As part of the 2014 batch of HASTAC Scholars' Visualization study group, I'm happy to summarize some thoughts by thinkers over the past decades who have written and described heuristics for good visual communication. First up, let's talk about Jacques Bertin & Jock Mackinlay.
Jacques Bertin was formally trained in cartography, and as you can imagine this line of specialization made him uniquely sensitive to the best practices for mapping data. His writings, notably condensed in the book Semiology of Graphics, present a rigourous manual to understand and communicate a visual language. Specifically, he tried to assign visual and perceptual analogues to graphically represent the features and relations between datapoints. As such, he effectively laid down a foundation for a *grammar* for visual communication.
Bertin held graphic representation to be one of two things, or both: namely, a mnemonic device (artifical memory) or a tool for discovery, wherein the perceived similarities and differences could inspire new hypotheses not obvious from a plain dataset.
The basic elements of graphic representation were given the term marks, and their core semantic/semiotic value is that they represent information other than themselves. For example, a small red square is just that, until it is used as a mark to represent a data point on a graph. Marks are typically the following: points, lines, areas, surfaces and volumes. That is to say, these are representations of data using the human capacity for spatial awareness.
Bertin subsequently describes different ways in which marks can be manipulated or modified to represent the data in a clear, internally consistent manner. These allowed modifications are termed visual variables. The original visual variables, as defined by Bertin, are:
- Position (proximity)
- Size (scale)
- Shape (dfferent shapes, tesselating shapes)
- Values (interpreted as brightness)
- Colour (interpreted as saturation)
- Orientation (alignment)
- Texture (fill or grain)
In modern times, Jock Mackinlay recommends motion (animation) as an 8th visual variable, given the capacity for digital dynamic representation. Jock Mackinlay, an eminent thinker in the field of computer visualization, has also greatly extended on Bertin's original marks and visual variables with recommended heuristics for variable priority depending on the type of data.
So what do marks and visual variables have that allows for effective data representation and interpretation? The theoretical basis for choosing a variable are:
marks and manipulations should be easily distinguishable
marks and manipulations should be cognitively easy to categorize in groups.
marks and manipulations should suggest relative weight or influence.
marks and manipulations suggest a hierarchy or intended reading direction.
marks and manipulations should be confined to a scale and range where all changes are easily distinguishable.
Speaking of modern-day visualizations, teh advantage of using and building on a grammar for visual communication & mapping should be obvious, right? Marks and visual variables are amenable for automated visual mapping! I envision with the rise of interest and techniques in (big) data analysis and other data mining methods, that visual grammatical heuristics will come in handy in generating automated readable syntheses of these results.