Variation in scatterplot displays

Scatterplots are typically constructed for the purpose of showing the association between two variables. We argue that such scatterplots should not vary in ways that are unrelated to the degree of association. Noting that the publication standards for the preparation of scatterplots are minimal, we review empirical literature that shows that a number of arbitrary features of scatterplots influence the inferences drawn by users. We discuss nine such features. We also review the literature that recommends how scatterplots should be prepared, and we examine 221 scatterplots published in recent journals. There was considerable variation among those published scatterplots. This article concludes with a call for further standardization by way of flexible guidelines.