Graph identification specifies the need to provide narrative information for identifying all encoding choices. This information is indispensable to the decoding process.
In addition, there is need to describe exactly the graph objective and any data particulars. The aim is to provide unbiased and clear descriptions that would achieve quick and unambiguous decoding.
Graphs without sufficient identification do not meet the American Statistical Association standards for graphical presentation.
Types of identification
Jacques Bertin (1967) distinguishes between two types of graph identification:
- Internal identification gives information about the choice of encoding tools. It is an indispensable piece of information, particularly when encoding complex and multidimensional data. Internal identification is typically addressed through an external legend or an internal on-the-plot identification of encoding tools.
- External identification provides information about the graph objective, the variables employed and source of data. It is independent on the choice of visual implantation or retinal variables. External identification includes titles, captions, notes and axes labels.
In addition, I find it useful classifying another form of identification:
- Direct identification directly identifies data values that may be useful for resolving uncertainty when in relative comparison or table-look up questions, as well as identifying specific events in the data that provide useful context for decoding.
Think of direct identification as idiosyncratic to each graph that may or may not be needed, in varying degrees, but internal and external identification are always needed in every graph.
Regardless, always bear in mind that graph identification is not data and therefore its visual prominence should be kept subtle.
Below is a conceptual example of a data graph with complete and sufficient identification. External identification includes a title describing the graph objective, a subtitle providing more context, a note acknowledging the data source and disclosing any data exclusions, axes titles with units of measurements, and axes labels. Internal identification is achieved through a legend that identifies the visual implantations. Direct identification is minimal in this graph with only the identification of ‘some important value’.