were unfamiliar with VR, therefore the TLX was used as a measurement to compare the subjective load of watching a 3D‐video on an HMD compared to a more common task such as watching a 2D video on a tablet. To investigate whether the participants in the experimental conditions showed differences in the subjective cognitive load, a univariate ANOVA with the factor 'Group' and the dependent variable 'Average Cognitive Load' was conducted.</p> <p>Levene's test indicated no differences (<emph>F</emph>(<reflink idref="bib1" id="ref12">1</reflink>, 46) &lt; 1, <emph>p</emph> = 0.463) in the error variances between the two experimental groups. The ANOVA did not reveal significant differences (<emph>F</emph>(<reflink idref="bib1" id="ref13">1</reflink>, 46) = 1.477<emph>, p </emph>= 0.231<emph>, η<sups>2</sups></emph> = 0.031) in subjective load between participants in the 3D‐video group (<emph>M</emph> = 8,49, <emph>SD</emph> = 3,22) and participants in the 2D‐video group (<emph>M</emph> = 7,34, <emph>SD</emph> = 3,34). Results can be seen in Figure 7.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/58I/01mar22/bjet13162-fig-0007.jpg?ephost1=dGJyMNXb4kSepq84yOvqOLCmsE6epq5Srqa4SK6WxWXS" alt="bjet13162-fig-0007.jpg" title="7 Averaged NASA TLX scores, separated by group. Error bars indicate 95% confidence intervals" /> </p> <p></p> <hd id="AN0155323232-34">DISCUSSION</hd> <p>This study investigated differences in vocational education using instructions through either a 2D video presented via tablet or a 360° video presented via HMD and found no effect of the presentation modality on test performance, but differences in total fixation duration. Therefore, the primary hypothesis was able to be partially confirmed since the effect was only significant for the instructor. The second hypothesis, which predicted a higher learning outcome in the 360° nVR video condition compared to the 2D‐video, must be rejected since no difference was found. The study revealed several differences in attention between the 2D‐video and nVR groups: participants in the nVR group fixated on the instructor both significantly more and significantly longer than the 2D‐video group. This led to a significantly higher total fixation duration on the instructor in the nVR group. A similar pattern was not seen in engine or table fixation patterns.</p> <p>Regarding Hypothesis 2, the results were contrary to the expected outcome. Participants were able to learn content from an unfamiliar field with both a tablet and a HMD in the same way: participants in both groups performed surprisingly well based on their backgrounds and the assumed complexity of the learning material. Regardless of if they had watched the 2D or nVR video, they scored very high on the knowledge recall test. However, these findings are in line with the meta‐analysis from Moro and colleagues (2021) and a recent study by Ulrich et al. (2021) found similar increases in learning outcome for education done via 360° videos, via 2D videos and via face‐to‐face training, which is in line with the results of this study as well.</p> <p>Speaking from a strictly performance‐based perspective, the results of this study show that tablets with 2D‐video content are just as effective for instruction as 360° nVR videos presented on an HMD. Since tablets are much more readily available and cost‐effective compared to setups utilising HMDs, 2D video‐based vocational education could be a viable short‐term solution until fully interactive VR content becomes accessible to a broader audience.</p> <p>As predicted, participants watching the 360° VR video rated the video experience as more engaging and felt more present—indicated by higher self‐report scores on the IPQ scales—compared to participants in the tablet group.</p> <p>Participants in the 360° video condition showed increased attentional focus on the instructor as indicated by a higher number of fixations and a longer total fixation duration on the instructor compared to the 2D‐video group. This is in line with previous research (Albus et al., 2021; Birmingham et al., 2008; Rubo &amp; Gamer, 2021), which found increased attention in VR for socially relevant stimuli, which is the case for the instructor in the presented content. Chen et al. (2018) found higher learning satisfaction using VR in automotive vocational education settings, as well as higher learning outcome. While this study did not find any increase in learning outcome compared to 2D videos, the presented results do indicate that there is higher engagement with the learning material even in non‐interactive VR scenarios, which could lead to higher learning satisfaction. However, this was not investigated in this study and should be confirmed in future research.</p> <p>In a short poll conducted after the primary experiments, both groups rated the importance of new educational technology as high, an opinion shared by many VR research groups and tech companies. As answer to an open question about problems and challenges regarding the usage of VR in education many participants stated that they themselves, as the next generation of teachers, are not well prepared because it was lacking in their own education. Another issue mentioned by several participants was that there is a lack of content for VR‐based education (see also Jensen &amp; Konradsen, 2018). While this still poses a challenge for many vocational institutions, companies are emerging, which aim to provide content for different levels of education. The recent surge in remote teaching due to the COVID‐19 pandemic highlighted the need for educational concepts, which are both engaging and remote. VR can fill this gap, and many companies such as edify (https://<ulink href="http://www.edify.ac/">www.edify.ac/</ulink>) or Labster (https://<ulink href="http://www.labster.com/">www.labster.com/</ulink>) strive to create virtual laboratories or other collaborative spaces for use with VR. Others such as the WDR, a German media company provide free classroom materials (https://www..wdr.de/schule/digital/unterrichtsmaterial/virtualreality‐100.html) tailored for school children, which utilise 360° videos that could be played also on a tablet or smartphone. Without a doubt, the next years will see an increase in available materials for all levels of education, which makes the scientific investigation of their effectiveness for the learner even more relevant.</p> <hd id="AN0155323232-35">Notes on data quality in HMD‐based and mobile eye‐tracking research</hd> <p>Due to the novelty of the deployed design, it is necessary to elaborate on the utilised eye tracking devices and their comparability between HDM‐integrated and standard mobile eye tracking devices. This is largely based on the Tobii environment used in this study. The principles can, however, also be applied to other eye‐tracking systems.</p> <p>On a surface level, VR provides two major experimental advantages over standard experimental settings. First, the participants' visual field is fully controllable, leading to higher experimental control and participant engagement (Bacca et al., 2015; Fox et al., 2009). Second, any head or body movement is counterbalanced by the HMD technology by default. This leads to a higher quality of the data compared to mobile eye‐tracking systems. However, we will focus more on the temporal dimension of eye‐tracking.</p> <p>One possible flaw in this study lies in the differences between the deployed eye‐tracking systems. While both systems were developed by the same company, there are differences between HMD integrated eye‐tracking and standard mobile or stationary eye‐trackers. One crucial difference lies in the automatic data segmentation and rendering in VR. Since the participations field of view is not only fully observable, but also controllable by and shared across all participants, problems of data loss that might occurred in the recording and analysis of mobile eye‐tracking data do not occur in VR at all. This is visible in both in the higher homogeneity of data collected by the VR‐based system and also in the temporal dimension, since video presentation will be matched frame‐by‐frame between participants, as can be seen in Figure 2.</p> <p>Mobile eye‐tracking introduces some significant differences compared to the traditional stationary eye‐tracking paradigms. Whereas the latter requires a stationary participant, often with a headrest or even head fixation, mobile eye‐tracking will often involve the participant moving their head or entire body over the course of the recording. This poses a significant challenge for area‐of‐interest based research, since the visual field and thus the areas of interest might move over the course of the experiment. On the other hand, mobile eye‐tracking enables the investigation of learning in more natural settings since the learner can remain in the normal learning environment and is obstructed less by the recording gear, which makes it much more viable for field studies conducted in many fields of educational research.</p> <p>In this study, participants were seated with their attention focussed on a screen in front of them. Therefore, some of the problems described above could be alleviated. However, while the participants were seated a fixed distance from the tablet (60 cm), screen size differences between participants due to body or head movement during data recording could not fully be accounted for despite instructions to move as little as possible.</p> <p>While great care has been applied in the structuring, filtering and the analyses of all data, there are still more chances for human error to occur in manual analysis compared to fully automated, standardised analysis. However, we are confident that these potential oversights, while problematic, should have no major impact on the present results.</p> <hd id="AN0155323232-36">Limitations</hd> <p>This study utilised a single knowledge test administered briefly after the educational content. To strengthen potential insights into long‐term knowledge acquisition, a longer delay or a form pre‐/post‐design would have been preferable. While this study incorporated a delay period of 15 minutes, this period might have proven insufficient to have a significantly effect on knowledge recall. On the contrary, it could be hypothesized that the learning abilities of high‐performing participants such as university students are more suited to short‐term recall such as the one applied in this study—even without prior knowledge on the subject. This is, however, strictly hypothetical and warrants further investigation.</p> <p>However, all participants were pre‐scanned for knowledge of car mechanics and confirmed to have no former professional experience. Furthermore, the deployed knowledge test was designed by a vocational education school as part of a collaborative project and was kept close to exams taken by mechanics‐in‐training as part of their vocational education. The difficulty of the knowledge test was deemed sufficiently high to exclude random guessing as a successful test strategy. In this light, the apparent ceiling effect seems surprising. However, it should be noted that the subjects in this study were undergraduate students enrolled at a public university compared to the usual target group of such exams who rarely have university‐level education backgrounds. As no comparative study was conducted with entry‐level vocational students, generalisation at this point is difficult and further research with the target group is necessary. While this study incorporated a delay period of 15 minutes, this period might have proven insufficient to have a significantly effect on knowledge recall. On the contrary, it could be hypothesized that the learning abilities of high‐performing participants such as university students are more suited to short‐term recall such as the one applied in this study—even without prior knowledge on the subject. This is, however, strictly hypothetical and warrants further investigation.</p> <hd id="AN0155323232-37">Implications for further research &amp; outlook</hd> <p>One crucial area of education where VR holds promise is in transferring knowledge from the theoretical to the practical domain. Previous research indicated that transfer learning is possible utilising VR environments, for instance for driving performance (Wallet et al., 2009) and in school educational knowledge acquisition (Meyer et al., 2019; Parong &amp; Mayer, 2018). So far, however, there are only a few studies which investigate learning effects of VR on real world scenarios. At present, the common approaches seem more focussed on knowledge recall than actual task performance. This is possibly due to the increased complexity in study designs that would need to utilise both virtual and real‐world facilities. Nevertheless, it would be a promising approach to combine VR‐based training scenarios with the follow‐up real‐world task performance. This could yield promising insights into the capability of VR to prepare vocational students for their later jobs in the real world.</p> <p>In addition, this study showcases the possibility to apply eye‐tracking methodology to VR research, which opens a variety of possibilities for further research on attention, cognitive load and other correlates of eye‐tracking data which can now be collected within VR. The speed of technological progress likely will not slow down in the near future, at the time of writing 6DoF VR HMDs with integrated eye‐tracking are already available at a consumer level in form of the HTC Vive Pro Eye (High Tech Computer Corporation, New Taipei, Taiwan). From the content side, both private and state‐directed content providers for 360° video and fully immersive virtual content begin to become more commonplace all over the world. The authors hope that the present paper offers a first insight using eye‐tracking into the differences and similarities between traditional educational 2D and 360° videos.</p> <hd id="AN0155323232-38">ACKNOWLEDGEMENTS</hd> <p>We want to express our gratitude to Mr. Udo Petruschkat and the Berufsbildungszentrum Märkischer Kreis (bbz), a vocational education institute in Iserlohn, Germany, for their assistance in producing of the instructional videos, willingness to share their facilities as well as providing this study with exam questions. Furthermore, we want to thank Mrs. Julia Herzhauser and Mr. Felix Assel who assisted in data collection as part of their Bachelor thesis.</p> <hd id="AN0155323232-39">CONFLICT OF INTEREST</hd> <p>The authors report no conflicts of interest.</p> <hd id="AN0155323232-40">ETHICS STATEMENT</hd> <p>This study was reviewed and approved by the Ethics committee of the University of Kaiserslautern (Application #72018). All collected data was anonymised before analysis. Informed consent was gathered from the participant before study participation, and the signed consent form is stored separately from other forms. Participant data is grouped using labels (VR1, VR2, ...) and cannot be traced back to individual participants after analyses.</p> <hd id="AN0155323232-41">DATA AVAILABILITY STATEMENT</hd> <p>The files used for quantitative data analysis are available upon request from Mr. Felix Hekele (felix.hekele@sowi.tu-kl.de) due to the large size of the exported eye‐tracking raw data. This does not include the original videos for each participant as recorded by Tobii to ensure participant anonymity, but all raw data extracted from them.</p> <ref id="AN0155323232-42"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref1" type="bt">1</bibl> <bibtext> Funding information This research was conducted as part of the 'InKraFT' project funded by the German Federal Ministry for Education and Research (BMBF, funding number 01PE17003)</bibtext> </blist> </ref> <ref id="AN0155323232-43"> <title> REFERENCES </title> <blist> <bibtext> Albus, P., Vogt, A., &amp; Seufert, T. (2021). Signaling in virtual reality influences learning outcome and cognitive load. Computers &amp; Education, 166, 104154. https://doi.org/10.1016/j.compedu.2021.104154</bibtext> </blist> <blist> <bibl id="bib2" type="bt">2</bibl> <bibtext> Armougum, A., Orriols, E., Gaston‐Bellegarde, A., Joie‐La Marle, C., &amp; Piolino, P. (2019). Virtual reality: A new method to investigate cognitive load during navigation. Journal of Environmental Psychology, 65, 101338. https://doi.org/10.1016/j.jenvp.2019.101338</bibtext> </blist> <blist> <bibl id="bib3" type="bt">3</bibl> <bibtext> Autorengruppe Bildungsberichterstattung. (2018). Bildung in Deutschland 2018: Ein indikatorengestützter Bericht mit einer Analyse zu Wirkungen und Erträgen von Bildung [Education in Germany 2018: An indication‐based report analyzing the effects and benefits of education]. wbv.</bibtext> </blist> <blist> <bibl id="bib4" type="bt">4</bibl> <bibtext> Bacca, J., Baldiris, S., Fabregat, R., &amp; Graf, S. (2015). Mobile augmented reality in vocational education and training. Procedia Computer Science, 75, 49 – 58. https://doi.org/10.1016/j.procs.2015.12.203</bibtext> </blist> <blist> <bibl id="bib5" type="bt">5</bibl> <bibtext> Baethge, M., Solga, H., &amp; Wieck, M. (2007). Übergänge aus allgemein bildenden Schulen in die Berufsbildung. In Berufsbildung im Umbruch. Signale eines überfälligen Aufbruchs [Vocational education in Upheaval. Signals of an overdue awakening] (pp. 37 – 58). Friedrich‐Ebert‐Stiftung.</bibtext> </blist> <blist> <bibl id="bib6" type="bt">6</bibl> <bibtext> Bhoir, S. A., Hasanzadeh, S., Esmaeili, B., Dodd, M. D., &amp; Fardhosseini, M. S. (2015, June). Measuring construction workers' attention using eye‐tracking technology. Proceedings of the Canadian Society for Civil Engineering 5th Int./11th Construction Specialty Conference, Canada, 222, 1 – 10.</bibtext> </blist> <blist> <bibl id="bib7" type="bt">7</bibl> <bibtext> Birmingham, E., Bischof, W. F., &amp; Kingstone, A. (2008). Gaze selection in complex social scenes. Visual Cognition, 16 (2–3), 341 – 355. https://doi.org/10.1080/13506280701434532</bibtext> </blist> <blist> <bibl id="bib8" type="bt">8</bibl> <bibtext> Chen, Y. F., Luo, Y. Z., Fang, X., &amp; Shieh, C. J. (2018). Effects of the application of computer multimedia teaching to automobile vocational education on students' learning satisfaction and learning outcome. EURASIA Journal of Mathematics, Science and Technology Education, 14 (7), 3293 – 3300. https://doi.org/10.29333/ejmste/91245</bibtext> </blist> <blist> <bibl id="bib9" type="bt">9</bibl> <bibtext> Cheng, Y., &amp; Huang, R. (2012). Using virtual reality environment to improve joint attention associated with pervasive developmental disorder. Research in Developmental Disabilities, 33 (6), 2141 – 2152. https://doi.org/10.1016/j.ridd.2012.05.023</bibtext> </blist> <blist> <bibtext> Dyer, E., Swartzlander, B. J., &amp; Gugliucci, M. R. (2018). Using virtual reality in medical education to teach empathy. Journal of the Medical Library Association, 106 (4), 498 – 500. https://doi.org/10.5195/jmla.2018.518</bibtext> </blist> <blist> <bibtext> Fox, J., Arena, D., &amp; Bailenson, J. N. (2009). Virtual reality: A survival guide for the social scientist. Journal of Media Psychology, 21 (3), 95 – 113. https://doi.org/10.1027/1864‐1105.21.3.95</bibtext> </blist> <blist> <bibtext> Frederiksen, J. G., Sørensen, S. M. D., Konge, L., Svendsen, M. B. S., Nobel‐Jørgensen, M., Bjerrum, F., &amp; Andersen, S. A. W. (2020). Cognitive load and performance in immersive virtual reality versus conventional virtual reality simulation training of laparoscopic surgery: A randomized trial. Surgical Endoscopy, 34 (3), 1244 – 1252. https://doi.org/10.1007/s00464‐019‐06887‐8</bibtext> </blist> <blist> <bibtext> Freina, L., &amp; Ott, M. (2015, April). A literature review on immersive virtual reality in education: State of the art and perspectives. Proceedings of the International Scientific Conference of eLearning and Software for Education (eLSE), Romania, 1 (133), 133 – 141.</bibtext> </blist> <blist> <bibtext> Grassini, S., &amp; Laumann, K. (2020). Questionnaire measures and physiological correlates of presence: A systematic review. Frontiers in Psychology, 11 (349), 1 – 21. https://doi.org/10.3389/fpsyg.2020.00349</bibtext> </blist> <blist> <bibtext> Hart, S. G. (2006, October). NASA‐task load index (NASA‐TLX); 20 years later. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, Los Angeles, CA, 50 (9), 904 – 908.</bibtext> </blist> <blist> <bibtext> Hart, S. G., &amp; Staveland, L. E. (1988). Development of NASA‐TLX (Task Load Index): Results of empirical and theoretical research. In P. A. Hancock &amp; N. Meshkati (Eds.), Human mental workload (pp. 139 – 183). North Holland.</bibtext> </blist> <blist> <bibtext> Huang, W. (2020). Investigating the novelty effect in virtual reality on stem learning (Doctoral dissertation). Retrieved from ProQuest Dissertations &amp; Theses Global database. (UMI No. 27959999).</bibtext> </blist> <blist> <bibtext> Huang, W., Roscoe, R. D., Johnson‐Glenberg, M. C., &amp; Craig, S. D. (2021). Motivation, engagement, and performance across multiple virtual reality sessions and levels of immersion. Journal of Computer Assisted Learning, 37 (3), 745 – 758. https://doi.org/10.1111/jcal.12520</bibtext> </blist> <blist> <bibtext> Jeelani, I., Han, K., &amp; Albert, A. (2018). Automating and scaling personalized safety training using eye‐tracking data. Automation in Construction, 93, 63 – 77. https://doi.org/10.1016/j.autcon.2018.05.006</bibtext> </blist> <blist> <bibtext> Jensen, L., &amp; Konradsen, F. (2018). A review of the use of virtual reality head‐mounted displays in education and training. Education and Information Technologies, 23 (4), 1515 – 1529. https://doi.org/10.1007/s10639‐017‐9676‐0</bibtext> </blist> <blist> <bibtext> Jeong, J. B., Lee, S., Ryu, I. W., Le, T. T., &amp; Ryu, E. S. (2020, October). Towards viewport‐dependent 6DoF 360 video tiled streaming for virtual reality systems. In Proceedings of the 28th ACM International Conference on Multimedia, Seattle, WA (pp. 3687 – 3695).</bibtext> </blist> <blist> <bibtext> Meppelink, C. S., &amp; Bol, N. (2015). Exploring the role of health literacy on attention to and recall of text‐illustrated health information: An eye‐tracking study. Computers in Human Behavior, 48, 87 – 93. https://doi.org/10.1016/j.chb.2015.01.027</bibtext> </blist> <blist> <bibtext> Merchant, Z., Goetz, E. T., Cifuentes, L., Keeney‐Kennicutt, W., &amp; Davis, T. J. (2014). Effectiveness of virtual reality‐based instruction on students' learning outcomes in K‐12 and higher education: A meta‐analysis. Computers &amp; Education, 70, 29 – 40. https://doi.org/10.1016/j.compedu.2013.07.033</bibtext> </blist> <blist> <bibtext> Meyer, O. A., Omdahl, M. K., &amp; Makransky, G. (2019). Investigating the effect of pre‐training when learning through immersive virtual reality and video: A media and methods experiment. Computers &amp; Education, 140, 103603. https://doi.org/10.1016/j.compedu.2019.103603</bibtext> </blist> <blist> <bibtext> Morales, T. M., Bang, E., &amp; Andre, T. (2013). A one‐year case study: Understanding the rich potential of project‐based learning in a virtual reality class for high school students. Journal of Science Education and Technology, 22 (5), 791 – 806. https://doi.org/10.1007/s10956‐012‐9431‐7</bibtext> </blist> <blist> <bibtext> Moro, C., Birt, J., Stromberga, Z., Phelps, C., Clark, J., Glasziou, P., &amp; Scott, A. M. (2021). Virtual and augmented reality enhancements to medical and science student physiology and anatomy test performance: A systematic review and meta‐analysis. Anatomical Sciences Education, 14 (3), 368 – 376. https://doi.org/10.1002/ase.2049</bibtext> </blist> <blist> <bibtext> Nobrega, F. A., &amp; Rozenfeld, C. C. D. F. (2019). Virtual reality in the teaching of FLE in a Brazilian public school. Languages, 4 (2), 36 – 49. https://doi.org/10.3390/languages4020036</bibtext> </blist> <blist> <bibtext> Olsen, A. (2012). The Tobii I‐VT fixation filter [White Paper]. Tobii® Technology. https://<ulink href="http://www.tobiipro.com/siteassets/tobii&amp;#8208;pro/learn&amp;#8208;and&amp;#8208;support/analyze/how&amp;#8208;do&amp;#8208;we&amp;#8208;classify&amp;#8208;eye&amp;#8208;movements/tobii&amp;#8208;pro&amp;#8208;i&amp;#8208;vt&amp;#8208;fixation&amp;#8208;filter.pdf">www.tobiipro.com/siteassets/tobii&amp;#8208;pro/learn&amp;#8208;and&amp;#8208;support/analyze/how&amp;#8208;do&amp;#8208;we&amp;#8208;classify&amp;#8208;eye&amp;#8208;movements/tobii&amp;#8208;pro&amp;#8208;i&amp;#8208;vt&amp;#8208;fixation&amp;#8208;filter.pdf</ulink></bibtext> </blist> <blist> <bibtext> Parong, J., &amp; Mayer, R. E. (2018). Learning science in immersive virtual reality. Journal of Educational Psychology, 110 (6), 785 – 797. https://doi.org/10.1037/edu0000241</bibtext> </blist> <blist> <bibtext> Reichenberger, J., Pfaller, M., &amp; Mühlberger, A. (2020). Gaze behavior in social fear conditioning: An eye‐tracking study in virtual reality. Frontiers in Psychology, 11, 35. https://doi.org/10.3389/fpsyg.2020.00035</bibtext> </blist> <blist> <bibtext> Reinhard, R. T., Kleer, M., &amp; Dreßler, K. (2019). The impact of individual simulator experiences on usability and driving behavior in a moving base driving simulator. Transportation Research Part F: Traffic Psychology and Behaviour, 61, 131 – 140. https://doi.org/10.1016/j.trf.2018.01.004</bibtext> </blist> <blist> <bibtext> Rubo, M., &amp; Gamer, M. (2021). Stronger reactivity to social gaze in virtual reality compared to a classical laboratory environment. British Journal of Psychology, 112 (1), 301 – 314. https://doi.org/10.1111/bjop.12453</bibtext> </blist> <blist> <bibtext> Schubert, T., Friedmann, F., &amp; Regenbrecht, H. (1999). Embodied presence in virtual environments. In R. Paton &amp; I. Neilson (Eds.), Visual representations and interpretations (pp. 269 – 278). Springer. https://doi.org/10.1007/978‐1‐4471‐0563‐3_30</bibtext> </blist> <blist> <bibtext> Schubert, T., Friedmann, F., &amp; Regenbrecht, H. (2001). The experience of presence: Factor analytic insights. Presence: Teleoperators and Virtual Environments, 10 (3), 266 – 281. https://doi.org/10.1162/105474601300343603</bibtext> </blist> <blist> <bibtext> Slater, M. (2003). A note on presence terminology. Presence Connect, 3 (3), 1 – 5.</bibtext> </blist> <blist> <bibtext> Slater, M. (2018). Immersion and the illusion of presence in virtual reality. British Journal of Psychology, 109 (3), 431 – 433. https://doi.org/10.1111/bjop.12305</bibtext> </blist> <blist> <bibtext> Tobii® Technology. (2012). Determining the Tobii I‐VT fixation filter's default values method description and results discussion [White Paper]. Author. https://<ulink href="http://www.tobiipro.com/siteassets/tobii&amp;#8208;pro/learn&amp;#8208;and&amp;#8208;support/analyze/how&amp;#8208;do&amp;#8208;we&amp;#8208;classify&amp;#8208;eye&amp;#8208;movements/determining&amp;#8208;the&amp;#8208;tobii&amp;#8208;pro&amp;#8208;i&amp;#8208;vt&amp;#8208;fixation&amp;#8208;filters&amp;#8208;default&amp;#8208;values.pdf">www.tobiipro.com/siteassets/tobii&amp;#8208;pro/learn&amp;#8208;and&amp;#8208;support/analyze/how&amp;#8208;do&amp;#8208;we&amp;#8208;classify&amp;#8208;eye&amp;#8208;movements/determining&amp;#8208;the&amp;#8208;tobii&amp;#8208;pro&amp;#8208;i&amp;#8208;vt&amp;#8208;fixation&amp;#8208;filters&amp;#8208;default&amp;#8208;values.pdf</ulink></bibtext> </blist> <blist> <bibtext> Ulrich, F., Helms, N. H., Frandsen, U. P., &amp; Rafn, A. V. (2021). Learning effectiveness of 360 video: Experiences from a controlled experiment in healthcare education. Interactive Learning Environments, 29 (1), 98 – 111. https://doi.org/10.1080/10494820.2019.1579234</bibtext> </blist> <blist> <bibtext> Violante, M. G., Vezzetti, E., &amp; Piazzolla, P. (2019). Interactive virtual technologies in engineering education: Why not 360° videos? International Journal on Interactive Design and Manufacturing, 13 (2), 729 – 742. https://doi.org/10.1007/s12008‐019‐00553‐y</bibtext> </blist> <blist> <bibtext> Wallet, G., Sauzéon, H., Rodrigues, J., &amp; N'Kaoua, B. (2009). Transfer of spatial knowledge from a virtual environment to reality: Impact of route complexity and subject's strategy on the exploration mode. Journal of Virtual Reality and Broadcasting, 6 (4), urn:nbn:de:0009‐6‐17577. https://doi.org/10.20385/1860‐2037/6.2009.4</bibtext> </blist> <blist> <bibtext> Yarbus, A. L. (1967). Eye movements during perception of complex objects. In A. L. Yarbus (Ed.), Eye movements and vision (pp. 171 – 211). Springer. https://doi.org/10.1007/978‐1‐4899‐5379‐7</bibtext> </blist> </ref> <aug> <p>By Felix Hekele; Jan Spilski; Simon Bender and Thomas Lachmann</p> <p>Reported by Author; Author; Author; Author</p> </aug>
Header	DbId: eric DbLabel: ERIC An: EJ1326762 AccessLevel: 3 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0
IllustrationInfo
Items	– Name: Title Label: Title Group: Ti Data: Remote Vocational Learning Opportunities--A Comparative Eye-Tracking Investigation of Educational 2D Videos versus 360° Videos for Car Mechanics – Name: Language Label: Language Group: Lang Data: English – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Hekele%2C+Felix%22">Hekele, Felix</searchLink><br /><searchLink fieldCode="AR" term="%22Spilski%2C+Jan%22">Spilski, Jan</searchLink><br /><searchLink fieldCode="AR" term="%22Bender%2C+Simon%22">Bender, Simon</searchLink><br /><searchLink fieldCode="AR" term="%22Lachmann%2C+Thomas%22">Lachmann, Thomas</searchLink> – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="SO" term="%22British+Journal+of+Educational+Technology%22"><i>British Journal of Educational Technology</i></searchLink>. Mar 2022 53(2):248-268. – Name: Avail Label: Availability Group: Avail Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us – Name: PeerReviewed Label: Peer Reviewed Group: SrcInfo Data: Y – Name: Pages Label: Page Count Group: Src Data: 21 – Name: DatePubCY Label: Publication Date Group: Date Data: 2022 – Name: TypeDocument Label: Document Type Group: TypDoc Data: Journal Articles<br />Reports - Research – Name: Subject Label: Descriptors Group: Su Data: <searchLink fieldCode="DE" term="%22Vocational+Education%22">Vocational Education</searchLink><br /><searchLink fieldCode="DE" term="%22Distance+Education%22">Distance Education</searchLink><br /><searchLink fieldCode="DE" term="%22Video+Technology%22">Video Technology</searchLink><br /><searchLink fieldCode="DE" term="%22Eye+Movements%22">Eye Movements</searchLink><br /><searchLink fieldCode="DE" term="%22Auto+Mechanics%22">Auto Mechanics</searchLink><br /><searchLink fieldCode="DE" term="%22Technology+Uses+in+Education%22">Technology Uses in Education</searchLink><br /><searchLink fieldCode="DE" term="%22Learning+Modalities%22">Learning Modalities</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+Simulation%22">Computer Simulation</searchLink><br /><searchLink fieldCode="DE" term="%22Instructional+Effectiveness%22">Instructional Effectiveness</searchLink> – Name: DOI Label: DOI Group: ID Data: 10.1111/bjet.13162 – Name: ISSN Label: ISSN Group: ISSN Data: 0007-1013 – Name: Abstract Label: Abstract Group: Ab Data: This study utilises a novel approach to investigate the effectiveness of different learning modalities by combining video-based learning with eye-tracking. An excerpt taken from a vocational education instruction for car mechanics was videotaped using two different cameras: a standard 2D video camera and a professional 360° camera. The video recorded with the 2D camera was presented on a tablet, with a fixed angle, whereas the video recorded with the 360° camera was presented as non-interactive 3DoF virtual reality (nVR) environment using a head-mounted display. In both conditions, participants' fixation patterns were recorded and analysed in conjunction with a set of standardised questionnaires. Participants (N = 48) were randomly assigned to either the 2D-video group or the nVR group, with 23 participants in the 2D-video and 25 participants in the nVR group. The task of the participants in both groups was to watch the educational video while wearing an eye-tracker and then complete a standardised test on the presented content. The eye-tracking data indicated that participants in the nVR group showed longer total fixation durations on the instructor, but not other areas of interest, compared to the 2D video group. The standardised test indicated no differences in learning outcome between the groups. Implications from the current study as well as limitations and a outlook for further research will be discussed. – Name: AbstractInfo Label: Abstractor Group: Ab Data: As Provided – Name: DateEntry Label: Entry Date Group: Date Data: 2022 – Name: AN Label: Accession Number Group: ID Data: EJ1326762
PLink	https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1326762
RecordInfo	BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1111/bjet.13162 Languages: – Text: English PhysicalDescription: Pagination: PageCount: 21 StartPage: 248 Subjects: – SubjectFull: Vocational Education Type: general – SubjectFull: Distance Education Type: general – SubjectFull: Video Technology Type: general – SubjectFull: Eye Movements Type: general – SubjectFull: Auto Mechanics Type: general – SubjectFull: Technology Uses in Education Type: general – SubjectFull: Learning Modalities Type: general – SubjectFull: Computer Simulation Type: general – SubjectFull: Instructional Effectiveness Type: general Titles: – TitleFull: Remote Vocational Learning Opportunities--A Comparative Eye-Tracking Investigation of Educational 2D Videos versus 360° Videos for Car Mechanics Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Hekele, Felix – PersonEntity: Name: NameFull: Spilski, Jan – PersonEntity: Name: NameFull: Bender, Simon – PersonEntity: Name: NameFull: Lachmann, Thomas IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 03 Type: published Y: 2022 Identifiers: – Type: issn-print Value: 0007-1013 Numbering: – Type: volume Value: 53 – Type: issue Value: 2 Titles: – TitleFull: British Journal of Educational Technology Type: main
ResultId	1