ISOTYPE Visualization

Working Memory, Performance, and Engagement

Collaborators

Steve Haroz - Psychology, Northwestern University

Robert Kosara - Tableau Research

Steven Franconeri - Psychology, Northwestern University

PDF

ISOTYPE Visualization PDF
An ISOTYPE visualization made by Otto and Marie Neurath. (via brainpickings.org)

The Question

Imagine that you are tasked with conveying data about industrial production rates over time to a lay audience. It needs to be quickly understood and easily remembered. You could use a simple standard bar chart. But another option is to transform the abstract numbers into concrete objects, depicting increasing car production as an increasingly taller stack of car icons.

This technique of stacking icons to represent data is called ISOTYPE visualization, and it has been around for a over 100 years.

Some visualization purists might bristle at such ‘chart junk’, seeking the minimalism of unadorned data. We studied whether the richer depictions of ISOTYPE visualizations actually carry benefits to a viewer.

We conducted five experiments on over 132 participants pitting ISOTYPE visualizations against more standard unembellished bar graphs. Here is a summary of our recommendations and the rationale behind them.

1. Stacked icons allow you to encode data as both length and quantity

The human visual system relies on different approaches for judgments of length (like the height of a tree) and quantity (such as counting the fruit on the branches).

Chart-making tip: Splitting a continuous bar into a stack of items allows the visual system to code both length (the height of the stack) and quantity (the number in the stack). We found that this stacking benefits immediate recall for the data in a briefly presented graph. Because the visual system is incredibly efficient at quantifying groups of 1-5 objects, we recommend splitting stacks into no more than 4-5 segments – we found no additional benefit for splitting for larger collections.

Caveat: This benefit only held for short-term recall of data, not long-term recall, and there was no evidence that it sped up data extraction. Critically, we found no evidence for impairments due to splitting – at the very least, it doesn't hurt.

Stretched

Stacked

2. Icons aid memory and are more engaging

Long-term memory is aided by additional information that ‘hooks’ to what you want to remember – you’ll better recall what you had for breakfast yesterday after someone reminds you of the name of the restaurant. Using real visual objects to encode data seems to lead to a mild improvement in remembering that data – you could better recall that daily production was 5 cars if you can recall that there was a stack of 5 cars.

Chart-making tip: Using pictographic icons can help with long-term memory of information.

Caveat: The benefit to long-term memory is small, and there's no benefit to speed or working memory.

Chart-making tip: We also found that ISOTYPE style charts were more engaging – when given a choice to view graphs of that style vs. standard bar graphs, our viewers had a clear initial preference for ISOTYPE.

Stretched

Stacked

3. Images that are not used to depict data are a distraction

We added images to the background and labels, which did not depict data. Short-term memory retention for the data dropped by 40%, and the speed of finding information dropped dramatically.

Chart-making tip: Images can be used to depict data, but background images or pictographic labels could merely distract.

Stretched

Stretched

Stacked

Try the different chart settings:

Start by clicking "Refresh"

Graph Type:





Color:



Category

List pictographs

Max Value: 5