Motivated by both the goal of developing a real-time training simulator for arthroscopic surgery and the goal of developing an accurate model of cloth for automatic garment handling, we are developing particle-based techniques for the mechanical and visual modeling of complex non-rigid materials. Using this approach, a material is represented by a large collection of particles interacting with each other according to simple physical laws operating on a microscopic level. A simplified version of our cloth model has been implemented in Thinking Machine's *Lisp for the CM-2. Each particle is mapped to a processor and is connected to its four nearest neighboring particles. An energy can be calculated for each particle based on its location and the locations of its neighbors. The total energy of the model may be minimized using a parallel Metropolis algorithm to produce a final cloth configuration.