The interpretation of graphs and tables

Undergraduates solved statistical reasoning problems based on data presented in a variety of graphs and tables. When assessing relative probabilities, undergraduates were equally successful at answering the questions regardless of how the data were displayed. When making data-based causal inferences, accuracy decreased and students were quite sensitive to differences in the data display: data in percentages produced more accurate responses than frequencies; graphs elicited better problem-solving strategies than contingency tables; and pie charts yielded the most consistently high accuracy.;In order to solve these problems, people must decide which of the quantities in the data display are relevant to the problem at hand and how these quantities should be combined to solve the problem. When people are uncertain about what is relevant, they may look to the data display to guide their problem solving. In this case, excess information in the data display may add to the difficulty of the problem. When someone is adept at a particular type of reasoning, s/he knows how to identify and ignore irrelevant information in the data display.;These studies show that graph interpretation is distinct from graph decoding, and that graph interpretation skill is not simply a function of the graph (or table) type. It is a complex interaction between the data display format, the type of problem to be solved and the problem-solver's facility with the reasoning underlying the particular problem type. Although these students can decode graphs and tables, compose and compare ratios, the format of the data display influenced their ability to solve some problems. These students had relatively little difficulty using tables and bar graphs and frequency data to solve probability problems, but had considerable (yet variable) difficulty using the same data display types when solving causal inference problems. The students were adept at comparing ratios in the probability task, but generally less successful at comparing ratios in the causal inference task. The results suggest that a major source of difficulty in graph and table interpretation for the students lies in the translation of both the problem statement and the data display into appropriate and compatible mental representations.