In this study, we investigated the representation effect in comprehending graphs. Many previous studies have confirmed the representation effect. In Experiment 1, we replicated the representation effect but using a set of graphs each of which is completely identical from the other in its perceptual characteristics. Participants drew a specific aspect of information from a line graph depending on the relation of x-axis and z-legend of the graph. In Experiment 2, participants were given a context for problem solving. The result showed that participants read a graph based on a given context; and the representation effect in comprehension was limited. In Experiment 3, participants generated a line graph by themselves. The result showed that they did not necessarily generate a consistent graph with a given context, and in comprehension the requirement of a context has very strong effects similarly as in Experiment 2.