being representations of the properties that they differentiate or detect. In accomplishing such differentiations, which are necessary for the subsequent selections of a recursively
self-maintenant system to be appropriately sensitive to the environment, the system does thereby create informational relationships, informational and perhaps causal
correspondences, with whatever properties it has detected. But, whereas the interactive model needs only that those detections be in fact of properties that are relevant to further indications of interactive potentiality, and does not require that those properties be represented, it is precisely such detections or differentiations that are standardly construed
as being representations of whatever properties they have differentiated — especially when those differentiations have occurred via the passive processing of inputs such as allegedly occurs in perceptual systems (Bickhard & Richie, 1983). Correspondence models construe representation as looking backward in time, down the input stream, to some privileged
locus of that input stream, such as the surface of a table reflecting light into the eye. The
interactive model construes representation as future oriented, to the potentialities for further
interaction that are afforded by the current environment, and to the conditional webs of potential interactions that can extend indefinitely in space and time.
Correspondence models of representation suffer from a host of fatal problems (Bickhard, 1993, 1996; Bickhard & Terveen, 1995). I will outline only a few of them here. The first is that such models cannot account for their own foundations. It is clear that not all informational or causal correspondences are representations. In fact, hardly any
are: the universe is full of such correspondences, since, for example, every single instance of every causal law yields such a correspondence. So one of the tasks facing a correspondence model is to specify what special kind of correspondence is representational
is to model what makes alleged representational correspondences so different from all
the others. There are multiple candidates for these extra criteria to be imposed on the class of correspondences per se (Fodor, 1990, 1998; Stich & Warfield, 1994), but the point that
I wish to focus on here is that none of them provide a naturalized model of the representational content, the mental content, for these alleged representations.
Representational content is that which specifies for the system what a representation is supposed to represent. It is the normative aspect of representation. It is what makes it possible for a representation to be in error: if a content of “cow” is predicated of what is in fact a horse, then that predication is false. Correspondence models are in a serious bind: if the special correspondence exists, then the representation exists, and that representation is
correct, but if the special correspondence does not exist, then the representation does not
exist, and it cannot be false. Information semantics, and other correspondence models,