University of Alberta Dictionary of Cognitive Science: Semantics

Semantics deals with the relationship between representations and the world. Anything which can said to be a representation--which could be said to stand for, represent, point to, indicate, mean, refer to, or in some way be about something else--has semantic relations to that something else. Semantics is what makes the word Coffee9 mean that smelly muddy brown hot liquid that people drink.

A representation's semantic properties are those properties the representation has in virtue of the sort of relationship the representation has with a part of the world. So when we talk about what object (the thing in the world) represents, or whether the representation is a true representation of its object or whether it's a highly inaccurate representation of that object, or whether it misrepresents that object, we're talking about the representation's semantic properties.

The problem is that if cognitive scientists define the essence of cognition as processes operating on representations, then any process which operates on a representation has no access to that representation's semantic properties. Fodor9s (1990) Formality Condition maintains that any process which operates on a representation can only operate on the representation's nonsemantic or formal properties.

The idea, then, is that if a process which operates on a representation is to be sensitive to the semantic properties of the representation, such as what object it represents, then that representation9s semantic properties must somehow be mirrored in the representation's syntactic properties. So my cow representation must be fairly complex, and somehow 3contain2 formal descriptions of all the properties I ascribe to cows, so that processes which operate on this representation (such as those which allow me to utter 3Cows give milk,2) can operate on those properties.

But whether the properties I ascribe to cows in such formal descriptions are true of cows is inaccessible to those processes. Whether what I believe is true or not is a semantic property of that representation9s relationship with the world. And semantic properties like truth are transparent to the processes that operate on my representations. Perhaps the best we can hope is that the formal properties of all my representations are consistent, and form a coherent network of beliefs that facilitate my acting successfully in my environment. Whether these are true or not is inaccessible to the brain-processes which operate on those representations. (Hence what Fodor (1980) calls 3Methodological Solipsism2)

References:

Fodor, J. (1980). 3Methodological Solipsism Considered as a Research Strategy in Cognitive Psychology2. Behaviour and Brain Sciences, 3(1), 63-109.
Fodor, J. (1978). 3Tom Swift's Procedural Grand-mother2. Cognition, 6.

(Both of these are reprinted in Fodor (1981).Representations, Brighton U.K.: The Harvester Press. pp204-224 and pp225-256.)