Machine learning from data, neuro-fuzzy information processing, approximate reasoning and genetic and evolutionary computation are all aspects of computational intelligence (also called soft computing methods). Soft computing methods differ from conventional computing in that they are tolerant of imprecision, uncertainty and partial truths. These characteristics can be exploited to achieve tractability, robustness and low solution costs when the solution to a complex (in machine terms) problem is required. The principal constituents of soft computing include: Neural Networks, Fuzzy Logic and Probabilistic Reasoning Systems. Genetic Algorithms (GAs), Evolutionary Algorithms, Chaos Theory', Complexity Theory and parts of Learning Theory all come under Probabilistic Reasoning Systems. Hybrid systems can be designed incorporating 2 or more aspects of soft computing that are more powerful than any of the components used in a stand alone fashion. A unified framework is needed to implement and manipulate such systems. Such a framework will allow for easy visualisation of the underlying concepts and easy modification of the resulting computer models. In this thesis, an investigation of the major aspects of computational intelligence has been carried out. The main emphasis has been placed on developing an object-oriented framework for architecting computational intelligence systems. Object models for Neural Networks, Fuzzy Logic Systems and Evolutionary Computation systems have been developed. Software has been written in C++ to realise sample implementations of the various systems. Finally, practical applications and the results of using the Neural Networks, Fuzzy Logic systems and Genetic Algorithms developed in solving real world problems are presented. A consistent notation based on the Object Modelling Technique (OMT) is used throughout the thesis to describe the software architectures from which the computer implementation models have been derived.