Data structures, especially in classification problems, pattern recognition, and pattern classifiers impact the realization of classification algorithms, and especially induce their architectures and the nature of the classification results. This chapter discusses a number of approaches in which a granular description of classification data impacts a form of the classifier. It shows that the concept of information granularity plays a pivotal role in the ensuing construction of the classifiers. One may refer to the pioneering study by R. Bellman, R. Kalaba, and L. Zadeh pointing at the role of abstraction in pattern recognition. The chapter elaborates on the design facet of the problem, which concerns a selection of the level of allowable granularity. This directly relates to the structure description and a way of selecting a suitable level of granularity. Fuzzy clustering delivers a holistic view of the experimental data in terms of information granules represented as fuzzy sets.