Graphs in corporate annual reports are a double-edged sword. While they offer the potential for improved communication of accounting information to users, the preparers of the annual reports can easily manipulate the graphs for their own interests. For over a decade, the empirical financial graphics literature has focused on examining company reporting practices. A particular concern has been measurement distortion, which violates a fundamental principle of graph construction. Unfortunately, it is not yet known whether observed levels of measurement distortion are likely to affect users' perceptions of financial performance. This study uses an experimental approach to address this issue. Pairs of graphs are shown to establish the level of difference that is just noticeable to graph readers. Six levels of "distortion" are investigated. Results indicate that if financial graphs are to avoid distorting the perceptions of users, then no measurement distortions in excess of 10% should be allowed. Users with lower levels of financial understanding appear to be most at risk of being misled by distorted graphs. Further research will be necessary to investigate whether this impact upon perceptions subsequently affects users' decisions in specific contexts.