In a systematic review, 194 studies on learning from animation were analysed. The analysis covers the learning domains, the representational characteristics of the animations, the assessed perceptual and cognitive achievements, and the assessment formats. Research on learning from animation focuses on assessing conceptual at the neglect of kinematic mental models. This is in contrast to an important rationale for making use of animations: that it needs to be learned what animations can specifically display, namely, how change in space and time occurs. This might explain why meta-analyses which compared the effectiveness of animations and static pictures found merely small overall effect sizes in favour of animations. To confirm this hypothesis, one meta-analysis was re-analysed with a new moderator. It encodes whether the features of the displayed changes were relevant to learning. Learning from animations was significantly more successful than learning from static pictures, if the features of the displayed changes had to be learned.