Depiction

6. Interpreting pictures and recognizing things

The task of characterizing seeing-in is one of the most challenging in this area. Given this, and scepticism about some of the central notions which any such characterization must deploy, there have been attempts to clarify in other ways the thought that depiction is peculiarly visual. For example, some have attempted to analyse depiction using the epistemic resources required to interpret it. The idea is that what is special about pictures, as opposed to words, is what one needs to know to interpret them. For language, one must know the conventions governing the words’ use, conventions specific to individual words. Understanding pictures, in contrast, requires a rather different sort of knowledge. Flint Schier, for example, claims that one must know what the depicted item looks like (1986). Moreover, this is in essence all one needs to know to interpret pictures, provided that one has a general competence with depiction, that is, the ability to interpret any pictures at all. Since this is so, Schier speculates that a depiction of something engages our ability to recognize that thing in the flesh, and that it is definitive of depiction to do so. Pictures thus piggyback on the everyday workings of our visual systems, exploiting the resources normally deployed to identify, say, a goat as before me, to allow a flat surface also to convey to me the thought of a goat. More recently, Dominic Lopes has defended a similar position, developing it in interesting ways (1996).

The recognitional view has many advantages. In particular, it allows for accommodating and explaining the features of depiction noted above in §3. Pictures engage our abilities to recognize what they depict, but those abilities are relative to points of view. I may be able to recognize you seen from the front, while being unable to recognize you from behind. So what individual pictures must engage is the ability to recognize what they depict from a particular point of view - hence the perspectival nature of depiction. Further, visual recognitional abilities are engaged in clusters, rather than singly - I recognize you because I recognize a person with such and such features standing before me. So if a picture is to engage a visual recognitional a bility of mine, it will engage others at the same time. Thus there is no depiction which is not fairly complex in content. Finally, since visual recognitional abilities are abilities to recognize the visible properties of things, the complex content of any picture will be composed of the sorts of features that have a visual appearance. So pictures always show us how things look.

However, there are at least two forms of attack to which the recognitional account of depiction is vulnerable. First, it is unclear that the explanations it offers are of the appropriate modal strength. It is presumably a contingent matter that our recognitional systems work as they do. Had they worked otherwise, some of the features of depiction that Schier and Lopes take themselves to explain would not have obtained. For instance, had our recognitional abilities not been engaged in clusters, then a picture could engage our ability to recognize one thing - a goat, say - without engaging our ability to recognize any other properties or kinds. The recognitional view thus implies that it is a contingent fact that pictures cannot represent something without representing it as having a range of properties. We might wonder whether this consequence of the view is true. Could anything count as depiction if it allowed for depicting a goat tout court ? (See Hopkins 2003.)

Second, as Schier himself admits, it is tempting to think that there is a deeper explanation for why depiction engages our recognitional capacities - that pictures do after all resemble what they depict. If this line can be made good, explanations can be provided, not only for the features of §3, but for the recognitional view too, along with the facts that inspired it, concerning the epistemic resources depiction requires. However, the problems facing the resemblance view were severe. Can solutions be found for them?