In the spring of 1988, Fritz Strack, Leonard Martin, and Sabine Stepper published a study in the Journal of Personality and Social Psychology that seemed, at first, like a trick played on the participants by their own faces. The experiment was simple. Participants were asked to hold a pen in their mouths while rating the funniness of Gary Larson's Far Side cartoons. The critical manipulation was in how the pen was held: one group held it between their lips, which prevented the muscles used in smiling from contracting; the other group held it between their teeth, which forced those same muscles into a position functionally similar to a smile. Neither group was told anything about the pen's purpose or about facial expression. Yet the results were consistent with a prediction derived from the facial feedback hypothesis: participants who had unknowingly been caused to simulate smiling rated the cartoons as significantly funnier than those who had been caused to simulate the opposite expression. The face, apparently, was not merely reporting an emotion. It was helping to construct one.
The study was elegant and the implication unsettling to anyone committed to the standard picture of the mind as a kind of central executive that processes inputs, generates internal states, and then issues outputs through the body. On that picture, emotional experience is computed first, and the body's response follows. Strack, Martin, and Stepper appeared to have reversed the arrow. The body moved first; the experience followed. This was not quite a new idea — William James had argued in 1884 that we feel afraid because we run, not that we run because we feel afraid — but the experimental demonstration seemed to nail the point in a way that introspection and philosophy never could. The study became one of the most cited in social psychology, a touchstone for what would emerge over the following two decades as a major reformulation of the cognitive sciences: the theory of embodied cognition.
Twenty years later, in 2008, Lawrence Williams and John Bargh published a study in Science that extended the argument from emotion to social cognition. Participants were briefly asked to hold either a cup of hot coffee or a cup of iced coffee while riding an elevator with a confederate. They were then asked to rate the character of a person described in a biographical sketch. Those who had held the warm cup rated the stranger as significantly warmer — more generous, more sociable — than those who had held the cold cup. Physical thermal sensation, which has nothing logically to do with personality assessment, had nevertheless shaped a social judgment. The hand had reached into the mind and adjusted the thermostat on how another person was seen. Together, these two studies defined a research program. The body was not a vehicle for the mind. It was part of the mind.
Classical Cognition vs. Embodied Cognition: A Comparison
| Dimension | Classical (Disembodied) Cognition | Embodied Cognition |
|---|---|---|
| Cognitive architecture | Mind as symbol-processing system; brain operates like a computer executing amodal representations | Mind as sensorimotor system; cognition emerges from perception-action loops distributed across brain and body |
| Role of the body | Input device and output device only; peripheral to genuine cognition | Constitutive of cognition; bodily states, postures, and movements shape and partly construct thought |
| Language processing | Sentences decoded through amodal, syntactic rules; meaning stored as abstract propositions | Language comprehension activates perceptual and motor simulations; understanding "kick" activates motor regions for kicking |
| Emotional states | Emotions are internal states computed by neural circuits; bodily expression follows from the state | Bodily states and expressions partly constitute emotional experience; altering the body alters the emotion |
| Memory | Stored as abstract symbolic records independent of the original sensorimotor experience | Memory is re-enactment; remembering an event involves partially re-running the perceptual-motor simulation of it |
| Perception-action coupling | Perception and action are separate modules; perception feeds into cognition which produces action plans | Perception and action are deeply coupled; perceiving an object activates affordances for acting on it |
Intellectual Lineage: Who Influenced Whom
The embodied cognition program did not emerge from experimental psychology alone. Its deepest intellectual roots lie in phenomenological philosophy, and tracing those roots clarifies what makes the position genuinely radical rather than merely a collection of interesting priming effects.
Edmund Husserl, writing in the early decades of the twentieth century, had argued that perception was always perception-from-a-perspective, shaped by the embodied position and attentional orientation of the perceiver. His student Maurice Merleau-Ponty extended this into a comprehensive philosophy of embodiment in Phenomenologie de la Perception (1945). For Merleau-Ponty, the body was not an object in the world that the mind happens to occupy. It was the very medium through which a subject has a world at all. The body is not experienced from outside; it is the lived standpoint from which all experience is structured. A hand reaching for a glass does not consult a mental map of space; it knows where the glass is in the way that a skilled musician knows where the notes are — through a kind of practical, pre-reflective bodily knowledge he called the body schema. This was philosophy, not psychology, and for decades it circulated largely among humanists and phenomenologists rather than cognitive scientists.
The bridge from phenomenology to cognitive science was built, in large part, by the linguists George Lakoff and Mark Johnson. Their 1980 book Metaphors We Live By made the audacious argument that abstract thought is not merely accompanied by metaphor but constituted by it, and that these conceptual metaphors are themselves grounded in basic bodily experience. The concept of argument, for instance, is structurally organized by the image schema of spatial conflict — arguments are won and lost, positions are defended and attacked, claims are knocked down. The concept of time is organized by the image schema of movement along a path. Anger is organized by the experience of physical pressure in a container. This was not a literary observation. Lakoff and Johnson argued that the bodily image schemas — container, path, force, balance, up-down orientation — were the fundamental building blocks from which all abstract conceptual structure was assembled, including formal reasoning, moral judgment, and scientific theory.
The implications were elaborated and radically extended in their 1999 book Philosophy in the Flesh, which declared that the mind was "inherently embodied," that reason was not the disembodied, context-independent faculty that Western philosophy from Plato to Kant had assumed, and that the entire tradition of Anglo-American analytic philosophy rested on a false picture of how cognition works. These were large claims. Critics argued, with some justice, that Lakoff and Johnson's framework was more interpretively flexible than empirically testable — any conceptual domain could be mapped onto some bodily image schema post hoc, making the theory difficult to falsify. The linguistic evidence was real but the link from linguistic structure to cognitive architecture required independent empirical support that the linguistic data alone could not provide.
That support came from an unexpected direction: cognitive neuroscience and experimental psychology. Lawrence Barsalou, a psychologist at Emory University (later at the University of Chicago), published a landmark theoretical paper in 1999 in Behavioral and Brain Sciences introducing perceptual symbol systems theory. Barsalou argued, against the dominant connectionist and classical frameworks, that cognitive representations are not amodal — that is, they are not stripped of the sensorimotor content of the original experience and re-encoded in a neutral symbolic format. Instead, Barsalou proposed that the mind stores and retrieves knowledge using partial re-activations of the perceptual and motor states that were originally active during the relevant experience. To think about a hammer is to partially re-run the visual, tactile, and motor simulations associated with hammers. Concepts are not definitions stored in a mental dictionary. They are flexible, context-sensitive simulations run in sensorimotor systems.
Barsalou's framework was not merely speculative. Neuroimaging data were beginning to accumulate that supported the core claims. Studies showed that reading action verbs activated motor cortex regions involved in executing those actions (Hauk, Johnsrude, and Pulvermuller, 2004, Neuron). Reading emotion words activated regions associated with experiencing those emotions. The convergence between phenomenological philosophy, Lakoff and Johnson's linguistic theory, and Barsalou's perceptual symbol systems model produced a coherent, multi-level program that the cognitive sciences could not dismiss as merely humanistic speculation.
Antonio Damasio's somatic marker hypothesis, developed throughout the 1990s and articulated in Descartes' Error (1994) and The Feeling of What Happens (1999), provided another independent strand of convergent evidence. Damasio's research on patients with damage to ventromedial prefrontal cortex — particularly the famous case of Phineas Gage and the more extensively documented case of the patient "Elliot" — showed that individuals who lost the capacity to integrate bodily signals into decision-making became profoundly impaired in practical reasoning and social judgment despite retaining normal intelligence and factual knowledge. The somatic marker hypothesis proposed that the body's affective signals — subtle physiological states that have been associated with previous outcomes — serve as heuristic guides in decision-making under uncertainty. Far from being noise that rational cognition filters out, the body's signals are an indispensable part of the computational process.
Giacomo Rizzolatti and his colleagues at the University of Parma discovered mirror neurons in the macaque premotor cortex in a series of papers in the 1990s (including Rizzolatti, Fadiga, Gallese, and Fogassi, 1996, Cognitive Brain Research). These neurons fired both when the monkey performed a specific hand action and when the monkey observed another individual performing the same action. The discovery suggested a neural mechanism for understanding the actions of others through internal simulation — the observer's motor system partially enacts what it perceives, and this motor enactment is the substrate of comprehension. Vittorio Gallese and Alvin Goldman developed this into a simulation theory of mind-reading: we understand others not by applying a theoretical model but by running motor simulations of their behavior in our own sensorimotor systems. The social world, on this view, is understood through embodied resonance rather than through inference from a detached observer's position.
The Empirical Foundation
Case Study 1: Glenberg and Kaschak — Action-Based Language Comprehension (2002)
Arthur Glenberg and Michael Kaschak published what became known as the Action-Compatibility Effect (ACE) in 2002 in Psychonomic Bulletin and Review. Participants were asked to judge whether sentences made sense, and to respond by moving a joystick either toward or away from their bodies. The critical manipulation involved the directionality of actions described in the sentences: some sentences described transfer of objects toward the reader ("Andy delivered the pizza to you"), while others described transfer away from the reader ("You delivered the pizza to Andy"). The finding was that participants were faster to respond when the direction of the joystick movement was compatible with the direction implied by the sentence. Understanding a sentence describing incoming motion was easier when the response required an incoming movement, and vice versa.
This was not a subtle statistical effect. It was a direct demonstration that comprehending language activates the motor systems used to perform the described actions — and that these motor activations are obligatory, not optional, occurring even when participants were just trying to determine whether a sentence was grammatical. Comprehension, Glenberg and Kaschak argued, is not the extraction of an abstract proposition from a string of words. It is the construction of an action simulation. The effect has been replicated across multiple laboratories and extended to a range of linguistic constructions, though the magnitude of effects varies considerably across studies and populations.
Case Study 2: Niedenthal et al. — Embodied Emotion (2005)
Paula Niedenthal and colleagues, in a series of studies published across several papers and synthesized in a major review in Science in 2007, demonstrated that emotional understanding is grounded in the perceiver's own bodily and emotional systems. In one key paradigm, participants were asked to judge whether two facial expressions depicted the same or different emotions. When participants were required to hold a pen between their teeth — preventing smiling — they were slower and less accurate at recognizing happy expressions than participants who held the pen in their fingers or did not hold a pen at all. The muscle contraction that is part of smiling, when mechanically blocked, impaired the recognition of smiles in others.
Niedenthal's framework, which she called Embodied Simulation of Emotion, proposed that we understand the emotional states of others by partially replicating those states in our own facial, postural, and visceral systems. This is not a deliberate process. It is automatic. When the body is constrained so that it cannot simulate, emotional recognition suffers. The implications extend beyond the laboratory: if emotional understanding depends on bodily simulation, then anything that disrupts the body's capacity to resonate with another's expression — chronic pain, medication that blunts motor responses, Botox injections that prevent facial movement — may also impair aspects of emotional social cognition.
Case Study 3: Zhong and Liljenquist — The Macbeth Effect (2006)
Chen-Bo Zhong and Katie Liljenquist published a study in Science in 2006 that remains one of the most striking demonstrations of the intersection between physical experience and moral cognition. In their paradigm, participants who were asked to recall a past moral transgression subsequently expressed stronger desire for cleaning products — soap, toothpaste, antiseptic — than participants who recalled a morally neutral past behavior. In a follow-up experiment, participants who were allowed to clean their hands with an antiseptic wipe after recalling a transgression subsequently showed reduced willingness to help with a good cause, compared to participants who had recalled the transgression but were not allowed to clean.
The pattern suggested that physical cleanliness and moral purity are not merely metaphorically linked in language — "clean conscience," "dirty hands," "washing away one's sins" — but cognitively linked at the level of representation. The physical sensation of being clean partially satisfies the psychological need to feel morally clean, and vice versa. Zhong and Liljenquist named the phenomenon the Macbeth effect after Shakespeare's Lady Macbeth, who obsessively washed her hands after instigating murder. The study was remarkable for what it implied about the depth of embodied metaphor: abstract moral self-evaluation was sharing computational machinery with the sensory system that monitors bodily cleanliness.
Case Study 4: Williams and Bargh — Warm Coffee and Social Warmth (2008)
The Williams and Bargh (2008) study in Science mentioned in the opening is worth examining in greater detail. The warm-cold temperature manipulation was not incidental to social cognition — it directly activated a conceptual schema that had been built, the researchers argued, from early developmental experience in which physical warmth and social warmth co-occurred. A held and comforted infant experiences both physical warmth and the social warmth of caregiving simultaneously, repeatedly, across the critical developmental period. These co-occurring experiences, Lakoff and Johnson's framework would predict, generate an embodied conceptual metaphor in which warmth and social positivity share representational structure.
The study reported two experiments. In the first, the brief elevator encounter with either a hot or iced coffee cup influenced personality judgments of a described stranger. In the second, participants who were told they had been evaluated as having a "warm" or "cold" personality were more likely to choose a reward for a friend versus a reward for themselves after the "warm" condition — the physical warmth prime, in other words, appeared to activate prosocial motivations. The study was widely cited and highly influential.
Limits, Critiques, and Nuances
The embodied cognition program has not escaped serious challenge, and a responsible account of the field must engage directly with three distinct categories of criticism: philosophical, methodological, and empirical.
The philosophical critique targets Lakoff and Johnson most directly. Their claim that all abstract thought is grounded in bodily image schemas is, as several philosophers of language noted, extremely difficult to evaluate empirically. The theory has enormous post hoc explanatory range — given any abstract conceptual domain, a resourceful interpreter can identify an image schema that appears to organize it — but this flexibility makes it poorly equipped to generate specific, falsifiable predictions. Philosophers of mind including Jerry Fodor and Zenon Pylyshyn pointed out that the mere existence of metaphorical language grounded in bodily experience does not demonstrate that bodily experience constitutes the conceptual structure itself. Linguistic convention and cognitive architecture are not the same thing. The inferential gap between "English speakers use spatial metaphors for argument" and "all reasoning is grounded in sensorimotor systems" is wide, and Lakoff and Johnson did not close it by identifying the gap as their opponents' philosophical prejudice.
The methodological critique is more technically specific and ultimately more damaging to particular lines of evidence. The replication crisis in social psychology hit the embodied cognition literature with particular severity. In 2016, a massively collaborative project — the Many Labs 2 project, led by Richard Klein and colleagues and published in Advances in Methods and Practices in Psychological Science in 2018 — attempted to replicate 28 effects across 125 samples in 36 countries. Several embodied effects failed to replicate. Eerland, Guadalupe, and Zwaan (2011) had published a study showing that leaning slightly to the left caused people to give lower numerical estimates — a left-lean affecting leftward-on-the-mental-number-line representations. The Many Labs 2 replication found no evidence for this effect across its large international sample.
Most significantly for the field, the Strack, Martin, and Stepper (1988) facial feedback study — the cornerstone of embodied emotion research — was directly targeted by a 17-laboratory replication project organized by E.-J. Wagenmakers and colleagues, published in Perspectives on Psychological Science in 2016. The replication used a critical modification: because the original study was conducted before participants could be observed by cameras, the replication had cameras present (a consequence of recording the study on video). The replication found essentially no evidence of the original effect. The weighted average effect size across the 17 laboratories was near zero, and the confidence interval excluded the original effect size. Strack offered the explanation that camera presence disrupted the natural facial feedback mechanism through self-conscious awareness of being observed. Wagenmakers and colleagues suggested that the original effect was a false positive. The debate has not been fully resolved, but the replication failure drew necessary attention to the fragility of many priming effects in social psychology, including embodied ones.
The distinction between strong and weak versions of embodied cognition is essential for evaluating what the field can still credibly claim. The strong version holds that abstract cognition is constitutively embodied — that the body is not merely influencing cognition but is part of the cognitive process itself, such that cognition could not occur without the relevant bodily involvement. The weak version holds that body states typically influence cognition, as contextual factors, without being necessary to it. Most of the priming paradigms — warm coffee affecting social warmth ratings, posture affecting emotional tone — are evidence for the weak version at best. They demonstrate that bodily states can influence cognitive outputs without demonstrating that the body is part of the cognitive machinery in any deep architectural sense. The strong version remains philosophically interesting but empirically under-supported by the experimental literature.
The mirror neuron story similarly requires qualification. Initial enthusiasm over their discovery led to sweeping claims about their role in language, autism, social cognition, and cultural transmission. Subsequent research has complicated the picture considerably. Mirror neurons were originally discovered in macaques, and while there is neuroimaging evidence for mirror-like properties in human premotor regions, the direct recording studies that established the macaque phenomenon have not been replicated in humans under equivalent conditions. Furthermore, the inference from "a neuron fires both when I act and when I observe you act" to "motor simulation is the basis of social understanding" requires several additional steps that are not securely established. Critical reviews by Hickok (2009, Trends in Cognitive Sciences) argued that the evidence for simulation-based understanding of actions was more equivocal than the popular account suggested, and that abstract semantic representations could account for much of the same data.
What survives these critiques is not nothing. The Glenberg and Kaschak ACE effect has a reasonable replication record. The neuroimaging evidence for motor involvement in language processing has been confirmed across multiple laboratories using different methodologies. Niedenthal's basic findings on embodied emotion recognition are supported by converging evidence from clinical populations as well as experimental manipulations. Damasio's somatic marker hypothesis is grounded in clinical neuropsychology with strong ecological validity. The thesis that cognition is entirely independent of bodily states — the position embodied cognition was reacting against — is almost certainly false. The thesis that cognition is constitutively and architecturally embodied in the strong sense remains contested.
What the Research Shows
The most defensible conclusion from three decades of embodied cognition research is that the classical picture of cognition as purely abstract, amodal symbol manipulation is empirically inadequate. Perception is not passive reception of signals but active anticipation shaped by motor systems and prior action. Language comprehension activates sensorimotor simulations that are not epiphenomenal but functionally relevant to processing. Emotional experience involves bodily states as components rather than merely correlates. Decision-making integrates somatic signals in ways that purely deliberative accounts cannot explain.
These are not small revisions to the classical picture. They imply that a cognitive science which models the mind solely through its computational relations to abstract symbols and ignores the body will systematically mispredict human cognition in important domains. At the same time, the field has learned through painful replication failures that the priming effects used to demonstrate embodied influences on cognition are smaller, less reliable, and more methodologically sensitive than the initial wave of research suggested. The Macbeth effect, the warm coffee effect, the leaning-left effect — each of these requires far more careful examination before it can serve as foundational evidence for strong claims about embodied architecture.
The field is in a productive transitional moment. The intellectual framework provided by Merleau-Ponty, Lakoff and Johnson, and Barsalou remains stimulating and generative. The empirical research it inspired has identified genuine phenomena. But the program requires more rigorous experimental methods, pre-registration of hypotheses, and careful separation of the strong and weak claims about embodiment's role in cognitive architecture. The mind is not imprisoned in the body, looking out through sensory apertures. But neither is the body a mere passenger. The question now is to specify, with real precision, where the boundary falls and what it means.
References
Strack, F., Martin, L. L., & Stepper, S. (1988). Inhibiting and facilitating conditions of the human smile: A nonobtrusive test of the facial feedback hypothesis. Journal of Personality and Social Psychology, 54(5), 768-777.
Williams, L. E., & Bargh, J. A. (2008). Experiencing physical warmth promotes interpersonal warmth. Science, 322(5901), 606-607.
Lakoff, G., & Johnson, M. (1980). Metaphors We Live By. Chicago: University of Chicago Press.
Lakoff, G., & Johnson, M. (1999). Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. New York: Basic Books.
Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22(4), 577-660.
Glenberg, A. M., & Kaschak, M. P. (2002). Grounding language in action. Psychonomic Bulletin and Review, 9(3), 558-565.
Zhong, C.-B., & Liljenquist, K. (2006). Washing away your sins: Threatened morality and physical cleansing. Science, 313(5792), 1451-1452.
Niedenthal, P. M. (2007). Embodying emotion. Science, 316(5827), 1002-1005.
Rizzolatti, G., Fadiga, L., Gallese, V., & Fogassi, L. (1996). Premotor cortex and the recognition of motor actions. Cognitive Brain Research, 3(2), 131-141.
Wagenmakers, E.-J., Beek, T., Dijkhoff, L., Gronau, Q. F., Acosta, A., Adams, R. B., Jr., ... & Zwaan, R. A. (2016). Registered replication report: Strack, Martin, and Stepper (1988). Perspectives on Psychological Science, 11(6), 917-928.
Damasio, A. (1994). Descartes' Error: Emotion, Reason, and the Human Brain. New York: Putnam.
Hauk, O., Johnsrude, I., & Pulvermuller, F. (2004). Somatotopic representation of action words in human motor and premotor cortex. Neuron, 41(2), 301-307.
Frequently Asked Questions
What is embodied cognition?
Embodied cognition is the view that cognitive processes — perception, memory, language comprehension, emotion — are fundamentally grounded in and shaped by bodily experience. Unlike classical cognitive science, which treats the brain as a disembodied processor, embodied cognition holds that the body's sensorimotor systems actively participate in thought. Lakoff and Johnson's 1980 'Metaphors We Live By' and 1999 'Philosophy in the Flesh' are the foundational philosophical statements.
What did the pen-in-mouth study show?
Strack, Martin, and Stepper (1988) found that participants holding a pen between their teeth (inducing a smile-like muscle state) rated Far Side cartoons as funnier than those holding it between their lips (preventing smiling). This facial feedback effect supported the idea that the body contributes to emotional experience. However, a 2016 large-scale multi-lab replication (Wagenmakers et al.) found no evidence of the effect, illustrating the replication challenges facing embodied cognition research.
What is the Macbeth effect?
Zhong and Liljenquist (2006) found that recalling an unethical deed increased desire for cleansing products (soap, toothpaste), and that physical cleansing reduced moral distress — suggesting physical and moral purity share representational structure. This 'Macbeth effect' became a flagship embodied cognition finding, though subsequent replications have produced mixed results.
How does embodied cognition explain language comprehension?
Glenberg and Kaschak's (2002) ACE (Action-sentence Compatibility Effect) showed that sentences implying movement toward or away from the reader affected response time when participants moved their hand toward or away from themselves. Hauk, Johnsrude, and Pulvermuller (2004) showed in fMRI that reading words for arm, leg, and face actions activated corresponding motor cortex regions — supporting simulation-based accounts of language comprehension.
What are the main criticisms of embodied cognition?
Lakoff and Johnson's framework has been criticized for unfalsifiability — any metaphor can be post-hoc labeled as embodied. The replication crisis hit embodied social priming hard: Many Labs 2 found near-zero replication of Macbeth effect and other social priming effects. Hickok's (2014) 'The Myth of Mirror Neurons' challenged the mirror neuron interpretation. Researchers now distinguish strong embodiment (body constitutes cognition) from weak embodiment (body influences cognition) — only the latter has strong empirical support.