Demonstratives on Pictures

Kaplan’s theory of demonstratives and deicticals Kaplan (1989a,b) can be briefly stated as follows. Expressions of this kind depend for their interpretation on the context of utterance and in a context of utterance they refer directly to whatever they refer to. Direct reference in turn consists in two properties. The first property is the absence of a Fregean sense. The context does its work once and for all and the reference is not influenced by a counterfactual circumstance in which something else is pointed at (in the case of demonstrative) or somebody else is speaking (in the case of deictical ”I”). The second property is the rigidity of the reference: if the reference is direct, it is the same in all possible worlds. It is rigid in a sense slightly stronger than Kripke’s (Kripke, 1972), because even in worlds where the object does not exist, the reference is still to the object referred to in the context of utterance. Demonstratives or deicticals can fail to refer when the utterance is abnormal. Consider e.g. a demonstrative without an accompanying pointing gesture (an incomplete demonstrative, in the terminology of Kaplan) or one with a pointing gesture that fails to point to anything. For deicticals we need slightly more imagination: consider an inscription in the beach sand of I am the greatest or an utterance coming from outside of space-time of I am here now. In such cases we have a defective context of utterance. One class of counterexamples to this theory is important in the context of dialogue systems incorporating the facility of pointing in a graphical representation visible on the computer screen: demonstratives used to refer to objects represented in a picture. Kaplan discusses his pointing to a picture of Spiro T. Agnew hanging on the wall behind him which people have put up in place of Carnap’s portrait, while saying: that man is the greatest philosopher of the 20th century. That is not yet a counterexample to Kaplan’s theory, as pointing to Carnap’s portrait is a way of pointing to Carnap himself, and pointing to a portrait of Agnew is pointing to Agnew. But pictures can lie: something can seem to be a portrait of somebody without actually being a portrait of anybody and we can portray objects in pictures which do not actually exist, have never existed and will never exist. It seems even feasible to draw impossible objects such as the golden mountain or the most perfect island (but not the square circle). One can stand in front of a picture of Pegasus and say things such as: That is a winged horse, That horse does not exist or even You do not exist. It would seem that all of these are meaningful and even true, while Pegasus could not even turn out to exist, following the reasoning of Kripke about unicorns. And this entails that also in these cases that stands for something. The something is not the picture or a graphical object in the picture because of these it is clearly the case that they do exist and that they are not winged horses. A similar case arises with spoken language dialogue systems. Suppose you ask it for the departure of the train to Groningen and it says: I do not know what you are saying. Who is the I in the system’s utterance? Is it your computer (the vocaliser), the computer program, your CPU, the programmer, the copyright owner of the system? Are you being lied at? And what if it says: I do not have that information here right now? Yet, the whole situation seems remarkably like calling an information telephone number with a human operator. Also in this case we seem totally unconcerned with the identity of the voice on the other side.