A neural net model describing the non-linear interactions between axonal spikes is presented. It reconciles aspects of pattern recognition (as action of an associative memory) with those of spike synchronization and phase locking. The stability of the synchronized state is studied in detail.