We often rely on visualizations to help us detect weirdness: places where our data didn’t do what we expected. However, there are many common situations where visualizations are poor weirdness detectors: they either do not make weirdness visible, or they cast perfectly normal data in a strange light. In this talk, I will discuss my efforts to measure and improve the ability of visualizations to present weirdness. In the first half of my talk, I’ll discuss graphical perception of data quality issues like outliers and missing data in simple univariate and bivariate visualizations. In the second half, I’ll discuss work on detecting and presenting abnormality and uncertainty in maps.