The results of a study of the relation between a number of mechanical properties of poly(ethylene terephthalate) (PET) yarns and their physical structure are presented. The relation was studied on a set of 295 drawn yarn samples, resulting from an exceptionally large variety in process conditions applied. The size of this experiment offered a unique opportunity to study the property–structure relation more extensively than ever before. Therefore, our understanding of the mechanical properties could be enriched with several new insights. A selection of seven yarn properties has been investigated. Five of them have been derived from the stress–strain curves and the remaining two are shrinkage and shrinkage force. The physical structure has been described with a set of five, statistically selected, parameters. Most of the yarn properties could very well be described in terms of these structure parameters. This description was performed by means of an artificial neural network, ANN. The type of calculation is completely naive, i.e., without any specific mathematical formulation for the relation concerned. The fitting results have been translated into physical aspects related to the well-known molecular two-phase model. The practical importance of a good physical understanding of yarn properties is that the essential possibilities and impossibilities of combinations of properties can far more easily be surveyed and understood. As a result, the efficiency of process developments can be substantially improved.