Shedding Light on the Graph Schema

Shedding Light on the Graph Schema Raj M. Ratwani (rratwani@gmu.edu) J. Gregory Trafton (trafton@itd.nrl.navy.mil) George Mason University Naval Research Laboratory different types (i.e. bar graphs and line graphs). In the pure blocks, it is thought that because the stimuli are of the same type, each stimulus primes or activates the next and thus people are quick to respond to the stimuli. However, stimuli only prime or activate other stimuli that rely on the same mental representation. Thus, in the mixed blocks, because the stimuli are of different types, they may rely on different representations and result in a slower response as compared to the pure blocks. There are several other interpretations to mixing costs, but this interpretation is very prevalent (Los, Abstract The current theories of graph comprehension have posited the graph schema as providing us the necessary knowledge to interpret any graph type. Yet, little is known about the nature of the graph schema, and no empirical data exist showing that there actually is a graph schema. In experiment 1 we show evidence that a graph schema does exist, and that graph schemas are not specific to each and every graph type. In experiments 2 and 3 we show that there is a different graph schema for typical and atypical graphs. We interpret these findings as evidence for a prototypical graph schema. Introduction When looking at Figure 1a and attempting to read-off the number of Widgets in Tray B, how does one have the necessary knowledge to be able interpret this specific type of graph? Given the large number of graph types (e.g. bar, line, dot, scatter, box plot, etc.) and the fact that the same symbol can represent completely different information in each of these graph types, how do we activate and use the information specific to each graph type? For example, the “dot” in a scatter plot as compared to the “dot” in a box plot mean very different things and in order to be able to interpret these different graph types, we have to be able to activate the appropriate knowledge. The current theories of graph comprehension solve this problem by positing a “graph schema.” Pinker (1990) suggests it is the graph schema that allows us to recognize specific types of graphs and allows us to find the desired information in a graph. Lohse (1993) suggests that the graph schema contains standard, learned procedures for locating information in the graph. Thus, when reading-off specific information from a graph, the current theories would suggest the following operations: (1) Early visual processes construct all possible relationships among graph elements, (2) Build a propositional representation of the graph, (3) Activate graph schema, (4) Devise the conceptual question, (5) Associate location of bar with each tray, (6) Associate each bar with values for each tray (7) Devise the conceptual message (Carpenter & Shah, 1998; Lohse, 1993; Pinker, 1990; Trickett, Ratwani, & Trafton, under review). The graph schema (step 3) is central to all current theories of graph comprehension. What is interesting is that there has been no empirical work to establish whether a graph schema really exists, and, if so, what its features are. Our research goal was to use the mixing costs paradigm (Los, 1996, 1999) to investigate the nature of graph schemas. In the mixing costs paradigm there are blocks of pure stimuli, composed of items of the same type (i.e. all bar graphs), and blocks of mixed stimuli, which are composed of items of Figures 1a-d. (clockwise from upper left corner) Graphs used in experiments 1-3: bar graph, line graph, dot chart, scatter plot. By using the mixing costs paradigm, we will be able to show which graphs share a similar mental representation. This internal representation, we believe, is what most graph comprehension theorists call the graph schema. We will describe in some detail exactly what we think a graph schema is in the general discussion. Assuming a graph schema or representation does exist, there appear to be several possibilities as to how the graph schema accounts for our ability to interpret different graph types. First, the schema may be graph specific; each and every graph type may have its own unique mental representation (when we say representation, we mean internal, mental representation). For example, a bar graph may have its own representation and a