Biases at the level of study design, conduct, data analysis and reporting have been recognized as major contributing factors to poor reproducibility. Authors from Turkey, UK and Germany (including PAASP partner Martin C. Michel) now add another type of bias to the growing list of biases, “perception bias”. Based on examples from the non-scientific literature they illustrate how data presentation can be technically correct but create biased perceptions by choices related to the unit of measure or scaling of the y-axis or graphs. For instance, one study outcome can lead to three entirely different interpretations depending on the choice of denominator used for normalization. The authors suggest that scientists should carefully consider whether their choice of graphical data representation may create perceptions that steer readers towards one of several possible interpretations rather than allowing them a neutral evaluation of the findings.