Meaning, Interpretation and Semantics

ion that is supposed to be characteristic for every empirical science. Physics studies movement on a frictionless plane, or the properties of a perfect vacuum; analogously, semantics deals with meaning and its properties on an equally abstract, yet empirical level. Or so it is assumed. But whereas physics, abstraction not withstanding, produces empirically testable predictions, it is far from clear that the same holds for semantics. Do we predict intuitions? What would that mean? Or are our formal theories intended as descriptions of the actual processes that occur in natural language users? If so, how are we to test them? And given the fact that these theories deal mainly with structural aspects of meaning and not with lexical content,5 are they not hopelessly incomplete? A less naive reaction might be dubbed the engineering view: here semantics is taken to be a much more modest enterprise. Instead of viewing it as a theory (in a strict sense) about what meaning is and how language users operate with it, one refrains from such far-reaching claims and merely requires that semantics provide a formally and empirically adequate specification of the input–output conditions. Whatever the actual underlying processes and meaning itself (which could be some kind of cognitive structure, or a system of neural networks), turn out to be, an account of them is to be in accordance with the results of a semantic theory, but the latter Note finally that once people started questioning the central role of the subject different views on the nature of language arose along with it. A prominent example is provided by the later works of Wittgenstein. This distinction often figures in a defense of the empirical status of semantics. The idea is that a clear distinction can be made between structural semantic operations and lexical content, where the structural operations are invariants over the varying contents of judgments. Such invariants can then be studied at the level of a single language, or at that of all natural languages. Or one can look even more generally at all forms of symbolic reasoning devices, which leads to one definition of what logic is all about. However, vis à vis natural language semantics the following observations may shed some doubts on the viability of the distinction. First of all, it is framework-relative, i.e., it depends (partially) on the formal languages one has decided in advance to use. For example, if one uses a sortal first-order logic different aspects will be regarded as structural semantic operations than if one uses a type-theoretic framework. This does not mean that the choice is a matter of mere convenience, but it does show that what structural semantic operations are, is not a purely empirical issue. Secondly, at a higher level, the distinction is interest-relative. (Cf., its origins: logic) It depends on an independent, theoretical decision which features of meaning one is interested in in the first place. In logic one is concerned with reasoning and hence one studies those aspects of meaning which are invariants with respect to that. But how could such a decision be made in advance for natural language? Again, this is not to say that the distinction is not useful or even necessary to make. The point is that it is not empirical, which means that it can not be appealed to in a defense of the empirical status of semantics.