7. Images as sign systems. The many structural semiotic theories are hardly mentioned in this book, despite a fairly extensive literature that includes Fernande Saint-Martin and the Belgian Groupe µ. The Swedish scholar Göran Sonesson, author of a number of books on systematic visual semiotics, is excluded from these Seminars. [16] Partly that is because both North American and some German scholarship rejects systematic semiotics, and partly it is because performative, open, and contextual readings have become central in art history.

8. Images as defective sign systems. This argument is usually assigned to Nelson Goodman, and especially his argument against naturalism. In the effort to capture «the crucial difference between pictorial and verbal properties,» he argues, representation is «disengaged from perverted ideas of it as an idiosyncratic physical process like mirroring, and is recognized as a symbolic relationship.» [17] The notion of a defective or incomplete system is crucial to this sense of what an image is: «In painting and sculpture, exemplification is syntactically and semantically dense. Neither the pictorial characteristics nor the exemplified properties are differentiated; and exemplified predicates come from a discursive and unlimited natural language.» [18] Goodman has an unresolved position in some contemporary discussions of the image, and of the texts on this opening list, he is the one most likely to be almost adopted: «almost» because the authors who most believe him, including Tom Mitchell in these Seminars, are also the ones least likely to use his theories in any detailed way. [19]

9. Images as a genus, composed of individual species. Goodman’s theories divide images into different kinds, and so do many others. In general, theories that try to divide images do not get much further than the distinction between naturalistic images and their proposed counterparts, which are normally named diagrams, notations, or graphs. [20] Thomas Sebeok’s Signs: An Introduction to Semiotics, for example, begins with Peirce’s triad icon, index, and symbol. Sebeok then comments: «the neglect of diagrams is particularly incomprehensible in view of the fact that they loomed large in Peirce’s own semiotic research.» [21] I think the Seminars reflect the general tenor of the literature in that they are less interested in the actual divisions than in the idea of dividing. Aside from a small recent literature on diagrams, most discussion on whether images are divisible into types has centered on the word/image dichotomy—and some form of that distinction is assumed even in philosophic texts interested in the image, such as Nancy’s essay «Distinct Oscillation» (i.e., between word and image). [22]