Over the last few months, I have referred in many articles to the use of eye-tracking as a technique to assess usability. Now it’s time to set things down and tell you everything from the very beginning. But to understand how Eye-tracking works, I first need to illustrate how our visual system works.

note: You may find a more detailed italian version here

Introduction to eye movements
We all live in a complex multisensorial environment which offers large amounts of information. Our brain comes into contact with this environment through sense organs. Of all these, sight is definitely the sense which allows us to interact the most with the surrounding environment. Through our eyes (and eye movements) we can obtain information about the spatial and material nature of the external world; we can make out the exact shape and identity of the elements within our view and register their relative positions.

It doesn’t take much to figure out we can’t see at a full 360° angle (at least not if we keep our heads still). In fact, we are limited to about 120°. Our visual field is divided into three regions: foveal, parafoveal and peripheral regions. This division is due to the anatomic structure of our eyes and responds to the different resolution with which the retina processes visual impulses. The retina’s sensitivity, indeed, is not homogeneous, its resolution being particularly high in a limited area called fovea. As the fovea is the only, small area which allows us to distinguish shape details, colours etc, we should always move our eyes to line the fovea up with what catches our eye. The dynamics of eye movements depends on a selection process based on visual attention. There are several eye movements but the most important ones are:

1) saccades: movements which place the content of the visual impulse in the area of the eye with the highest resolution, the fovea;

2) fixations: “pauses” on different elements (they include three types of small movements: nystagmi, drifts and microsaccades)

Saccade movements (the fastest movements the human body can make, whose length ranges according to the visual task from 20 to 50 milliseconds) and fixations alternate so that the eye can move from a particular point of an image or a visual scene to another, but be careful: during a saccadic movement we don’t get any new information because the visual sensitivity is reduced (saccadic suppression). It’s only during fixation that we get new information from the external world.

If we apply everything we’ve said until this moment to a monitor, we can conclude that:
Looking at a monitor from a “normal” distance of 50cm, our visual system is able to retrieve detailed information only in the region focused on the fovea. Then, thanks to eye movements, we start creating our “illusion” of space, quickly gathering so many different pieces of the mosaic, which are areas smaller than 100 pixels at a time.

Strategies for the visual exploration of the Web (according to Krug)
Now that we have established that nobody can see an entire page at the same time, but only a bit at a time, we can discover and explore a new world: the world of the exploration methods exercised by users.

Even when the Eye-tracking was an expensive lab device, it was evident that knowing what the user focuses his attention on could represent an important breakthrough. Our gaze (or eye movements) is strictly related to our attention system; the processing of what we see and what is transmitted to our brain is an integral part of the cognitive process.

It’s clear that knowing exactly where users focus their attention is an extraordinary way to check the effectiveness of an interface or to make planning choices. An “easy” experiment, for example, can show how some planning solutions do not respond to the exploration strategies adopted by the users.

The greatest Gurus of usability have known these things for ages. Even good old JN – who has always avoided mentioning eye-tracking – has experienced it…the only problem he has now is that with the “certain proof” eye-tracking can offer, I don’t know how many people will still be willing to pay his expensive fees of heuristics/witchcraft.

But JN is not the only one who knows and remains silent. Bruce “TOG” Tognazzini even registered a patent in which eye-tracking was used as an “input device” but he never mentioned it in relation to the assessment of websites.
Well, the truth is that up to 18 months ago, eye-tracking was seen as a helmet put on users’ heads, with two cameras on it, and probably, only a few people (even among Gurus) know that now things have completely changed.

To the left you can see how the designer expects the user to “explore” his website. To the right you can see what happens in reality, according to the nature of our visual system. Of course, if the user really wants to read the rest, his eyes have to move in a much more regular way but the point is deciding to read. Before that, there’s only the phase of the visual exploration.

What to measure and with how many users (experimental setup)
With Eye-tracking, the experimental setup used to validate, for example, the interface of a website is not so different from the one used with other methodologies. First of all, it is important to define what to test and what questions the client wants to find an answer to. Working on these aspects, you can get a first idea of a scenario and a first list of tasks you want to test, etc.

At this point, you can determine (together with the client) the ideal composition of the group of sample users:
- if the test is a validation test of an existing GUI, the task is not too fragmented and the target does not have more than two subgroups, 10/12 people can be enough.
- if the test is still under design, 4/6 people are enough, but in this case the test has to be repeated at least twice.

Let’s now focus on the real test, which consists of two phases: the MEASUREMENT and the ASSESSMENT phases.

It’s the client who decides – partly, at least – what to measure (for example, I want to understand if my purchase procedure works) but in many cases this is not enough. The experimenter should be supported by an experienced designer; they should work together to observe all the elements which interact with what they want to measure. During this phase, the experimenter and the designer decide how to proceed with the test: they decide whether it is possible to assign simple observation tasks to the users or if it’s necessary to draw up a list of objectives, tasks, suggestions, similar methods of registration etc.

Regardless of how it is organized, the measurement phase determines the clear superiority of Eye-tracking over other methods:
- it raises 30/100 samples per second of the position of the user’s eye;
- it is able to distinguish the saccade movement from the fixations
- it calculates the length of every single fixation

But it also allows the experimenter
- to review the entire FOOTAGE of the eye-movements of the subject (even step-by-step);
- to create graphic outputs in “false colours” which underline the way the attention of the single subject (or aggregates) was distributed.

All this happens without helmets, visible cameras, experimenters in lab coats, etc. The most modern Eye-trackers (i.e. the machines that measure eye-movements) are no different from normal LCD monitors. The advantage is huge because, for the first time, it is possible to measure the usability of a GUI directly from a video, without interfering with the user’s system. Not only are modern eye-trackers the most objective existing tests: they are also the most ecological form of measurement/test.

A team for Eye-tracking
Faced with this incredible resolving technology, the analysis team needs to move a step forward. On the one hand, the fact that you can see how the user has moved through the page gives many certainties. On the other hand, the fact that you can create experimental setups which lead to concretely useful and objectively “true” results is possible only if the team is made up of people who have a solid background in neurosciences, psychology and cognitive processes.

The ideal team is therefore made up of an experimenter who has experience in tests and has a background in psychology (or at least who knows a lot about the visual system and the cognitive process) and of a surveyor who is a GUI expert.
With such a team it’s easier to give the client clear information and to effectively redesign the site, if necessary.

NOTE: Personally, I would advise against assigning the validation of a GUI to a neuroscientist or psychologist who has never designed an interface, let alone to a lab technician, however good he may be. Eminent experts à la Ralph Molich (the man who used to say “I don’t make projects, I take measurements”) seem anxious to pass off as half-surveyors half-sorcerers, but to me they only look out of time and, worst of all, culpably alien to the client’s business and to his way of communicating.

Comparison with other validation methods
Call it a slogan of sorts, but “when it comes to usability, I trust what the users see”. On the one hand, this means that if tomorrow I can’t see the footage of all the visual/cognitive process, I will miss the certainties I have today. On the other hand, as Eye-tracking allows to point out objectively the process which leads to the mistake, I could assert my experience as a designer. In other words, ET is not a machine which wants to replace the expert; on the contrary, it is the first and useful instrument with which the antiguru wants to fight the stalwart guru’s witchery.

Even if Italy has more usability experts than well-designed websites, usability-evaluation methodology and technologies deserve more in-depth analysis.

Look at the diagram and ask yourself: do I prefer a method that allows me to reach goals where usability measurement and evaluation are both implemented using subjective criteria? Or do I prefer a method which tends to evaluate objectively both these dimensions? Well, the answer is clear and even if we can’t always affirm that “objective” is better than “subjective”, we have reason to believe this is often the case.
WARNING: this DOES NOT MEAN that tomorrow we’ll do nothing more than “record” eye-movements and toss the figures to the marketing gurus. It would be like giving the results of a CAT scan to a plumber (a non-random example of a barely-educated, filthy-rich man who is often caught red-handed swindling his customer).

ET needs to improve a bit in terms of lab-practice standardisation and to acquire that formal rigidity that until yesterday allowed non-objective methodologies to contribute to design and usability. As Eye-tracking takes on, all this will become natural and at that point Eye-tracking may well be considered a method potentially more effective than those used in the past.

There’s nothing more exciting than putting yourself in the shoes (and in the eyes) of your user. Therefore, I wish ET could be used not only in firms and labs able to afford an Eye-tracker but also (maybe with the contribution of outsourcing services) in those smaller companies, universities and firms that, given the current conditions of the Web, are anxious to turn over a new leaf. In the meanwhile, its price will go down – as is always the case with technology – and at that point I really want to see who still wants to go on designing “with his eyes closed”.