In “Representational Genera,” Haugeland built an account of representational kinds around the structures of their contents. Icons, which include pictures, images, graphs, and diagrams, differ from logical representations like linguistic expressions because iconic contents are essentially patterns of dependent and independent variables. Such structures are absent in language. Recording is a witless, replayable process that Haugeland deployed to calm worries about his content-wise account of representational genera. This paper elevates recording from the supporting role Haugeland gave it to star of the show.
The main claim defended here is that some kinds of representation—pictures, images, graphs, diagrams—are modeled by recording processes, while others, like languages, are not. Extensionally, this distinction is close to Haugeland’s, but it is intensionally subtler, more plausible, and, as I hope to show, more useful. This approach abandons Haugeland’s goal of distinguishing representational kinds exclusively in terms of their contents, but there are many advantages to doing so.