Traditional views conceive graphing as knowledge represented in students' minds. We show in our critique that such views lead to a common assessment problem of how to account for variations in performance across contexts and tasks, and a common attribution problem that locates difficulties in students' deficient cognitive apparatus. Grounded in recent research of scientists at work and everyday cognition, this article provides an alternative perspective that conceives of graphing as observable practices employed to achieve specific goals. This perspective highlights the nature of graphs as semiotic objects, rhetorical devices, and conscription devices. This shift in perspective dissolves problems with assessment and inappropriate attribution of student difficulties. The plausibility and fruitfulness of the new perspective is illustrated in three ways. First, we show that successes and failures of various graphing curricula become understandable in terms of the presence or absence of social dimensions of the practice. Second, we show how our perspective necessitates new assessment practices. Third, we show how our practice perspective on graphing led us to different learning environments and to new foci for conducting research in student‐centered open‐inquiry contexts. © 1997 John Wiley & Sons, Inc. Sci Ed 81:91–106, 1997.