We present a new approach for modeling plants which allows for adaptation to geometrical and material properties of a simulated environment. The form of a plant is derived from the paths of one or more automata which begin as seeds and progressively explore the environment. The automata sense their surroundings by casting rays and using intersection information to influence subsequent patterns of growth. In this way the power of well-known ray tracing techniques is applied to a problem of geometric modeling. By seeking locations which satisfy various environmental criteria such as proximity to surfaces and avaiability of light, the automata extend the repertoire of simulated plant growth to include emulation of phenomena such as clinging vines and heliotropism. This approach to database amplification can be used in conjunction with previous methods such as particle systems and graftals.