Abstract

I suggest that the representational content of a scientific model is determined by a ‘key’ associated with it. A key allows the model’s users to draw inferences about its target system. Crucially, these inferences need not be a matter of proposed similarity (structural or otherwise) to its target, but can allow for much more conventional associations between model features and features to be exported. Although this is a simple suggestion, it has broad ramifications. I point out that it allows us to re-conceptualise what we mean by ‘idealisation’: just because a model is a distortion of its target (in the relevant respects, and even essentially so), this does not entail that it is a misrepresentation, even with respect to the features it distorts. Rather, we should focus on interpreting the distorted aspects of such models non-literally. I investigate various ways of doing so, and demonstrate that for at least some idealised models, the result is that they are not misrepresentations after all, thereby diffusing various puzzles associated with their use in science.