Visual metaphors in the form of still or animated pictures have been used in user interfaces with the hope of enhancing learning and use of computer applications. This paper studies animated metaphors with the intent to understand how they relate to human cognition and how their quality can be measured. We present a model of the relationships within metaphors, suggest lightweight evaluation techniques based on this model, and test these techniques in an empirical investigation. The results indicate that a lightweight analysis based on still images and made by domain-aware but metaphor-unaware judges can be used as a first step in deciding which metaphors are worthy of further study, and to direct animation efforts to overcome the most crucial problems. Furthermore, the results show that animation may increase or decrease the quality of a metaphor by considerable amounts; hence the final evaluation must be based on actual use of fully implemented metaphors. The results also confirm earlier suggestions to use rich metaphors and provides evidence that richness of the still image is important for the effectiveness of animation.