The pneumatic splicing technique was developed in the 1960's to overcome the deficiencies of knotting of textile yarns, but there has been little research into splicing because it is a complicated process, very difficult to observe. A splice is formed very quickly, using highly turbulent air within a closed environment. Direct observation is almost impossible. The poor understanding of the mechanism has led to ad-hoc solutions, and can yield unpredictable splice characteristics. A better understanding of the splicing process is long overdue. The objective of the research is the evolution of a theoretical model which would offer a better understanding of the splicing process than existing models, and which would be capable of being used as a predictive design tool for future splicers. The research programme presented in this thesis demonstrated the feasibility and efficiency of the Taguchi design of experiment method, as applied to the splicing process. The Taguchi method simplified the research, and facilitated the identification of the design parameters most relevant to splice quality. Mechanical aspects of experimental splicers were altered to establish how design features affect splicing performance. The effects of commonplace process parameters such as air pressure and blast duration were examined in detail. These experiments demonstrated the importance of air mass flow rate as a fundamental parameter in splicing; they also revealed the existence of a phenomenon which was called "scaling". Scaling describes the change in performance of a splicer, when it is confronted with yarns of a size which are outside its normal envelope of performance. Other techniques were used in conjunction with Taguchi; these included flow visualisation, computational fluid dynamics simulations, and the use of scanning electron microscopes. These approaches have been used together, to optimise splicer parameters and to investigate the airflow within the splicing chamber - the component which ultimately determines the splice structure. These techniques were also used to model the airflow within a wide range of commercially available splicing chambers. The validity of the visualisation and simulation methods are discussed, and a degree of confidence in the accuracy of the outcomes is established. By combining simulation images with the conventional experimental results, it was possible to develop an improved theoretical model for the splicing process. The model proved sufficiently robust to explain a number of phenomena in splicing previously poorly understood. The new model has been used as a prediction tool for splicer design characteristics. It is likely that the model will permit the creation of a new generation of high-performance splicing tools.