The thesis exemplifies, illustrates and argues for an epistemology of Cognitive Engineering (CE). Others have proposed an ontology of the discipline. To give cogency to this ontology of CE with its own practices and method of validation, both radically distinct from Psychology and Cognitive Science, separate streams of science and technology have been asserted. Thus, design knowledge is not scientific knowledge, and one cannot specify the former with the latter: design problems with their own specific requirements define their kind of knowledge. However, where the need is for the design or evaluation of a system which mimics human cognitive behaviour, how can it be met without scientific (or descriptive) knowledge? The project of designing or evaluating Natural Language Dialogue (NLD) systems presents just such a need. Two solutions are required: (i) to show how a transition might plausibly be made from scientific (or descriptive) knowledge of linguistics to explicit design (or prescriptive) knowledge of NLD systems, differentiating resources which are usable from those which are not: an NLD framework; and, (ii) to endorse this transition with an argument for the epistemological validity underpinning the move, and to provide a general foundation for the relationship between science and technology (comprising applied science and engineering): a foundational framework. The thesis responds to both needs: promoting, criticising, and supplementing linguistic theories as the basis for the framework to satisfy (i). And to fulfil (ii), concepts adapted from Speech Act theory are combined with elaborated versions of key terms in the ontology of CE in order to argue for the commonalities of science and engineering. These arguments are situated via polemical disputes about the status, with respect to science, of Human Computer Interaction (HCI) and design disciplines more generally.