Note that the work of Holmqvist et al. A taxonomy of eye movement measures that is oriented toward user experience and the intensity of user engagement was outlined by Bojko through measures of attractiveness. This taxonomy contains two main categories: measures of attraction and measures of performance. The first group has area noticeability measures, area interest measures, and emotional arousal measures.
Area noticeability measures, for example, includes the percentage of participants who fixated on an area of interest; the number of fixations prior to first fixation on an area of interest; and the time to first fixation on an area. The second group has mental workload, cognitive processing, target findability, and target recognisability measures. Furthermore, Bojko emphasises the importance of reporting the percentage of participant focus on areas of interest or screens when analysing data to preclude any possibilities of distortion.
For example, if a study has ten participants, there is a possibility that only one participant gazed at or fixated on an area of interest. In Tobii Studio, two types of data are collected through the software: numeric data and visualisation. Numeric data can be divided into three groups: fixation, visit, and click. Fixation data include: fixation duration, fixation count, and percentage fixated, among others. Examples for gaze measures are: visit duration and visit count. Click measures consist of time to first mouse click and mouse click count, among others.
All collected data are based on an area of interest defined by the researcher conducting the analysis. Tobii visualisations include gaze-plot, heat map, and cluster, which are all based on fixation and saccades data. The following classification Table 1 of the eye movement measures was used in the present study.
However, even though several studies have shown significant differences between the styles of search activities e. Furthermore, most existing studies relied on system-side performance measures e. A task scenario had been developed, in consultation with Dr Accomazzi and CfA staff members, by Dr Michael Kurtz, an esteemed astrophysicist who has been on the system development team for over two decades.
The first two tasks were self-directed, and participants were asked to search for their own topics and publications. The remaining three tasks were based on utilizing the visualisation tools author network, metrics, and paper network in the system. The scenario was pilot-tested by two users, and revised based on their feedback and that of librarians working in the Center for Astrophysics Library. Twenty usability test and eye-tracking sessions took place at Harvard Wolbach Library between April 18 and May 8, Each participant attended the research session individually. The sample size of the study was not as high as planned due to the limited availability of astrophysicists on site during the time of the research study, however, very rich and extensive eye movement data on thousands of interface objects was obtained.
In total, participants examined 6, individual interface objects across five tasks. Overall, the participants produced 81, fixations and 30, visits, averaging more than 4, fixations and 1, visits per person. All participants came in for an individual session.
Each session was led by a researcher acting as moderator for the session. One or two researchers observed. The session took place in the Collaboration Space an open space on the ground floor, away from the reference desk of the Center for Astrophysics Library. Each participant began by taking a minute cognitive test. Immediately following the cognitive style test, a pre-session interview was conducted during which the participant was asked about their past user experience with the ADS, the types of searches they conducted, and whether they considered themselves a visual or verbal person.
After the pre-session interview, after being calibrated by the eye-tracker, participants were asked to perform five tasks. The task performance made up the actual usability testing session.
As described earlier, the first two tasks focused on searching their own topics and their own publications, with Task 2 asking them to explore three visualisation tools: word cloud, metrics, and author network. Tasks 3 through 5 were common tasks for all participants, with Task 3 connecting the visualisation of search results to author network, Task 4 to metrics, and Task 5 to paper network.
Table 2 outlines the gist of the five tasks. See the Appendix for the full task scenario. Upon completing their tasks, participants did a brief retrospective task analysis session where they reviewed their eye movements for a randomly chosen task for three to five minutes and articulated their thoughts. Lastly, the researchers interviewed the participants to discuss their likelihood in using individual visualisation tools, what additional features they considered useful, and how they compared ADS 2. Thus, during one research session, the participant took the cognitive style test, moved on to the usability test which included a pre-session interview, a task scenario performance and a retrospective task analysis, and finished with a post-session interview.
Excluding the cognitive test time, each session time averaged 67 minutes, with task performance time averaging 35 minutes. Eye-tracking data were coded in Tobii by scenes and segments. On each scene, various areas of interests were specified. Tobii metrics then generated various counts and measures based on those areas. Given that our research question 4 specifically addresses the differences between cognitive styles and their eye behaviour in viewing various interface elements and visualisation objects, we grouped the areas of interest into four categories including ACT buttons, tabs, menus and links to perform an action , TXT textual objects , IMG image or visualisation objects and HELP help information.
- Cognitive Publications.
- Eye Tracking: The Complete Pocket Guide.
- Photorealism in Computer Graphics (Focus on Computer Graphics).
- Essays in the Metaphysics of Modality.
- Dissipative Structures and Weak Turbulence!
In addition to research question 4, this coding of visualisation objects served as the basis of our examination for research question 2 as well as research question 3. Note that in performing statistical analysis of group differences, non-parametric Mann-Whitney U tests were used, as the eye-tracking data were not necessarily normally distributed. The research study included twenty participants: three were female and seventeen male. Half of the participants were native English speakers. Research length ranged from five to forty years, averaging sixteen years.
Seventeen participants worked at Harvard Center for Astrophysics, and of the remaining three, one was a visiting scientist from Argentina, one from Yale, and one from Massachusetts Institute of Technology. The demographic characteristics of these participants reflect the study population of astrophysicists, according to Dr.
Prof. Dr. Reinhold Kliegl
Accomazzi, Dr. Kurtz, Christopher Erdmann Head librarian of the Center for Astrophysics Library , and other Center staff members personal communication, April 11, Some of the research projects which participants were working on include: impact of embedded B-star wind on the Ophiuchus cloud; analysis of the variability of C IV lines in TW Hya; and characterising shock waves in the solar corona, and their ability to accelerate protons to high energies. The average self-rating of familiarity and proficiency with the system was 5.
Smith et al. Another participant commented:. Most participants had used ADS to search for a specific paper as a known item, but some participants also conducted subject searches related to their topics. The most frequently mentioned search type is author search, followed by title search, keyword search, or subject search. Note that in our design of the tasks for the eye-tracking scenario, there were tasks for author search and subject search. The first two tasks were self-directed: participants were given a chance to search for their own topic and their own work using keyword and title search options.
We did not observe any problems with tasks that were unfamiliar or out of the ordinary to participants, and we believed that our tasks were the typical search tasks that participants performed when using the system for their scholarly work. Overall, imagers and analysers were popular groups among participants. In subsequent data analysis, comparisons were made between imagers and non-imagers non-imagers included both bimodal and verbalisers , analysers and non-analysers non-analysers included both intermediate and wholistic styles , analytic-imager and non-analytic-imagers.
Analytic-imager participants are participants who tested both as an imager and an analyser. Non-analytic-imager participants included six intermediate bimodal participants, two intermediate imagers, two analytic bimodal participants, one wholistic imager, and one intermediate verbaliser. The cognitive style groups used in statistical analysis for this study are also listed in this table.
Eye tracking research in cognitive psychology
Note that due to the small sample number in each category, the generalisability of the study results would be limited. This means that, in this participant group, the imagers tend to be younger and have less research experience. Even though there were no differences in terms of task time performance, participants who had used ADS 2. However, participants who were not familiar with ADS 2. It appears that participants without earlier experiences relied more on HELP than those who had such experiences.
Table 4 shows various differential patterns between the two groups based on a series of Mann-Whitney U tests. Note that the Mann-Whitney U test is a non-parametric test, which was used to analyse the eye-tracking data-sets, as eye movement patterns were not normally distributed.
Such impacts may manifest through two areas: 1 time-based performance measures, and 2 types of interface elements or visualisation objects. Time-based performance measures include such quantitative measures concerning time as overall task time, percent of task time, time to first click, time to first fixation, and time from first fixation to click. In terms of overall task performance, no significant differences were found between performance indicators such as completion difficulties, query revision frequency, or self-rated proficiency scores for a detailed report on task performance through usability measurements, see Tang, There were no differences between analysers and non-analysers in their percent of task time on types of interface objects.
Interestingly, even though analyser-imager participants were found to have spent a significantly shorter overall time on their tasks than non- analyser-imager participants, they spent a higher percentage of their task time on IMG objects and HELP objects. Meanwhile, the measures concerning time and activities are also reported here as they relate to performance and efficiency.
It took a significantly longer time for analysers to move from their first fixation to a mouse click when they viewed an IMG object. Analyser-imager participants were quicker to get their first fixation on the TXT or ACT objects than non-analyser-imager participants. Osaka University Researchers at Osaka University developed a quantitative method for identifying individuals with autism by analyzing temporo-spatial gaze patterns, which could help experts diagnose the issue earlier.
Products and services Tobii Pro offers eye tracking systems for psychology and neuroscience studies in a controlled research setting, such as a lab, as well as examining human behavior in real-world environments, like in an office or home. Publications Research paper database. Talk, A. Graded expression of source memory revealed by analysis of gaze direction. Gwizdka, J. Davis, R. Riedl, J. Randolph Eds. Cham: Springer International Publishing.
Related The Minds Eye: Cognitive and Applied Aspects of Eye Movement Research
Copyright 2019 - All Right Reserved