Neural networks of PDP (parallel distributed processing), models of computation are based on mathematical models of neural processes. The authors explore the application of PDP to simple spatial reasoning in computer-aided design. Knowledge that provides a mapping from performances to design descriptions can be encoded into PDP systems in the form of learned patterns. The implementation of a simple PDP pattern associator is described. The system is shown to facilitate a rudimentary kind of associative or analogical reasoning. It also facilitates reasoning where only incomplete information is available. There are important concerns in design reasoning