The theory of direct perception holds that competent observers are able to detect optical patterns that specify the relative mass of colliding balls. Heuristic theorists, on the other hand, claim that judgments of relative mass are based on variables that do not specify relative mass. We contrasted these views with an experiment in which participants were given feedback on their ratio-scaled estimates of the relative mass of simulated colliding balls. Correlations between judged relative mass and various kinetic and kinematic measures of the collisions revealed that (1) judgments of relative mass become more accurate with feedback, (2) different observers use different variables, (3) during training, many observers change which variables they use, (4) before training, observers tend to use nonspecifying variables or combinations thereof, (5) after a minimal amount oftraining, at least some observers seem to detect mass-specifying information, and (6) the judgments do not support a generalization of the heuristic model of Proffitt (1989, 1994). These findings suggest that direct perception of relative mass is a skill that can be developed through appropriate training.Kinematics is the branch of mechanics that is concerned with aspects ofpure motion, such as position, velocity, and acceleration. Kinetics (or dynamics), in contrast, is the branch of mechanics that deals with motion by taking into account such variables as force, mass, and friction. Because kinetic properties of the environment are important for the guidance of action, it would be beneficial for animals to perceive them visually, and indeed, humans, for instance, have been shown to perceive kinetic properties (e.g., Michotte, 1963;Runeson & Frykholm, 1983). The fact that the optical array consists only of kinematic variables creates a problem: How are kinematic variables exploited in the perception of kinetic properties?The nature ofthis exploitation has been debated by proponents ofthe theory ofdirect perception (Runeson, 1995; Runeson, Juslin, & Olsson, in press; Runeson & Vedeler,The Netherlands Organization for Scientific Research (NWO) is gratefully acknowledged for funding this project. This research was conducted while D.MJ. was supported by a grant from the Foundation for Behavioural and Educational Sciences of this organization (575-12-070), awarded to C.F.M. S.R.'s participation and his laboratory resources were supported by the Swedish Council for Research in the Humanities and Social Sciences (HSFR). We also thank David Gilden, Geoffrey Bingham, and Harold Sedgwick for helpful reviews on earlier versions of this article. Correspondence concerning this article should be addressed to D. M. Jacobs, Faculty of Human Movement Science, Vrije Universiteit, van der Boechorststraat 9, 1081 BT Amsterdam, The Netherlands (e-mail: djacobs@fbw.vu.nl). , 1994Todd & Warren, 1982). The theory ofdirect perception assumes that kinetic properties of the environment are specified by kinematic patterns. Smart perceptual mechanisms are hypothesized for the...