data, where graphics are m ore advantageous, particularly if the dataset is large. This is because diagrams or graphic representations make relations in the data explicit and direcdy perceivable. They can be understood as a form o f external memory, which places inform ation that may be needed for future inference together. Hence, they appear to “im prove recall for presented data, [and] reduce short-term mem ory loads during problem solving” (Casner 1991: 118). In other words, “visualization can act as an extension o f cognitive processes, augmenting working m em ory by providing visual markers for concepts and by revealing structural relationships between problem com ponents” (Ware 2000: 335).
However, even when a display successfully encodes the relation between elements, it m ight still no t help to solve the problem at hand. It has been previously m entioned that, in the field o f Inform ation Visualization, some work has been carried out to characterise the tasks supported by techniques for data investigation (Casner 1991; Z hou and Feiner 1998, 1996; W ehrend 1990; W ehrend and Lewis 1990; K napp 1995). These taxonomies attem pt to break up problem s into tasks that can be supported by visual representations according to the type o f data that they represent and, in some cases, also consider the work processes their potential users might need to support. These taxonomies “interface high-level presentation intents” (Zhou and Feiner 1998: 392), such as inform ing the user about a relation between attributes, with low-level visual techniques, such as linking and brushing between views. They also proved useful in a geovisualization context for defining user tasks that need support w hen exploring unknow n data with a vague goal. Furtherm ore, they enabled the evaluation o f the tool used for such purpose.
However, these typologies are incom plete or at least too general to describe geovisualization tasks. In Inform ation Visualization, maps have no t been considered as representations that can be interacted with and used as exploratory tools for reducing the search for inform ation and facilitating knowledge construction. O n the contrary, they are mostly seen as providing
the context in which the analysis o f some attributes— other than coordinates— may
be meaningful. T he fact that “spatially extended region[s] or object[sj” (W ehrend 1990: 16) are little m ore than a backdrop that gives context to analysis is no t surprising since the data with which Inform ation Visualization is concerned are typically abstract and non- geographical. Thus, position or location o f elements in a graph representing abstract data may be relevant— along with their arrangement or position relative to other elements in the view— for depicting similarities between data items. However, relative position and arrangement are interpreted as attributes from which a display is constructed (Wehrend 1990).
Chapter 5 Refining Task Characterization
These ‘attributes’ may be used to im prove the design o f visual displays by making them m ore effective in the sense o f facilitating the perception o f the inform ation they attem pt to convey, as well as increasing their efficiency by speeding user searches for inform ation. In contrast, w hen representing georeferenced inform ation, the space o f the display is usually reserved for depiction o f geographic space (M acEachren et a l 1998). In addition, it is well know n that in geography “everything is related to everything else, but near things are m ore related than distant things” (Tobler 1970: 236) and that attribute patterns and relations will tend to m anifest in space accordingly. Hence, spatial relations and relative positions between geographic entities are no t attributable to the semantics o f a visual graphic designed to facilitate perception as in the case o f Inform ation Visualization displays. Rather, they are attributable to spatial processes that may exist in the actual physical space that the map represents. Hence, m aps as visualizations o f spatial inform ation can indeed portray inform ation that gives access to details about attributes in a dataset as in aspatial visualizations, b u t relative position or location o f geographical entities has a m eaning in itself over and above facilitating perception.
This chapter will therefore describe the third and final evaluation which is aimed at providing evidence in support o f the need for a task typology that reflects the specificity o f spatial tasks. T he evaluation was conducted using the same environm ent used for the second study (Chapter 4), but was slightly im proved to take into account some o f the usability problems and identified limitations. The experimental design and the purpose o f the evaluation are discussed in §5.2. T he tasks considered are detailed in §5.2.4 and the results are presented in §5.2.5. T he data used for this experim ent is from the 2000 U.S. Census o f Population for the area o f Los Angeles, as described in m ore detail in Chapter 6.
5.2
E x p e r i m e n t a l d e s i g n
5.2.1 E x p e r i m e n t Goa ls
There were two main goals for this third set o f experiments. The first was to investigate the differences between tasks found in the previous chapter. It has been previously shown that tasks w here a cognitive visual operation was perform ed relative to a spatial feature took distinctively longer than similar tasks w here the relative spatial com parison was n o t required from users. In order to inspect this difference in m ore detail, each participant com pleted two sets o f four questions. Therefore, eight tasks were examined to consider w hether a finer