Loss networks are a class of resource allocation models which have proved useful in the study and design of communication networks, including cellular mobile networks, integrated services digital networks, database structures and multiprocessor architectures. This paper is concerned with estimating performance measures of loss networks such as congestion probabilities. Although many of these measures have explicit formulae, it often requires too much processor time to perform the calculations, making it necessary to use approximation techniques. Speciically we compare the Erlang xed point approximation with the recently-proposed Markov random eld method. The principles behind both will be presented. Simple examples are chosen to demonstrate each method's application, and comparison is made in terms of relative accuracy, computational time and complexity.