Spatiotemporal coordinates (more generally, theoretical structures which vary under symmetries) do not, it is often thought, represent objective features of the physical world; instead, they are typically taken to encode certain facts about the system or structure doing the representing (“the observer”). Why don’t they represent anything objective (if, indeed, they don’t)? How do they come to represent something subjective (if, indeed, they do)? And what does all this tell us about the nature of representation in physical theories? This paper seeks to answer these questions.