Options include:
- The knowledge speakers of a language have to possess to speak the language fluently;
- The logical and more generally mathematical relations between utterances (e.g., entailment, probabilification, etc.);
- How the brain implements linguistic competence.
Each conception brings along with it criteria of adequacy for linguistic theorizing, thus making some data relevant or irrelevant. Perhaps the answer will differ between semantics and pragmatics, as the prevalence of psycholinguistic research in pragmatics and the paucity thereof in semantics suggests.
The issues we propose discussing resemble those concerning formal frameworks in other areas of philosophy, such as formal epistemology and decision theory: how worried should we be if nothing like credences or utility functions are psychologically realized, for example? Hopefully answering our questions, questions that concern a central area of philosophical and linguistic research, will shed light on the relation between formal models and psychological realizations of those models more broadly.