It is a commonplace that the natural sciences ‘enlarge’ our world, that they enable us to reach outside the narrow circle of our sense-knowledge to more complex domains of many layers. But it is not so simple to specify in what this enlargement consists. Most would say that we come to know of the existence of myriads of entities of which our ancestors knew nothing. But how exactly do we do that? And how reliable can such knowledge-claims be? Questions like these continue to divide philosophers of science. Conventionally, the main division is said to lie between ‘realists’ and their critics (‘anti-realists’). But there are, notoriously, almost as many realisms as realists. And the critics of realism represent a wide variety of philosophical positions. So boundary-lines shift, and differences that at first sight appeared fundamental vanish as the debate continues. Still, there is in the end a genuine disagreement here, and it concerns the most important philosophical question that can be raised about the significance of the natural sciences: what quality of understanding do they afford of the underlying structures of the world around us?

After a preliminary discussion of the tangle of differences separating realism and anti-realism, I shall argue for a broadly realist answer to this last question, taking care to avoid the overstated versions of realism on which critics have too often focussed, versions that they have in some instances themselves created.