How people perceive (or misperceive) trends influences their political attitudes (e.g., Bartels, 2008). In particular, processing time series on climate change is a challenging task: cognitive biases often lead to misinterpretation of climate trends and, consequently, to a lower willingness to take action against global warming (for a review, see Hardy & Jamieson, 2017). In this paper, we build on peak-end-rule theory (e.g., Kahneman et al., 1993) and the literature on endpoint bias (e.g., Jamieson & Hardy, 2014). We investigate what we define as cyclic variation bias, that is, the misperception of an underlying trend due to regular oscillations around that trend.

In a pre-registered between-subjects experiment included in a probabilistic survey in the city of Berlin (N = 1,833), we exposed participants to a simulated yet highly realistic visualization of declining water levels in a large Berlin lake. All experimental groups were exposed to exactly the same trend, but we manipulated the time window to vary the season at the end of the graph and the number of years shown. Varying the cyclic phase (i.e., the visualization of seasons) in the graph had a significant effect on participants' perception of the trend. This effect was attenuated when participants were exposed to a three-year instead of a two-year time series. Exposure to seasonal variation also correlated with support for climate action, although this effect did not reach statistical significance. We discuss the implications of the cyclic variation bias for communication about climate change and other politically relevant phenomena.