[en] Particle-in-cell (PIC) models are widely used in fusion studies associated with energy research. They are also used in certain fluid dynamical studies. Parallel computation is relevant to them because (1) PIC models are not amenable to a lot of vectorization - about 50% of the total computation can be vectorized in the average model; (2) the volume of data processed by PIC models typically necessitates use of secondary storage with an attendant requirements for high-speed I/O; and (3) PIC models exist today whose implementation requires a computer 10 to 100 times faster than the Cray-1. This paper discusses parallel formulation of PIC models for master/slave architectures and ring architectures. Because interprocessor communication can be a decisive factor in the overall efficiency of a parallel system, we show how to divide these models into large granules that can be executed in parallel with relatively little need for communication. We also report measurements of speedup obtained from experiments on the UNIVAC 1100/84 and the Denelcor HEP