This work deals with the problem of obtaining numerical approximation to solutions of a system of partial differential equation using successive relaxation. A feature of the backward constraints idea is that a multistep method should be chosen so that there is a wide separation of the moduli of the roots of the characteristic equations. This, with the application of the appropriate number of artificial conditions initially and terminally ensures stability. The research presented is an attempt to investigate how well backward constraints method performs as an integrator for partial differential equations. We have shown that the backward constraints method is stable and highly accurate for a method of lines solution of model linear partial differential equations. We compared the backward constraints method with the well-known Crank-Nicolson method and we proved that for high accuracy, the backward constraints method is better than Crank- Nicolson.