Visualizations are the currency of scientific communication, but have been largely ignored in bibliometric studies. We classified more than 8 million figures from PubMed into coarse figure types and related the distribution of visualization types to scientific impact, breaking down the results by field of study. Remarkably, we find a significant correlation between scientific impact and the use of visual information, where higher impact papers tend to include more diagrams, and to a lesser extent more plots and photographs. We also find that the effect is modulated by whether citations come from within one's own field or from outside one's field, suggesting new strategies for communicating results across disciplinary boundaries. To explore these results and other ways of extracting this visual information, we developed a "figure first" search engine that quickly identify related figures; we hypothesize that figure-oriented search can improve precision and recall for certain search tasks and present some preliminary results. In ongoing work, we have visualize large collections of related figures as a map that can be used to understand changes in visualization over time and across fields. Finally, we have recently extended our classifier with a new unsupervised approach that can accommodate arbitrary class labels without providing additional training data; this technique allows fine-grained classification for precise searches and test-set generation. We use these results to articulate a new research agenda --- viziometrics --- to study the organization and presentation of visual information in the scientific literature.