Cheng, HaoWang, JunhongWang, YunLee, BongshinZhang, HaidongZhang, DongmeiBorgo, RitaMarai, G. ElisabetaSchreck, Tobias2022-06-032022-06-0320221467-8659https://doi.org/10.1111/cgf.14560https://diglib.eg.org:443/handle/10.1111/cgf14560Combining data visualizations, animations, and audio narrations, data videos can increase viewer engagement and effectively communicate data stories. Due to their increasing popularity, data videos have gained growing attention from the visualization research community. However, recent research on data videos has focused on animations, lacking an understanding of narrations. In this work, we study how data videos use narrations and animations to convey information effectively. We conduct a qualitative analysis on 426 clips with visualizations extracted from 60 data videos collected from a variety of media outlets, covering a diverse array of topics. We manually label 816 sentences with 1226 semantic labels and record the composition of 2553 animations through an open coding process. We also analyze how narrations and animations coordinate with each other by assigning links between semantic labels and animations. With 937 (76.4%) semantic labels and 2503 (98.0%) animations linked, we identify four types of narration-animation relationships in the collected clips. Drawing from the findings, we discuss study implications and future research opportunities of data videos.10.1111/cgf.14560527-53913 pages