As we have seen, our eyes can only perceive a very small area with a high resolution. The saccadic movements, however, let us inspect different portions of space. Our brain then reconstructs the information, making us confident that we clearly see a larger amount of space.

The sequence of inspected areas is called visual queries, and in this post we will have a deeper discussion on them.

A simplified information flow model

As explained in Colin Ware’s textbook, the neural activity involved in our sight process is very involved, but for our purposes we can build a simplified model, which will help us in designing visualizations. In this model we have two information flows, a bottom-up flow and a top-down one. In the bottom-up flow the information provided by our eyes is collected, filtered and elaborated. The elaboration goes through many steps. In the first step, the retinal signal is converted into features like color, size and shape. At each subsequent step we construct pattern of increasing complexity. Some of the most relevant patterns emerging in this phase will be discussed in the post about Gestalt psychology. These patterns are then stored into our visual working memory.

The visual working memory is considered the main bottleneck of our attention, as it can only store few objects at each time and only for a very small amount of time, typically few seconds or less. An experimental evidence of this fact can be found in the so-called “Door study”, by Simons and Levin, which can be seen here. The study shows that, when we talk to a stranger, we are not even able to recognize his/her own face, and if our speaker is exchanged we often don’t even notice it (unless of course the original speaker and the new one are too dissimilar). For this reason, it is generally recommended not to use interaction whenever possible, as interacting takes time, and during this time some important information may get lost, so we may miss some important pattern 1.

In the top-down flow we direct our attention, so our eyes, depending on our available information and expectations, as well as depending on our task. At the lower level, our attention makes us focus on the elementary signals we are looking for: if we are looking for our car, and our car is red, our attention will let us focus on red objects.


Our model for vision.

I would like to stress you with the fact that signals considered as interesting are stronger have a very important consequence:

Our perception is strongly biased by what we expect.

Other aspects in visual perception

There is also another important aspect that enters into the game, which is our culture, and this fact is well known by user experience designers. As an example, our visual queries are generally influenced by our writing system: in most western cultures we write from left to right and from top to bottom, and this makes us start looking for new information at the top left of a web page, so UI designers usually put there the most important informations. On the other hand, in other Countries like the ones in the Middle East, the writing goes from right to left. Due to this, companies like Google or Microsoft usually reverse the user interface on the left-right plane for languages like arabic, and they put the most relevant information on the right.

Also experience plays a major role in what we perceive: if you are new to a specific task, this will require you more attention, so it will be more likely for you to ignore other information source.

What to do, what to avoid

Since the exact structure of the query is subjective and depends on the task, there is no recipe to determine how one will perform visual queries. There are however few questions that you can ask yourself when making a visualization:

  • What task should the visualization accomplish?
  • Who will be the audience?
  • What could be a reasonable way to perform this task from a cognitive point of view?
  • Is there any visualization which could make this task easier, from a cognitive point of view?

Moreover, you should avoid adding distracting elements to your visualizations.

  • Any decoration may distract your user, so keep your visualizations as clean as possible.
  • Avoid fancy fonts, they are designed to be visually attractive.
  • Avoid useless images, especially the ones depicting human shapes, as they strongly attract our attention.
  • Make less relevant information less visible, use soft colors for them or shades of gray, reduce the width of less important lines like axis, ticks or grids.

Conclusions

We discussed some model of visual perception, and we have seen that our capacity to store information is the main bottleneck in what we see. What we see is determined by many factors, and the most relevant is attention. We finally gave some suggestion to improve our data visualizations by carefully driving the reader’s attention.

Suggested readings

  1. As a fun fact, this memory bottleneck is used in many magic tricks, as the one here. ↩