Past research has shown that many scientists, when asked to interpret unfamiliar graphs that have nevertheless been culled from introductory undergraduate courses in their own field, experience problems and cannot give the standard answer accepted in the field. Yet, these same scientists turn out to be highly competent when it comes to graphs from their immediate domain of research. In this research, which is based on ethnographic studies among scientists and technicians, I show how graph interpretation in one biology laboratory initially required tool (computer) mediation. After scientists had become familiar with the phenomenon, data collection, and resulting graphs, they interpret the latter correctly without requiring prior transformation. Furthermore, in the course of their work, they established what they understood to be a one-to-one correspondence between graphs and some aspect of the natural world. As a result of scientists' embodied laboratory work, talk about the graphical representation and talk about the object represented are often indistinguishable. The process of developing competency in graph use is equivalent to that of a tool that becomes transparent to the consciousness of its user; I describe this process in terms of activity theory as a transition from tool-mediated action to operation.