What does it take for a scientific model to represent? Models, as an integral part of scientific practice, are historically and contextually bound in their application. Practice-oriented debates in recent philosophy of science have emphasised how models can be said to act as representations in practice, e.g., as “mediators” between theory and data. However, the questions that remain open are: what exactly is it that is represented by a scientific model, and in which sense can we speak of a representation? I argue that the proper object of representation in scientific practice is not a model as such but, rather, the entire experimental system in which a model is an active part. The implication of my claim is a strongly historicized perspective on the capacity models to represent: the capacity of a model to represent must be judged against the individual life of an experimental system. In support of my argument, I turn to modelling issues in fragrance chemistry regarding the molecular basis of odors. Explanations of irregularities in the pursuit of so-called structure-odor relations provide an interesting example to analyze the modelling strategies that inform the notion of chemical similarity.