Statistical charts complement textual reports by visualizing overall patterns or relations in the data. However, layout algorithms may place charts far from their associated text. Such distant placement can cause reading difficulties, or worse, a misinterpretation. We conducted an eye-tracking experiment comparing reading behaviors in two proximity levels: Placing text and chart on the same page, versus placing them on two different pages. The results indicate that the proximity influences text-reading stronger than chart-reading behavior. We discuss design implications for text-chart layout algorithms and practices. This paper and all data and materials are freely available at https://osf.io/xunt9.