Methodology: to offer a dissertation in two halves; the first of which being a taxonomic account of essential doctrines from four of the most characteristic empiricist scientific anti-realists. Locke and Berkeley represent classical empiricism, and Bas van Fraassen and Nancy Cartwright represent the moderns. The taxonomy addresses primarily three sets of issues, namely causation, theoretical entities and inference to the best explanation, stressing the plurality of concerns and approaches possible within an anti-realist framework. The second half then applies some of the lessons learned from the study of earlier forms of anti-realism and offers a critique of inference to the best explanation and its role in science. The aim is to set up a form of empiricist anti-realism which is heedful of the anti-metaphysical programme of positivism, but which is more accommodating to inferences to theoretical entities and less absolutist in its strictures on representation and reference. Part Two addresses various topics within empiricist responses to science, ranging from a critique of uses of epistemic privilege in the justification of inference to the best explanation, to an attempt to apply ideas from artificial life-style emergent simulations of living behaviour as a model of the inference to theoretical entities. Finally, the thesis concludes with an attempt (inspired by van Fraassen's 'Gentle Polemics'), to relate (and thereby criticise), the realist's demand for an explanation of the success of science to anthropic arguments of similar form used to explain the simplicity of nature.