As recounted by Vortmann et al. (2021) it was observations of eye movements during reading that spurred the “scientists’ fascination for human eye gaze behavior,” who began the studies of eye movements already in the 19th century. Yet, unlike then, thanks to modern eye tracking systems and eye tracking technology we can now gather and analyse a variety of data relating to eye movements. This data includes differences in eye movements that we are often quite unaware of ourselves, such as gaze point and patterns, or eye fixations and saccades. And yet, if analysed correctly and meticulously, it can provide us with important knowledge, becoming if not the window to the soul, then certainly the window to what piques our interests – and more.
Now, you might think that it all sounds very complex. It is certainly compelling to think that our eye movements can be measured and used to gain new insights into understanding visual attention, and even cognitive processes. But how exactly is it done, and what for? Let us investigate.
Eye trackers – what is needed to use them?
Contrary to what one might imagine, from the perspective of the user gathering data on eye movements requires neither the extraordinary perception skills of Sherlock Holmes nor the elaborate equipment found in futuristic hard sci-fi. All you need are the right tools – and the knowledge on where to find them. But do not worry – eye tracking data recording does not require a set-up from A Clockwork Orange.
Although there are various technologies and eye tracking devices, one of the easiest and most innovative solutions is webcam-based eye tracking. How does it work? It is simple! In terms of hardware and software requirements, all that is needed to collect eye tracking data within the RealEye system is a PC/ Laptop/ Mac with a webcam and Chrome/ Edge/ Firefox browser. No additional eye tracking device is needed, not even eye tracking glasses – in terms of how data is acquired, screen-based eye tracking is enough. You can just sit back and create a study of your choice. And hey – you can even run the study on your smartphone! With RealEye Dashboard setting it up is a piece of cake. Just remember that for achieving the highest eye tracker data accuracy it is important for the participants of the study to take heed of proper calibration! Head and eye position is important for the accuracy of eye tracking measurements.
Eye tracking metrics – how is the gaze measured?
Human factors in computing are important – and thanks to advanced eye tracking solutions data collection is easy from the perspective of the user. And yet, it does not mean that the process behind it is that simple itself. As already mentioned, the use of eye tracking can provide us with a number of different data on specific eye movements that can then be analysed and interpreted. And yet, as we continuously learn, the human gaze is more complex than it may seem. After all, as a philosophical concept, it was even one of the key components of Michel Foucault’s The Birth of the Clinic.
But lest we stray too far from eye tracking research – what elements of eye gaze behaviour are important for eye trackers? One of the basic terms you should know is gaze point or gaze position – a specific location on a visual source where a person’s gaze is focused. That one was not too difficult – so let us delve deeper. According to eye movement research conducted by Santhoshikka et al. (2021) and Thomas (2023) fixations and saccades, eye blinks, and pupil dilation are amongst some of the most commonly collected eye tracking measurements. Out of these, perhaps the most crucial are fixations – series of gaze points – and saccades – movements of a person’s eyes between fixation points. Another important factor relates to pupillometric data – measurements of changes in pupil size dependent on various stimuli.
This gaze data combined, especially if the eye tracking accuracy is high, can help us learn more about a person’s reactions and responses to different visual stimuli.
Data Collection and Modelling
Now that we know how many factors eye movements consist of, let us consider possible ways of extracting data so that it can be analysed and interpreted. Raw data is not enough. As rightly noted by Duchowski (2007, p. 136) “Although intuitively (…), it is possible to guess where the subject happened to be paying attention in the environment (…), it is not possible to make any further quantitative inferences about the eye movement data without further analysis.” And certainly not qualitative, one might add! Let us then consider different frameworks and data sets and thus delve deeper into what analytics consist of.
Mathematical frameworks: Mathematical frameworks are some of the most crucial tools in identifying gaze patterns. Eye tracking data can be represented i.e. with the help of cartesian coordinate system – especially useful in analysing two-dimensional visual stimuli, such as in the case of screen-based tests – and polar coordinate system – with regard to circular displays or radial layouts.
Data sets: This is also where data sets come into play. Patterns of eye movements may vary depending on the circumstances. This is why it is only prudent to also distinguish various types of methods in which data can be acquired. One of the most common is pairs data set – often used in statistical analysis to compare responses to two different stimuli. Two of the data sets that are perhaps most useful for analysing eye tracking data are the alignment-task data set and the pairs-task data set. The former is based on the results of a task of aligning certain elements of visual stimuli by a participant of a given study; the latter is something of a combined data set, in the sense that it combines elements of both previous data sets – aligning given features of pairs of, for example, images. But do not feel confined to these two data sets – there are plenty of options to explore!
Presented data sets and frameworks can then be valuable in visualising the results of a given study and learning more about peoples’ cognitive reactions to and perception of visual information. What should be duly noted, however, is that it is always prudent to have access to both raw data and data to which preprocessing was applied. As noted by Madeira dos Santos (2021, p. 19) data preprocessing and feature extraction is “the phase in which raw data is processed and features are extracted from it. These features serve as input to the model (…).” It is important to remember that eye tracking can also – and certainly shall – evolve with time. Therefore, new perspectives in data modelling and feature extraction will likely emerge, and the data so far gathered may well be also applied and analysed in new ways.
Visual analytics
Once a sufficient amount of data on gaze behaviour is collected – the real fun begins. There are various ways of presenting the results of a study and its analysis. Arguably one of the most useful and user-friendly ones is visual analytics. Burch (2021, p. 75) notes that “the combination of computers with their analytical power and humans with their perceptual abilities and strengths make visual analytics a powerful tool (…).”
Heat maps: One of the most suggestive visualisation techniques is heat maps. And yes, here you have likely guessed correctly – they are extremely useful tools for visual depiction of gaze points representing the results of participants whose data was collected. To put it more simply: heat maps show you which part of a given image caught the attention of a user, as well as when and for how long. And they do so in a very easy-to-understand manner – do you remember the good old “hot or cold” game? It works similarly to heat maps – warmer colours represent the collected data on the areas of an image that the user focused on the most, and colder colours represent those that gained less attention.
Areas of Interest: This kind of visual analysis of eye tracking data is also useful for identifying so-called AOIs, Areas of Interest – they are a highly effective tool for calculating gaze patterns with regard to a particular area of an item of your choosing. The list of metrics that can be extracted is long and includes average time to first fixation, time spent looking at an area, fixation duration and number, revisits, number of gazes, covered gaze distance, and many more. Feel free to learn more in this handy glossary!
Additional Benefits: What can be especially useful if you wish to conduct a larger study with a diverse group of participants – that is to gather a person-independent data set – or to simply explore more detailed person-dependent data – for example through replaying gaze and fixation plots for individuals – is the possibility to also extract both the analyses of the data gathered on individual participants of a study or particularly selected subgroups. This kind of narrowing down and testing data subset allows for analysing particular results in a way that can be in many respects significantly better than all statistical summary – not taking anything away from its general usefulness, of course.
All of these analytics can be simply accessed through RealEye Dashboard – user-friendly solutions like this one can feature engineering for eye tracking of the highest value, as well as state-of-art analysis software.
Application Fields and Further Research
As one can well imagine, information learned thanks to eye trackers also has a number of application fields. According to Burch (2021, p. 192-196), these can include human-computer interaction; physical disablement; visual analytics; medicine; marketing or product and design or web usability; immersive analysis; gaming, entertainment, sports; education; car driving or automotive; and aviation or military sectors. The possibilities seem to be endless – and they likely are.
A review of eye-tracking research clearly shows that it is a rapidly growing field. Especially the proceedings of the SIGCHI conferences, devoted to exploring human-computer interaction, and the outcomes of the ACM Symposium on Eye Tracking Research & Applications show how much interest there is in this subject, how far it has come, and how much there is still to learn. Also the research conducted by Cognitive Systems Lab is worth keeping an attentive eye on – such as a recent original research by Vortmann et al. (2021), providing new insights into the possibilities of optimising the accuracy of attentional state classification based on eye tracking data.
But there are plenty more fascinating new directions to follow, including such interdisciplinary approaches as combining neuroimaging data with the information provided by eye trackers within the field of psychology (Bendall et al., 2019), and even delving deeper into the field of translation studies (Walker, 2021). And if you wish to learn more about the data on which a given research is based – do remember to always check the data availability statement, as especially within academia such data is often accessible upon request. And there is always something new to learn – isn’t it exciting?
Bibliography
Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems. (2023). Association for Computing Machinery, New York, NY, USA. https://dl.acm.org/doi/proceedings/10.1145/3544548
Bendall, R. C.A., S. Lambert, A. Galpin, L. P. Marrow, S. Cassidy. (2019). “Psychophysiological indices of cognitive style: A triangulated study incorporating neuroimaging, eye-tracking, psychometric and behavioral measures.” Personality and Individual Differences (144). https://www.sciencedirect.com/science/article/pii/S0191886919301436
Burch, M. (2021). Eye Tracking and Visual Analytics. River Publishers.
Duchowski, A.T. (2007). Eye Tracking Methodology. Theory and Practice. Springer.
Madeira dos Santos, P. D. (2021). Using eye-tracking data to study models of attention and decision-making. [Master’s thesis, NOVA University Lisbon].
Santhoshikka, R., R. Laranya and C. Harshavarthini. (2021). “Eye Tracking and Its Applications.” IARJSET (8). https://iarjset.com/wp-content/uploads/2021/08/IARJSET.2021.8824.pdf
Thomas, J. (2023). “Using eye tracking to validate cognitive processes in high stakes assessments.” [NCME conference paper published on ResearchGate].
Vortmann, L-M., J. Knychalla, S. Annerer-Walcher, M. Benedek and F. Putze. (2021). “Imaging Time Series of Eye Tracking Data to Classify Attentional States.” Frontiers in Neuroscience (15). https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2021.664490/full
Walker, C. (2021). Eye-tracking study of equivalent effect in translation: the reader experience of literary style. Palgrave Macmillan.
https://www.realeye.io/features/online-analysis-dashboard
https://www.realeye.io/features/online-webcam-eyetracking
https://www.realeye.io/features/online-webcam-eyetracking/eye-tracking-devices-range-of-technology
.png)
