Let us set some general views about the task of semantics.20 Semantics can be viewed as an empirical science which tries to model actual linguistic facts, it can be conceived of as a less ambitious representation, an anti-realistic one, or it can be seen to aim at producing deductive theories which seem to be just inspired by the data. After that, we will discuss how these views on the goal of formal semantics stand concern our discussion about the sources of semantic normativity.
The formal semanticist’s tasks
Semantics can be viewed as an empirical science, when its object is supposed to be found in actual linguistic facts, and its aim is to predict, explain or de- scribe those phenomena. The source of observation — the facts of semantics — are judgments of (non-) validity of certain inferences or schemata thereof. This source certainly relies on the semanticist’s intuitions, where the generality and correctness of an individual’s insights are not an evident affair. Observation is theoretically laden, and this also applies to the semanticist. If one keeps a pro- cessual, interpretive view on meaning, the input is still given by intuitions but these pertain to communication as an activity wherein meaning is attained and does not lie underneath.
A more modest standpoint that avoids some of the difficulties in which the empirical view on semantics incurs only sets for the discipline an ‘engineering task’. The goal is not to achieve a theory of meaning and actual linguistic oper- ation, but rather — in the spirit of ‘saving the phenomena’ — to produce formal theories that the apparent agents’ processes should agree with. The theories are supposed to be neutral with respect to the actual facts, in the spirit of anti-realism in philosophy of science.
A seemingly even less ambitious turn is taken by those who see semantics as a deductive science, much as mathematics or other formal disciplines. In this case, intuitions do not come as an input but rather constrain the descriptions of structures that the theories might give. This approach sets for the discipline the goal of constructing (or discovering) and studying theories for the sake of obtaining deductively consequences related to a certain realm of the world that is trying to be modeled. Modeling here is not however a matter of getting the world as it is, but instead not to infringe its boundaries. Nonetheless, the fact that the adequacy criterion is given by something external to the theories (‘reality bites’) normally seems to crawl into this minimalistic claim.
The task and its implications concerning the source of semantic normativity
Both an engineering driven and an empirically oriented conception of the task of formal semantics have a truly contrastive relationship with facts about natural language. Semantic normativity, as we have seen, requires that the source of reasons for semantic judgments of real linguistic agents may not be found in isolation, defined over only the speaker or the hearer in a dialogue. As we already argued in 2.3.2–3, if a theory in formal semantics aims at giving explanations of certain phenomenon in natural language interpretation, reasons for the correct predictions should also have a justificatory force.
A theory with anti-realistic aims may provide a model of interpretation with- out claiming these are processes explanatory of actual phenomena, and yet the model should not misrepresent the requirements of justification, if only because judgments of validity that are anti-realistically modeled are theoretically laden.21 This is not to claim that formal semantics has to purport to give a normative theory of justificatory judgments of semantic (in)correctness. Formal semantics may well try to remain silent about the normative dimension of interpretation. However, the proposed explanations or descriptions of interpretation should not conflict with a possible justificatory use of such reasons. A semantic theory that explained or modeled certain phenomena in natural language would predict cer- tain correct uses which, if unacceptable when submitted to the kind of test that can provide reasons for semantic judgments, they would result in an exercise with little or no interest for what semantics indeed would set itself to account for.
The case of the deductive ambition seems at first to eschew these duties. If the deductivist succeeds in remaining free from talk about ‘reality which bites’, it would be possible for it to ignore the conditions that would constrain the representation of a source of reasons for normative judgments. However, as soon as these theorists talk of evidence not merely inspiring but also constraining their theorizations, an improper representation of the source could amount to a failure in the theory’s minimal considerations of material adequacy. Thus, even if the successful deductivist could in principle remain careless about the representation of the sources of semantic normativity, in practice formal semanticists might not be able to avoid this constraint.
Let us briefly argue here why an agnostic position might be in principle ac- ceptable while an ‘atheistic’ one may not. Suppose the formal semanticist does not want to take any commitments with respect to modeling the justification of linguistic interpretation, with the normative aspects of the semantics/pragmatics interface. By an agnostic position we mean a semanticist who may not aim at modeling normative phenomena of natural language in a descriptive or explana- tory manner. This position nonetheless does not run against basic aspects in normative features of natural language semantics. An atheistic would instead deny that formal semantics should care at all about normative phenomena; i.e., misrepresentations would become allowed. With a clear cut distinction between natural science and other disciplines, such an atheistic would seem to deny any interest to the normative conditions we have discussed here. While the agnostic simply avoids potential conflict, this atheistic iconoclastically demurs any rele- vance of normative aspects in semantic phenomena.
Note that the career prospects for the agnostic and the atheist differ radically. In case the formal semanticist abstains from theorizing about the normative as- pects of the phenomena they explain or model, his attitude would let him model the behaviour of actual agents and not just artificial ones. Insofar as the material adequacy conditions for semantic normativity described in 2.3.2 are admitted as such, the atheistic obliteration of normative aspects of their data runs against an actual feature of actual agents in dialogues. Unless the atheistic is willing to give up his task to be one related to human speakers who actually use the natural languages for which a formal theory is devised, it seems that the price of his attitude is rather high.
Let us turn back toISPnow. In view of the arguments above, we can consider what this theory’s ambitions are and examine how it represents the source of semantic normativity of the data it models.