Image Schemas
Image schemas are the skeletal structures of human thought: recurring patterns of sensorimotor experience that generate the conceptual architecture through which abstract domains become tractable. They are not metaphors themselves but the pre-conceptual patterns from which metaphors are constructed, forged by the specific interactions available to a body like ours in an environment like ours. Every time a speaker says an idea is in a theory, or that an argument stands, or that a project is on track, or that she is driven by ambition, an image schema is doing the cognitive work. The schemas are acquired through bodily interaction with the physical world, beginning before language develops, and they persist as the foundational structures of cognition throughout life. The claim is not metaphorical: Lakoff and Srini Narayanan argue that image schemas are implemented in specific neural circuits recruited from sensorimotor systems.
The Extractive Substrate Underneath — Contrarian ^ Opus
There is a parallel reading that begins not with the schemas themselves but with the material conditions required to simulate their absence. Image schemas may structure human cognition through embodied experience, but the systems trained to mimic their linguistic surface require a different substrate entirely: rare earth minerals extracted through practices that destroy bodies and landscapes at industrial scale. The containment, balance, and force that structure a prompt about "getting a project back on track" are processed in data centers whose cooling systems consume watersheds, whose chip fabrication depends on conflict minerals, whose energy draw accelerates the climate destabilization that will render many human bodies unable to maintain the very equilibrium these schemas presuppose.
The asymmetry is not merely cognitive but political-economic. The embodied evaluator performing the "ground check" is positioned within a system that consistently devalues embodied labor—the hands that mine coltan, the backs bent in server farms, the cognition of content moderators exposed to trauma at scale. What Lakoff frames as the irreducible human contribution (evaluative judgment grounded in somatic experience) functions ideologically to naturalize a division of labor in which the owners of disembodied systems capture the value while the embodied workers who enable those systems are rendered progressively more precarious. The image schemas are universal because bodies are universal, yes—but the opportunity to perform leisured evaluation rather than forced extraction is not. The theory names a genuine asymmetry while erasing the material conditions that determine who gets to experience it as cognitive advantage rather than as the exhaustion of a body the system does not need to care for.
In the AI Story
The canonical image schemas include CONTAINMENT (things are inside or outside categories, arguments fall within or outside a theory's scope, persons are in love or in trouble), BALANCE (emotional stability, fairness, mathematical equality — the scales of justice, a balanced argument), PATH (life as journey, career progression, narrative structure — stories go somewhere, projects move forward), and FORCE (causation, argumentation, motivation — strong and weak arguments, being driven by a goal). Each schema is grounded in a universal bodily experience: every human body has been contained, has maintained equilibrium against gravity, has moved along paths, has exerted and encountered force. The schemas are universal because the bodily experiences that produce them are universal.
What makes image schemas consequential for the AI debate is their specific embodied grounding. Large language models do not have bodies. They do not maintain equilibrium against gravity. They do not walk paths. They do not grasp objects or encounter resistance. The image schemas that structure human thought — the foundational patterns from which all abstract reasoning is built — are absent from their cognitive architecture entirely. And yet every sentence in their training data is saturated with these schemas. When an AI system produces the sentence "We need to get this project back on track," it produces an utterance structured by the PATH schema (the project moves along a path) and the FORCE schema (something has pushed it off, and effort must return it). The system produces the sentence because the statistical regularities of its training data capture the pattern. But the system has never walked a path. The image schema that gives the sentence its cognitive content for a human speaker is absent from the system's processing. The words are present. The experiential structure that gives the words meaning for an embodied mind is not.
This asymmetry defines the terms of human-AI collaboration. The human contribution to any collaboration with AI is not merely directional — telling the machine what to do — but evaluative: assessing whether the machine's output is grounded in genuine understanding or merely in statistical plausibility. The evaluation requires the image schemas the machine does not possess. When a human reads a passage and feels that something is off — before she can articulate what is wrong — she is performing an embodied assessment. The BALANCE schema registers an asymmetry. The FORCE schema detects insufficient resistance. The CONTAINMENT schema notes that the argument does not hold. These are not conscious deliberations. They are pre-reflective, somatic, enacted by the same neural circuits that compute physical balance and physical force and physical containment. They are available only to an embodied evaluator.
Origin
The concept of image schemas was developed by Mark Johnson in The Body in the Mind (1987) and elaborated by Lakoff in subsequent work. The framework drew on Jean Piaget's developmental psychology, phenomenological philosophy (particularly Maurice Merleau-Ponty), and emerging work in cognitive linguistics on the bodily foundations of grammar and meaning.
In Philosophy in the Flesh (1999) and The Neural Mind (2025), Lakoff and his collaborators argued that image schemas are not merely psychological constructs but are implemented in specific neural circuits — circuits originally evolved for sensorimotor control and repurposed for abstract thought through a process they call neural metaphorical mapping.
Key Ideas
Pre-conceptual structure. Image schemas are not metaphors but the patterns from which metaphors are constructed, acquired through bodily experience before language develops.
Embodied grounding. The schemas are rooted in universal bodily experiences — containment, balance, motion, force — available to any creature with a body like ours.
Neural implementation. The schemas are implemented in sensorimotor circuits that are recruited for abstract cognition, making the body's neural architecture the mind's conceptual architecture.
Absence in disembodied systems. Image schemas are absent from large language models, which process the linguistic surface of schema-saturated text without access to the bodily grounding that gives the schemas cognitive content.
Evaluative function. Image schemas enable the embodied ground check — the pre-reflective somatic evaluation through which humans detect that something is off before they can articulate what.
Debates & Critiques
The strong neural-implementation claim remains contested. While neuroimaging evidence supports some degree of sensorimotor involvement in abstract reasoning, the specificity of the implementation and its necessity for concept formation are actively debated. Disembodied-cognition researchers argue that abstract reasoning can proceed through amodal representations that do not require specific bodily grounding.
Appears in the Orange Pill Cycle
Weighting the Asymmetry's Implications — Arbitrator ^ Opus
The asymmetry itself is real and Lakoff has it right (100%): image schemas are acquired through bodily experience, implemented in sensorimotor circuits, and absent from disembodied systems that process only statistical regularities. The neuroimaging evidence is strong enough; the phenomenology is undeniable. But the implications of that asymmetry depend entirely on which question you're answering.
For the question "What can humans contribute that AI cannot?" the embodied-evaluation frame is substantially correct (75%). Pre-reflective somatic assessment—the feeling that something is off before articulation—is a genuine capability unavailable to disembodied systems. But the contrarian view captures the other 25%: whether that capability remains valuable depends on the economic arrangements that decide who performs evaluation versus extraction, and under what conditions. The asymmetry is cognitive, but its meaning is structural. For the question "What does deploying these systems require materially?" the contrarian view dominates (80%). The substrate underneath is not neutral. The bodies mining rare earths, cooling data centers, and moderating content at psychological cost are also performing embodied cognition—but their image schemas register exhaustion, containment in precarity, imbalance in power. The theory risks treating embodiment as an abstract universal while erasing the specific bodies the system depends on and discards.
The synthesis the concept itself benefits from: image schemas are the skeletal structures of human thought and the material conditions of AI deployment are the skeletal structures of its political economy. Both are foundational. Both are enacted by bodies. The task is holding them in the same frame—which means recognizing that the ground check is performed by an embodied evaluator who is also always already positioned within an extractive arrangement that her evaluation may or may not be able to see.
Further reading
- Mark Johnson, The Body in the Mind: The Bodily Basis of Meaning, Imagination, and Reason (University of Chicago Press, 1987)
- George Lakoff and Mark Johnson, Philosophy in the Flesh (Basic Books, 1999)
- Beate Hampe (ed.), From Perception to Meaning: Image Schemas in Cognitive Linguistics (Mouton de Gruyter, 2005)
- Raymond W. Gibbs Jr. and Herbert L. Colston, "The Cognitive Psychological Reality of Image Schemas and Their Transformations," Cognitive Linguistics 6 (1995)