Over on EagerEyes, Robert Kosara recently asked, what happened to ISOTYPE? It’s a good question, with a multi-faceted answer. Here’s a few facets, mostly focused on inherent limitations of the Isotype design approach:
1. What’s an Isotype line chart?
Isotype examples are typically bar charts and maps, as they show counts of things – i.e. what we call unit visualizations. Extending it to other types of visualizations is non-trivial. Certainly scatterplots shouldn’t be too hard (e.g. fruit, animals), but what about line charts. Certainly timeseries data is important to plot for many analyses but what’s the Isotype answer for line charts? Typically Isotype reduces the timeseries down to a few time periods and draws them as their typical stacked bar. As such, Isotype is an approach to creating charts, but Isotype is not a comprehensive system for all types of visualizations.
Some have tried, for example, see the Agricultural Outlook Charts from the USDA in the 1950’s. Some of the bar charts in these publications are heavily influenced by Isotype, such as the coins in the Figure 1 left. However the line chart in Figure 1 right struggles with icons, instead the icons are limited to identifying the line, and indicating the trend with the horse representationally and quantitatively heading down-hill.
Figure 2 shows another 1950’s publication heavily influenced by Isotype: Midcentury White House Conference on Children and Youth, A Chart Book. On the left is a bar chart that could almost have been lifted straight out of Isotype, wonderfully clear. On the right is a pie-chart infused with Isotype icons, where only by luck the thin wedges of the pie fit it smaller icons assigned to them (and strictly speaking, the pie segment with “both parents” should have repetition of that icon through out the area, but that wouldn’t quite work either.
2. What’s the icon for GDP or CPI?
Icons for concrete real-world objects can be easy to design, as the real-world object can be the basis for the icon, e.g. people, fruit, animals, tractors, and so on. It gets trickier when some of those categories are visually similar: Isotype never created separate icons for wheat, barley and rye, for example. And Tufte’s log animal chart has three rather similar looking small furry mammals, which can be only be definitively decoded by referring to the labelled scatterplot on the previous page (exercise for the reader to now go find the previous page:-)
After that, icons get difficult to design for abstract objects. What would the icons be for financial data asset classes, such as: stock, bond, credit default swap, collateralized debt obligation, repo, option, future and a forward? Even concrete real-world entities can be misinterpreted, such as the famous dogcow.
Making simple, expressive icons requires design effort. Gerd Arntz‘s wonderfully expressive icons helped drive the success of Isotype, but are beyond the design capabilities of the average non-designer. Gerd could create an icon for the abstract concept of unemployment with a human icon, looking down, at rest, hands in pocket: it’s brilliant, but not easy to design especially with such clean, clear graphical shapes that can be easily printed.
3. And what about the axes (and the values)?
Perhaps most audacious move of Isotype is the removal of the numeric axes. Isotype charts are beautiful with their clean depiction of icon stacks and graphical cues. Removing the numeric axes is brilliant because you can easily and visually compare ratios of different stacks, e.g. one stack of icons is twice as long as another stack of icons.
But what if you want to know the values?
It’s a common task to want to know what number a stack of icons represents. Unfortunately, Isotype makes it hard. You have to first count the number of icons. Then you have to find the legend, where it tells you how many items that one icon represents. So, for example if I look at Eheschliessungen in Deutschland (Die Bunte Welt, 1928, page 42), I see that in 1919-1922 there are 8 marriage icons and each icon represents 400,000 marriages, so 8 x 400,000 = 3.2 million marriages. That’s math. That’s cognitive effort.
Furthermore, when the design system uses a small number of icons, it’s not very precise. Isotype tends to use full icons or nice fractions such as 1/4 or 1/2 of an icon. The prior stack of 8 icons could easily represent 3.1 million or 3.3 million.
If the chart had a numeric axis, you could just scan it and estimate the number directly – much easier. Or you could put the number directly in the chart. In Figure 3, the same marriage chart from Isotype is replicated with US data the Midcentury White House Conference on Children and Youth, with the addition of quantitative values at the end of the icon stack:
4. Good Isotype is hard
Often simple designs are the result of hard work. Simplicity takes effort. In The transformer: principles of making Isotype charts (Hyphen 2009), Marie Neurath’s first hand account describes the design task of transforming data into an Isotype representation (what we might now refer to as encoding). Marie explains a myriad of design decisions made in different charts to get the desired reading of the result. For example, coffins are replaced with tombstones to address the issue of relative size of adjacent icons and potential misinterpretation. Or, doubling with width of an adjacent bar so that relative portions can be perceived. And so on. These are non-obvious design solutions, arrived through a design process to achieve a good effect that may seem obvious in retrospect. (Unfortunately, image copyright status is uncertain).
Has Isotype really disappeared?
The prior four points are focused on Isotype’s limitations that make it hard for Isotype to extend more generally across data visualization. I don’t even address points such as modernism (which both Robert Kosara have both previously talked about) or technical changes (by the late 1950’s phototypesetting became the norm, but the technology tended to soften edges and fine detail, so crisp icons may have been difficult to reproduce- I discuss some of these other factors, which also limit the use to text, in my forthcoming book Visualizing with Text).
That said, I believe that unit visualizations have inherited the legacy of Isotype. And there are some fantastic unit visualizations:
- Photos: Instead of the effort of designing icons, why not use photographs of the entities of interest? A favourite is Münzkabinett’s interactive piles of coins (Gortana et al).
- Shapes: Less expressive than photos, simple shapes are a good choice for unit visualizations with hundreds of thousands of items, such as SandDance.
- Labels: I’m personally interested in the use of labels in unit visualizations, such as this example of the passengers on the Titanic. I’m not the first, there are more compelling examples such as Maya Lin’s Vietnam Memorial, which in turn was based on earlier lists of casualties.
- Physical units: Physical visualizations are well suited to using units. These include specifically designed physical unit visualizations using concrete scales, as well as examples in the real world where the units can be perceived individually or as part of a whole, such as the fields of WWI crosses.