We all think visually, some of the time. For the most part we are unaware that we are visual thinking, as in the case that we thread our way through a crowded room, or when we arrange groups of chopped vegetables and other ingredients on a countertop in preparing to cook a complicated dish.  Graphic designers may think consciously of design decisions, analyzing which combination of colors and forms most clearly conveys the required information, but they are in the minority.

We also use visualizations as thinking tools. Many people who work creatively, construct rough and ready sketches, which are often unintelligible to anyone else, that somehow help them organize their thoughts. Those abstract squiggles hint at different ways of partitioning physical space or the space of ideas or even both together.

Visualizations are tools for reasoning about data and to be effective they must support the activities of visual thinking.  Part of this is ensuring that data is mapped to the display in such a way that informative patterns are available to resolve visual queries concerning the cognitive task. This requires matching the graphic representation with the capabilities of human visualization.  For example, correlations between variables should be visually easy to see and commonly searched for symbols should be more distinctive than those that are rarely sought out.  In addition, interactions must be designed to support an efficient visual thinking process. Visual analytics is an example of distributed cognition and cognitively efficient interactions require that perceptual and cognitive processes in the brain of the analyst must be efficiently linked to computational processes in a computer. For example, data points representing companies can be shown simultaneously in a map view and in a scatter plot view; the technique of brushing can be applied so that points on the map, when selected, are highlighted in both views. This can support reasoning about the growth of industries related to geographic regions, but to be cognitively efficient the brushing effect should ideally appear in less than a tenth of a second.

Visual Thinking Design Patterns (VTDPs) are based partly on a prior construct developed by Ware [1] and called visual thinking algorithms (VTAs).  VTDPs represent a broadening of this original concept with a change in emphasis. VTDPs are a method for describing the combined human-machine cognitive processes that are executed when interactive data visualizations are used as cognitive tools.

VTDPs take their inspiration from Alexander’s design patterns [2] intended for architects and from design patterns developed for software engineering [3].  Although considerable research has shown that perceptual and cognitive principles can be applied usefully to the design of interactive visualization, this knowledge is only applied in practice if a particular designer has taken an interest in the relevant research.  VTDPs are intended to provide an accessible structured method for combining knowledge about interaction methods and visualization designs together with cognitive and perceptual principles.

• Like their precedents, VTDPs are intended to describe best practice example solutions to design problems where interactive visualization is an intended component. 
• VTDPs provide a method for taking into account perceptual and cognitive issues especially key bottlenecks in the visual thinking process, such as limited visual working memory capacity.  They also provide a way of reasoning about semiotic issues in perceptual terms via the concept of the visual query.
• VTDPs incorporate the common set of interactive techniques used in visualization and suggest how they may be used separately or in combination. 

VTDPs should be understandable with a modest amount of training, and it should be possible to incorporate them into common design practices. VTDPs are potentially applicable to all visualization design problems, including design of presentation materials, interactive training materials and analytic tools.  The following section concentrates on a single application domain — the design of visualization tools used in visual analytics.

Like software engineering and architectural design patterns, VTDPs are not modules and are not re-usable.  The demands of analysis are almost infinitely varied due to the enormous variety of data types and analytic problems.  As a result, any modularization of VTDPs would necessarily restrict the domain.  Nevertheless, implementations of VTDPs can take advantage of modular components.  For example, map display software may allow for extra magnifying windows which can be used to support the pattern comparisons in a large information VTDP space.

1. C. Ware, Information Visualization, Perception for Design, Third Edition.  Morgan Kaufman. 2012.
2. C. Alexander, S. Ishikawa, and M. Silverstein, A pattern language. Oxford University Press. 1977.
3. E. Gamma, R. Helm, R. Johnson, and J. Vlissides. Design Patterns: Elements of Reusable Object-Oriented Software. Addison-Wesley. 1995.