Mark Turner and the Overthrow of the Semiotic Conduit in Cognitive Science
For much of modern linguistics, applied pedagogy, and classical cognitive science, language has been organized around a durable but increasingly unstable assumption: the Conduit Metaphor (Reddy, 1979).
This framework presupposes that linguistic expression operates as a downstream encoding of fully specified, pre-linguistic mental content. Cognition is thus conceived as prior, complete, and internally legible; language functions as a transmissive channel.
Within this architecture, the mind is implicitly treated as a self-contained computational system, and language as a neutral representational conduit.
Over the past two decades, however, converging evidence from Conceptual Blending Theory (Fauconnier & Turner, 2002), the Extended Mind Hypothesis (Clark & Chalmers, 1998), and embodied-enactive approaches (Varela, Thompson & Rosch, 1991) has rendered this model increasingly insufficient as a general theory of linguistic cognition.
What emerges is not a refinement of the expression model, but a reconfiguration of its underlying cognitive ontology.
1. Cognitive Semantics as Dynamic Integration Across Non-Isomorphic Mental Spaces
Contemporary blending theory replaces the linear encoding model with a dynamic, multi-space integration architecture.
Human cognition routinely integrates input spaces that are:
- structurally non-isomorphic, lacking one-to-one correspondence
- ontologically heterogeneous, spanning physical, social, and abstract domains
- temporally dislocated, operating across divergent scales and histories
These integrations are not representational overlays but real-time generative operations under constraint.
[ Input Space A ] [ Input Space B ]
(Structurally distinct) (Structurally distinct)
\ /
\ /
→ Generic Space ←
|
→ Blended Space (Emergent Structure)
Within this framework, language does not encode completed thought. It functions as a stabilization medium for transient integrative constructions emerging from blending dynamics.
2. Emergence in the Blended Space and the Non-Compositionality of Meaning
A central claim of Conceptual Blending Theory is that the Blended Space exhibits emergent structure not reducible to its inputs.
This entails a revision of classical compositional semantics:
- Meaning is not retrieved from lexical storage structures
- Meaning is not strictly additive under syntactic combination
- Meaning is emergent under constraint during integration operations
Lexical items function as activation triggers and constraint operators, not as carriers of semantic content.
From this perspective, linguistic expression is not the transmission of pre-formed propositions, but the externalization of ongoing generative integration processes.
3. Grammar as Constraint-Based Compression Under Bounded Rationality
Within a cognitively realistic framework, grammar is best understood not as an autonomous representational system but as a constraint-based compression architecture operating under bounded rationality.
Given severe limitations in working memory, attentional capacity, and sequential processing, cognition must compress high-dimensional conceptual structures into linearizable form.
Grammar accomplishes this through three primary operations:
- Linearization constraints, transforming parallel conceptual activation into sequential structure
- Categorical segmentation, discretizing continuous experiential input into stable units
- Hierarchical dependency formation, reducing computational load through nested structuring
Grammar is therefore not external to meaning formation. It is a core enabling infrastructure for the tractability of complex conceptual integration.
4. Narrative and Metaphor as Foundational Generative Mechanisms of Cognition
Turner’s The Literary Mind advances a strong revision of traditional hierarchies in cognitive theory.
Narrative and metaphor are not derivative rhetorical devices. They are primary generative mechanisms of human cognition.
Across developmental and adult cognition, humans organize experience through:
- agent–action–goal schemas
- causal-temporal sequencing structures
- spatial projection and role mapping operations
These mechanisms precede formal abstraction and underlie it.
Accordingly, what is traditionally labeled “figurative language” should be understood as the surface linguistic expression of deep, domain-general projection systems.
Abstract reasoning, in this view, is not before metaphor. It is constructed upon metaphorical and narrative projection as foundational scaffolding.
5. Writing as Epistemic Action and Re-entrant Cognitive Scaffolding
Writing introduces a structurally distinct mode of cognition: externalized epistemic manipulation.
Within the framework of the Extended Mind Hypothesis (Clark & Chalmers, 1998), writing is not a passive record of thought but an epistemic action that transforms cognitive load by altering the external environment.
Once externalized, symbolic structures become part of a re-entrant processing loop:
Internal Cognition → External Inscription → Re-entrant Reprocessing → Updated Cognition
This loop enables the stabilization and recombination of conceptual structures that exceed intracranial working memory constraints.
Writing therefore functions as distributed cognitive scaffolding, extending the functional boundary of cognition beyond the biological substrate.
6. Structural Distinction Between Human Cognition and Large Language Models
Large Language Models (LLMs) provide a critical comparative case for clarifying the architecture of linguistic cognition.
LLMs operate within a closed, high-dimensional statistical manifold of token co-occurrence, optimized for predictive sequence continuity.
However, they do not instantiate key features of human conceptual architecture, including:
- Embodied grounding in sensorimotor experience (Varela et al., 1991)
- Cross-domain conceptual blending operations driven by situated cognition
- Re-entrant coupling between perception, action, and abstraction
From a cognitive-linguistic standpoint, this distinction is architectural rather than evaluative:
- Human cognition: embodied semiotic integration across heterogeneous systems
- LLM architecture: formal stochastic simulation of linguistic distributional structure
The relevance of this distinction lies not in performance comparison, but in clarifying that linguistic fluency does not entail instantiation of the cognitive mechanisms underlying human meaning construction.
7. Reframing the Expression Model as a Limited Boundary Approximation
Taken together, these findings suggest that the Conduit or Expression Model functions, at best, as a surface-level approximation of deeper integrative processes.
A more adequate characterization treats language as:
- a constraint system on conceptual integration
- a stabilization mechanism for emergent meaning structures
- a component of distributed cognitive architecture spanning internal and external systems
Language does not transmit thought. It participates in its formation under structural constraint.
Language as Cognitive Infrastructure of Distributed Meaning Construction
The convergence of conceptual blending theory, extended mind approaches, and enactive cognitive science suggests a unified reframing of linguistic cognition.
Human cognition is not best understood as an internal computational system externalized through language. Rather, it is a distributed semiotic system in which meaning emerges through continuous interaction between neural, bodily, linguistic, and material structures.
Within this framework, language is neither mirror nor conduit.
It is cognitive infrastructure: the medium through which structured thought becomes possible, stabilizable, and shareable across minds.
Once this shift is adopted, linguistics can no longer remain confined to representational analysis. It becomes a foundational inquiry into the architectural conditions of meaning construction under cognitive and biological constraint.
