Prior beliefs of viewers are a significant factor of bias when working with visualizations. This belief bias can easily be evoked by axis labels and context, leading to an influence on a viewers interaction with and analysis of said visualization. Even for low level visual analytic tasks, viewer beliefs show themselves as an influence, where one would otherwise hope viewers would be entirely objective. Through a user study, I show how belief triggering axis labels can influence a viewer’s correlation estimates of scatter plots when compared to the same plots with neutral X/Y labels. Viewers first reported their beliefs in six variable pairs that are used as labels later. The viewer’s baseline correlation estimates of plots with X/Y labels were gathered, and compared to the estimates of the same plots, but with belief triggering labels (i.e. Income/Happiness). When viewers believed in the relation of the variables, they increased their correlation estimates. Meanwhile when they believed them not to be related, they decreased their estimates. Further more, this effect extends to the viewer’s gaze behavior. Viewers with different belief strengths showed different levels of fixation on outliers, increasing or decreasing from their baseline dependent on belief strength. They also generally showed different shifts in attention distribution across the plots. The results reaffirm the possible effect of belief on even low level analytic tasks, and suggest an effect on viewer perception akin to selective attention or increased salience of features that align with existing belief. These findings can aid researchers working on visualizations techniques and dynamic visualization systems that aim to recognize and mitigate such cognitive biases.