Readings 04: Encoding
This week, the topic is Encodings. The Visual channels to which we can map data. These can be thought of as the building blocks from which visualizations are constructed. We’ll read about different encodings, and hopefully get a sense of why you might choose one over the other. And you’ll look at some standard designs and try to understand how they are put together from encodings.
Unfortunately, I don’t have a way to let you read the original source where the idea of basic encodings/visual variables were introduced (see Bertin’s Books (Semiology of Graphics)).
The primary readings are three chapters that discuss the different encodings, and a classic paper they all refer to:
Marks and Channels (Chapter 5 from Munzner’s Visualization Analysis & Design) (Munzner-05-MarksAndChannels.pdf 0.4mb)
A nice discussion of the main encodings, with information of how they differ and how to choose.
Arrange Tables (Chapter 7 from Munzner’s Visualization Analysis & Design) (Munzner-07-ArrangeTables.pdf 0.6mb)
Position encodings are extra important and potentially more complex, so they get their own chapter. This chapter is particularly interesting because Munzner shows us how to break down a lot of standard (and some not so standard) charts into basic encodings. (note that we’ve skipped over Chapters 4 and 6 - we’ll come back to these).
Basic Principles of Visualization (Chapter 5 of The Truthful Art) (theTruthfulArtCh5.pdf 10.2mb)
In some ways, this is redundant with Munzner - but I like it as a different perspective, less formal and less academic. It provides some thoughts on how to make practical use of the research literature (which we will look at).
Cleveland and McGill. Graphical Perception and Graphical Methods for Analyzing Scientific Data. Science 229(4716), 1985. (online library) (ClevelandMcGill85.pdf 1.3mb)
This paper is referred to by Munzner, Cairo, and, well, everyone else. It’s the first rigorous attempt to understand how people perform at reading encodings. I think it’s important to see the original paper, so you know what they are talking about.
There are many more recent papers that continue the tradition of trying to rigorously and empirically determine what works and doesn’t work. It’s become a whole genre. We’ll see more when we talk about evaluation and perception. See Heer&Bostock (optional, below) for a more modern take on this paper.
Optional:
Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design. Jeffrey Heer, Michael Bostock ACM Human Factors in Computing Systems (CHI), 203–212, 2010 PDF (607.4 KB)
This paper is interesting since it recreates most of Cleveland and McGill as a Mechanical Turk study, with a much broader population. The presentation is much more modern (and easier to interpret). This could be a replacement for the original.
Visual Representation from Semiology of Graphics by Sheelagh Carpendale. Lecture slides/notes.
A lot of the idea of encodings come from [Bertin]( Bertin’s Books (Semiology of Graphics)), but it’s too hard to read the original sources. Sheelagh Carpendale (a well known Vis professor) provides a great discussion in her slides that mix modern examples with Bertin.
Automating the Design of Graphical Presentations of Relational Information by Jock Mackinlay, ACM Transactions on Graphics, 1986.
In an amazing 1986 system, Jock Mackinlay tried to automatically create charts from data. One of the key insights was to think about visualizations in terms of the basic encodings, which let him reason about these basic building blocks. His intuitions of what encodings were better/worse for different tasks was the beginning of trying to formalize this. He based this on his intuitions - but experiments show that he wasn’t too far off.
In 1986 he was systematizing the design of visualizations - he had to have a systematic way to design, since he wanted to do it automatically! His approach is exactly what we are doing: considering visualizations in terms of encodings that can be reasoned about. Amazingly far ahead of its time.