The Citadel Itself: Defending Semantic Internalism

Abstract

Semantic internalism is the view that linguistic meaning amounts to forms of conceptual instructions, and that the process of forming linguistic representations does not involve reference to extra-mental entities. Contemporary philosophy of language remains predominantly externalist in focus, having developed systems of extensional reference which depart from classical rationalist assumptions. I will defend semantic internalism using a broad range of case studies, accruing what I see at the most convincing arguments in its favour. Particular focus will be placed on exemplar cases such as natural kind and artifactual terms.Copredication via inherent polysemy will be used as a strong source of evidence for internalism, countering the received view of the externalist character of meaning. Overall, my aim is to comprehensively defend internalism against its critics and to push the exploration of linguistic content and meaning "back into the brain".

The paper was boring and long: A reply to Michel and Löhr’s theory of predicate ordering in copredication

“…Going beyond this, in Murphy (2021c) it was shown that what I called the "persistence conditions" for copredication (the conditions under which certain complex polysemous nominals can sustain coherent sense relations in a brief narrative which "takes away" certain senses until the object referred to is reduced in sense number) are also correlated significantly with semantic complexity, not sense frequency. This provides further evidence for the centrality not of prediction, but of semantics (see Murphy 2023bMurphy , 2024a.…”

Natural language syntax complies with the free-energy principle

“…The blanket states can be partitioned into sensory states and active states. External states affect internal states only through sensory states, while internal states affect external states only through active states (Murphy, 2023a ). The implicit circular causality is formally identical to the perception–action cycle (Fuster, 2004 ).…”

“…1 MERGE serves a similar role to an elementary function, as in the theory of computability (e.g., the zero function, the identity function), in that it is meant to be non-decomposable. Putting many subsidiary technical details aside, these sets are then ‘labeled’ and given a syntactic identity, or a ‘head’ (Frampton & Gutmann, 1999 ; Leivada et al, 2023 ; Murphy et al, 2022 , 2023 ; Murphy, 2015a , b , c , 2023a ; Woolnough et al, 2023 ), based on which element is most structurally prominent and easiest to search for (i.e., Z in the structure {Z, {X, Y}}). 2 Labeling takes place when conceptual systems access the structures generated by syntax.…”