The aim of this study is to compare and contrast two Associative Memory (AM) model?s application to the domain of character recognition. The two AM models in question are One-Shot (OSAM) and Exponential Correlation Associative Memories (ECAM). We discuss if and how these AM models implement the concepts of recurrence, learning and domains of attraction. We identify how these concepts affect the suitability of each model to tackle the problems presented in this application domain. The problems identified in our study are variation in training set size, effect of noisy data, and effect of symbol transformation. Our study highlights both conceptually and experimentally the aspects of each model that make them suitable to distinct sub-domains of character recognition.