Pre-attentive attributes

Title design by Sakina Salem

When I was training to be a graphic designer, I had never thought that in only a couple of years I would be creating data visualisations for Code for Africa and Pesacheck. The job called for skills in Adobe CC but I had taken journalism and statistics as electives in college and maybe it was just meant to be. So now I approach data visualisation as a graphic designer. I think about readability, colour, font and whether my message is clear. While preparing for a training session, I found out that there exists a phrase for the choices we make in presenting our charts: pre-attentive attributes.

Pre-attentive attributes are contrasting characteristics given to a piece of data in a chart or visualisation so that our brain is drawn to it first before we pore over the rest of the information. Our brain tends to skim over things that look familiar so when something is markedly different, the brain perks up.

The brain has also been conditioned over the years to expect certain things. For example, if a faucet is labelled red, it has hot water. Pre-attentive attributes take advantage of this conditioning to ensure the brain processes the data quickly. This same conditioning can sometimes let us down when presented with information that goes against what our brains have come to expect. We shall see an example of such a chart and its implications in the “Orientation” section below.

Data can be overwhelming, repetitive and complex. Moreover, people have mere milliseconds to spend on data we want them to see. Pre-attentive attributes can help you make your point loud and clear.

There is a myriad of pre-attentive attributes one can use to improve the clarity of their visualisation. The 4 attributes we will discuss are some of the most commonly used in data visualisations.

• Length
• Hue and contrast
• Orientation
• Spatial grouping

Length

Length is an attribute that efficiently points out the most important piece of information. Bar charts and column charts are most frequently used because lengths are just so easy to compare. For example, in one glance, could you tell which colour has the biggest share in the pie chart below?

Fig 1: Can you tell which share is the largest?

Now consider this column chart. It is remarkable how much difference there is between the longest and second-longest bars and yet it was so difficult to see that same difference in a pie chart.

Fig 2: Length is a more effective attribute for comparisons

So does this mean that pie charts are bad? Stay tuned for an article on that.

Hue and contrast

For generations now, there are some things we understand just by its colour. Colour even transcends language. We now associate red to heat, falsehoods, stop, and anything that can lead to danger. Similarly, blue is almost always used to denote cold and green is used to denote truth. We see it so often in our daily lives, that if the colours on our faucets were switched today, we would be in serious danger of burns! We can use this to our advantage in data visualisation to highlight information that requires urgent attention.

Fig 3: Our brain makes associations with certain colours

However, not all information can be denoted by these colours. There are several occasions when chart colours need to follow a brand colour scheme. A good understanding of contrasting hues can still help create charts that highlight the most important information. To spotlight a particular piece of data on a bar chart, let all other bars be the same colour while your highlighted data is of a noticeably lighter shade (darker colours recede, brighter colours pop!).

Flourish.studio is one of the data visualisation platforms we use at Code for Africa. I made the highlighted bars in the chart below using “Custom Overrides” under “Chart Styles”. This chart highlights the fact that only two out of fourteen reported sources of water in Kenya’s Kitui county are available directly at home and a very small percentage of the population has access to it.

Fig 4: Use hue to differentiate important pieces of data

Note: When using colour, especially similar hues, we should be mindful of colour blindness and other accessibility issues. Online tools such as https://www.webaccessibility.com/ are available to check the accessibility of your links.

Orientation

Among all the chaos of life, a well-oriented chart is a sight for sore eyes. There are a few ways that orientation can be used to improve our visualisations. First, we must ask ourselves, does the data have to be in that particular order? For example, let’s look at Fig 5 which shows the population of different countries in alphabetical order. Now, if our goal was to show that the beginning letter of a country name does not affect their population, then we’ve succeeded.

Fig 5: A column chart in alphabetical order is not usually ideal — Source

However, chances are we want to show the country with the highest population and the countries that follow. So why not provide that information upfront? If the chart was organised by the highest population from left to right or better yet, from top to bottom (allowing the names to be horizontal) like in Fig 6, suddenly its a no-brainer which country is the most populated. You could use it in conjunction with hue and contrast to point out the position of a specific country to see where it stands.

Fig 6: Use an ordered bar chart and contrasting hues to make your point clear — Source

Note: Not all cultures and languages read from left to right. We should be mindful of our audience when creating charts so as not to mislead them.

Apart from providing information upfront, we must also consider what people have come to expect. How do you interpret the chart below when you first see it?

Fig 7: Line chart showing gun deaths in Florida — Source

For most people, the first thing they would take away from this chart is that the number of deaths decreased sharply after the enactment of the ‘stand your ground’ law which allows the use of deadly force in a potentially fatal event. However, the y-axis is upside down!

Our brain almost always denotes a dip in the line as getting closer to 0 and a rise as getting away from 0. We get this from nature, where sea level is at 0 and mountains are positive values above sea level. While all the information in the chart is true, we have to pay close attention to interpret it correctly. Constituents and legislators alike might make hasty policy decisions based on a chart they have perceived incorrectly.

While the designer had every right to choose this orientation, it was perhaps not the most responsible decision for this particular topic which has such far-reaching consequences.

Spatial grouping

Birds of a feather… flock together!

Objects that are close or similar to each other are perceived as belonging together. In data visualisation, this attribute is particularly helpful in maps to show concentrations in different areas and can be used particularly well with the hue attribute.

Fig 8: A visual using hue and spatial grouping to connect the cases to the region

Fig 8 shows the spread and impact of disease in Uganda. The data showed different districts affected each year and the number of reported cases and deaths. In this particular case, there was no overlapping region for each time period so I was able to show data in the bar charts belonging to different areas in Uganda. So the cases in the red bars occurred in the red regions and so on. Note the colour spectrum starting with red to denote danger and moving to yellow to show a decline. This shows that as time passed, the disease was contained in a small region and had the fewest cases.

Conclusion

In hindsight, all of these ideas seem intuitive. Yet, you’d be surprised how often we miss the opportunity to use them and create more effective and impactful visualisations. At the same time, we have seen the implication of using some of these attributes incorrectly.

In a time when people are confronted with so much information, we can help our audience get the information they need upfront and without obscurity. Pre-attentive attributes can be the difference between data lost in the sea of digital information and facts that incite change.