Linking and brushing are interaction techniques. They can be used to enhance the work with scatterplot matrices, parallel coordinates and many other InfoVis techniques [19,30].
Brushing means selecting a subset of the data items with an input device (mouse). This is usually done to highlight this subset, but it can also be done to delete it from the view or to de-emphasize it, if the user wants to focus on the other items.
Brushing is most interesting in connection with linking. For instance in a scatterplot matrix, the user could brush some points in one plot. This causes the brush effect (highlighting, etc.) to be applied on those points in the other plots that represent the same data items. Figure 2.7 illustrates this. The brush has been applied to the second plot of the first row. In this case it highlights the brushed glyphs by changing their color and shape. This is useful for analysis over more than two dimensions, to find conditional dependence. As an example, one could want to find out whether A correlates with B when only cases where C and D are in a particular region are taken into account.
Linking can not only be applied to multiple scatterplots. All kinds of multiple view systems [2] can be linked [8]. Multiple view systems allow the same dataset to be viewed with different visualization techniques, such as scatterplots and parallel coordinates. Using more than one visualization technique is often an advantage because different methods have different strengths and weaknesses, so one may find an interesting relationship in one view that is not obvious in another. Being able to brush one of the views and see the brushed cases highlighted in another considerably improves the analysis.
