Autocorrelation associative memories (autocorrelators) are feedback neural network architectures that store bipolar patterns. Hete-rocorrelation associative memories (heterocorrelators) are feedback neural network architectures that store bipolar pattern pairs. The relationship between autocorrelators and heterocorrelators is examined. A review of the encoding algorithms, recall operations, stability proofs, and capacity arguments of first-and higher-order autocorrelators, commonly referred to as Hopfield associative memories (HAM’s), and first-order heterocorrelators, commonly referred to as bidirectional associative memories (BAM’s) is presented. Higher-ordered BAM’s and first-and higher-ordered intraconnected BAM’s are then introduced. The encoding, recall, and stability procedures of each are discussed. Finally, simulation results that compare the storage capacity and storage efficiency of three heterocorrelators, the heteroassociative HAM, the BAM, and the intraconnected BAM, are presented. © 1990 IEEE