What measures the eye-tracker. Part two

In the previous post, we settled on the fact that the result of the work of the eye-tracker is to record the sequence of fixations of your view on the displayed images. Now we need to somehow convert this sequence into such a form so that it can be conveniently analyzed.

Each fixation has several characteristics: time (when they looked), coordinates (where they looked) and duration (for how long they looked). Well and, of course, for analysis it will be important to know exactly who looked and what was shown to him at that time.

Now we will think about how to display this information to us so that the analyst could draw some conclusions based on the data on viewing the materials.

')
The simplest and most informative is recording a session. We just make a video in which we show everything that happened during testing. That is, we impose on the picture that was shown to the respondent, the record of fixations of his gaze and that’s all. With this approach, the researcher in detail sees the whole process of testing. There are only two drawbacks to this method - the time it takes to view the materials is equal to the testing time and the need to review the testing progress of each respondent separately.


In order to shorten the viewing time, you can build a gaze plot. An image that was shown to the respondent is taken, and circles are marked on it, marking the fixation. The size of the circle symbolizes its duration, and in the circle itself we write its sequence number so that the sequence of fixations can be traced. This method is also very informative, but, unfortunately, is applicable only for short time intervals. Since the number of fixations grows very quickly (several pieces per second), the picture very quickly becomes completely unreadable even for one respondent, and if there are several of them, then something can only be understood at an interval of a couple of seconds ...

In order to get rid of the shortcomings of the two previous methods (long viewing and littering), you can build a “heat map”. At the same time, those places that were looked at more are painted in more “hot” colors, and where less - in “cold” (or, alternatively, a dark mask is applied to the image, which becomes more transparent in those places where they looked more). The main drawback of the “heat map” is that information about the sequence of fixations is lost — you cannot say where you looked at earlier and later.


Nevertheless, despite the loss of such important information as a sequence of views, it was precisely heat maps that became most prevalent due to their clarity and colorfulness. Well, in order to smooth out the loss of information about the order of viewing, you can break all the time into some segments and build separate maps for each such segment.

But now we come to one of the subtle points. I think that when reading the story about the heat map, you took for granted the phrase about the areas “which you looked at more”. But let's think: what does it mean to “look more”?

There are at least two parameters: the number of fixations in a given area and their duration. And the maps based on these parameters will differ somewhat (and sometimes quite significantly!).

There is one more parameter that is used when building heat maps for several respondents - the relative viewing time. Such a map shows what percentage of the viewing time each respondent spent on a particular area.

Let's imagine that we tested two respondents by showing them a picture with two areas. The first one looked at this picture for 10 seconds, and he had 10 fixations of 200 ms each recorded in the first area, and he never looked at the second one. Well, the second respondent looked at the picture for 20 seconds, and made 8 fixations of 500 ms in the second area, without ever looking at the first one. In this case, if we build a heat map according to the number of fixations, then the first area will be “hot” (10 fixations against 8); if the map is constructed according to the absolute viewing time, the second region will win (4 seconds against two); and if you build a map according to relative time, the areas will be of the same color - both respondents spent 20% of their time.

Example heat map for the same results.
By the number of fixations	In absolute time	By relative time

Therefore, when in some article or somewhere else you are shown beautiful heatmaps and they say that on the basis of such and such conclusions are made - it is worth thinking about what data on this map is shown and for what time interval. Just for reference - in my post about the difference in the perception of sites, I used maps for the number of views built over the entire time the page was viewed. But for other types of cards, the picture is similar ...

There are other, much more exotic ways of presenting information , but because of their exoticism, they are much less common.

Well, in the next post we will try to figure out what we can tell all these cards about and how to put them practically into our website, flyer or program ...