Data visualizations are designed to help observers make judgements about relationships and patterns in collections of data. They are supposed to help audiences understand and reason about data. When they confuse or lead audiences to make erroneous judgements, they are failures or, worse, deceptions.
As is so often the case with new technology, enthusiasm accompanied popularity.This enthusiasm spawned many beautiful and influential examples of data visualization. It also spawned exuberant failures, confusions, and abuses.
George Cram
In George Cram’s wonderfully-named Unrivaled Family Atlas of the World we see some examples of those early exuberant failures.
This bar chart is centre-aligned, curved to fit on the page and coloured uses only three colours in repetition. The lengths of arcs are notoriously difficult for the human eye to compare. Psychologists had just started describing and cataloging such visual challenges and illusions in the late 1800s in their efforts to study human perception and cognition.
At first glance this might appear to be an Isotype. It is not. It is just an area graph. Each icon represents a country, with the size of the icons corresponding to the relative value of the relevant variable. Comparing the area of one pink cow to another is extremely difficult, beyond the simple observation that one is bigger than another. How much bigger is impossible to say.
The Canadian edition of Cram’s Atlas was published in Toronto and contained several charts specific to Canada. This example uses squares of different sizes (again in the same three-colour theme) to depict the relative sizes of various religious denominations as a proportion of the populations of four Canadian Provinces. The boxes are hard to compare. Worse yet, the areas of the boxes do not accurately scale with the data values they are supposed to represent. The largest box on the page is 3,500 times bigger than the smallest box but it represents a population 59,000 times as large.
These are confusing area charts with odd, irregularly subdivided regions. The only way to know that Austria, for example, had greater military expenditures in 1866 than Italy is to look at the numbers, which defeats the very purpose of a data visualization. As with other examples from Cram’s Atlas, the designers prized form over function. The shapes of the data visualizations allowed a greater number to appear on the page, each in interesting or attractive patterns. But consequently, the content was harder to interpret. In this respect, Cram was decades ahead of his time: designers today often appear more interested in shiny interfaces and fancy graphics than they are in easy-to-interpret data visualizations.
This chart was created by Francis Bowen for his Harvard University economics textbook. The horizontal axis shows months, and the vertical axis shows the price of gold. Unfortunately, this chart is arranged to show how the month, not the year, affects gold prices, and that does not seem to have any pattern. Because of the design choice to make month of the year the main variable, basic and more interesting questions like, ‘which year had the highest or lowest gold prices?’ or ‘did gold prices generally go up from year to year?’ are not easily answered by looking at this confusing chart.
This pamphlet by William Mallock feels like a modern infographic and not in a good way. Its bold language, matched by giant red uppercase letters stating that one view is ‘The Popular Fallacy’ while the other is ‘The Real State of Affairs’, grab our attention. While an explanation of the data and the interpretations of the Wage Theory of Henry George (1839–1897) accompanies the chart, our focus is dominated by the chart’s glitz, and that overrides the value of the data. This is also a problem in the infographic-style visualizations in this exhibition (by Schmeall and Adams, for example) as well as so many modern infographics that follow in the footsteps of this shiny but uninformative style.
In this revamped 1939 version of the first textbook on data visualization for a general audience, Willard Brinton gave technical advice on how to create visualizations and explained the principles behind their construction (see also item 17). The general quality of this book makes it all the more disappointing to find that it includes a dual axis graph. In a dual axis graph, two or more variables are plotted on the same graph, each measured against a different vertical axis. Here the amount of electrical power used each year, the average wage of factory workers per hour, and the number average number of hours worked per week are all plotted on one graph. It is intended to show that as electrical power use increases, wages increase, and hours per week decrease, though the graph, however beguiling it may seem, cannot support such conclusions.
